Tracking and Control in Multi-Function Radar: by Joseph Mackay Butler

TRACKING AND CONTROL IN
MULTI-FUNCTION RADAR
by
Joseph Mackay Butler
A thesis submitted to the University of London for the Degree of
Doctor of Philosophy in Electronic Engineering
Department of Electronic and Electrical Engineering
University College London
August 1998
(LONDON)
ABSTRACT
The phased array multi-function radar is an effective solution to the requirement for
simultaneous surveillance and multiple target tracking. However, since it is performing
the jobs usually undertaken by several dedicated radars its radar time and energy
resources are limited. For this reason, and also due to the large cost of active phased
array antennas, it is important for the strategies adopted in the control of the radar to be
efficient. This thesis investigates and develops efficient strategies for multi-function
radar control and tracking. Particularly the research has focused on the use of rotating
array antennas and simultaneous multiple receive beam processing.
The findings of the research challenge the traditional view that three or four fixed
(static) array faces is the best antenna configuration for a multi-function radar system.
By developing novel methods for the comparison of systems utilising different antenna
configurations it is shown that a rotating array multi-function radar performs the
surveillance function with a greater efficiency in its use of radar time than a static array
system. Also, a rotating array system benefits from the ability to distribute the radar
resources over the angular coverage in a way that is impossible with a static array
system. A novel strategy is presented to achieve this, which allows the rotating array
system to better support the realistic situation of a high concentration of radar tasks in a
narrow angular sector.
It is shown that the use of broadened transmit beams coupled with simultaneous
multiple narrow receive beams can eliminate the compromise on radar beamwidth
between the surveillance and tracking functions that is associated with multi-function
radars. This technique would allow construction of multi-function radar systems with
narrow beamwidths, giving improved tracking performance, without extending search
frame times excessively.
Efficient tracking strategies for both static array and rotating array multi-function radars
are developed. They are applied through computer simulation to demonstrate tracking of
highly manoeuvrable targets with a narrow beam multi-function radar. Track robustness
is attained through the use of multiple beam track updating strategies at little cost in
terms of radar time.
2
To my Mother and Father.
ACKNOWLEDGEMENTS
Firstly, my thanks must go to Professor H.D. Griffiths. His help and supervision
throughout this research have been unending, and he has made my stay at UCL both
pleasant and memorable. Thanks too must go to Dr A.R. Moore and Mr W.A. Levett
from DERA Portsdown, who have aided and encouraged me since I first entered the
field of radar. Without their active support, completion of this research would not have
been possible. I would also like to express my gratitude to Professor K. Milne and
Professor R. Benjamin for their expert advice and stimulating discussions during their
visits to UCL.
Also at DERA Portsdown, thanks are due to G. Rhees, E. Brice, P. Donnely, and in
particular to M. Deighton and P. Clements whose help has gone beyond that which
could have been expected. At BAeDS my thanks go to W.Stafford and S.Noyes.
Finally, I would like to give thanks to the friends and colleagues who have contributed
either directly or indirectly to this thesis, and have put up with me throughout it.
Particularly thanks go to R. Bullock, G. Davidson, W. Lee, H. Read and L. Vinagre
from the Antennas and Radar Group at UCL, C. Andrew, R. Maunder, A. Ridehalgh and
F. Read, for their complete lack of interest in the field of radar and the healthy antidote
which they provided, my ever patient family E. Butler, J. Butler, E. Cooper and D.
Cooper, and last to V.A. Byram who has proof read this thesis, and has helped more and
put up with more than all of the above.
4
MULTI-FUNCTION RADAR TRACKING AND CONTROL
TABLE OF CONTENTS

1 INTRODUCTION 17

I.! Historical background 17

1.2 Objectives 18

1.3 Thesis Layout 19

1.4 Novel Aspects 21

2 AN OVERVIEW OF MULTI-FUNCTION RADAR 22

2.1 Background 22

2.2 Multi-function radar functions 23

2.2.1 Search 24

2.2.2 Plot confirmation 25

2.2.3 Track initiation 25

2.2.4 Track maintenance 26

2.2.5 Functional flexibility 27

2.3 Underlying theory of multi-function radar 27

2.3.1 Radar systems 27

2.3.2 Phased arrays 34

2.3.3 Multi-function radar system aspects 39

2.4 Benefits of using a multi-function radar 44

2.5 Penalties of using a multi-function radar 45

2.6 Static versus rotating systems 47

3 MESAR: AN EXAMPLE MULTIFUNCTION RADAR SYSTEM 49

3.1 Introduction 49

3.2 Array face and transmit/receive modules 51

3.3 MESAR adaptive beamforming 52

3.4 Radar control 53

3.5 Adaptive signal processing and waveform generation 54

3.6 The plot processor and tracker 55

3.6.1 Plot processing 55

3.6.2 Track filtering 55

3.6.3 Adaptive track update rate 56

3.6.4 Adaptive transmit energy 57

3.6.5 MTI velocity response matching 57

3.7 Trials results 57

3.7.1 Multi-function operation 58

3.7.2 Tracking 58

3.8 Summary 60
4 THE IMPACT OF VARIABLE BEAM WIDTH AND

MULTIPLE BEAM
PROCESSING ON MULTI-FUNCTION RADAR 61

4. I Introduction 61

4.2 Generating a broadened main lobe antenna pattern 62

4.3 Processing of simultaneous multiple receive beams 64

4.4 Calculation of the search function time 65

4.4.1 Static phased array multi-function radar system 66

4.4.2 Rotating phased array multi-function radar system 71
5

4.5 Search time in a noise limited environment 76

4.5.1 Dwell time compensation 79

4.6 Search time in a clutter limited environment 82

4.7 Impact on the tracking functions 85

4.8 Conclusions 86
5 RESOURCE MANAGEMENT IN STATIC AND ROTATING ARRAY

MULTIFUNCTION RADAR 89

5. 1 Introduction 89

5.2 Resource management and task scheduling objectives 91

5.3 Task scheduling in a static array multi-function radar 92

5.3.1 Background 92

5.3.2 The MESAR algorithm 93

5.3.3 The modified MESAR algorithm 97

5.3.4 Simulation architecture 98

5.3.5 Scheduling using a simple two sector surveillance system 100

5.3.6 Scheduling using the MESAR surveillance volume 104

5.3.7 Plot confirmation latency using the MESAR scheduler 109

5.4 Task scheduling in a rotating array multi-function radar 110

5.4.1 Background 110

5.4.2 An algorithm for task scheduling in a rotating multi-function radar system 114

5.4.3 Beam search patterns with a rotating multi-function radar 118

5.4.4 Results with the rotating multi-function radar task scheduler algorithm 119

5.4.5 Other resource management issues for a rotating multi-function radar 123

5.5 Penalty functions and fuzzy logic for efficient scheduling 124

5.5.1 The use of fuzzy logic 125

5.6 Conclusions 126

6 ADAPTIVE TRACKING IN MULTIFUNCTION RADAR 129

6. 1 Introduction 129

6.2 Tracking of manoeuvring targets with a narrow beam adaptive tracker 130

6.2.1 Simplified analysis of tracking manoeuvring target with a narrow beamwidth 131

6.3 Allocation of resources in tracking for a static array system 134

6.3.1 Allocation of resources for tracking in MESAR 135
6.3.2 Parameter optimisation for allocation of resources in the ELRA adaptive tracking

system 136

6.3.3 Comparison of the two strategies 139

6.4 Development of an adaptive tracking system simulation model 140

6.4.1 Modelling of the target measurement process 140

6.4.2 Kalman filtering 141

6.4.3 Augmenting the constant velocity Kalman filter for manoeuvring targets 145

6.4.4 Adaptive update rate strategy 150

6.4.5 Adaptive transmitted waveform energy 152

6.4.6 Missed detections and declaration of lost track 152

6.4.7 Development of a multiple beam track update strategy 152

6.4.8 Development of a rotating array track update strategy 154

6.4.9 Validation and testing of the adaptive tracking system simulation model 158

6.5 Results 159

6.5.1 A summary of the results collected 159

6.5.2 Single beam static array multi-function radar tracker 161

6.5.3 Multiple beam static array multi-function radar tracker 167

6.5.4 Single beam rotating array multi-function radar tracker 171

6.5.5 Multiple beam rotating array multi-function radar tracker 177

6.5.6 Comparison of static and rotating array tracker performance 182

6.6 Conclusions 184
6
7 A ROTATING, NARROW BEAM WIDTH MULTI-FUNCTION RADAR
SYSTEM 186
7. I Introduction 186
7.2 Requirement 187
7.3 Antenna configurations considered 188
7.4 Analysis of time budget and free space detection performance 189
7.4.1 Surveillance 189
7.4.2 Tracking 192
7.4.3 Summary 196
7.5 Conclusions 197

8 CONCLUSIONS AND FURTHER WORK 199
8.1 Summary 199

8.2 Evaluation of static array, rotating array and variable beam pattern multi-function radar 201
8.3 Resource management of static and rotating array multi-function radar 203
8.4 Adaptive tracking in multi-function radar 206

9 REFERENCES 208
APPENDIX A: THE BENCHMARK TRACKING PROBLEM 214
APPENDIX B: STATIC ARRAY TRACKER MONTECARLO

SIMULATION RESULTS 219
APPENDIX C: ROTATING ARRAY TRACKER MONTECARLO

SIMULATION RESULTS 227
7
LIST OF FIGURES
Figure 2-1 Functions of a multi-function radar for naval air defence ____________________________ 24
Figure2-2 The closed loop multi-function radar tracker_______________________________________ 26
Figure2-3 A simplified block diagram of a typical radar ______________________________________ 28
Figure 2-4 Monopulse antenna patterns and error signals. (a) overlapping antenna patterns, (b)
difference antenna pattern, (c) sum antenna pattern, (d) error signal _______________________ 31
Figure 2-5 Weighting functions for minimum monopulse estimation error, and the associated beam
patterns 32
Figure 2-6(a,b) A 2 dimensional Taylor weighting function, and its associated antenna pattern ______ 33
Figure 2-7 A Bayliss weighting function in one dimension and a Taylor weighting function in the other,
andits associated antenna pattern ___________________________________________________ 33
Figure 2-8 Generation of a phase front with no relative change in phase between each element ______ 36
Figure 2-9 Generation of a phase front with a constant phase change between adjacent elements 36
Figure2-10 Change in effective aperture with scan angle______________________________________ 37
Figure 2-1J(a,b) Beam broadening, loss in signal to noise ratio, and loss in monopulse accuracy as a
functionof scan angle ___________________________________________________________ 39
Figure 2-12 A cross section through a hemispherical search volume, showing sectored coverage with
regionnumber and frame time ______________________________________________________ 41
Figure 2-13(a-e) Angular and temporal variation in multi-function radar detection performance_____ 41
Figure2-14 A digital beamformer ______________________________________________________ 43
Figure3-1 The MESAR multi-function radar______________________________________________ 50
Figure 3-2 A simplified block diagram of the MESAR system _________________________________ 51
Figure3-3 The thinned MESAR array ___________________________________________________ 51
Figure 3-4 A simplified block diagram of the MESAR digital adaptive beamformer________________ 53
Figure 3-5 Adapted and unadapted beam patterns in the presence ofjammers____________________ 58
Figure 3-6 The MESAR PPI display, showing a recently initiated track_________________________ 59
Figure3-7 Adaptive update rate tracking ________________________________________________ 59
Figure3-8 Track splitting function________________________________________________________ 60
Figure 4-I(a,b,c,d) Synthesis of a broadened main lobe antenna pattern using Milne's method, showing
broadening of the scan pattern main lobe by (a) a factor of 2, (b) afactor of 4, (c) afactor of 6, (d)
a factor of 8. Blue line shows the scan pattern, red line shows the desired pattern, green line shows
therealisable pattern. ____________________________________________________________________ 63
Figure 4-2 Example tx and rx beam formations using simultaneous multiple beam processing _______ 64
Figure4-3 The UV space coordinate system and equations_____________________________________ 67
Figure 4-4 (a) A visualisation of UV space, showing beam broadening, (b) the same visualisation
showinga triangular lattice of beams_________________________________________________ 67
Figure 4-5 The relationship between the uv and real world coordinate systems, and the associated
equations 68
Figure 4-6(a, b) A comparison of lines of constant azimuth and elevation for (a) a vertical array, and (b)
anarray inclined 15° back from vertical_____________________________________________ 69
Figure 4-7 Beam directions in UV space, using MESAR as an example _________________________ 69
Figure 4-8 (a) MESAR coverage limits in UV space, (b) Beam dwell time plotted as height above the UV
plane 70
Figure 4-9 (a) Number of beams, and (b) average beam dwell times as a function of beamwidth. Crosses
represent answer from exact calculation of all beam directions and dwell times, the line represents
theestimate from the UV method outlined above ________________________________________ 71
Figure 4-10 A visualisation of the rotating UV space, and the associated equations ________________ 72
Figure 4-11 Lines of constant azimuth and elevation in the rotating UV space, for a vertical array 73
Figure 4-12 (a, b, c) Lines of constant azimuth and elevation in the rotating UV space, for arrays inclined
backfrom vertical (a)30°, (b)60°, (d)90° ____________________________________________ 73
Figure 4-13 Beam directions in rotating UV space, using a MESAR search configuration as an example74
Figure 4-14 (a) MESAR coverage limits in rotating UV space, (b) Beam dwell time plotted as height
abovethe rotating UVplane ______________________________________________________ 75
Figure 4-15 (a) Number of beams, and (b) average beam dwell times as afunction ofbeamwidth,fora
rotating array multi-function radar. Crosses represent answer from exact calculation of all beam
directions and dwell times, the line represents the estimate from the rotating UV method outlined
8
above 76
Figure 4-16 The example surveillance coverage volume for comparison of different multi-function radar
antennaconfigurations ________________________________________________________________ 76
Figure 4-17 (a,b,c) 0'-60° elevation search sector plotted in (a) UV space for a 4faced multi-function
radar scanning to ±450, (h) UVspace for a 3faced multi-function radar ±450, (c) Rotating UV
space, applicable to a 1,2,3 or 4 faced multi-function radar ______________________________ 77
Figure 4-18 (a,b) Number of beams required to search the specified volume as a function of beamwidth,
andthe comparison of the static and rotating systems ___________________________________ 78
Figure4-19 (a,b) Average dwell times and comparison______________________________________ 78
Figure 4-20 (a,b) Search function time and comparison ___________________________________ 79
Figure 4-21 (a,b,c) Signal to noise ratio and Probability of detection as a function of beamwidth, with
broadening of both the transmit and receive beams, using the nominal Ims waveform above - ie not
compensatingfor the loss in two way gain ___________________________________________ 80
Figure 4-22 (a,b,c) Search function time required to maintain a uniform detection performance with
broadenedtransmit and receive beams ________________________________________________________ 81
Figure 4-23 (a,b) Signal to noise ratio and Probability of detection as a function of beamwidth, with
broadening of only the transmit beam, using the nominal Ims waveform above - i.e. not
compensating for the loss in one way gain _____________________________________________ 81
Figure 4-24 Search function time required to maintain a uniform detection performance using a
broadened transmit beam and simultaneous multiple narrow receive beams__________________ 82
Figure 4-25(a,b) Contamination of Doppler filters by clutter __________________________________ 84
Figure 4-26 An example of the use of single beam and multiple beam updates in the surveillance and
tracking functions 86
Figure 5-1 Flow diagram of the MESAR scheduler algorithm_______________________________ 95
Figure 5-2 Degradation of surveillance detection performance ________________________________ 96
Figure 5-3 Static multi-function radar task scheduling simulation architecture ____________________ 100
Figure 5-4(a,b) Beam directions plotted in UV space, and beam dwell times plotted above the UVplanelOO
Figure 5-5 Surveillance time balances as a function of time, with no target detections _____________ 100
Figure 5-6(a,b) Achieved averaged occupancies and frame times over the simulation period _______ 100
Figure 5-7 Beam dwell times plotted above the UVplane for the 2 region surveillance volume______ 101
Figure 5-8 Surveillance time balances as a function of time, with no target detections _____________ 102
Figure 5-9 Achieved averaged occupancies and frame times over the simulation period ____________ 102
Figure 5-10 Surveillance, plot confirmation, track initiation and tracking timebalances with new target
detections every 2s 103
Figure 5-11(a,b) Total achieved occupancies ofeachfunction vs time, and achieved surveillance region
occupancieswith track occupancy vs time ______________________________________________ 104
Figure 5-12 Achieved surveillance frame times vs time _________________________________________ 104
Figure 5-13 MESAR search regions in standard configuration _______________________________ 104
Figure 5-14 Surveillance timebalances for the 9 regions of MESAR, with no detections ____________ 105
Figure 5-15(a,h) Achieved averaged occupancies and frame times over the simulation period. Solid lines
show desired values scaled to account for a desired surveillance occupancy of less that 100% - 106
Figure 5-16 Total achieved occupancies of each function vs time ______________________________ 107
Figure 5-17 Surveillance, plot confirmation, track initiation and tracking timebalances with new target
detections every 2s 107
Figure 5-18(a,b) Total achieved occupancies of each function vs time, and achieved surveillance region
occupancies vs time 108
Figure 5-19 Surveillance region frame times for MESAR, with track detections every 2s___________ 108
Figure 5-20 Look number scheduled as afunction of time for plot confirmation tasks (green x) and
surveillancetasks (red +) ____________________________________________________________ 109
Figure 5-21(a,b) Histogram of the delay in scheduling plot confirmation tasks. (a) the MESAR algorithm
scheduling looks, (b) the MESAR algorithm scheduling un-interruptible tasks_______________ 110
Figure 5-22 A comparison of static and rotating multi-function radars in the presence of a non-un q"orm
loadingwith azimuth _________________________________________________________________ 113
Figure5-23 A surveillance window of opportunity ________________________________________ 115
Figure 5-24 Flow diagram of the rotating scheduler algorithm_______________________________ 117
Figure 5-25 Surveillance beam directions plotted in real space coordinates ___________________ 118
Figure 5-26 Search order of surveillance beam directions for a rotating system _________________ 118
Figure 5-27 Generating a pseudo-random search pattern with a rotating system_________________ 119
Figure 5-28(a) Limit offorwa rd/backward scanning, assuming a two-way gain loss with cos 3 (çv), (b)
Maximum achievable occupancy within a narrow angular sector, compensation waveforms using a
CO5 1(çD) law 120
9
Figure 5-29 A comparison of the array pointing direction with the azimuth direction of surveillance and
tracking tasks 122
Figure 5-30 (a) Total desired occupancy of surveillance and tracking tasks as a function of azimuth
angle, (b) Surveillance and tracking occupancies used to service the desired occupancies shown in
(a) 122
Figure 5-31 Fuzzy windows of opportunity for a surveillance task _____________________________ 125
Figure 5-32 Fuzzy windows of opportunity for a tracking task _______________________________ 125
Figure 6-l(a,b,c,d,e) Maximum prediction error between a constant velocity prediction and a
manoeuvring target, with varying manoeuvre strength and range of manoeuvre start. (a)
0. Is, (b) Th,,d = 0. 1 7s, (c) Thj1 n,,=O. 33s, (d) ThI,d-O. 6 7s, (e) Thj,fld = l.33s _____________________ 133
Figure 6-2 Maximum prediction error between a constant velocity prediction and a target manoeuvring
witha lateral acceleration of 7g at a range of 40km ___________________________________ 134
Figure 6-3(a,b) Track update interval, and number of beams required for each track update as a function
of track accuracy requirement (V,,) and signal to noise ratio (SNR,,) ______________________ 138
Figure 6-4 Radar load as a function of track accuracy requirement (V,,) and signal to noise ratio (SNR,,)139
Figure 6-5(a) Signal to noise ratio as a function of target angle from beam boresight, (b) Accuracy of
monopulse estimate as a function of the target signal to noise ratio __________________________ 141
Figure 6-6(a,b) 3 dimensional and 2 dimensional plot of Benchmark Trajectory 1, showing position truth,
positionmeasured, and smoothed filter output ___________________________________________ 143
Figure 6-7(a,b) X-coordinate position and velocity errors for Benchmark Trajectory I ____________ 143
Figure 6-8(a,b) Y-coordinate position and velocity errors for Benchmark Trajectory 1 _____________ 143
Figure 6-9(a,b) Z-coordinate position and velocity errors for Benchmark Trajectory 1 _____________ 144
Figure 6-I0(a,b) 3 dimensional and 2 dimensional plot of Benchmark Trajectory 3, showing position
truth, position measured, and smoothed filter output ____________________________________ 144
Figure 6-II(a,b) X-coordinate position and velocity errors for Benchmark Trajectory 3 ___________ 144
Figure 6-12(a,b) Y-coordinate position and velocity errors for Benchmark Trajectory 3 ___________ 145
Figure 6-13(a,b) Z-coordinate position and velocity errors for Benchmark Trajectory 3 ____________ 145
Figure 6-14(a,b) 3 dimensional and 2 dimensional plot of Benchmark Trajectory 1, showing position
truth, position measured, and smoothed filter output ___________________________________ 147
Figure 6-15(a,b) X-coordinate position and velocity errors for Benchmark Trajectory 1 ____________ 147
Figure 6-16(a,b) Y-coordinate position and velocity errors for Benchmark Trajectory 1 ____________ 147
Figure 6-17(a,b) Z-coordinate position and velocity errors for Benchmark Trajectory 1 ___________ 148
Figure 6-18(a,b) Manoeuvre detection flag for Benchmark Trajectory 1 _______________________ 148
truth, position measured, and smoothed filter output ___________________________________ 148
Figure 6-20(a,b) X-coordinate position and velocity errors for Benchmark Trajectory 3 ____________ 149
Figure 6-21(a,b) Y-coordinate position and velocity errors for Benchmark Trajectory 3 ____________ 149
Figure 6-22(a,b) Z-coordinate position and velocity errors for Benchmark Trajectory 3 ____________ 149
Figure 6-23(a,b) Manoeuvre detection flag for Benchmark Trajectory 3 ________________________ 149
truth (blue crosses), position measured (red circles), and smoothed filter output (green plus signs)151
Figure 6-25(a,b) Track update interval used for tracking of Benchmark Trajectory 1 ______________ 151
Figure 6-26(a,b) X and Z coordinate (azimuth and elevation) position and velocity errors for Benchmark
Trajectory 3 151
Figure 6-27(a,b,c,d) Beam update strategies, and their periods of use _________________________ 153
Figure 6-28(a,b) Multiple beam update tracking of Benchmark target trajectory 1 using a signal to noise
ratio of 25dB and a desired accuracy of 1/40" of a beamwidth. (a) Plan view of track showing
position truth (blue crosses), position measured (red circles), and smoothed filter output (green plus
signs), (b) Number of beams as afunction of time __________________________________ 153
Figure 6-29(a,b) Multiple beam update tracking of Benchmark target trajectory 1 using a signal to noise
ratio of 12dB and a desired accuracy of 1/20th of a beamwidth. (a) Plan view of track showing
position truth (blue crosses), position measured (red circles), and smoothed filter output (green plus
signs), (h) Number of beams as afunction of time ____________________________________ 154
Figure 6-30 Quantisation of track update directions to array azimuth broadside, shown for a twin faced
arraymulti-function radar ________________________________________________________ 155
Figure 6-31(a,b) Different conditions for the quantisation of track updates, shown fora twin faced array
multi-function radar. (a) Conditions for a current track update ahead of array azimuth broadside,
(h) Conditions for a current track update behind of array azimuth broadside ________________ 155
Figure 6-32(a,b) Rotating array update strategy tracking of Benchmark target trajectory 1 using a signal
to noise ratio of 25dB and a desired accuracy of 1/40" of a beamwidth. (a) Plan view of track
showing position truth (blue crosses), position measured (red circles), and smoothed filter output
10
(green plus signs), (b) Track update interval as a function of time__________________________ 157
Figure 6-33 Angle of track updates away from array azimuth broadside for the rotating array update
strategy (red squares). Overlaid is the manoeuvre flag (blue circles), showing where the tracker
perceiveda manoeuvre _______________________________________________________________ 157
Figure 6-34(a,b) X and Z coordinate (azimuth and elevation) position and velocity errors for Benchmark
Trajectory I using the rotating array track update strategy ______________________________ 158
Figure 6-35(a,b) Total radar time required to maintain track on Benchmark target trajectory 6, and (b)
the percentage of successful tracks, with a beamwidth of 2.5° __________________________ 163
Figure 6-36(a,b) Average track update interval required to maintain track on Benchmark target
trajectory 6, (b) Achieved smoothed track accuracy for Benchmark target trajectory 6, excluding
periods of manoeuvre, with a beamwidth of 2.5° _______________________________________ 163
Figure 6-37 Percentage of radar time required for tracking Benchmark target trajectory 6 using only a
20dB signal to noise ratio instead of a 29dB signal to noise ratio, with a beamwidth of 2.5° 163
the percentage of successful tracks, with a beamwidth of 0.5° __________________________ 165
periods of manoeuvre, with a beamwidth of 0.5° _____________________________________ 165
20dB signal to noise ratio instead of a 29dB signal to noise ratio, with a beamwidth of 0.5° 165
Figure 6-41 Ratio of radar time required to track Benchmark trajectory 6 with a 0.5°beamwidth to that
with a 2.5°beamwidth, for a system employing single beam track updates _________________ 166
Figure 6-42(a,b) Total radar time required to track Benchmark target trajectory 6, and (b) Percentage of
successful tracks, as a function of radar beamwidth_____________________________________ 166
the percentage of successful tracks, with a beamwidth of 2.5° ___________________________ 168
trajectory 6, (b) Average number of beams required for each track update, with a beamwidth of
2.5° 168
Figure 6-45(a,b) Achieved smoothed track accuracy for Benchmark target trajectory 6, excluding periods
of manoeuvre, and (b) Percentage of radar time required for tracking Benchmark target trajectory
6 using only a 20dB signal to noise ratio instead of a 29dB signal to noise ratio, with a beamwidth
of 2.5° 168
the percentage of successful tracks, with a beamwidth of 0.5° ____________________________ 169
0.5° 169
6 using only a 20dB signal to noise ratio instead of a 29dB signal to noise ratio, with a beamwidth
of 0.5° 169
Figure 6-49 Ratio of radar time required to track Benchmark trajectory 6 with a 0.5°beamwidth to that
with a 2.5° beamwidth, for a system employing multiple beam track updates ________________ 170
Figure 6-50(a,b) Total radar time required to track Benchmark target trajectory 6, and (b) Percentage of
successful tracks, as a function of radar beamwidth, using multiple beam track updates _______ 171
Figure 6-51 A comparison of the total radar time required to track Benchmark trajectory 6 with the static
andmultiple beam track update systems ____________________________________________ 171
the percentage of successful tracks. Single faced rotating system with a beamwidth of 2.5° 173
periods of manoeuvre. Single faced rotating system with a beamwidth of 2.5° _____________ 173
20dB signal to noise ratio instead of a 29dB signal to noise ratio. Single faced rotating system with
a beamwidth of 2.5° 173
11
periods of manoeuvre. Single faced rotating system with a beamwidth of 0.5° _____________ 174
20dB signal to noise ratio instead of a 29dB signal to noise ratio. Single faced rotating system with
the percentage of successful tracks. Twin faced rotating system with a beamwidth of 2.5° 175
Figure 6-59(a.b) Average track update interval required to maintain track on Benchmark target
periods of manoeuvre. Twin faced rotating system with a beamwidth of 2.5° ______________ 175
20dB signal to noise ratio instead of a 29dB signal to noise ratio. Twin faced rotating system with a
heamwidth of 2.5° 175
the percentage of successful tracks. Twin faced rotating system with a beamwidth of 0.5° _____ 176
periods of manoeuvre. Twin faced rotating system with a bea,nwidth of 0.5° _______________ 176
20dB signal to noise ratio instead of a 29dB signal to noise ratio. Twin faced rotating system with a
heamwidth of 0.5° 176
trajectory 6, (b) Average number of beams required for each track update. Single faced rotating
system with a beamwidth of 2.5° 178
Figure 6-66(a,b) Achieved smoothed track accuracy for Benchmark target trajectory 6. excluding periods
of manoeuvre, and (b) Percentage of radar time requiredfor tracking Benchmark target trajectory
6 using only a 20dB signal to noise ratio instead of a 29dB signal to noise ratio. Single faced
rotatingsystem with a beamwidth of 2.5° ___________________________________________ 178
system with a beamwidth of 0.5° 179
of ma,weuvre, and (b) Percentage of radar time required for tracking Benchmark target trajectory
6 using only a 20dB signal to noise ratio instead of a 29dB signal to noise ratio. Single faced
rotatingsystem with a bearnwidth of 0.5° _________________________________________ /79
the percentage of successful tracks. Twin faced system with a beamwidth of 2.5° _____________ 180
trajectory 6, (b) Average number of beams required for each track update. Twin faced system with
6 using only a 20dB signal to noise ratio instead of a 29dB signal to noise ratio. Twin faced system
with a beamwidth of 2.5° 180
Figure 6-73 Total radar time required to maintain track on Benchmark target trajectory 6, and (b) the
percentage of successful tracks. Twin faced system with a beamwidth of 0.5° 181
trajectory 6, (b) Average number of beams required for each track update. Twin faced system with
a heamwidth of 0.5° 181
6 using only a 20dB signal to noise ratio instead of a 29dB signal to noise ratio. Twin faced system
with a heamwidth of 0.5° /81
Figure 6-76(a,b) Comparison of the radar time required to track Benchmark target trajectory 6. (a)
Single faced array system, (b) Twin faced array system _________________________________ 182
Figure 6-77(a,b) Comparison of the radar time required to track Benchmark target trajectory 6 with a
single faced rotating array multiple beam system and a single faced static array multiple beam
12
system (a) Absolute radar time required by each system, (b) ratio of the radar time used by to
rotatingsystem to the static system _____________________________________________________ /33
Figure 6-78(a,b) Comparison of the radar time required to track Benchmark target trajectory 6 with a
twin faced rotating array multiple beam system and a single faced static array multiple beam
system (a) Absolute radar time required by each system, (b) ratio of the radar time used by to
rotatingsystem to the static system ________________________________________________________ 183
Figure 7-1 Requirement search volume for the static and rotating array systems___________________ 187
Figure 7-2(a,b) Simultaneous multiple beam configurations for (a) Regions 2 and 3, (b) Region 1.
Dashed line shows transmit beam pattern, filled circles show possible receive beam positions_ 191
13
LIST OF PRINCIPAL SYMBOLS AND ACRONYMS
Symbol Definition Units

A leam_ruv area of beam in rotating UV space m2
A I,eamuv area of beam in UV space m2

A area of surveillance sector in rotating UV space m2
Asectoruv area of surveillance sector in UV space m2
azmax maximum azimuth extent of surveillance sector m2
Ar effective aperture area of the antenna on receive m2
az azimuth radians
az11. minimum azimuth extent of surveillance sector radians
B bandwidth of receiver Hz
bw antenna 3dB beamwidth radians
cn receiver correction factor
DERA Defence Evaluation and Research Agency

dt dwell time in each beam direction S
E(ø) antenna gain in direction 0 dB

ci elevation radians
elm, maximum elevation extent of surveillance sector radians
elfl,) minimum elevation extent of surveillance sector radians
E weighting function for each element
ERP Effective Radiated Power w
E. smoothed, normalised residual in the x coordinate, between the target
predicted and measured position
f1 Doppler frequency Hz
GaAs Gallium Arsenide
G, gain of the antenna on receive dB
G gain of the antenna on transmit dB
H measurement matrix relating target state to target measurements
1MM Interactive Multiple Model
JPDA Joint Probabalistic Data Association
ix normalised distance between the target predicted and measured position

Jy normalised distance between the target predicted and measured position
Jz normalised distance between the target predicted and measured position

k Boltzmann's constant JK'
K Kalman filter gain matrix
MESAR Multi-function Electronically Scanned Adaptive Radar
MHT Multiple Hypothesis Tracker
MoD Ministry of Defence
14
MTI Moving Target Indication
n number of pulses integrated
N number of elements, or subarrays in the array
n_beams number of surveillance or tracking beams
0 occupancy of the task or function %

OCCsearch occupancy of the search function %
P, covariance of predicted target state
covariance of smoothed target state
P
Pay average power W
Pr received power W
PRF Pulse Repetition Frequency Hz
PR! Pulse Repetition Interval S
Pt transmitted power W
Q plant noise term
R covariance of measured target state
r, R slant range of the target m

RADAR RAdio Detection And Ranging
RCS Radar Cross Section m2
R, maximum detection range m
Rx receive
S received signal power at the radar W
S mm minimum detectable signal W

SNR, snr signal to noise ratio dB
snro signal to noise ratio on array broadside dB
snr signal to noise ratio at a scan angle of ( dB
time S
T receiver temperature K
TAS Track And Scan
till function time of the task or function 5
tfr frame time of the task or function S
Ty time taken by the pulse to travel to the target and return S
TWs Track While Scan

Tx transmit
U coordinate of UV space antenna coordinate system
Ur coordinate of rotating TJV space antenna coordinate system

V coordinate of UV space antenna coordinate system
th V
VII signal from the channel
coordinate of rotating UV space antenna coordinate system
WI) complex weight applied to a subarray or element
x, forecast target state
15
smoothed estimate of target state
x
res residual between predicted target position and measured target position m
Yles residual between predicted target position and measured target position m
z measured target state
Zres residual between predicted target position and measured target position m
(1) antenna rotation rate radians/s
target transition matrix
abeam solid angle of beam steradians
sector solid angle of surveillance sector steradians
standard deviation of the tracker prediction error radians
standard deviation of the measurement error radians
o element spacing m
phase difference m
scan angle away from array normal radians
radiated wavelength m
smoothing constant for tracker residual estimate
03 3dB beamwidth radians
oc, beam width at scan angle of p radians
monopulse accuracy radians
monopulse accuracy at scan angle p radians
16
1 INTRODUCTION
1.1 Historical background
A Radar (Radio Detection And Ranging) transmits radiowaves and measures the
returned energy scattered by distant targets. Radar technology was invented some time
before the beginning of the second world war when the benefit of the wireless
technology to the air defence of Great Britain became apparent to the Ministry of War.
There ensued a rapid period of development throughout the war, in which radar played a
vital part in detecting both air and sea borne targets for the Royal Air Force and Royal
Navy, making a significant contribution to the war's outcome.
Throughout World War II, as radar systems were developed with increasing
functionality and performance, so the enemy's ability to counter the technology
increased. This situation has continued, resulting in the invention of new techniques
such as jamming, chaff, decoys and stealth. Also, the requirements asked of the radar
have increased, such as the number of targets that must be tracked, and the precision
with which this task must be carried out. The result is highly sophisticated modern day
radars, which remain a vital tool for both civil and defence applications.
Early systems, at the beginning of the war, could simply provide angle and range
estimates for detections made. Later, after the invention of the first plan position
indicator (PPI) displays, plot tables were used to enable skilled operators to maintain
track on targets.
The 1 960s saw the widespread development of phased arrays which allowed electronic
(rather than mechanical) steering of the antenna beam; some nineteen phased array
projects starting in that decade [Billam, 1985]. This technology, with the emergence of
digital control, allowed the automatic operation and control of phased array radars, and
the advent of the multi-function radar, in which the electronically scanned antenna is
multiplexed between many different tasks.
Since then, both microwave and digital technology have advanced at an extremely rapid
17
I Introduction
pace. New radars are emerging which are capable of adapting many aspects of their
operation to the scenario and environment in which they are operating, under computer
control. The flexibility provided by a software controlled phased array antenna
supported by adaptive waveforms, signal processing and beamforming allows the
modern multi-function radar to modify almost any aspect of its radar performance on a
task to task basis. Such flexibility permeates all aspects of multi-function radar design
and gives the potential to optimise performance and efficiency to meet weapon system
design needs.
Whilst the concept of phased array multi function radars has been around for many
years, the realisation of their full potential has been constrained by the availability of
technology at realistic prices. It is only relatively recently that technology has progressed
to a point where the software and processing power, and the antenna components have
both become readily available to provide multi-function radars with the full flexibility
that electronically adaptable phased arrays can offer.
1.2 Objectives
The aim of this work is to develop and investigate novel techniques for the control and
tracking of a multi-function radar, with the objective of improving the overall efficiency
and performance of the sensor. The emphasis will be particularly on improving the
performance of the tracking function.
With the capability of performing multiple functions with a single sensor come
drawbacks. Firstly, the total radar time-budget from this single sensor must be shared
between each of the functions. This means that radar time is at a premium and in many
practical scenarios, less radar time is available than is ideally required. Techniques that
are efficient in their use of radar time are thus crucial, and this forms a significant focus
of the work.
Secondly, a single sensor performing all of the functions normally undertaken by several
other sensors, each dedicated solely to a single function, must inherently contain
compromises. The most significant of these in the multi-function radar is the beamwidth
that is used for the search and tracking functions. This normally results in a broader
beamwidth being used than is ideal for the tracking function to ensure the search is
18
I Introduction
performed adequately. This is also addressed within the study.
It is likely that multi-function radars will exploit rotation of the array antenna, to avoid
the need for the three or more phased array faces that are required for full hemispherical
coverage. This will give a significant advantage in terms of cost. The study aims to
establish the system implications of this, both for the tracking and surveillance
functions, and to determine efficient techniques for control of the rotating array multi-
function radar. This should enable a valuable comparison of the efficiency and
performance of static and rotating systems.
New multi-function radars offer a wide variety of adaptivity and flexibility that may be
controlled in real time, possibly as a response to one or more previous radar detections.
In order to track targets with a wide range of dynamic capability reliably and efficiently,
the algorithms for tracking and control must be highly adaptive and rapidly responsive.
Techniques developed aim to exploit the many degrees of freedom that future multi-
function radars will bring, such as adaptable update rate, antenna beamwidth,
transmitted energy and frequency. Findings will build upon the research undertaken for
the MESAR (Multi-function Electronically Scanned Adaptive Radar) experimental
multi-function radar, and it is hoped that this work will result in improved techniques
and algorithms being implemented in that radar.
1.3 Thesis Layout
Chapter 2 gives an introduction to some of the background topics pertinent to Multi-

function radar. This includes a short definition of a multi-function radar and the
functions it may be expected to undertake, as well as a brief introduction to some of the
underlying theory of radar, phased arrays and multi-function radar.
Chapter 3 then gives a detailed introduction to the MESAR experimental multi-function

radar, to which much of this research will be applicable. MESAR is an advanced
experimental active phased array radar which has been developed collaboratively
between the Defence Evaluation and Research Agency and British Aerospace Defence
Systems (formerly Siemens Plessey Systems) since the early 1980s. The fully functional
radar was completed in early 1993, and has been used as a test-bed for experimentation
and demonstration of techniques. This chapter indicates the performance and capability
19
I Introduction
of a current, state of the art multi-function radar system, and gives a background to what
has been achieved already and what flexibility remains to be harnessed.
Chapter 4 investigates the impact of two digital beamforming techniques that will be
available in the future version of the MESAR radar, MESAR II. These are the
generation of variable beamwidth, and simultaneous multiple receive beam processing.
Novel methods are developed to allow an accurate analysis of the search time- budget
for a sectored surveillance volume, for both static and rotating array multi-function
radars. Two important conclusions are drawn out from the analysis made. Firstly it is
shown that the traditional view that 3 or 4 fixed array faces is the best antenna
configuration for a multi-function radar is not correct, and that rotating array systems
can perform the surveillance function a great deal more efficiently. Secondly, it is
demonstrated that the inherent compromise in beamwidth normally associated with a
multi-function radar does not necessarily hold with a simultaneous multiple receive
beam system. The search and track beamwidth could be de- coupled and each be
optimised to their task.
Chapter 5 addresses the topic of resource management and task scheduling of a multi-
function radar. It covers the objectives of resource management and task scheduling in
multi-function radar, and presents figures of merit by which a scheduling and resource
management system may be judged. It also covers the complications that arise with
rotation of the array antenna, and presents a novel algorithm that copes with these
complications. Models are presented that have been developed for both static and
rotating array multi-function radars based around a task scheduler centered architecture,
which are used to evaluate both scheduler performance against the proposed objectives.
Chapter 6 presents a detailed investigation of a static and rotating array adaptive

tracking of manoeuvring targets. The objective of the work presented is to develop an
effective tracking strategy against manoeuvring targets that is efficient in its use of the
multi-function radar's resources. Models of both static and rotating array adaptive
trackers are developed, and then applied through Monte Carlo simulation to determine
an efficient signal to noise ratio to aim for when updating tracks. It is shown that
multiple beam track updates are necessary to enable efficient and robust tracking, and a
strategy for multiple beam update tracking is suggested. The impact of tracking using a
narrow beamwidth is assessed, and successfully demonstrated through simulation.
20
I introduction
Chapter 7 gathers together the most significant conclusions of chapters 4, 5 and 6, and
presents a short assessment of a narrow beamwidth, rotating array multi-function radar
which utilises simultaneous multiple beam processing. Its radar time-budget resources
and detection performance are compared to a static array system, similar to the current
MESAR system. The viability of the multiple beam rotating array system is
demonstrated, and it is shown to have to have very favourable characteristics.
Finally chapter 8 presents the conclusions of the work so far, and suggest further work
that would be beneficial in the area of this g/vc.
1.4 Novel Aspects
The novel aspects of this thesis are contained in chapters 4, 5, 6 and 7, and are
summarised here. In chapter 4, a new analysis of multi-function radars with different
array antenna configurations and of simultaneous multiple beam processing has been
made, utilising a novel method for accurate estimation of the surveillance resource
requirements. Chapter 5 proposes a set of objectives for the resource management of a
rotating array radar, which may be used to judge the effectiveness of future strategies.
The existing MESAR scheduling algorithms are validated against these objectives. A
novel rotating array scheduler has been developed and assessed, which takes advantage
of a previously unrecognised ability of rotating array multi-function radar systems.
In chapter 6 a novel multiple beam, rotating array adaptive tracker is developed. This,
along with a static array tracker based on that used in the MESAR radar, are applied in a
Monte Carlo simulation to determine the most efficient operating parameters. The result
is a comparison of the tracking performance and efficiency of static and rotating array
multi-function radars. The capability of such systems operating with narrower
beamwidths than those currently used is demonstrated against the benchmark tracking
problem developed by Blair and Watson [Blair et al, 1994].
Chapter 7 demonstrates the viability of a narrow beamwidth multi-function radar

system, and the benefits such a system would give.
21
2 AN OVERVIEW OF MULTI-FUNCTION RADAR
2.1 Background
This chapter gives an introduction to some of the topics pertinent to Multi-function

radar. The functions of a multi-function radar for naval air defence are described in
section 2.2 to give a short background and a 'time-line' of a multi-function radar's
operation from surveillance through to missile interception.
Section 2.3 then gives a brief introduction to some of the underlying theory of radar,
phased arrays and multi-function radar, which is relevant to this thesis. Finally, in
sections 2.4 and 2.5, the benefits and drawbacks of using a multi-function radar rather
than several specialised radars are discussed.
The multi-function radar is equivalent to a suite of radars, employed for some

application such as air defence. To fulfil its purpose it performs several different
functions which previously would have been undertaken by many different, dedicated
radars. The exact functions that are undertaken are obviously dependent upon the
application, but at least, the multi-function radar will provide search coverage and
concurrent tracking of multiple targets.
The multi-function radar utilises one or more phased array antennas to give one or more
electronically steerable beams, in azimuth, elevation or both. These beams are
multiplexed between the different tasks in different directions under automatic digital
control, so that the total radar time is shared amongst the functions.
At its simplest level a multi-function radar consists of the following:
i. a phased array antenna, (including the associated beamforming and

microwave transmitting and receiving hardware)
ii. signal processing
iii. radar control computer
iv. waveform generator
22
2 An Overview of Multifunction Radar
It is the coupling of an electronically steerable beam, adaptive waveform generation and

signal processing, with real time automatic computer control that gives the multi-
function radar such power.
Many different types of antennas are utilised in multi-function radars. The antenna
normally consists of one or more planar phased arrays, although cylindrical or spherical
arrays have been used. It may be mechanically rotated or use fixed or positioned arrays,
relying solely on electronic scanning to give the volume coverage.
It is essential to realise at the outset the importance that time plays in a multi-function
radar. The multi-function radar is trying to do the job of many radars, each of which
could dedicate all of its radar time to its own task whether it be short range surveillance
or tracking for example. For this reason it is imperative that the multi-function radar's
time is not wasted but is distributed as efficiently as is possible. Barton, [Barton, (1988)]
recognises this, and gives it as the second of two critical design factors in multi-function
radar;
. Choice of radar frequency, high enough to provide the narrow tracking beam, and low
enough to permit rejection of clutter during volume search modes.
Budgeting of radar time, to provide the dwell times necessary for clutter rejection in
both search and tracking modes.
Billam [Billam, 1997a] discusses the issue of radar time in multi-function radar
specifically in a recent paper. It is one of the dominant factors in the system design of a
multi-function radar, and thus any techniques that are efficient with radar time are of
prime importance.
2.2 Multi-function radar functions
The functions that a multi-function radar must undertake are specific to its application.
Multi-function radars are utilised in land, sea and air based platforms. To date they are
primarily used for military applications due to the high costs. A ship-borne or land-
based multi-function radar may take over the functions of surface search, short, medium,
and long range search, multiple target tracking, weapon control and aircraft control. An
airborne multi-function radar may also take over the functions of navigation or terrain
23
2 An Overview of Mull, unction Radar
mapping/avoidance.
The primary application of this research study is naval air defence; thus this will be used
as an example here, and often inherently assumed throughout the rest of the thesis.
Nevertheless, research presented is often still pertinent to land and air based multi-
function radar.
Figure 2-1 below shows the functions that a multi-function radar performing naval air
defence may undertake. The main functions are surveillance, tracking and missile
guidance. A summary of these functions is given in sections 2.2.1 to 2.2.4 below. Other
functions may well be needed, such as target recognition with specific requirements
such as the need for long radar waveforms or very precise scheduling.
Hh P
(FIRE CONThOL (MultIple chaIm.i)
hdcoo,e Teccin
Tt'J
SoarthlTrack Gadaoco Heacng
aIat0fl
Raped Conn
OO Search
SO f
l ___________
EVEILAN
Hh Speed HonZon
Figure 2-1 Functions of a multi-function radar Jhr naval air defence
2.2.1 Search
The aim of the multi-function radar during search, is to survey a given volume of space
within a given frame time and with a given detection performance, set by the operational
requirement. It must do this with the lowest number of false alarms possible.
Additionally, if the search function is being used for track while scan purposes,
requirements in the positional accuracy of surveillance detections may be included.
In practice this requirement is often very difficult to achieve due to restrictions on the
time-budget available to undertake the search function. Various strategies may be
adopted with the aim of using the radar time optimally. This is discussed later in section
2.3.3.1. To minimise the radar time-budget for surveillance, the search volume will
almost certainly be divided into several sectors, with the waveforms and frame times for
each sector that are matched to the performance required.
24
2 An Overview of Mu1tfunction Radar
2.2.2 Plot confirmation

A significant advantage of the multi-function radar is that it can rapidly confirm
detections that are made during the search function, in order to distinguish them from
false alarms due to thermal noise. This function is termed plot confirmation and is
essential in minimising the reaction time of the multi-function radar.
Generally, a 'look back' is used with a waveform that has similar parameters to the
original surveillance waveform that made the detection, to ensure that the target is
detected again. The interval between the surveillance detection and the look back needs
to be minimised to avoid the target decorrelating and hence not being detected. Typical
look back times are between lOms and lOOms.
Additionally the plot confirmation look back may be used to attempt to rapidly resolve
an ambiguity in range or velocity by modifying the PRF of the waveform.
This technique is only used on detections that do not associate with existing tracks in the
system. On each surveillance detection the radar must measure the position of the
detection and compare this with the expected positions of targets that it is already
holding in track (plot association). If the detection correlates with a predicted target
position, then that information may be used as a plot to update that track. If not then a
look back would be requested.
2.2.3 Track initiation

Upon a successful confirmed detection, track initiation is undertaken to establish a track
of sufficient accuracy (in terms of position and velocity) that it may be handed to the
track maintenance function. The aim of track initiation is to establish a track of a given
accuracy in the shortest time possible.
The combination of the plot confirmation and track initiation functions provides the
earliest reaction times and target threat assessment for weapon system needs. In
comparison to a suite of dedicated radars the delay between initial detection can be
significantly larger. In practice several detections are required from a dedicated
surveillance radar (to fulfil an n out of m criterion) in order to obtain the low false alarm
rates that are required before cueing the tracking radar. This requires several sweeps of
the surveillance antenna, which is likely to take several seconds. The dedicated tracking
radar must then perform a search to acquire the target, since the surveillance radar that
25
has cued it has a broader beam. This whole process may be achieved in the quickest
time by a multi-function radar that may direct the beam at will.
2.2.4 Track maintenance

The multi-function radar is capable of undertaking two basic means of tracking; track-
while-scan (TWS) and track-and-scan (TAS). Track-while-scan simply feeds plots from
the surveillance function to the track extractor, so that the tracks are updated at the
surveillance rate, as with conventional rotating surveillance radars (although the plots
may not be updated at regular intervals with a multi-function radar, due to the random
nature of the scan). Track-and-scan, sometimes referred to as de-coupled tracking,
updates tracks independently, and at a rate higher than the surveillance rate. This
therefore implies an additional load on the radar time-budget.
In general track-while-scan would only be used against low priority targets, or targets
for which the surveillance update rate would give sufficient accuracy. The majority of
targets of interest to a weapon system are likely to have the potential to manoeuvre or
may be closing rapidly. It is most likely that they will have high update rate or accuracy
requirements, and will therefore be tracked using track-and-scan. Thus it is track-and-
scan that is the main consideration of this study.
ARRAY SIGNAL
COMMUNICATIONS PROCESSING
Jr
PLOT
RADAR TASK POSITION
SCHEDULER EXTRACTOR
4
Jr
RADAR CONTROL TRACK
AND WAVEFORM * EXTRACTOR
SELECTION
Figure 2-2 The closed ioop multi-function radar tracker
The multi-function radar tracker is a closed ioop, as shown in Figure 2-2. For successful
multi-function radar target tracking the inter-related functions of plot to track
association, track filtering and prediction, and radar assignment must all work
seamlessly. If, for example, an incorrect waveform, or an update interval that is too
long, is chosen on the basis of the tracker output, then the next update will be of poor
quality, or may fail to illuminate the target.
26
2.2.5 Functional flexibility

The enormous flexibility of the multi-function radar system also allows it to be
reconfigured for several different applications. For example a ship borne multi-function
radar could be configured for medium range air defence as its main role. By changing
the configuration of the search volume, the digital waveforms selected, the digital signal
processing applied, and the control algorithms, the same radar could be applied to long
range air defence, or to ballistic missile defence. This may be incorporated as a simple
software, or menu driven option. Perhaps even more significantly, a major benefit of the
multi-function radar is that it can in principle be rapidly reconfigured for unforeseen
applications and eventualities.
2.3 Underlying theory of multi-function radar
A multi-function radar is a specific, and advanced type of radar. As such the well
understood theory from radar and phased array systems is applicable, but also, as has
been pointed out by Billam ([Billam, 1994]), new theory is developing pertinent purely
to multi-function radar systems. This is exemplified by the two books published on this
subject in recent years purely on multi-function radar system aspects ([Billeter, 1989],
[Sabatini & Tarantino, 19941). In some respects it could be said that the system level
theory of multi-function radar - its application and control - is somewhat behind the
capability that the various technologies involved in a multi-function radar suite can
potentially offer. This may lead to a highly sophisticated and adaptable system being
utilised in an essentially conventional way, simply by virtue that a well developed theory
is not available at the radar design stage.
This section introduces the areas of radar theory and phased array theory relevant to this
study, before giving a background to some of the system aspects of multi-function
radars.
2.3.1 Radar systems

The subject area of radar is vast, and well covered in many texts. Notable introductions
to the subject are given by Barton, Levanon and Skolnik ([Barton, 1988], [Levanon,
1988], [Skolnik, 1981]). The purpose of this section is to give only the broadest account
of a complete radar system, and to present a brief introduction to the techniques and
27
equations that are applied within the scope of this study.
A highly simplified block diagram of a typical radar is shown in Figure 2-3 below.
miUer
crowave feed Oscil

exer
Antenna Receiver __________
Signal.......
processing
y
Display
Figure 2-3 A simplified block diagram of a typical radar
A pulsed oscillator feeds a low power RF signal to a transmitter. This signal is

amplified, and possibly modulated. The resultant signal is passed to the antenna for
radiation into space. The antenna directs the transmit energy, in the desired direction. A
target, illuminated by the energy radiated from the antenna, will scatter the energy in all
directions. In general, only a very small amount of the energy will be re-radiated back in
the direction of the antenna.
This small amount of energy, which competes with the energy reflected from other
sources such as the land or sea, or external interference sources, re-enters the antenna.
Received signals are amplified and generally mixed down to a lower frequency in the
receiver before being fed to the signal processor.
The benefits of a highly directional antenna are clear. Firstly, both the energy that is
transmitted, and that which is received, is maximised in the required direction.
Secondly, the interference from other directions is minimised. Lastly, a narrow main
beam formed by a highly directive antenna allows a high angular measurement and
resolution capability.
The signal processor has the job of trying to recover the wanted information, if any,
from the signal it is fed. It must determine, by varied means, whether a target is present
or not in each radar resolution cell. It must extract information about that target, such as
range, echo signal strength, angular position or Doppler velocity, If a target is detected,
its range may be determined from the fundamental radar equation [2-1].
28

[2-1]
R=
2
where R is the slant range of the target

c is the speed of propagation
Tr is the time taken by the pulse to travel to the target and return
Barton [Barton, 1988] gives a list of the major problem areas in radar, which, although
recognised in the early days of radar, remain only partially solved today.
i. adequate signal to noise ratio in free space

ii. clutter reduction
iii. interference reduction
iv. signal selection
v. measurement accuracy
vi. size, weight, cost and reliability
It is useful to have these problems in mind when considering multi-function radar, since
high performance must be attained in every one of these areas to ensure a good level of
multi-function radar performance.
Often problems in one of these areas will affect the multi-function radar system in more
complicated ways than a conventional radar. For example, poor clutter cancellation will
lead to false detections, and subsequently to false track initiations. This, in turn, will
lead to less radar time being devoted to other functions, such as surveillance, causing a
drop in detection performance.
2.3.1.1 Radar range equation

The range equation is one of the most fundamental equations in radar theory. It allows
the prediction of the range at which a target of given radar cross section may be
detected, given basic information about the radar, the target, and the propagation path of
the radiation through space. One form of this equation allows the determination of the
signal strength given the range of a target:
a ( Ar
[2-2]
s=(PG)(_ I I
29
where S is the received signal power at the radar from a spherical target, of cross section
P, is the power of the transmitter
G, is the gain of the antenna on transmit
c is the radar cross section
A r is the effective aperture area of the antenna on receive
R is the slant range of the target
The first factor of the three in equation [ 2-2], is the effective radiated power in the
direction of the target. The second is the fraction of this power that is intercepted by the
spherical target. The final term is the fraction of the power, scattered by the spherical
target, that is intercepted by the radar receiving antenna. This is more readily expressed
as below.

- FGGt2cy [2-3]
- (4ir)3R4
where Gr is the gain of the antenna on receive

) is the wavelength of the transmitted signal
The physical relationship that the signal received varies with the inverse of R4 means
often tremendously high powers and highly directional antennas are required to detect
targets that are far away. As air targets become faster and more stealthy it becomes
increasingly difficult to detect them at sufficiently long ranges using radar.
Equation [ 2-3] may be rearranged to give the range at which a target may be detected,
given the minimum signal strength that is sufficient for detection.
4IJGiGr;t2a [2-4]
Riiiax =
(4r)3
where Rm L. is the maximum achievable detection range

Smin is the minimum detectable signal
In practice, the detection of a target depends not only on the signal strength returned
from that target, but also on other signals with which the target signal is competing.
Thus Smjn is determined also by signals such as the noise generated in the receiver or
noise from external sources such as jammers, or the residue of uncancelled clutter. This
leads to advanced techniques for eliminating clutter and external interference.
2.3.1.2 Monopulse radar

For a simple radar system, that employed no angular processing techniques, the angular
measurement accuracy that would be obtained is of the order of the beamwidth of the
30

2 An Overview of Mult(function Radar
radar. For target tracking this measurement accuracy is not sufficient, even when using a
narrow pencil beam. To obtain a higher accuracy several methods exist. The most
common of these is monopulse.
Monopulse extracts information about the angular location of a target within the radar
main beam by comparing signals received in two or more simultaneous antenna beams.
This technique may be used on just a single pulse, hence the name monopulse. Excellent
accounts of monopulse are given in, amongst others, Skolnik ([Skolnik, 1981]), and
further detail may be found in Rhodes ([Rhodes, 1959]).
Amplitude comparison monopulse employs two overlapping antenna patterns to obtain

the angular error in one coordinate. These patterns may be generated either by two
adjacent feeds with a reflector antenna, or in the case of a phased array, by correctly
summing the outputs of each element. Figure 2-4, (reproduced from [Skolnik, 1981]),
shows clearly the principle of monopulse.
/ /\ ,/'\\
/ '1
1\i
/ \i
I
I
J\
I / \/j
/ ki
/ f\ / V
4 e 4 e
(a) (b)
4 0
(c) (d)
Figure 2-4 Monopulse antenna patterns and error signals. (a) overlapping antenna patterns, (b)
difference antenna pattern, (c) sum antenna pattern, (d) error signal
The summation of these two overlapping patterns will generate the sum beam, . This is
the standard beam that would be used for transmission, and for detection and range
measurement in receive. The difference of these two patterns will generate the
difference beam, A. This beam contains the information upon reception which gives the
angular position of the target. The signals received from the sum and difference patterns
are combined in a phase sensitive detector to produce the error signal. if measurement of
angle is required in both dimensions then, generally, two difference beams are formed as
well as the sum beam, one for azimuth and one for elevation angle estimation.
31
The amplitude of the error is given by the ratio of the difference channel to the sum
channel. The sign of the difference signal is measured by comparison of the phase
between the sum and difference channels. Although the error signal is not linear, it can
be accurately calibrated to estimate the target position.
In practice, with a phased array the elements are summed using different sets of weights
to generate the beams. Barton [Barton, 1988], gives the optimum monopulse angle
estimator for a noise only environment as that shown in Figure 2-5 below. These
weighting functions are optimal in the sense that they give the maximum efficiency in
the sum channel and maximum slope in the difference channel. The penalty associated
with these weighting functions is the high sidelobes that are generated in the associated
antenna patterns.
It
081
061
O4
02[
-0
Figure 2-5 Weighting functions for minimum monopulse estimation error, and the associated
beam patterns
Sidelobes as high as these can rarely be tolerated in any real application; therefore other
weighting functions are employed, which always give a lower slope in the difference
channel. The weighting function chosen depends upon the requirements for the sidelobe
levels and the loss that can be accepted in main beam gain. An excellent paper by Harris
[Harris, 1978], gives the main features and methods of calculation of many of the
important weighting functions. Two of the most commonly used weighting functions
used to generate a sum beam pattern are the Dolph-Chebyshev weighting function
[Dolph, 1946] and the Taylor weighting function ([Taylor, 1955], [Villeneuve, 1984]).
32
2 An Overview of Multfunction Radar
These patterns give the minimum broadening of the main beam for a given sidelobe
level. The Dolph-Chebyshev function gives an antenna pattern with all sidelobes of
equal level. Taylor weighting function is an extension of this, to give an antenna pattern
with only the first n sidelobes equal, thereafter decaying as sin(x)/x.
A two dimensional Taylor weighting function applied to a planar phased array is shown
in Figure 2-6. The associated antenna pattern is also shown. The weighting function
gives near in sidelobes of -40dB, broadening the main beam by a factor of 1.25.
20
Figure 2-6(a,b) A 2 dimensional Taylor weighting function, and its associated antenna pattern
In the case of an active phased array, it is unlikely that the same weighting function will
be used on transmit as in receive. Usually the transmitlreceive modules require that a
uniform weighting function is used when transmitting. Therefore, a heavier weighting
function must be used in receive to ensure a low two-way sidelobe antenna pattern is
obtained. This results in active phased arrays often operating with high antenna
sidelobes in transmission, and a significantly broader beam on receive than in transmit.
I
I
Figure 2 . 7 A Bayliss weighting function in one dimension and a Taylor weighting function in
the other, and its associated antenna pattern
The equivalent, ideal, weighting to a Taylor pattern for generating a difference beam is
the Bayliss [Bayliss, 1968] weighting function. This pattern generates the first
33
differential of the sum pattern that is obtained from the Taylor weighting function. A
planar phased array illumination function of a Bayliss weighting function in one array
dimension, coupled with a Taylor weighting function in the other dimension, and its
associated antenna pattern are shown in Figure 2-7.
Monopulse was originally conceived as a technique for dedicated tracking radars, and as
such was aimed at maintaining targets on the nose of the main beam, and null of the
difference beam, termed beam boresight. However, when a target manoeuvres, a lag is
created between the prediction that the track smoothing filter makes and the real
position of the target. The magnitude of this difference is dependent upon the strength of
the manoeuvre and the track update interval. As the lag increases and monopulse
estimates are made away from beam boresight, so the estimates degrade in accuracy. For
a multi-function radar, dividing its time between many targets, possibly with a strong
manoeuvre capability, this lag, and the consequent degradation in estimate accuracy, can
become significant.
Barton [Barton, 1988] gives the precision of a monopulse estimate close to beam
boresight as;
[2-5]
(&,
2...J(snr).(n)
where a is the monopulse accuracy (rads)

6 is the 3dB beamwidth
snr is the signal to noise ratio
n is the number of pulses integrated
2.3.2 Phased arrays

Skolnik (1981) defines a phased array radar as;
".. a directive antenna made up of individual radiating antennas, or elements,

which generate a radiation pattern whose shape and direction is determined
by the relative phases and amplitudes of the currents at the individual
elements."
By properly varying the relative phases, it is possible to steer the direction of the
radiation. The antenna elements themselves may be of any antenna type such as dipoles,
or open ended waveguides. The array is made up of a number of these elements, suitably
34
spaced with respect to one another.
The two most common array geometries are the linear array and the planar array. The
linear array consists of elements arranged in a straight line in one dimension. This will
generate a fan beam, steerable in one dimension. The geometrical extension of this is the
planar array, in which the elements are arranged to lie in a plane. This will generate a
beam that may be steered in two dimensions. The beam may be either a fan shaped
beam, or a pencil beam, depending on the shape of the array.
Less commonly, it is also possible to arrange the elements in a non planar fashion.
Arrays with elements arranged on the surface of a cylinder or on the surface of a sphere
have been made. An electronically scanned planar array with a narrow pencil beam that
may be scanned in two dimensions is assumed in this study. Typically this type of array
supports scanning to angles of up to 600 away from array broadside.
Phased arrays are split into active and passive arrays. In a passive array the power
generated by a single, high power transmitter is divided by the array beamforming
network. This network is also used in receive to recombine the received power. An
active array utilises a transmitting/receiving module for every array element. This
module usually contains the power amplification, the low noise receiver and the phase
shifter.
The active array can give a high effective radiated power (ERP). The transmit and
receive losses are often considerably less than those in a passive array. Also a high duty
cycle can be sustained, typically up to 30% or so, compared to around 10% for a passive
array. The cost of active arrays, to date, however is considerably greater than that of
passive arrays. Also the transmit sidelobes are often higher than is desired.
2.3.2.1 Electronic beam steering

A beam is steered electronically through control of the signal radiated by the individual
elements of an array to cause these signals to add coherently in phase in the desired
beam direction [Kahrilas, 1976]. If all of the elements generate their normal pattern,
with no relative phase shifts applied between each element, the composite phase front
will propagate normal to the array face. This is shown below in Figure 2-8.
35
Direction of main beam
Radiating elements - VV
Phase shifters —*
Beamformer network —*••
Figure 2-8 Generation of a phase front with no relative change in phase between each element
By introducing a constant, relative phase difference between adjacent elements, i.e. a

phase gradient across the aperture, the beam may be steered away from the normal to the
array face. This is shown in Figure 2-9.
Direction of main beam
Radiaflng:l:rnents -
Beamformer network —*
J
Figure 2-9 Generation of a phase front with a constant phase change between adjacent
elements
The phase difference between adjacent elements must be such that the signals from each
element add in phase in the desired direction. The phase difference required at each
element along a linear array is shown in equation [ 2-6] below. It is dependent on the
path length difference, &(n-1).sin(0) , that causes the signals from each element to add
in phase in the desired direction.
[2-6]
= ! ö(n l)sin(p) forn= 1,2,..,N
where Ac1 is the phase difference (rads)

S is the element spacing
q is the scan angle away from array normal
A is the wavelength
N is the number of elements in the linear array
For a planar array the phase shift in any given element is the sum of the phase shifts
36

required in each (orthogonal) dimension. The element spacing used within the array is
usually somewhat less than half of the wavelength in order to avoid the generation of
grating lobes.
The most common way to control the phase at each of the elements in a planar array is
to introduce a phase shifter into the feed path to each element. These are of various
types and technology. Very often digital phase shifters are employed which can control
the phase among discrete values in the interval (0 , 21t). For very large arrays or when a
large bandwidth signal is required a true time delay may need to be used instead of a
phase shift that is limited to 2ir.
The antenna pattern radiated from the array may be calculated by summing the signals
radiated from each of the antenna elements;
[2-7]
E(0) =
n0
where E(9) is the antenna gain in direction 9

E is the weighting function for each element
A is the radiated wavelength
6 is the element spacing
N is the number of elements in the linear array
2.3.2.2 Properties of phased arrays

A fundamental property of the phased array is the change in effective aperture with scan
angle. The effective aperture falls off with the cosine of the scan angle, (Figure 2-10).
Effective aperture
Physical aperture
Physicalaperture __________ __________

x
Figure 2-10 Change in effective aperture with scan angle
37
This causes an increase in the radar beamwidth, and a decrease in gain. The radar
beamwidth broadens with the reciprocal of the cosine of the scan angle.
[2-8]
= cos(q)
where 6is the beamwidth at scan angle of (

6 is the 3dB beamwidth on array broadside
The reduction in antenna gain that is caused by the reduced effective aperture is
proportional to the cosine of the scan angle. In a real system the antenna gain will fall
off even faster due to the effects of the non-isotropic element pattern and mutual
coupling. The total reduction in gain is generally closer to the cosine of the scan angle
raised to the power of 1.5, although this depends on the exact elements and geometry
used. Therefore the loss in the received signal power due to the two way reduction in
gain, and thus the loss in the signal to thermal noise ratio, is proportional to the cosine
cubed. This will be assumed throughout this study.
snr =snr0cos3(co) [2-9]
where snr ;, is the signal to noise ratio at a scan angle of (

snr0 is the signal to noise ratio for the same rcs target, on array broadside
The monopulse estimate accuracy when scanned away from broadside will be degraded
twofold, since the beam is broader with respect to the broadside case, and the signal to
noise ratio is lower with respect to the broadside case. The monopulse estimate accuracy
stated in [2-5], now becomes { 2-10] below.
[2-10]
__________ 03
[ snr ). (n) J [2snr. (n). coss]

where cy , is the monopulse accuracy at scan angle (rads)
From the above equations it is seen that if uniform performance, either in terms of
detection capability or in terms of tracking accuracy, is required with varying scan
angles, then the transmitted energy must be increased. This is discussed later in Chapter
5.
Figure 2-1 1 below, shows the variation in antenna beamwidth, signal to noise ratio and
monopulse accuracy, as a function of scan angle.
38
Loss to Sional to IIoIse Ratio

- Loss to Monopulsa Acct.wacy
Beamwldtt, Bloadening
I
I
20 40
Sos, to, 1,0,0 &o.d.. (d..․)
Figure 2-11(a,b) Beam broadening, loss in signal to noise ratio, and loss in monopulse
accuracy as a function of scan angle
2.3.3 Multi-function radar system aspects
Here, some of the background pertinent to multi-function radars will be introduced. A

detailed background to the subject may be found in one of the two texts that have been
published which are dedicated to multi-function radar systems by Billeter, and Sabatini
& Tarantino ([Billeter, 1989], [Sabatini & Tarantino, 1994]). There are also chapters in
several texts ([Barton, 1998], [Kingsley & Quegan, 1992]). Firstly three important
parameters are introduced; the frame time, function time and occupancy.
The multi-function radar's total radar time is shared between many tasks or functions.
Two parameters are used to define a job such as search of a surveillance region, or the
tracking of a target. The first parameter is the frame time, which, for surveillance, is the
time over which the search of a region is made. In a tracking task, the frame time may
be considered to be the track update rate.
The second parameter is the function time, which is the total radar time that is required
to complete a job. For surveillance this would be the sum of all of the dwell times in
each beam direction of a region. For a tracking job, this would be the dwell time of each
update.
Thus with these two descriptors of a job, the distribution of the multi-function radar's
load may be calculated. The percentage of radar time that is required to complete a task
or function is called the occupancy of the task. This is defined below;
39
[2-11]
tir
where 0 is the occupancy of the task or function

ti,, is the function time of the task or function
tjr is the frame time of the task or function
These parameters allow us to describe the distribution of the multi-function radar

resources throughout the different functions and tasks.
2.3.3.1 Search
An advanced multi-function radar has the scope to modify many of its radar parameters
on a beam by beam basis. Given this, it would be wrong to assume that the correct way
to perform the search function with this type of radar is to replicate the operation of a
conventional turning search radar. As Billeter {Billeter, 1989] states, the radar's data
rate, sensitivity and type of processing need not, and usually should not, be uniform over
the volume covered by the radar. The coverage volume to be surveyed by a multi-
function radar is usually divided up into sectors, bounded by range and angle. Within
these sectors the radar parameters may be adapted to suit the environment and the
expected targets. An example of sectored coverage for a naval air defence radar is
shown in Figure 2-12, below.
Regions are usually defined around an area in which the expected target characteristics
and environmental characteristics are, very broadly speaking, homogeneous. The radar
parameters within that region are matched to give a certain level of performance against
these characteristics. For example, region 1 in the above diagram, may be characterised
as searching for fast, pop-up targets in high levels of sea clutter. The radar parameters
chosen to match these could utilise a short surveillance frame time, with excellent
clutter processing. Another region may be characterised by searching for small cross
section targets where there are high levels of barrage noise jamming. In this instance,
waveforms giving a large transmitted energy and processing techniques to counter the
jamming may be employed. These two examples assume highly threatening cases, of
great interest to a weapons system. Equally, there may be benign regions where little
radar time should be spent.
40
0 !flr
• PuMsd
o
.. I,, "..., ( G ;,
Figure 2-12 A cross section through a hemispherical search volume, showing sectored coverage
with region number and frame time
The resultant search detection performance may vary with angle, and with time
throughout the operational use of the radar. Figure 2-13 below shows an example of the
angular and temporal variation in detection performance that may be provided by the
multi-function radar. It is important for efficient utilisation of the multi-function radar,
that the radar control computer allows this to occur. The radar beam scheduler must be
flexible enough to cope with these variations. This is discussed in chapter 5.
(.) dsIsctl,n b (b) The ,son
4.) This. is..Iu..I

(c) OflSI J.ImIlflg (d) J.mmig couristed
Figure 2-13(a-e) Angular and temporal variation in multi-function radar detection performance
The matching of the radar time budget, waveforms, signal processing and other radar
parameters to these cases is a crucial part of the multi-function radar system design. A
list of some of the parameters that may be adapted in the search function is given below.
i. Transmit power vii.Beam spacing

ii. Transmit waveform (number of pulses viii.Transmit beamwidth

and bursts, frequencies, pulse length) ix. Receive beamwidth(s)
iii.Doppler processing x. Beam search pattern
iv. CFAR thresholds xi. Plot confir,nation delay
v. Counter measure processing xii.Plot confirmation waveform and energy
41
vi. Frame time
Some parameters such as frame time or Doppler processing are likely to remain constant
within a region, while others such as the total transmitted energy may vary for each
beam position.
Optimisation of the multi-function radar parameters for the search function has been the
subject of several published papers, notably those by Billam [Billam, 1982, 1986,
1989a, 1992, 1994].
2.3.3.2 Tracking
When discussing tracking radar, Skolnik [Skolnik, 1981] defines two distinct types; a
continuous tracking radar, and a track while scan radar. The former supplies
continuous tracking data on a particular target, while the track while scan radar supplies
sampled data on one or more targets, derived from the surveillance function. The multi-
function radar tracker lies somewhere between these two types of radar.
This is reflected in the tracking update rates that a multi-function radar might supply.
Broadly speaking, a continuous tracking radar might supply update rates of the order of
100-10 Hz. A track while scan radar might use update rates of the order of 1 - 0.1Hz. A
multi-function radar might use update rates in the order of 10 - 1Hz.
The beamwidth used for tracking in a multi-function radar is likely to be broader than
desired. This is due to the necessity to perform the search with the same beamwidth. If a
very narrow beam is used, the time taken to survey the volume will become too large.
For a weapons system radar, with high track accuracy requirements, the multi-function
radar is immediately at a disadvantage, having to use both a broader beam and lower
track update rates than would be desired.
The tracker also needs to recognise that a variety of targets exist, some of greater
interest than others. Billeter [Billeter, 1989] states;
'The radar needs to maintain a track, even if the user system does not want the
information. The alternative is to continually re-detect the target. It is necessary to track
a target to ignore it."
42
Re-detecting a target and forming another track on it is likely to be more expensive than
maintaining track on it. Therefore, the multi-function radar must aim to track all targets
within its coverage volume. There may be very many targets, with only few that are of
genuine interest. The algorithms for tracker control must therefore attempt to minimise
radar time and energy on targets that are not of interest, and maximise it on targets of
interest. For this reason it is likely that both track while scan and track and scan will be
used within the multi-function radar. Targets of little interest to the weapons system,
referred to as low priority targets forthwith, may be tracked using track while scan, or
track and scan with a low demanded accuracy. Targets of interest to the weapons
system, termed high priority targets, for which accurate position and velocity
information is desired, are likely to take the majority of the multi-function radar tracking
resource allocation.
Optimising the performance of a multi-function radar tracker against rapid and

manoeuvring targets is covered in Chapter 6.
2.3.3.3 Digital beamforming

In a digital beamformer the signal from each element or sub-array feeds a receiver, the
output of which is digitised. The digital signals in each receiver channel are multiplied
by a complex weight and then summed as shown in Figure 2-14 below.
The antenna pattern may now be adapted in several ways; the main beam and the
antenna pattern nulls may be steered to any desired direction. The digital signal may be
summed together in many different ways, or if the signal information is stored, it may be
passed through the beamformer several times. This allows several antenna patterns to be
generated for the same received signal information.
Antenna elements
or sub arrays V
Receiver
ND converter
weigni'2,) Muttipher
______
SUM
Digitally Beamformed Signal
Figure 2-14 A digital beamformer
43
The array output is formed from the weighted sum of the element or sub-array outputs,
as in equation [2-12] below.
nN-I 2 12]
Array_ output =
where N is the number of elements or sub-arrays

V,,is the signal from the nth channel
W,,is the complex weight applied
C,, is a receiver correction factor
This technique is used in several areas such as sonar and communications, as well as
radar. The technique has several useful applications in radar. It may be used to place
nulls in the antenna pattern in the directions of interference sources such as jammers. It
may be used for superresolution processing in angle. Also, since the same information
may be summed in different ways many times, it is possible to generate simultaneous
multiple receive beams, and effectively search in several directions simultaneously
(assuming these directions have been illuminated with the transmit beam).
Many aspects of digital beamforming have been demonstratedlresearched using the

MESAR radar. Within the scope of this study the two most important aspects of digital
beamforming are the generation of a broadened antenna beam pattern to allow a
'variable beamwidth' mode of operation, and the processing of simultaneous multiple
receive beams. These are discussed in further detail in Chapter 4.
2.4 Benefits of using a multi-function radar
The primary benefit of a multi-function radar is the ability to steer the beam at will,
under automated computer control. From this capability, the multi-function radar may
be tailored to its operating environment in a way that is impossible with a mechanically
scanned antenna. As Billeter [Billeter, 1989] states,
"Spatial allocation, temporal allocation, environmental adaptation, and situational

adaptation are all related to the inherent flexibility of the beam-steering"
For the full flexibility of a multi-function radar to be realised, not only must the beam
steering be flexible, but so must the waveform generation, pulse compression, digital
signal processing, etc. Each may be required to perform a different operation at every
44
different pulse burst.
The process of hand-over from an initial surveillance detection to acquired track can be
minimised in the multi-function radar. Once a surveillance detection is made, the multi-
function radar can immediately confirm that this detection was from a real world object
rather than noise, without having to wait for another sweep of the antenna as in a
mechanically scanned system. A series of rapid updates may then be scheduled to enable
a good positional and directional estimate of the target to be made. This process can
require many sweeps of a mechanically scanned antenna, extending the track hand-over
time. It is essential that this is minimised if a rapid weapons system reaction time is
required.
The benefits may be summarised;
• Flexibility in resource allocation, spatially and temporally
• No collimation errors
• Minimisation of the search to track handover time
• Rapid dismissal of noise false alarms
• Non predictable operation for ECCM
• Variable track data rate
• Variable surveillance data rate
2.5 Penalties of using a multi-function radar
The multi-function radar by definition fulfils a combination of radar purposes. It

replaces radars which are dedicated to a single purpose, each of which has its own
optimised parameters for that purpose, such as beamwidth, frequency, energy and
waveform. If the optimal parameters used by the individual, dedicated radar systems, are
disparate from each other, then it is clear that the multi-function radar represents a
compromise in those respects.
This is the case in practice. A typical naval tracking radar may operate with a frequency
of around 10GHz, and a beamwidth of a fraction of a degree. This will give that tracking
radar great angular plot accuracy with a relatively small antenna size. The surveillance
radar that may be used to cue the tracking radar is likely to operate at a frequency of
around 1GHz, with a beamwidth of several degrees in azimuth, and possibly tens of
degrees in elevation. This gives the surveillance radar a smaller number of beams to
45
search the volume, thus allowing a greater dwell time per beam position for clutter
cancellation and signal integration, with a penalty of decreased angular plot accuracy. If
a multi-function radar is then adopted to perform both of the functions of search and
track, there are immediately compromises that must be faced.
If the multi-function radar in this example were to adopt the parameters of the tracking
radar, although its performance in that respect would be excellent, the time required to
survey the sky for targets with such a narrow beam would be prohibitively long making
it useless for the search function. Similarly, if the surveillance radar parameters were
adopted, the tracking performance would not be adequate with such a broad beam. A
compromise must be made on the parameters adopted for the multi-function radar that
means it is sub-optimal for both functions.
Furthermore, the multi-function radar has to divide its time between the two functions.
The tracking radar in the above example may dedicate all of its time to just one target,
where as the multi-function radar not only has to share its time between many other
targets, but also between the surveillance function. The multi-function radar therefore
has significantly less time for this function than would ordinarily be desired. This is
reflected in reality, by dedicated naval tracking radars providing update rates of tens of
Hz, e.g. 50Hz, whereas a typical naval multi-function radar would only support update
rates in the order of a few Hz for the tracking function.
The impact on the tracking function is thus twofold. Firstly the beamwidth is likely to be
broader than would ideally be desired. Secondly the track update rate that may be
supported is likely to be significantly lower than is desired. The outcome is that the plot
accuracy from a multi-function radar may not be as good as that from a dedicated
tracking radar, and that, due to the lower update rate, there is an increased probability
that the target may manoeuvre and not be illuminated at the next update.
Both the compromise in beamwidth and the tracking issues discussed above are
addressed within the scope of this study, in Chapters 4,6 and 7.
The penalties may be summarised;
• design compromises
• complex hardware, e.g. adaptive digital signal processing or beamforming
• complex software for control
46
. resource budget limitations
• limited spectrum
• vulnerability
• cost
Cost is a disadvantage both in terms of the research and development required for a
multi-function radar system, but also in terms of hardware. Phased arrays, especially
active phased arrays, are very expensive. Also, for each array face that is added, further
hardware and signal processing are also required. Often systems use only one or two
array faces to minimise cost, resulting in the multi-function radar having less than the
full 3600 instantaneous azimuth coverage. In this case the antenna must rely on
mechanical rotation to give the full azimuth coverage.
2.6 Static versus rotating systems
The choice between the use of a static or a rotating antenna for an multi-function radar
is most likely to be made on the basis of cost. In many cases, a system with enough faces
to cover the required search volume with no mechanical rotation is too expensive, or
unnecessary in terms of required performance. In these cases an antenna with one or two
phased arrays may be employed, and mechanical rotation ensures that the whole volume
may be covered over a certain time.
In a rotating multi-function radar there are some significant complications to the radar
control problem that do not exist in the case of a static face multi-function radar, aside
from the increased mechanical complexity. In light of this, it is worth pointing out that
there are some benefits to be accrued by rotating the array antenna.
Surveillance is essentially a repetitive, periodic function. The search of beam directions

may be synchronised with the rotation rate of a rotating array, so that the search
function is performed on array azimuth broadside. This minimises the losses associated
with scanning. Equally, in the tracking function, if updates on a particular target may be
performed on array azimuth broadside, then the minimum beamwidth will be obtained,
giving the maximum azimuth plot accuracy and resolution.
These factors can be very significant and are investigated further within Chapter 4,
which makes a quantitative comparison of a variety of static and rotating array antenna
47
configurations, whilst also assessing the impact of multiple receive beam processing in
those systems. A recent comparison of static and rotating multi-function radars has been
made by Billam [Billam, l997b], which concluded that a single faced rotating system
was the most efficient option.
Another advantage of a rotating system is the possibility of distributing the radar time
around the coverage volume in a way that is not possible with a static system. If a small
sector exists with a higher than normal load of radar tasks, then in a static system, the
one array face pointing in that direction will be more highly stressed than the other
faces. In a rotating system it is possible to share the extra load amongst all of the faces
to some extent. This is a technique that seems not to have been considered in the
literature. It is discussed in further detail in Chapter 5.
48
3 MESAR: AN EXAMPLE MULTIFUNCTION
RADAR SYSTEM
3.1 Introduction
This section gives an overview of the MESAR multi-function radar and the new
capabilities that will come with the development of MESAR II, since the research from
this study will be directly applied to that system. This chapter aims to give an indication
of the performance and capability of a current, state of the art multi-function radar
system, and to show what has been achieved and what flexibility remains to be
harnessed in an optimal way. It pays particular attention to the tracking and control
issues, describing the current MESAR tracker and the results that have been obtained
through aircraft trials, and the advances in MESAR II that may be used within the
tracker. A more detailed account of the system is given in many papers ([Moore, Levett
& Butler, 1995], [Billam, 1986, 1987 & 1989b], [Salter, 1988], [Moore & Salter,
1995]), from which this overview is formed.
MESAR is an advanced experimental active phased array radar which has been
developed collaboratively between the Defence Evaluation and Research Agency and
British Aerospace Defence Systems (formerly Siemens Plessey Systems). It was
developed through the late 1980s and early 1990s exploiting the recent advances in
digital and solid state microwave technology then available. The complete fully
functioning multi-function radar, Figure 3-1, was delivered to the DRA in early 1993,
and is now used as a test-bed for experimentation and demonstration of techniques that
will be applied to future multi-function radar systems.
The MESAR demonstrator radar was not intended to give the full radar performance
that would be expected of an operational multi-function radar, but to be capable of
demonstrating the system functionalities and control needed by such a radar. In fact, the
MESAR demonstrator has surpassed this initial requirement and has been able to
demonstrate several features with performance in excess of any conventional radar
system. As such, it is one of the most advanced radars of its kind in existence. The
LONDON
UNIV.
3 MESAR: An Example Multfunction Radar System
current MESAR experimental system incorporates the following features:
Figure 3-1 The MESAR multi-function radar
• A single solid state active phased array, with multiple sub-arrays. The array is thinned
due to cost constraints on the research programme.
• A single agile surveillance beam.
• Simultaneously available pencil, circular difference and sidelobe blanking beam shapes.
• Wide-band frequency agility.
• Sectored surveillance for efficient use of energy.
• Variable surveillance data rate.
• Selectable choice of pulse repetition interval, expanded pulse length and compressed
pulse length.
• Digital waveform generation giving a choice of pulse lengths and pulse compression
laws.
• Digital pulse compression.
• Digital adaptive beamforming giving anti-jamming capability for the sum, difference
and sidelobe blanking beams.
• Programmable, parallel signal processing.
• Variable track update rate.
• Computer control of all radar functions.
• Data recording for off-line analysis and processing
A simplified block diagram of the MESAR system showing the components is given in
Figure 3-2. After combination of element outputs into sub-arrays, the entire system is
digital, and therefore highly flexible.
50

_____ /RX
I I I I I I ]MODULE Lfl WI
1A/1l
_
II I DIGITAL
I BEAM PULSE
CONTROL g
7OESSION
TX/AX one per
sub-array
- L____J
-i SIGNAL
FiBRE PROCESSING
OPTIC
CONTROL
_____________ 4
ARRAY TASK PLOT I
SCHEDULER POSITION I
__________H EXTRACTOR
RADAR
WAVEFORM CONTROL TRACK
I GENERATOR AND EXTRACTOR I
Figure 3-2 A simplified block diagram of the MESAR system
A new system is now under construction, creatively called MESAR II, which will
include a new active array and signal processing. The tracking strategies that are
developed from this study will be tested directly in experimental trials using the
MESAR II radar.
3.2 Array face and transmitlreceive modules
MESAR has a single, static, active array face. It consists of nearly one thousand
radiating elements, although only one in six element positions is populated with an
active transmit receive module due to cost limitations, see Figure 3-3. This results in the
system having a lower output power, and higher antenna sidelobes than one using an
equivalent fully filled array.
Figure 3-3 The thinned MESAR array
51
3 MESAR: An Example Mull ifunction Radar System
The array is just over 1 .8m in diameter, with elements in a triangular lattice which
radiate horizontally polarised waves. Operating at around 3GHz this gives the system a
beamwidth of around 30
The transmit/receive modules contain a transmit power amplifier and a low noise
receive amplifier based on GaAs microwave integrated circuitry. The module also
contains a 4 bit digital phase shifter and a phase computation and control processor, so
that each module in the array can set its own phase shift. The phase computation
requires only the position of the element in the array, which is stored in ROM, and the
desired beam pointing direction and radio frequency, which are broadcast to all of the
modules over a fibre optic link. This distributed phase computation avoids the need for a
central phase computation engine and control system.
The new modules fabricated for MESAR H will allow several transmit and receive beam
shape configurations. These may be used for generating variable transmit and receive
beamwidths independently, thus allowing either broadened transmit/receive beams
giving variable beamwidth operation, or broadened transmit with narrow receive beams
to allow simultaneous multiple receive beam processing in the digital beamformer.
Each of the MESAR modules generates around a 2W output power, and can operate
with a mean duty cycle of around 30%. This gives the MESAR system a very low mean
output power of less than 100W. Therefore, in order to detect targets at any reasonable
distance, the system requires a large amount of processing gain, which it achieves both
through coherent and non coherent integration. The waveforms that are used in MESAR
are thus unusually long in comparison to those which would be expected in an
operational system.
3.3 MESAR adaptive beamforming
Digital adaptive beamforming in MESAR is used to provide an effective capability to

cancel jammers both in the antenna sidelobes, and in the skirts of the main beam, termed
adaptive nulling.
A block diagram of the MESAR digital beamformer is shown in Figure 3-4. The
element signals for groups of elements are combined using analogue beamforming into a
52
3 MESAR: An Example Multfunc:ion Radar System
number of subarrays. The signals from these subarrays are digitised, and the digital data
from each subarray channel, corresponding to all the range cells in one or more pulse
repetition intervals, is stored in a buffer. The weight computation unit now has the task
of calculating, in real time, the required phase and amplitude weights needed by each
subarray channel. The weights are chosen to allow the antenna gain characteristics in the
required look direction to be maintained, whilst simultaneously generating nulls in the
antenna pattern to adaptively cancel jamming and interference sources. These complex
weights are then applied to the data stored in the buffer.
--1 RxandADC ..- z
-
Sub array
receivers
a. U
.
5.6
a- -
Ax and ADC
Figure 3-4 A simplified block diagram of the MESAR digital adaptive beamformer
Three sets of weights are calculated in MESAR in order that sum, difference and
sidelobe blanking beams are generated. All three beams will then be protected from
jamming and interference sources.
In MESAR IL the digital adaptive processing is to be significantly enhanced to include

processing for super resolution techniques and for generating multiple receive beams.
Both of these functions may be used to aid the tracking function.
3.4 Radar control
Central to the success of MESAR is its radar control system. The radar control system is
designed to allow the multi-function radar to achieve the required performance from
each of its radar functions. The MESAR Radar Control process provides for the
following [Stafford, 1990]:
A priority structure for radar functions such that each function takes absolute priority
over all other functions with lower priorities.
. An algorithm that determines in real time the sequence in which radar tasks must be
53
3 MESAR: An Example Multifunction Radar System
performed. It also permits the dynamic variation of data rates and dwell times.
Adaptive waveform control according to environmental and ECM conditions.
The ability to select in real time, waveforms for both surveillance and tracking which
match the clutter and jamming conditions, and provide the required detection
performance.
Operating in real time the radar control process will be able to rapidly schedule 'look-
back tasks for plot confirmation and track initiation.
To keep the antenna fully occupied for efficient use of transmitter energy.
MESAR's radar control is covered further within Chapter 5.
3.5 Adaptive signal processing and waveform generation
The flexibility asked of MESAR impacts upon the design of the entire multi-function
radar system. Not only does it require a flexible control system, but also flexible
waveform generation and adaptive signal processing.
MESAR uses a digital, programmable waveform generator. Digital samples of the pulse
modulations for a variety of transmitted pulse lengths and frequency sweeps are stored
in RAM. On transmission, the samples for the chosen waveform are passed through a
D/A converter, filtered and frequency shifted. The transmit waveform may be changed
on a pulse by pulse basis if required.
On receive, a purpose built digital pulse compressor convolves the returns with the
sampled weighting function of the matched filter, which is suitably weighted to give low
range sidelobes. The filter coefficients are programmable.
The remainder of the signal processing, Doppler filtering, constant false alarm rate
processing, monopulse angle estimation and plot extraction, is performed on parallel
distributed array processor containing 1024 processors. The processor has to be
programmable so that the processing applied can be matched to the waveform
transmitted; the signal processing algorithms are coded in parallel Fortran.
54
3.6 The plot processor and tracker
This section describes some of the adaptive aspects that are already an integral part of
the current MESAR tracker.
3.6.1 Plot processing

The plot processor in MESAR [Noyes, 1990] uses a simple, nearest neighbour algorithm
in most cases. However, unlike most other phased array systems using de-coupled
tracking, it takes some account of the problems that may occur with closely spaced
targets.
When the multi-function radar makes an update on a known track, the multi-function
radar pencil beam will only illuminate the area of sky close to that track. This area of
sky may include parts of the association gates of other targets. However, it is unlikely
that the whole of the association gate for these other tracks will be included. The picture
with which to make the correct association in this circumstance is not complete after a
single update on a single track.
The effect of associating incorrect plots into a smoothed track is to degrade the track
accuracy, and this should be avoided. Therefore, when a track update gives a conflict in
association, the strategy adopted in the MESAR tracker is to avoid this by requesting an
early update on the conflicting track. Then the tracker will await the output from this
update, before making either association.
Data received from the surveillance function on a track is not fed into the track
smoothing algorithms, since MESAR does not currently use monopulse during
surveillance, thus the accuracy of the plots is significantly poorer.
3.6.2 Track filtering

The current version of the MESAR radar uses a single filter model for track smoothing.
This is a constant velocity Kalman filter, with a manoeuvre detector strategy, which
gives the filter an adaptive memory based upon the target manoeuvre conditions [Noyes,
1990]. This method allows a simple filter to be employed, with a relatively low plant
noise, giving a higher track accuracy in phases of constant velocity flight. In the case of
a manoeuvre the plant noise is increased somewhat to allow the filter to be more
55
adaptive to target manoeuvres.
Development of a simulation of the MESAR Kalman filter based tracker is covered in

Chapter 6.
3.6.3 Adaptive track update rate

The look back interval of a track in MESAR is adapted on an update by update basis
[Noyes, 1990]. The time of next update that is calculated is based on two factors;
i. there is a high probability that the target falls within the 3dB beam on the
next update. This value will depend on the azimuth and elevation smoothed
accuracy of the track, and on the amount that the track can possibly
manoeuvre in azimuth and elevation.
ii. there is a constant smoothed track accuracy in position or velocity.
Choosing an update time that is too large will mean that one of the above is not fulfilled.
Choosing and update time that is too small will mean the tracking function will increase
the occupancy of the tracking function over its optimal value.
Thus in MESAR, there are two times that are used when requesting a track update;
a. latest update time. This is the earliest of the two possible outcomes of (i)
above, depending on whether the azimuth or elevation manoeuvre capability
is larger. It represents the time by which an update must be made before
which there is a chance that the target will fall outside of the 3dB beamwidth,
and not be detected.
b. desired update time. This is the time which is given from (ii) above. It must
be less than the latest update time calculated in (a). If the update is not made
on time, then it is unlikely that the track will be lost until (a) above is
exceeded, but will mean the required track accuracy is not achieved.
Tracks will be scheduled using their desired update time. However, if the latest update
time is exceeded, then the track task is elevated in priority to the track maintenance
level, to ensure that it is scheduled at the earliest opportunity.
56
3.6.4 Adaptive transmit energy

Once a target is under track, unlike surveillance, we have some idea of the radar cross
section of the target we are looking for. We may therefore tailor our transmit waveform
to illuminate the target with only a sufficient amount of energy to give a desired
detection performance.
A smoothed estimate of the signal to noise ratio is built up using a simple a smoothing
filter with the measurements of signal to noise ratio from each update. This smoothed
estimate is maintained between 26dB and 29dB by adapting the number of MTI bursts
or pulses that are used in the transmit waveform [Stafford, 1991].
The objective of this is to ensure energy is not wasted on very large targets, or on targets
that have got close in range to the radar, which would normally give signal to noise
ratios in excess of useful levels.
3.6.5 MTI velocity response matching

During surveillance, no range rate information is available on the targets that are being
searched for. Therefore several moving target indication (MTI) bursts must be
transmitted in each beam direction to eliminate the blind velocities of each MTI filter.
In the tracking function however, an estimate of the target range rate is available. The
MTI filter chosen may therefore utilise weights that ensure the estimated target range
rate lies at the peak of the filter response [Stafford, 1991]. This gives two benefits;
the target receives the maximum gain through the MTI filter
. the amount of radar time required for each target is kept to a minimum
3.7 Trials results
The first set of MESAR trials were undertaken in 1988 when a number of receive only
tests were completed with the array in order to characterise the operation of the digital
adaptive beamformer ([Salter, 1989], [Moore & Butler, 1994]). A typical result is shown
in Figure 3-5, showing both adapted and unadapted beam patterns. This clearly
demonstrates the adaption of the antenna pattern to generate nulls in the direction of
jammers.
57
3 MESAR: An Example Mul:func:ion Radar System
Azimuth
Figure 3-5 Adapted and unadapted beam patterns in the presence offammers
An 18 month programme of trials commenced in early 1993. These were conducted at

the DERA antenna test range, Funtington and the MoD Test and Evaluation range at
West Freugh, Scotland, and were designed to test the radar performance in severe
environmental and jamming conditions. These are reported fully in [Butler, 1995]. An
overview of some of these results is given below.
3.7.1 Multi-function operation

Figure 3-6, below, is a photograph of the main radar display in the form of a Plan
Position Indicator, showing surveillance and multi-target tracking results. The following
information is displayed:
. Surveillance plots - the main detections from the surveillance functions. At this stage
surveillance plots will include false alarms.
• Confirmed track update - those plots which are confirmed as true targets
• Missed track update - a track update which failed to detect
• Slow tracks - those tracks which fall below a specified velocity threshold
Whilst this display is only a snap-shot in time, it demonstrates the multi-function

operation of MESAR.
3.7.2 Tracking
An established track is maintained by a series of updates, the rate of which can vary
according to the target range and tracker confidence in plot prediction of the tracked
object. Figure 3-7 shows the track plots obtained from two weaving aircraft. The
spacing of the track plot updates shows the adaptive nature of the MESAR tracking
function.
58

3 MESAR: An Example Mu1rfunction Radar System
Figure 3-6 The MESAR PPI display, showing a recently initiated track
20.0
15.0
I
10.0
5.0! I
.
5.5 0.0 5.0
Rang. ., km
Figure 3-7 Adaptive update rate tracking
Of special interest is the situation where a track splits into two. The benefit of
recogrnsing split tracks is the early detection of multiple targets as they resolve and the
ability to rapidly form tracks on the individual objects. Figure 3-8 shows a
demonstration of this function through the separation of a l000lb bomb from an aircraft
59
3 MESAR: An Example Multfirnction Radar System
flying at 32,000ft.
I
I
0 20 10 10 00 100 100
100.0, S.,.00
Figure 3-8 Track splitting function
3.8 Summary
MESAR is a fully functioning, active multi-function radar that already operates in real
time using an adaptive tracker and control system. It consists of a single static array
face, which is capable of generating only a single beam. Therefore it would be beneficial
to consider the impact of a rotating array antenna and multiple beam techniques on the
control and resource allocation of such a system.
Broadly speaking there are two main areas in which the tracker may be improved. The
first is to use more advanced track filtering and plot association methods such as the
Interactive Multiple Model (1MM) filtering methods, the various Joint Probabalistic
Data Association (JPDA) methods, or Multiple Hypothesis Tracking (MHT) techniques.
These issues are being addressed under the construction of MESAR II and the current
MESAR research programme, and will not be covered within this study. The second
area for improvement is that of utilising the new functionalities available such as
variable beamwidths, simultaneous multiple receive beam processing and super-
resolution techniques, as well as optimising the performance for a rotating system. This
is the area that is followed in this study.
60
4 THE IMPACT OF VARIABLE BEAMWIDTH AND
MULTIPLE BEAM PROCESSING ON MULTI-
FUNCTION RADAR
4.1 Introduction
For a multi-function radar designed to operate in severe levels of clutter, a fundamental

drawback is the compromise that must be made in the antenna beamwidth used for
surveillance, tracking and other functions.
For tracking a very narrow beam is required to give the maximum tracking
accuracy and target resolution in a multiple target environment.
For surveillance a broader beam is desired so that the time taken to perform the
search is minimised, especially in regions of high clutter, where long Pulsed
Doppler waveforms must be used to ensure adequate clutter cancellation.
The requirement for keeping the search occupancy within reasonable limits results in a
beamwidth that is broader than is ideal for the tracking function. Significant benefits
would be gained if a narrower beam could be supported without increasing the
surveillance load excessively.
Future derivatives of multi-function radars such as MESAR will offer the capability to
exploit digital control in shaping the transmit and receive antenna patterns. These will
allow the beamwidth to be broadened through a suitable phase or amplitude weighting
imposed at the array elements. Also multiple receive beams may be processed, either
through provision of multiple receive channels, steered at the element level, or through
the use of digital beamforming.
The inherent compromise with a multi-function radar on beamwidth does not

necessarily hold with a variable beamwidth system. Search and track beamwidth could
be de-coupled and each be optimised to their task. The impact of these variable antenna
patterns on the multi-function radar is discussed here, working to the assumption that, in
61
4 The Impact of Variable Beamwidth and Multiple Beam Processing on Multifunction Radar
general, the narrower the tracking beam is, the better the tracker performance will be, in
terms of resolution and tracking of closely spaced targets, performance in multipath and
jamming, and track accuracy. In particular, the impact of variable beamwidth and
simultaneous multiple receive beam processing on the search time-budget is assessed.
Sections 4.2 and 4.3 discuss the generation of broadened beams and the use of
simultaneous multiple receive beams. In section 4.4 novel methods of estimating the
search time-budget for both a static and a rotating multi-function radar are developed.
These are then applied in sections 4.5 and 4.6. In section 4.5 an accurate comparison of
static and rotating multi-function radar systems performing the search function is made,
which appears not to have been reported previously in the literature. Also in that section
the impact of using broadened transmit and receive beams and simultaneous multiple
beam processing is made in thermal noise only conditions. Section 4.6 then makes some
account for the effects of clutter. Finally, section 4.7 presents some ways in which
simultaneous multiple receive beams may be used for tracking in multi-function radar,
and the conclusions of the work are contained in section 4.8.
4.2 Generating a broadened main lobe antenna pattern
It may be desirable to broaden the main lobe of the antenna pattern in some
circumstances to enable a larger area of space to be illuminated than would occur with
the uniformly illuminated antenna. This may be performed in either transmission or
reception, and may be coupled with the use of simultaneous multiple receive beam
processing, discussed in the next section.
The field of pattern synthesis has been the subject of a great deal of research, and much
published literature exists on the subject. It is not an objective of this research to
investigate or present techniques for pattern synthesis. However, it is important to
demonstrate the feasibility of generating broadened main lobe antenna patterns since
much of the work in the following chapters depends upon it.
An effective solution to the problem of radiation pattern synthesis for an array antenna
has been given by Milne [Milne, 1987]. This method consists of selecting a 'scanning
function', which is a physically realisable antenna pattern with desirable characteristics,
such as low sidelobes. This antenna pattern may then be convolved with the ideal
62
4 The Impact of Variable Beamwidth and Multiple Beam Processing on Multfunction Radar
pattern that is desired. This will yield a physically realisable antenna pattern that
approximates the ideal pattern. The method is superior to methods based on the
Woodward synthesis technique [Woodward, 1947], since it utilises both the amplitude
and phase of the array illumination function. For an N element linear array, this gives
2N- 1 rather than N degrees of freedom.
Figure 4-1 below shows the results of a convolution of a scanning function based upon
the Dolph-Chebychev weighting function, used to obtain a uniform sidelobe level, with
several ideal broadened main lobe antenna patterns. A 50 element linear array was
considered, operating at a wavelength of 0. im, and with a half wavelength element
spacing. This resulted in a scanning function with a beamwidth of 2.76°, and a sidelobe
level of 40dB below the main lobe peak (shown by the blue line in each figure).
- -44) -.44, - -50 0 to 5.0 44) 5.0
(a) (b)
I, -.-.----.... -7 .—.-., ---------
a . 7 4) a - __ __
(c) (d)
Figure 4-1(a,b,c,d) Synthesis of a broadened main lobe antenna pattern using Mime 's method,
showing broadening of the scan pattern main lobe by (a) a factor of 2, (b) a factor of 4, (c) a
factor of 6, (d) a factor of 8. Blue line shows the scan pattern, red line shows the desired
pattern, green line shows the realisable pattern.
Ideal patterns consisting of a rectangle function over the desired angular sector were
chosen. The width of rectangle in the ideal pattern was related to the 3dB beamwidth of
the scanning function. Broadening factors of 2, 4, 6 and 8 times the scanning function
3dB beamwidth were chosen (shown by the red line in each figure). The physically
63
realisable patterns, (shown by the green line in each figure), are seen to retain the low
sidelobes whilst approximating the ideal pattern well.
This method is appropriate to both a passive array multi-function radar and also an
active array multi-function radar when in reception. However, in transmission the
amplitude of the active array must remain uniform, and the formation of the desired
pattern must be made using only a variation in the element relative phases. In this
instance, the broadened main lobe antenna pattern will suffer from the high sidelobes
associated with the uniformly illuminated active array antenna.
4.3 Processing of simultaneous multiple receive beams
A radar may observe multiple directions simultaneously, by illuminating a large area

with a broadened transmit beam, and generating multiple narrow receive beams to cover
that area, as shown in Figure 4-2, below. This is most commonly performed in 'stacked
beam' rotating surveillance radars, where all elevation angles of interest are surveyed in
one observation. The cost of the technology required to achieve this has prevented the
technique from becoming widespread in multi-function radars thus far.
rransmit
Receive
a-
Step ecan with broadened
Transmit beam
Receive with multiple Other possible beam groupings
receive beams
Figure 4-2 Example tx and rx beam formations using simultaneous multiple beam processing
Currently, real time digital beamforming at element level for arrays operating at the
desired radar frequencies are not practicable. Therefore, digital beamforming is usually
performed at subarray level, where the analogue output of many elements is combined
before analogue to digital conversion. Formation of multiple receive beams may either
take place at subarray level, through digital beamforming, or at the element level, as one
or more separate receiver channels.
64
The advantage of forming the beams at subarray level is the adaptive control over the
beam pattern clusters that are generated. Beams may be separated by different offsets,
and in different patterns if required. However, there are some constraints laid in forming
the beams at this stage in the receiver chain. Firstly, the receive beams formed must be
within the sub-array beam width to avoid degradation in gain. Secondly, beamforming at
sub-array level introduces a quantisation in the phase taper across the array, which can
degrade the sidelobe levels of the receive beam. These disadvantages are eliminated if
the beamforming is performed at element level, but at the expense of providing a
separate receive channel for each beam.
MESAR H will initially provide multiple receive beams at the element level. This will
be achieved through using the difference channel, thus only two simultaneous receive
beams may be generated. It is anticipated that multiple beamforming at sub-array level
will be added at a later date.
4.4 Calculation of the search function time
The search function time is the total radar time needed to perform a search of a given
region of space. It is determined by two factors; the number of beam directions that are
needed to fully cover the region of space, and the dwell time required in each beam
direction. From this parameter the search occupancy may be obtained by including the
frame time over which the search is to be completed. For a single face, equation [ 4-1]
gives the search occupancy.
search function_ time [4-1]

O CC CUT( h = -
search_ frame_ time
It is usual (see section 2.3.3.1) for the search volume to be divided into sectors, each
sector having a uniform detection performance with angle (in most conditions). This
section presents novel methods for estimating the search function time for a given
sector, for both a static and a rotating multi-function radar system. These methods are
compared to exact calculation of the beam lattice for every sector. The methodology is
then applied in the following sections 4.5 and 4.6.
65
4.4.1 Static phased array multi-function radar system

If the antenna characteristics were invariant with scan angle, the calculation of search
function time would be a simple matter. This is the case for a spherical array, where the
beamwidth and gain do remain invariant with scan. The search function time is given
simply by
i=n_heamx
[4-2]
function_ time = dt. = n_beams. dt
i=1
where dr is the dwell time in each beam direction

n_beams is the number of beams
since for all directions the same dwell time will give the same detection performance at
any angle. It is possible to avoid calculating each beam direction explicitly to derive the
number of beams. If it is assumed that the beams are arranged in a triangular, close
packed lattice, with their 3dB beamwidth contours overlapping, the number of beams
may be approximated by the ratio of the solid angle of the sector to be searched to the
solid angle of the beam (at its 3dB contours). Then a good approximation of the search
function time is given by equation [4-3] below.
sector
[4-3]
jimction_ time dt
beam
where Qector is the solid angle of the search sector

Qheam is the solid angle of the beam
With a planar array, however, the calculation is complicated by the change in effective
aperture with scan angle, varying both the beamwidth and detection performance. The
calculation is further complicated by the tilt of the array with respect to vertical. These
issues make determining the surveillance function using simply the real world
coordinate system difficult.
Fortunately, the calculation is made simpler if we apply a strategy often adopted by

antenna engineers for this problem, which is to use direction cosines in the antenna
coordinate system, often called UV space [Best, 1962].
This coordinate system has the useful property that the antenna beam pattern remains
invariant with the scan off angle from broadside. The coordinate system and equations
are shown below. (It is important to note the angles 0 and p do not correspond to
66
4 The Impact of Variable Beamwidth and Multiple Beam Processing on Mull jfunction Radar
azimuth or elevation in this case). The unit vectors u and v lie in the plane of the array.
The angle 9, measured from the normal to the array plane, represents the scan angle.
u = sin(0)cos(q)
14-4]
V
v = sin(6)sin(q,)
Figure 4-3 The UV space coordinate system and equations
A useful visualisation of the UV space transform has been described by Von Aulock
[Von Aulock, 1960]. This is shown below in Figure 4-4(a). Beams projected from the
plane below the unit sphere (the UV plane) onto the surface of the unit sphere broaden
exactly as does the antenna beam, scanned from array broadside. Once the beamshape is
known in UV space it is relatively simple to calculate a square or triangular lattice of
beam directions since the beam may be linearly translated in UV space without regard to
the scan angle of the beam, as shown in Figure 4-4(b). This projection is equivalent to
the orthographic projection [Chamberlin, 1947], used in the field of cartography for
projecting lines of latitude and longitude onto a sheet of paper.

U U
Figure 4-4 (a) A visualisarion of UV space, showing beam broadening, (b) the same
visualisation showing a triangular lattice of beams
To relate the UV coordinates of a beam to real space coordinates we must include a

rotation between the two coordinate systems to account for the tilt of the array from
vertical.
67
[451
v = zcos(a) - y sin(a)
w = ycos(a) + zsin(a)
Figure 4-5 The relationship between the uv and real world coordinate systems, and the
associated equations
Figure 4-5, above, shows the relation between UV space and real space for an array
tilted back cx° from vertical. The unit vector w is assigned to the array normal direction.
The transformations from UV space coordinates u, v, w to real world coordinates x, y, z,
are those in equations [4-5].
Therefore we may derive u and v from azimuth and elevation, and vice versa, using
equations [4-6] and [4-7].
u = cos(el)sin(az)
v = sin(el) cos(a) - cos(el) sin(az) sin(a) [4-6]
w = cos(el) sin(az) cos(a) + sin(el) sin(a) = fi - ( u2 + v2)
U
az = tan -I
l_(u 2 +v2)cos(a)_vsin(a)
J
_________ [4-7]
el = sin'(vcos(a) +jl_(u2 + v2) sin(a))
Figure 4-6, below compares lines of constant azimuth and elevation for an array which
is vertical, and an array which is inclined 15° back from vertical. We see that in the
tilted array case, lines of constant elevation are not horizontal as they are in the vertical
case, implying that to trace a line of constant elevation, the array scan angle is constantly
changing.
68

4 The Impact of Variable Beamwidth and Multiple Beam Processing on Multfinction Radar
06 •.. - 08 - - .......
06 0.6 . --------.. . -. _: •. -
0.4 - - -. - 0.4 :- .-------------.. -
0.2. 02 -. -
0.
-02 - .. -0.2. :------ .
-0.4 -;.. - . -04-
-05 -06
. j .
-06 -
-
-0.5 0 0.5 I -I -0.5 0.5
Figure 4-6(a,b) A comparison of lines of constant azimuth and elevation for (a) a vertical array,
and (b) an array inclined 15 °back from vertical
It is now a relatively simple matter to calculate the exact beam directions for a given
search volume using UV space. Figure 4-7 shows the calculated beam directions using
one of the MESAR search configurations as an example. The limits of this configuration
are ±45° in azimuth and 1.18° - 52.5° in elevation. An array broadside 3dB beamwidth
of 3.1° is assumed, with beams overlapped at 0.85 beamwidths. The exact number of
beams that are used for this search sector in MESAR is 595.
Since we now have a spatial system in which the beam remains invariant, we can apply
the same method as used for the spherical array above in equation [4-2], to estimate the
number of beams in the search volume, without explicitly calculating each beam
direction. This estimate is given by the ratio of the search sector in UV space to the area
of the beam in UV space. The area of the beam in UV space, Amuv, is given simply
by equation [4-8].
Figure 4-7 Beam directions in UV space, using MESAR as an example

=Jrsin 2 (6 3 ) [4-8]
The area of the sector to be searched, may be found by integration between the
69
4 The Impact of Variable Beamwidth and Multiple Beam Processing on Mulrfunction Radar
curves which define the sector in UV space, shown in Figure 4-8(a) below. The area is
given by [ 4-9], here expressed as the sum of three integrals of the three curves
bounding the sector in UV space.
II .1uj&ll 1
[el_mm Jaz=az_max
ec wr_ = 2 +[Uaz_max.ei_min Va max,eimmn

f [4-9]
[0z
[ __
- Uaz- x,el - - ,e1
- I fv z l
- L o iel=eLrnax)j
This integral may either be calculated numerically, or in this case it has an analytical
solution which may be evaluated. This gives a result within 8% of calculating the
number of beams exactly. This is sufficient to perform an initial analysis of the search
function time.
To obtain the search function time, we also need to include the dwell times. The dwell
time of each beam is given by the broadside dwell time, dt broadsjde, divided by cos3(0). If
this is plotted as a surface above the beam directions in UV space, where the height of
the surface represents the dwell time at that angle, then the volume under this surface
represents the total radar time for search. This is shown in Figure 4-8(b).
I3
-05 ..
7l
-1 -0.5 0 0.5 -1 -1
V
U U
Figure 4-8 (a) MESAR coverage limits in UV space, (b) Beam dwell time plotted as height
above the UVplane
We may calculate an average dwell time, from this surface, which when multiplied
70
4 The Impact of Variable Beamwidth and Multiple Beam Processing on Multfunc:ion Radar
by the number of beams, gives us the search function time.
el_max az_max dtbT,5j

1 [4-10]
dt,,=
$ j
A k,r 'iv el_mm az_mm
cos3(6)
This integral has no analytical solution, so it must be solved numerically.
I
j
Figure 4-9 (a) Number of beams, and (b) average beam dwell times as a function of beamwidth.
Crosses represent answer from exact calculation of all beam directions and dwell times, the line
represents the estimate from the UV method outlined above
Thus we can determine the area of the coverage sector and volume under the dwell time
surface by simply integrating the expressions for u, v and dtav , over the desired azimuth
and elevation limits. This then allows the calculation of the number of beams and beam
dwell times as a function of the beamwidth. Figure 4-9 compares the estimates given by
this method to those from calculating every search beam direction and dwell time
exactly, using the same MESAR search sector as an example. The method gives
excellent correlation to within a few percent. The search function time is simply the
product of the number of beams and the average dwell time down each beam.
4.4.2 Rotating phased array multi-function radar system

In this section a new spatial system is derived to allow the same methodology that is
used to determine the search function time for the static phased array system, described
above in section 4.4.1, to be applied to the rotating phased array system.
Assuming that the surveillance function is to be carried out on array azimuth broadside,
then scanning only occurs in the elevation plane. UV space is not an appropriate way to
represent the search sector in these circumstances, since it depends on the beam
broadening in both planes. An equivalent spatial representation may be derived in which
71
4 The Impact of Variable Beamwidth and Multiple Beam Processing on Multqunction Radar
the beam areas remain constant for an elevation only scanning system.
A visualisation of this spatial system is shown in Figure 4-10 below, with the associated
equations relating Ur and Vr to azimuth and elevation.
Ur = az. cos(el) [4-11]

VT = sin(el - a)
UT
az = [4-12]
::'•' .•.:.ç.••: (sin'(v) i-a)
el= sin'(v)+a
Figure 4-10 A visualisation of the rotating UV space, and the associated equations
A unit hemisphere with lines of constant azimuth and lines of constant elevation plotted
on its surface is shown. The lines of constant elevation are 'unwrapped' onto the surface
of a cylinder. Since they are unwrapped, no distortion of the azimuth occurs. The lines
on the cylinder are then projected onto a flat surface in the x-z plane. Comparing this to
the UV space projection above, where the surface of a sphere is projected onto a plane -
thus distorting space in 2 dimensions, in this case the surface of a cylinder is projected
onto a plane - causing a distortion in just one dimension - the elevation dimension.
This projection will be referred to as rotating UV space, and the earlier static projection
simply as UV space. This projection gives the desired result, that the beam area remains
invariant with scan angle. It is interesting to compare the projection of lines of constant
azimuth and elevation with a vertical rotating array, shown in Figure 4-11 below, to the
static vertical array, shown in Figure 4-6(a). It is clear that the lines of constant azimuth
do not become compressed at the extremes of the figure as they do in the UV space
projection, although the lines of constant elevation remain the same in both figures.
Although the image in Figure 4-11 looks rather like the orthographic projection above
and does look somewhat familiar, it has not been found in the cartography literature; the
most similar projections found being the Sanson-Flamsteed sinusoidal and Moliweid
[Adams, 1914], which are both equal area projections. Usually the desire in cartography
is to conserve either the area or the distance when projecting from the surface of a globe
onto a flat plane (it is impossible to conserve both). This projection conserves neither,
72
so is presumably of little interest in that field.
-1 -0,5 0 0.5 1 1.5
Figure 4-11 Lines of constant azimuth and elevation in the rotating UV space, for a vertical
array
Figure 4-12 shows the variation in the projection as a function of array tilt angle from
verticaL
0.8
0.6
0.4
0.2 -.
a.'- 0
,;
-0.2
-0.4
-'0.6 . .,
'-0,8 - '
-1.5 -I -05 0 0.5 ¶ 1.5 _;
(a) (b)
-1.5 -I -0.5 0 0.5 1 1.5

U.
(c)
Figure 4-12 (a,b,c) Lines of constant azimuth and elevation in the rotating UV space, for arrays
inclined back from vertical (a)30°, (b)60°, (d)90°
As in the static array case, we may use this spatial system to derive the exact beam
directions for a rotating radar, given the azimuth and elevation limits of the sector. For a
rotating system, it is likely that sectors will cover all azimuth directions. However to
enable comparison to section 4.4.1 above, we will consider the same MESAR search
73
configuration as before, with limits of ±450 in azimuth and 1.18° - 52.5° in elevation,
and a 3dB beamwidth of 3.1°, with beams overlapped at 0.85 beamwidths. Figure 4-13
shows the calculated beam directions in the rotating UV space. The exact number of
beams that are used for this search sector in MESAR is 657. As we would expect, the
rotating system requires a greater number of beams to search a given sector because the
beam area is smaller than in the static case, due to the absence of azimuth scanning.
Figure 4-13 Beam directions in rotating UV space, using a MESAR search configuration as an
example
Similarly, we may also estimate the number of beams in the search volume, without
explicitly calculating each beam direction. The estimate is now given by the ratio of the
search sector in rotating UV space to the area of the beam in rotating UV space. The
area of the beam in rotating UV space, is the same as that in the static form of
UV space;

=irsin2(03) [4-13]
The area of the sector to be searched, A t _Nv, may again be found by integration.
re,
=21
f Ur Se!
L el_mm
,el_max
ri [4-14]
= az_ max' - sin(2e1 - a) + el. cos(a) I
[2 Jel_mm
The result is simpler in the rotating case, having a simple analytical solution, since the
lines of constant elevation remain horizontal in rotating UV space, even when the array
is tilted. The area given by equation [4-14], is shown in Figure 4-14(a). As in the static
case, this approximation gives a result within 8% of calculating the number of beams
74
exactly.
I I
Figure 4-14 (a) MESAR coverage limits in rotating UV space, (b) Beam dwell time plotted as
height above the rotating UVplane
Figure 4-14(b) shows the dwell time in each beam direction plotted as a surface above
the beam directions in rotating UV space. As in the static case, the dwell time of each
beam is given by the broadside dwell time, dtbroadside, divided by cos 3 (9), but now, 9
depends simply on the elevation. Thus, the average dwell time, dtay, is given simply by
equation [4-15], below.
1eLmax
dtr,gJ$j
Ij
ei_n cos 3 (el - a)
dtc =
sin(el_ max— a) - sin(el _ mm— a)
(el_max
dtbT,S, 1
- sin(el_max— a) - sin(el_min— a)
j 5 cos 2 (el - a)ael')
)
¼Cf nun
= tan(el_ max— a) - tan(el_ mm— a)

.•.
sin(el_ max— a) - sin(el_ mm— a)
dtbr i3j(je 14-15]
Thus we can simply estimate the number of beams and beam dwell times as a function
of the beamwidth for a rotating array system. The estimates this method gives are
compared to calculating every search beam direction and dwell time exactly in Figure 4-
15 below, using the same MESAR search sector as an example. Again, the method gives
excellent correlation.
75
i_.. ,....,",.............
-.- _i _x -
_.,_ ;
W;
x -;
31
I I
j
2 5 4 5 7 I S I'.
Be.,,..dt' 145: B..w.? ,ç
Figure 4-15 (a) Number of beams, and (b) average beam dwell times as a function of
beamwidth, for a rotating array multi-function radar. Crosses represent answer from exact
calculation of all beam directions and dwell times, the line represents the estimate from the
rotating UI' method outlined above
4.5 Search time in a noise limited environment
The methods derived in sections 4.4.1 and 4.4.2 are now applied to a comparison of
different static and rotating multi-function radar antenna configurations. In this section
only targets whose detection is limited by thennal noise are considered. Clutter is
considered in the following section. From this an understanding of the effects of using
broadened beams and simultaneous beam processing will be obtained, as well as a
system level comparison of the multi-function radars.
Figure 4-16 The example surveillance coverage volume for comparison of different multi-
function radar antenna configurations
The configuration we will consider will be of a search of 03600 azimuth and 0600 in
elevation, with a multi-function radar tilted 200 back from vertical, shown in Figure 4-
16, above. Nominally, a ims dwell time is assumed on array broadside, which is
increased (assuming coherent integration of pulses) to offset a two way loss in gain of
the array that varies with the cube of the cosine of the scan angle (see section 2.3.2.2
above). Two static array antenna configurations will be considered; a 4 faced system,
requiring scanning in azimuth to ±450 , shown in Figure 4-17(a), and a 3 faced system,
requiring scanning in azimuth to ±60°, shown in Figure 4-17(b).
76

02 .•..• . . '.
0.e
: .t.....
04 ,. - ..
. ........ ......
0.2 .,
......................
0;
'-4 ...!
• ... . .... I............................
.02 .
-° .01 -02 -04 - 0 02 0

uJ U-,
0.2
0I
04
02 ,, - ,., * -
0 ........
-02'UT'TT ....:
-a'. -. : :
.
Figure 4-17 (a,b,c) 0 °-60 ° search sector plotted in (a) UV space for a 4 faced multi-
function radar scanning to ±45 ' (b) UV space for a 3 faced multi-function radar ±45 ' (c)
Rotating UV space, applicable to a 1,2,3 or 4 faced multi-function radar
Rotating systems with 1 - 4 faces will be considered. In the rotating case, the number of
beams and the dwell time required for each beam is independent of how many array
faces the multi-function radar has, since there is no azimuth scanning for the search
function. Thus one representation of the search sector in rotating UV space is all that is
required for all of the rotating multi-function radar systems, shown in Figure 4-17(c).
Figure 4-18, below, compares the number of beams in the total search volume as a
function of beamwidth, for the 3 and 4 faced static multi-function radar systems and the
rotating multi-function radar systems. Figure 4-18(b) shows the ratio of the number of
beams required in the static systems to the rotating systems. It is seen that the rotating
systems require a greater number of beams than the static systems to cover the example
search volume. Around 8% fewer beams are required by the 4 faced static system, and
15% less are required by the 3 faced static system. This is due to the extra beam-
broadening that occurs with azimuth scanning.
77
06
r—.41.c. do
loco doIIflccoog
07
06
I
05
j 04
02
0I
2 3 4 5 6 7 8 9 10
Bo*oh (dog) 800,w (dog)
Figure 4-18 (a,b) Number of beams required to search the specified volume as a function of
beamwidth, and the comparison of the static and rotating systems
Figure 4-19 compares the average dwell time required to search the volume, assuming a
nominal ims dwell time in the array broadside direction. At this stage no compensation
in transmit energy is made for the loss in gain in broadening either the transmit or
receive beams; thus the average dwell time remains constant with beamwidth. The ratio
of the average dwell times required by the static systems to those of the rotating
systems, shown in Figure 4-19(b), shows a dwell time over 25% greater is required by
the 4 faced static system, and over 70% greater than the 4 faced static system, to achieve
the same detection performance.
BOdo (doW B.4dth (dog)
Figure 4-19 (a,b) Average dwell times and comparison
The increase in average dwell time required by the static systems greatly outweighs the
fewer beams necessary for search. This is reflected in the total search function times for
the systems, shown in Figure 4-20(a) as a function of beamwidth. The comparison in
Figure 4-20(b) shows that the 3 and 4 faced static systems require 18% and 44%,
respectively, more radar time to perform the same search as the rotating system.
78
4 The Impact of Variable Beamwidth and Multiple Beam Processing on Mult (function Radar
t 6
4 laco doIrIc.I9g
2oo1at/Roa
I II08
3 4 5 6 7 8 9 tO
8owith (dog) Baca,jg, (dog)
Figure 4-20 (a,b) Search function time and comparison
Table 4-1 below summarizes the results from this comparison, using a 2.5° beamwidth
as an example.
Antenna No of beams Average beam Search function Search function time

Configuration (per face) dwell time time (per face) (total)
1 face rotating 3380 1.l7ms 4.Os 4.Os
3 face rotating 1127 1.l7ms 1.3s 4.Os
3 face static 963 2.O3ms 1.9s 5.7s
4 face static 774 1.55ms 1.2s 4.8s
Table 4-1 Summary of static and rotating multi-function radar search comparison for a 2.50
beamwidth
Table 4-1 presents two interesting results. Firstly, when comparing the 4 faced and 3
faced static systems with their rotating equivalents, it is found that the static systems
require 20% (4 faced) and nearly 50% (3 faced) more radar time to execute the search.
More significantly, when comparing the 3 faced static system with a 2 faced rotating
system, nearly the same amount of radar time must be dedicated from each of the 3 faces
as is required from just 2 faces in the rotating case. Thus this comparison suggests one
of the 3 faces could be eliminated if rotation was employed with very little detriment to
the search. The savings are thus significant, although they do vary depending on the
search volume and the array tilt angle. Interestingly these findings are in agreement with
a recent paper by Billam [Billam, 1997b], in which different static and rotating multi-
function radar's performance is compared.
4.5.1 Dwell time compensation

Compensation for the loss of gain associated with broadening the antenna patterns is
now considered, through coherent integration of pulses. First the case of broadening
both the transmit and receive beams is considered.
79
4 The Impact of Variable Bea,nwidth and Multiple Beam Processing on Multfunction Radar
Figure 4-21(a) below shows the decrease in search function time that occurs for the
multi-function radar systems as the beamwidth is broadened, with a nominal broadside
waveform of ims dwell time. Clearly since the broadside waveform is not being
increased to offset the loss in both transmit and receive gain, the signal to noise ratio
and probability of detection drop rapidly, Figure 4-21 (b,c).
3 4 5 6 7 8 9 10
Bearnw,dth (deg) 8W
Figure 4-21 (a,b,c) Signal to noise ratio and Probabilit y of detection as a function of
beamwidth, with broadening of both the transmit and receive beams, using the nominal Ims
waveform above - ie not compensating for the loss in two way gain
If it is now assumed that the broadside waveform is increased in length, to allow extra
pulses to be coherently integrated, then the impact on the search function time can be
seen. Figure 4-22 shows that in order to maintain uniform detection performance as the
transmit and receive beams are broadened, the search function time increases very
rapidly. It is obvious that using both broad transmit and receive beams would be an
extremely inefficient way to perform the search.
80

4 The Impact of Variable Beamwidth and Multiple Beam Processing on Mulrfunc:ion Radar
--4Im01
-- 3du 16 •-.. --
– -Aoi.&ig
IS . .
14 .•;.
S
I 12
I Io ... '. . ...........

I
2 2 4 5 5 7 5 9 10 "2 3 4 S 6 7 0 9 10
PryodsS.c6oi,
—4
0.9 39
0.9
07... .........
0.6
0.5 . . . . .
0.4 .
03.
02.
0.1 ..
7 4 5 6 7 5 9 10
SW
Figure 4-22 (a,b,c) Search function time required to maintain a uniform detection performance
with broadened transmit and receive beams
Figure 4-23 and Figure 4-24, below, now consider search function time when using
simultaneous multiple narrow receive beams coupled with a broad transmit beam.
I... . —- --
3 4 5 6 7 I 9 10 799 10
—4 l.csd
09 -. --
0.9
0.7
Figure 4-23 (a,b) Signal to noise ratio and Probability of detection as a function of beamwidrh,
with broadening of only the transmit beam, using the nominal lins waveform above — i.e. not
compensating for the loss in one way gain
81
Figure 4-23 shows the decrease in search function time, signal to noise ratio and pd that
occur as the transmit beamwidth is broadened, with a nominal broadside waveform of
ims dwell time.
When the broadside waveform is increased to compensate for the loss in transmit gain,
the search function time is found to be invariant with transmit beamwidth, in order to
provide uniform detection performance. This is the familiar result given by the search
radar equation, and is shown in Figure 4-24.
I 3lacedI*611C I
16
3 4 5 6 7 10 2 3 4 5 6 7 6 9 tO
Be,w,1th (deg) 6W
ProbeblSty of deleclieb
6W
Figure 4-24 Search function time required to nwintain a unbrm detection performance using a
broadened transmit beam and simultaneous multiple narrow receive beams
4.6 Search time in a clutter limited environment
Clutter is a notoriously difficult factor to predict through either mathematical modelling,

or through statistical calculations. Clutter, and the methods of separating it from targets
of interest, is one of the largest areas of theoretical and experimental research in the
radar field. Not only are the clutter statistics highly dependent on what is being
illuminated, and how it is being illuminated by the radar (in the sense of resolution and
PRF), but also on the filtering and processing that occur upon reception within the radar.
82

For this reason, it is only sensible in a system level analysis such as this to incorporate
the effects of clutter in their broadest terms. Here, the effects of variable beamwidth
operation and simultaneous multiple beam processing on a multi-function radar with
MTI and Pulsed Doppler processing are shown using only relevant elements of the radar
equation.
Table 4-2 compares the relevant parts of the radar equation for the three different types
of system under consideration; a single narrow pencil beam system, a variable
beamwidth system in which both the transmit and receive beamwidths have been
doubled, and a simultaneous multiple beam system in which the transmit beamwidth is
doubled, but narrow receive beams are used. No clutter is assumed in this case; the
target competes with thermal noise only. Nominally a 10 beamwidth is assumed for the
narrow beamwidth system, which is assigned a nominal gain of 1. Thus the broadened
1116th
beam system has a two way gain of of the narrow beam system, and the multiple
beam system has a two way gain of ¼, for a doubling of beamwidth. It is assumed in this
case that the search function time available is Is. Therefore, the dwell time available in
each beam direction is simply 1/number of beams.
System Narrow Beam Broadened Beam Multiple Rx Beams

Beamwidth 1° Tx / 1° Rx 2° Tx / 2° Rx 2° Tx / 1° Rx
Transmit Gain, 1 1/4 1/4
Receive Gain, Grx I ¼

Number of Search Beams 4000 1000 1000
Dwell Time per Beam, t,j 1/4000 s 1/1000 s 1/1000 s
Search Time I s 1 s 1s
SNR G txGrxt,j 1/4000 1/16000 1/4000
Relative SNR 4 1 4
Table 4-2 Radar equation comparison of narrow beam, broadened beam, and multiple beam
systems, assuming noise limited detection
The result is that for the same search function time, the broadened beam system has a
quarter of the signal to noise ratio of the narrow beam system, whilst the multiple beam
system maintains the same signal to noise ratio.
If clutter is now considered, the story is somewhat different. Resolution in the Doppler
domain is governed by the dwell time, td. It is this that will define the detection
performance in the presence of strong clutter. This is shown in Figure 4-25(a) below,
reproduced from Barton [Barton, 1988]. He gives the contamination of the velocity
spectrum as per equation [4-16] below.
83
Figure 4-25(b), below, plots the percentage of filters contaminated by sea clutter as a
function of the number of pulses for a 3GHz system. Here an unambiguous range of
30km is assumed, and a clutter spectral width of 1 .4m/s, corresponding to sea state 5
according to the DERA Naval Environment Clutter, Attenuation and Propagation
Specification [Money, 1995].
I I
I
N.ro p.i
0 1 2... N
Figure 4-25(a,b) Contamination of Doppler filters by clutter
2V,, [4-16]
Lv = Ec + -
N
where 4v is the velocity spectrum that is contaminated

4c is the clutter spectral width
Vh is the unambiguous velocity
N is the number of pulses per Doppler burst
The case of clutter limited target detection is shown in Table 4-3 below. In this instance
td remains fixed in all three systems, at a nominal value of lms. For the narrow beam
system this results in an increase in search time from ls to 4s, but with what may be
regarded as an excess of signal to noise ratio; it exceeds the signal to noise ratio required
for detection by a factor of 4 compared to the noise limited case. The system with
broadened transmit and receive beams can complete the search in the same is frame
time with the ims dwell time, but has a relative signal to noise ratio of just 1, as in the
noise limited case.
The multiple receive beam system has the same is search time as in the earlier case of
noise limited detection. It also retains the same relative signal to noise ratio of 4, as in
the earlier case. Thus there is no penalty in using the simultaneous multiple receive
beams in the clutter limited case. The same signal to noise ratio is achieved as was in the
noise limited case. However, the longer dwell time coupled with multiple beams, allows
the clutter cancellation requirement to be met with a search time of is, compared to 4s
84

for the narrow beam system. Hence multiple beam processing may be regarded as an
extra degree of freedom for the multi-function radar, trading off energy transmitted in a
given direction for time, and therefore allowing longer waveforms to be transmitted
without increasing the function time.
System Narrow Beam Broadened Beam Multiple Rx Beams

Beamwidth 1° Tx / 1° Rx 2° Tx / 2° Rx 2° Tx / 1° Rx
Transmit Gain, 1 ¼ ¼
Receive Gain, Gro 1 ¼ 1
Number of Search Beams 4000 1000 1000
Dwell Time per Beam, td 1/1000 s 1/1000 s 1/1000 s
Search Time 4 s 1 s 1s
SNRoc G 0 Gt, 1/1000 1/16000 1/4000
Relative SNR l 1 4
Table 4-3 Radar equation comparison of narrow beam, broadened beam, and multiple beam
systems, assuming clutter limited detection
4.7 Impact on the tracking functions
Generally, higher performance may be obtained for the tracking function by using a
narrower beam. This is particularly the case for current multi-function radars which use
relatively broad beams in the region of 2° to 3°. It has been shown in the previous
section that the surveillance function time, theoretically, does not increase with narrower
beams, if multiple receive beam processing is possible; the compromise in beamwidth
that is normally associated with an air defence multi-function radar ([Barton, 1988],
[Moore & Salter, 1995]), is no longer valid. Thus multi-function radars may benefit
from the narrower beamwidths normally associated with dedicated tracking radars,
giving benefit in terms of resolution, plot accuracy and performance in jamming and
multipath.
However, the narrower beamwidth, coupled with the low track update rates associated
with multi-function radars generates a problem in tracking manoeuvring and rapidly
crossing targets. It is possible that the target will not be illuminated at the next track
update if the target manoeuvres or if azimuth/elevation crossing rate certainty is low.
Multiple beam updates should therefore be considered for the tracking function. Figure
4-26 below shows how the use of multiple beams could be envisaged for tracking and
surveillance.
85
Narrow bsem lurch
Targs
Mat
or Hot msrrcsrNnn9 tlr5Se
Hr Speed Horizon Search
Figure 4-26 An example of the use of single beam and multiple beam updates in the surveillance
and tracking functions
The results of sections 4.5 and 4.6 suggest that, assuming perfect coherent integration of
pulses, in principle, we may use simultaneous multiple beams for tracking wherever we
wish, without detriment to the detection performance. In a noise limited situation the
increase in dwell time required as the transmit beam is broadened, will be exactly offset
by the loss in gain, hi a clutter limited situation, there may be an advantage to using
multiple beam track updates, since the waveform length may be determined by the
Doppler resolution required to separate the clutter from the target, rather than the target
from noise.
4.8 Conclusions
A novel method of calculating the search function time for a static radar has been
developed. This gives analytical expressions which may be quickly evaluated for a range
of beamwidths and search sectors, without explicitly calculating each beam direction
and dwell time. This will allow surveillance configurations of a multi-function radar to
be evaluated rapidly using simple equations. The method is based on estimating the
number of beams and the dwell times required in each beam direction in UV space, in
which the beam area remains invariant.
By creating a similar projection to the UV space projection, involving a transformation

onto the surface of a 3D cylinder, another spatial system may be produced in which the
beam area of a rotating multi-function radar system remain invariant. This projection is
somewhat akin to the Sanson-Flamsteed sinusoidal and Mollweid projections used in
the field of cartography. The above methodology for estimating the beam directions and
dwell times, and thus the search function time, may be applied in this spatial system for
86
4 The impact of Variable Beamwidth and Multiple Beam Processing on Multfunction Radar
the rotating system.
The above methods for static and rotating multi-function radars have been compared
with the process of calculating beam directions and dwell times exactly, and have found
answers to within a few percent. The accuracy decreased somewhat as the beamwidth
increased (to around 8% for a 100 beamwidth), due to the exact number of beams
required to cover the edges of a given region.
A comparison of the search function for different configurations of static and rotating
array multi-function radars has been made. It was found that the static systems require
20% (4 faced) and nearly 50% (3 faced) more radar time to execute a complete search of
the same volume. When comparing the 3 faced static system with a 2 faced rotating
system, it was found that nearly the same amount of radar time must be dedicated from
each of the 3 faces as is required from just 2 faces in the rotating case. Thus this
comparison suggests one of the 3 faces could be eliminated if rotation was employed
with very little detriment to the search. These figures will vary to some degree,
depending upon the exact volume that is to be surveyed, and the tilt angle of the arrays.
The impact of using a broadened beam both in transmit and receive on the multi-
function radar system has been evaluated. For any significant broadening of the beam
there is a heavy penalty to be paid both in the tracking and surveillance functions.
The variation of surveillance search time has been determined as a function of transmit
beamwidth for target detection limited both by thermal noise and by clutter, assuming
multiple narrow receive beams. When detection is limited by thermal noise the result is
as predicted by the familiar search radar equation; the search time is invariant with
transmit beamwidth, since a wider beam requires a longer dwell time to regain the signal
to noise ratio.
When target detection is limited by clutter, the search time may be decreased by
broadening of the transmit beam. This is because the dwell time in that instance is
determined by the resolution required in the Doppler domain, governed by the length of
the transmit waveform.
Although the study has been aimed at investigating the effects of using broadened
beams, it is not implied that systems should necessarily use broader beams than they
currently do. Quite the reverse. The implication from these arguments is that systems
87
4 The Impact of Variable Bearnwidth and Multiple Beam Processing on Multifunction Radar
may be built with narrower receive beamwidths than those in use currently, whilst still
retaining the clutter cancellation levels required. This can be achieved within the time-
budget constraints of the multi-function radar, through using broadened transmit beams
and multiple receive beams. This would allow the benefits of a dedicated tracking radar
to be realised, such as excellent monopulse accuracy, resolution of closely spaced
targets, and resistance to main beam jamming.
88
5 RESOURCE MANAGEMENT IN STATIC AND
ROTATING ARRAY MULTIFUNCTION RADAR
5.1 Introduction
To provide the functions of volume surveillance, horizon surveillance, track

maintenance, missile uplink, etc., the multi-function radar must rapidly switch the beam
between many tasks. It is the job of the multi-function radar resource manager to ensure
that the most appropriate radar tasks are chosen as time progresses and to allocate the
multi-function radar resources in a sensible manner. This chapter covers the resource
management and task scheduling of a multi-function radar.
There are two objectives. Firstly, to investigate resource management and task
scheduling in multi-function radar, and particularly the complications that arise with
rotation of the array antenna. Secondly, to develop models of static and rotating array
multi-function radars based around a task scheduler centred architecture, which may be
used to evaluate both the scheduler performance and tracker algorithms.
A background to the areas of resource management and task scheduling is given below.
Section 5.2 presents a set of objectives that are proposed here to be used for multi-
function radar resource management and scheduling systems. Section 5.3 then covers
task scheduling in a static multi-function radar. Here a model of a multi-function radar
centred on the task scheduling algorithm used in the MESAR radar is developed, and
the validation and results from that model are presented. In section 5.4 task scheduling
in a rotating multi-function radar is covered. In this case, no suitable algorithms or
strategies were at hand for implementation into a computer model. Therefore a novel
algorithm was created. An analysis of the complications due to rotating the array
antenna is presented, along with an algorithm which successfully copes with those
complications. Results from the development of a second model of a multi-function
radar, centred on the rotating scheduler algorithm, are presented. Section 5.5 discusses
ways in which the performance of both the static and rotating scheduler algorithms may
be improved by using penalty functions or fuzzy logic, and section 5.6 contains the
89
5 Resource Management in Static and Rotating Array Mutt (function Radar
conclusions of the work.
The distinction between the functions of resource management and task scheduling is
somewhat blurred and to some extent varies from system to system, so a definition that
will be adhered to in this chapter is given below.
In order for efficient multi-function radar resource management to be achieved two

distinct processes must be completed;
i. All tasks must be ranked in a priority order. Note that the priorities of tasks
may change throughout an engagement.
ii. Tasks must be formed into a timeline for the multi-function radar to peiform.
This is the specific job of the task scheduler.
The first of these processes, ranking the functions in priority order, is usually
determined a priori using the expert knowledge of an engineer; however the priorities of
tasks may be elevated or lowered if necessary under automatic control algorithms. This,
together with the calculation of desired time for tasks to be scheduled and the amount of
radar time required, may be defined as the job of resource management. The resource
manager may also include macroscopic rules for the radar's management and allocation;
for example, the manner in which surveillance is degraded in the case of an overload of
tasks. These rules may take information from several areas of the sensor, and also from
outside of the sensor. The resource management algorithms must allow robust multi-
function radar operation in all scenarios that may be encountered, and allow a tactically
sensible degradation of performance if the multi-function radar becomes overloaded
with tasks.
The task scheduler is a subsystem of the multi-function radar resource manager. It is fed
tasks, with their desired times for scheduling, their relative priority, and possibly other
constraint information. At any time the task scheduler will have a large list of tasks,
each of which has a different priority and timing constraints pertaining to it. The task
scheduler should try to maximise the use of radar time, whilst ensuring that the radar
time is also given to the most important of these tasks throughout the duration of a
scenario.
The formation of the timeline may be done in advance over a small time interval, as
90
5 Resource Management in Static and Rotating Array Multifunction Radar
suggested by Sabatini & Tarantino [Sabatini & Tarantino, 1994], or it may be decided
in real time, task by task, as in MESAR ([Stafford, 1990], [Wray, 1992]). A drawback of
using the former approach is that, although it may be simpler from the computation and
processing aspect, it is difficult to accommodate jobs which rapidly appear, such as plot
confirmation jobs, without throwing away or delaying others. In practice all scheduler
algorithms must accommodate a small queue of the order of microseconds to allow for
hardware setup times, for example in the phase shifters.
The scheduler's task in constructing this timeline is complicated due to the constraints
which apply to each task;
i. tasks vary in the criticality of the time period in which they must be
scheduled. Some may have a small window of opportunity which must be met
for the task to be successful, while others may have much looser time
constraints.
ii. tasks differ significantly in length.
iii. tasks may become suddenly necessary or urgent, or may become unnecessary.
iv. tasks may have to adhere to some constraint such as close to array broadside
operation in a rotating system.
5.2 Resource management and task scheduling objectives
The resource manager is central to the effective control of the multi-function radar. If it
does not perform its job correctly, every aspect of multi-function radar performance may
be sub-optimal. In this section a set of objectives for a resource manager are presented,
against which its performance may be measured.
The following broad objectives are suggested for resource management and task
scheduling.
i. schedule each task as near to the requested time as possible.

ii. schedule each task as close to array broadside as possible.
iii.schedule each task to minimise the radar idle time.
iv. schedule each task to maxim ise the tactical benefit of the multi-function
radar.
91
5 Resource Management in Static and Rotating Array Multijunction Radar
It is important to note that these are conflicting objectives and it is highly unlikely that
all can be met for any sustained period of time. For example, it may be much more
important to schedule a particular task at exactly the right time, than to constrain it to
array azimuth broadside. Therefore these objectives must be reflected in the constraints
with which a task is specified by the resource manager.
The first three may be viewed as primarily the responsibility of the task scheduler. The
fourth objective then is the responsibility of the resource manager. It does not lend itself
to use as a figure of merit, but it may be broken down into measurable objectives if we
are to use it as figure of merit. These objectives will depend on the mission of the multi-
function radar system and also the stage through the mission. A set of objectives for a
naval air defence system are proposed below;
i. achieving a surveillance detection performance

ii. achieving a surveillance data rate, for track while scan
iii.achieving rapid track initiation
iv. achieving a tracking accuracy
v. maintaining missile communications
The relative importance of these at a given time will vary. For example, achieving the
required detection performance for surveillance is imperative at the early stages of an
engagement scenario. However, if the channels of fire of the weapon system are
saturated, then the emphasis is less on detecting new targets, and more on retaining
those already detected.
5.3 Task scheduling in a static array multi-function radar
5.3.1 Background
In a static multi-function radar, the constraint of scheduling tasks such as surveillance
close to array broadside that must be adhered to in the rotating case does not apply. This
significantly simplifies the task scheduling problem, and hence this is the case that was
addressed first.
Two scheduler algorithms were considered; firstly the original MESAR scheduler
92
algorithms ([Chestnutt, 1984], [Birch, 1989], [Stafford, 1990]). These are the algorithms
currently implemented in the MESAR system, although in the system itself they are now
in a modified form. Subsequently a modified version of these algorithms was generated
involving primarily a change to the time balance strategy.
It should be noted that since MESAR design note 242 [Birch, 1989] was written the
algorithms implemented in MESAR have been further developed and thus the design
note is somewhat out of date. Therefore an explanation of the current algorithm
implemented in MESAR is given below in section 5.3.2.
5.3.2 The MESAR algorithm

The MESAR static task scheduling architecture centres around a prioritised list of all of
the radar jobs. The following priority levels exist;
7. track maintenance (highest priority)

6. plot confirmation
5. track initiation
4. track update
3. surveillance
2. slow track map/surface picture
1. receiver calibration (lowest priority)
A description of what constitutes a job, a task, and a look in MESAR is given below for
clarification. A job may be surveillance of a sector or region, maintaining track on a
specific target, or track initiation of a target. It will typically consist of several tasks, for
example, searching a single surveillance beam position, performing one track update, or
performing one track initiation update. Each task will typically consist of several
activities which are non coherently integrated to give a detection. A look is one or more
activities from a task that are transmitted coherently by the radar.
These looks are scheduled one at a time, and may be interleaved with looks from other
tasks. Only when all of the looks that make up a task have been completed is the whole
task said to have been scheduled. The only advantage in breaking tasks into smaller
looks is that the scheduling delay of other tasks may be minimised. Thus if tasks are
already small enough, this is unnecessary.
A timebalance scheme controls the scheduling of tasks on the job table [Birch, 1989].
93
Every job has a timebalance associated with it, which tells the scheduler how much
radar time is owing to that job at that time. Only radar time that will actually be
available to a job will be credited to its account. A positive timebalance indicates that
the job is in need of radar time in order to complete a task. A zero timebalance means
this job has had exactly the right amount of radar time, thus far. Finally, a negative
timebalance means that this job does not need any radar time at this time.
When a look is scheduled from a task, the look time that has just been used is shared out
amongst all the other jobs on the radar job table. The time balances of all other tasks are
incremented by a fraction of the look-time in proportion with their own desired
occupancy. The highest priority levels receive their shares of the look time first, and
then each lower level receives its share. Lower levels only receive increments if the look
time has not been shared out completely to the higher levels. Therefore, if the radar is
overloaded, i.e. the desired occupancy of all the jobs in the job table is greater than
100%, then lower level timebalances will not be incremented, and thus the tasks from
these levels will be delayed.
When all the looks that make up a task have been scheduled and the task is complete,
the job's timebalance is decremented by the total dwell time of the task.
A new job may be inserted onto the job table with a negative timebalance, in order that
it is not scheduled for some time. In this case all other job time balances must be
incremented by a proportion to ensure the timebalances always have a sum of zero. This
guarantees that timebalances are incremented relative to the radar resources available.
A flow diagram of the MESAR algorithm is shown in Figure 5-1, and a description is
given below.
94
4r
.STPRIORlLEVEL
Is a task underway
on this level?
NO
Go down to the next Is there a job with a
positive TB?
NO
NO Is this the lowest level !ooshobwiththe
Schedule a look from the
Increment all timebalances
If this is the last look in the

task, decrement this job's
timebalance
Figure 5-1 Flow diagram of the MESAR scheduler algorithm
Starting from the highest priority level;
i. If a job is already underway on that level, then that job is chosen for
scheduling of looks. This means that tasks from each level will be completed
sequentially, rather than many tasks from the same level being interleaved,
and thus drawn out in time.
ii. If no task is underway then the task is chosen with the most positive time
balance.
iii. If no task has a positive time balance then move down to the next level and
repeat (i)-(iii)
iv. If no task has a positive time balance on the job table, then choose the
surveillance task with the least negative time balance.
v. Schedule one look from the chosen task, and increment all other time
balances by a fraction of this task.
vi. If that was the last look of a task decrement this job's time balance by the task
dwell time.
95
5 Resource Management in Static and Rotating Array Mu1tfunction Radar
In this way, looks are scheduled from a task one by one. Scheduling of looks from a task
may only be interrupted by jobs at a higher level on the priority table. The timebalances
are in effect simply used to measure when a task is ready for scheduling. Once looks
from a task on a particular level have begun being scheduled, that task is always chosen,
regardless of its own and other tasks' timebalances, unless it is interrupted by higher
priority jobs.
The algorithm only allocates radar time to the lower priority levels if it is available, i.e.
unrequired by higher priority levels. Since surveillance is the lowest priority level in
MESAR, the occupancy of the surveillance function, and thus the surveillance frame
times, increase and decrease as the tracking load changes. This is shown in Figure 5-2,
below. The multi-function radar may be specified to meet its surveillance detection
performance when say, surveillance has an 80% occupancy. If there are no tracking
tasks, surveillance will thus exceed its detection performance. Conversely, if the system
has more tracking tasks than the 20% headroom allowed, surveillance will fail to meet
its detection performance requirement.

Benign (no tracks) Overloaded (many tracks)
tJ- - /
Figure 5-2 Degradation of surveillance detection performance
Timebalances appear, in practice, to be used in a simpler way than their rather

complicated calculation warrants. This is reflected in the recent modifications to the
MESAR scheduler which simply schedule tracking tasks on their desired update time,
disregarding the timebalance completely. This should have no effect on the scheduling
of tracking tasks, but is simpler.
The radar job controller, maintains the radar job table and acts as an interface with the
rest of the radar control processing such as the tracker. It inserts and deletes jobs as is
necessary, for example a plot confirmation job request as a result of a successful
surveillance detection. In this way the task scheduling algorithms themselves can be
96
kept simple, since much of the complexity specific to each function's operation resides
in the radar control processing.
5.3.3 The modified MESAR algorithm

The modified algorithm involves primarily a change to the timebalance scheme. In
MESAR the timebalances are a 'bank account' of radar time for each job. They represent
how much radar time is owed to a job at any instant, and in overload situations, how
much radar time can be given to that job by the radar. In practice, however, the
timebalances are only used to determine whether a task is ready for scheduling - i.e. if
the job has a timebalance greater than or equal to 0 then this job is ready.
In the modified version of the algorithm the timebalances simply represent how early or
late tasks from a job are. The timebalance units are now seconds, which makes the
values themselves meaningful; if a task does not need to be scheduled for is, its
timebalance is -is. This simple change will not affect the performance in any way. It
merely allows for easy understanding of the scheduler operation.
Two versions of the algorithm have been coded. The first follows the MESAR algorithm
closely, allowing tasks to be broken down into smaller looks, and following the MESAR
algorithm rules for their scheduling. The second is a simplification which schedules
tasks as whole units. The simplified version operates more quickly, and may be
applicable if the simulation is going to be of a high power radar representative of an
operational radar. The waveforms in this case are likely to be considerably shorter than
in MESAR and therefore will not delay tasks significantly.
The scheduler now uses simply the task look back interval (the time between successive
tasks - e.g. the track update interval, or the surveillance beam revisit time), and the task
time (the dwell time of the task) to control the scheduling of tasks.
A zero timebalance indicates this is the exact time for the job to be scheduled. A
positive timebalance indicates the job is now late, and a negative indicates the job is not
ready to be scheduled yet. As time passes the timebalances are continually incremented.
The jobs timebalance is now only decremented or reset when that job has been
scheduled.
Jobs may be inserted with a negative timebalance, so that they are not first scheduled for
97
some time. This would happen immediately after a surveillance task for example, when
a plot confirmation task would be inserted onto the job table with a timebalance of -0. is.
Once a task has been scheduled one of two things may happen to the job timebalance
which are subtly different in their result;
i. the timebalance is decremented by the task look back interval.

ii. the timebalance is reset to minus the task look back interval, so that the next
task will not occur until the desired time has elapsed.
In the first case, if the task was late then it is possible that the timebalance of that job
would still be positive after it was decremented. Therefore more tasks may be scheduled
for that job straight away. This is useful in the context of surveillance for example,
where if the search of a region is running late due to overload, the search may catch up
by searching several beams very rapidly.
This is not a useful property for all functions however. When updating a track for
example, there is little benefit accrued from scheduling two or more track updates in
rapid succession. In this instance the timebalance should be reset to minus the task
interval, so that all track updates are spaced with the look interval. (Note that the look
interval used could be modified to be shorter or longer if required).
Currently the algorithm resets the timebalance of tracking jobs. Surveillance and track
initiation job timebalances are decremented by the task look back interval, so that if they
are running late, they may catch up by scheduling several looks. Neither situation
applies to plot confirmation jobs since they only exist for the period of one task.
5.3.4 Simulation architecture

The task scheduling simulation architecture is shown in Figure 5-3 below. It has been
made intentionally very similar to the architecture used in MESAR where a job
controller looks after every job on the radar job table, from its initial request until
completion. It is the job controller's responsibility to ensure such tasks as the selection
of beam direction, and selection of waveform parameters such as PRF, frequency,
number of bursts etc. are calculated. If the job is successful, the job controller may need
to trigger a new job request, or delete ajob from the radar job table.
98
JOB REQUESTS JOB CONTROLLERS

PfilORflY TABLE
Plot conlbnatbi job request slot conllrmetion job con

Plot confirmation
Track mNiplIon job request Track initiation Track initiation job controller
Track udate
Track job request Track job controller
Suiverilance
Surveillence job request SCHEDULER Survellence lob controller
GIVES TIME TO JOB
CONTROLLERS
BASED UPON
OCCUPANCY,
DWELL TIME AND
PRIORITY LEVEL OF
EACH JOB
Figure 5-3 Static multi-function radar task scheduling simulation architecture
5.3.5 Scheduling using a simple two sector surveillance system

Initially investigation was limited to a simple two sector surveillance system for both
validation of the algorithms and ease analysis of results.
The following surveillance volume was constructed, consisting of 2 regions, both of

which spanned a 900 azimuth sector. Region 1 spanned 00200 in elevation, and region 2
spanned 20°-50° in elevation. The parameters of the regions specific to the scheduler are
shown in Table 5-1 below. Figure 5-4 shows the beam directions and dwell times.
Region Broadside Occupancy No of beams Frame time Function time Look Interval
Dwell time
1 ims 20% 283 1.415s O.283s 5ms
2 2ms 80% 343 0.8575s O.686s 2.5ms
Table 5-1 Scheduler specific parameters of the 2 region surveillance volume
• Io•
2..
1 1.1
Figure 5-4(a,b) Beam directions plotted in UV space, and beam dwell times plotted above the
UVplane
Initially no compensation in the dwell time was made for the loss of gain due to
scanning. Thus the broadside dwell times quoted for each region apply to eveiy beam
direction within the region. The beam directions are plotted in UV space in Figure 5-
99
S
5 Resource Management in Static and Rotating Array MultfuncIion Radar
4(a) and the flat dwell time surface is plotted above the UV plane in Figure 5-4(b).
The simulation allows targets to enter the surveillance volume with a set probability of
detection. Initially this was set to 0 so that no detections were made, and only
surveillance of the 2 regions took place. Surveillance tasks were not broken down into
looks in this case. The results of the simulation are shown in Figure 5-5 and Figure 5-6,
below.
Figure 5-5 Surveillance time balances as a function of time, with no target detections
Figure 5-5 shows the timebalances of each region for a few milliseconds at the
beginning of the simulation. The time balances of both regions start at 0. Each time a
task is scheduled that region's timebalance is decremented by the look interval. With
these simplistic dwell times the timebalances are clearly periodic, region l's
timebalance being decremented every 5ms and region 2's timebalance being
decremented every 2.Sms.
Figure 5-6(a,b) Achieved averaged occupancies and frame times over the simulation period
Figure 5-6 above shows the achieved occupancies and frame times of the surveillance
regions over a 20s period. These are calculated as an average over an interval of time, in
this case, is. The solid lines show the desired values, and the crosses plot the achieved
100
5 Resource Management in Static and Rotating Array Multfunction Radar
values. In this simple case the occupancies and frame times are achieved with great
precision, as would be expected.
A more realistic surveillance volume was constructed by allowing the surveillance dwell
times to increase with scan angle, to compensate for the loss in gain with coherent
integration of pulses. As elsewhere, the two way gain is assumed to fall with cos3((p).
The same two regions were used, covering O02Oo and 20°-50° in elevation. Table 5-2
shows the scheduler parameters of the new regions. The function time of each region is
increased over that shown in Table 5-1, since the dwell times in each beam direction are
all greater than the broadside dwell time. Correspondingly this increases both the frame
time and the look interval, since the desired occupancies of the two regions has
remained fixed.
The dwell time surface, shown in Figure 5-7, now shows two distinct bowls as the beam
dwell times increase towards the limits of the two regions.
I..
Figure 5-7 Beam dwell times plotted above the UVplane for the 2 region surveillance volume
Again the probability of detection was set to 0, so that no targets were detected during
the simulation. Neither were surveillance tasks broken down into smaller looks. The
surveillance regions were searched in a pseudo-random order. The results of the
simulation are shown in Figure 5-8 and Figure 5-9 below.
101
Figure 5-8 Surveillance time balances as a function of time, with no target detections
V V - V
i=1
V U V
I
I
Till'. il..,.,1.
Figure 5-9 Achieved averaged occupancies and frame times over the simulation period
Figure 5-8 shows the timebalances over the simulation interval. Again they can be seen
to be periodic, related to the frame times; however even for this simple two region
system they show a more variable nature due to the variation in dwell time for each
beam surveyed. The achieved occupancies and frame times, shown in Figure 5-9, also
meet their targets very accurately. At all times a 100% total system occupancy is
achieved.
The impact of introducing tracks into the system was then analysed. At time t=2s, a
target was introduced into the simulation. This target was given a probability of
detection 1; thus it was detected in surveillance and confirmed by plot confirmation. A
track was then initiated, and subsequently track updates were scheduled every is. The
waveform used for plot confirmation, track initiation and track maintenance was that of
the surveillance waveform which detected the target. At time t=4s, another track was
introduced, and so on, a new track being introduced every 2s for the 120s duration of the
simulation. The results are shown in Figure 5-10 to Figure 5-12 below.
Figure 5-10, below, shows some of the timebalances for the first few seconds of the
simulation. The process of scheduling plot confirmation, track initiation and track tasks
102
can be clearly seen here. At time t=2, the first plot confirmation job request is made, and
a task inserted into the radar job table with a time balance of -0. is. As surveillance
carries on, this task's timebalance increases until it becomes positive, at which point it is
scheduled. Subsequently 10 track initiations can be seen, the track initiation time
balance being decremented by the look interval (0. is) each time a task is scheduled.
Finally at time t=3. 1 a track job request is made, and track updates commence at time
t=4.ls.
From time t=2 onwards, the surveillance timebalances oscillate around an increasingly
positive value, rather than zero. This is indicative of the surveillance jobs becoming
later and later as far as the scheduler is concerned. This occurs because the surveillance
regions were specified to use 100% of the multi-function radar system occupancy. Thus
as tracking tasks are introduced and require occupancy, the total surveillance occupancy
decreases. The value around which the surveillance timebalances oscillate indicates how
far behind schedule surveillance is in seconds.
Th,. (.
Figure 5-10 Surveillance, plot confirmation, track initiation and tracking timebalances with
new target detections every 2s
Figure 5-1 1(a), below, shows the total achieved occupancy of each function. As more
tracks are introduced into the system, the track occupancy increases and thus the lower
priority surveillance occupancy decreases correspondingly. Again it is important to note
that a 100% total system occupancy is achieved at all times. Figure 5-11(b) shows the
decrease in surveillance region timebalances as the tracking occupancy increases. It
shows the two regions occupancies decrease in proportion with their desired
occupancies - i.e. the occupancy of region 2 decreases four times faster than region 1.
The frame time of the regions is plotted in Figure 5-12.
103

- TcSsl Suivsllines I p I=:

- TcI.I Tr.ddrig I
- TcuiTmckIiiit I
- TPIotConI I
Ta
0,
Figure 5-11(a,b) Total achieved occupancies of each function vs time, and achieved
surveillance region occupancies with track occupancy vs time
Figure 5-12 Achieved surveillance frame times vs time
5.3.6 Scheduling using the MESAR surveillance volume

MESAR in its standard surveillance configuration has a total of 9 surveillance regions.
The region spatial limits are given in Table 5-3, below, and are shown in Figure 5-13.
The scheduler specific parameters of the MESAR regions are given in Table 5-4.
I
t
Figure 5-13 MESAR search regions in standard configuration
104 -
5 Resource Management in Static and Rotating Array MuItfunction Radar
Region Mm Range Max Range - Mm Elevation Max Elevation

____________ (Kin)., (1Cm) degree) , (degrees)
1. 2.0 10.0 0 7.4
___________ 2.0 10.0 7.4 52.0
3 10.0 25.0 0 2.7
4 10.0 25.0 2.7 5.1
5 10.0 25.0 5.1 14.3
6 10.0 25.0 14.3 52.0
_____________ 25.0 52.0 0 2.7
____________ 25.0 52.0 2.7 7.4
9 25.0 52.0 7.4 23.4
Table 5-3 MESAR search region parameters
Again, initially, the probability of detection was set to 0, so that no targets were detected
during the simulation. A maximum look time of 5ms was specified. Thus every task that
was scheduled, was scheduled in looks of no more than 5ms. This was necessary since
the MIESAR dwell times can be extremely long as it uses integration of pulses in order
to overcome its very low power. Many beams have dwell times in excess of lOOms, the
longest beam dwell time being over 200ms. The surveillance regions were searched in a
pseudo-random order, as in MESAR.
Figure 5-14 Surveillance timebalances for the 9 regions of MESAR, with no detections
Figure 5-14 shows the timebalances of the 9 surveillance regions from the first 5s of the
105
5 Resource Management in Static and Rotating Array Mult(function Radar
simulation.
In this instance, the timebalances oscillate around an increasingly negative value. This is
due to the total surveillance occupancy being just 77%, thus surveillance tasks are being
scheduled earlier and earlier. The frame times achieved are therefore shorter than the
desired frame times.
Figure 5-15(a) shows the achieved occupancies of each of the 9 regions. In this case, the
solid lines representing the desired occupancy of each region, have been scaled up to
account for the fact that the total surveillance desired occupancy is less than 100%,
although the achieved occupancy with no tracking tasks will equal 100%. A great deal
more oscillation around the desired values is seen in this instance. This is primarily due
to the averaging interval (5s) being small with respect to the frame times of the regions.
Over a complete frame time, the desired occupancies are achieved, but since the
scheduler works on a fixed look interval basis, the achieved occupancy fluctuates
throughout the frame time period.
Figure 5-15(b) shows the achieved frame times. Again, there is considerable fluctuation
around the desired values over the short averaging period chosen, although the frame
times are met over a complete scan of a region.
aJ
Rmon3I
4I
*
:
,.
5J
S I .;:.;.;l
00
o .i
•..
T]' RuonI
R.on9
I
IRson1

01
Rso
- . a - I I

•0 0
Figure 5-15(a,b) Achieved averaged occupancies and frame times over the simulation period
Solid lines show desired values scaled to account for a desired surveillance occupancy of less
that 100%
Figure 5-16 shows the total occupancies of each function. Obviously the occupancies of
all tracking related functions are zero. The plot only serves to show that surveillance has
expanded to use 100% of the multi-function radar system occupancy, and that 100%
utilisation of the multi-function radar is again achieved.
106
Survefflanc.
- Tr.ckmg I
Tm©k k*uIICfl I
- PIccon1
ci
Figure 5-16 Total achieved occupancies of each fwzction vs time
Tracks were then introduced as in section 5.3.5. Tracks were introduced every 2s for the
duration of the simulation, and the same waveform was used for plot confirmation, track
initiation and track maintenance as that of the surveillance waveform which detected the
target. The results are shown in Figure 5-17 to Figure 5-19 below.
Figure 5-17 shows the timebalances for the first seconds of the simulation. Initially the
surveillance time balances become increasingly negative, since surveillance expands to
use the full 100% system occupancy. Then at t=2s, quite the reverse occurs, and
surveillance becomes completely shut out, its occupancy decreasing to zero briefly,
whilst plot confirmation and track initiation are carried out on the detection.
Interestingly the waveform which detected the target had a dwell time in excess of the
look back interval requested for plot confirmation and track initiation.
Figure 5-17 Surveillance, plot confirmation, track initiation and tracking timebalances with
new target detections every 2s. QitJ UL' 44ké WsQitA C)',WflkL l4fl'1fM
Thus even though the job was scheduled immediately after the surveillance waveform
which detected the target, it was still late. This can be seen in the plot confirmation
timebalance. The plot confirmation job was inserted into the radar job table with a
positive timebalance, indicative that the job was already late. Also the track initiation
107

timebalances become more and more positive, (unlike those for subsequent track
detections), showing that the desired update rate cannot be achieved.
Figure 5-18(a) shows the achieved occupancies of each function. The surveillance
occupancy falls away as time goes on, to meet the tracking occupancy demands, which,
by t=85 fully occupy the radar, squeezing out surveillance completely. At time t=45s
there is a particularly high track initiation load which is met by lowering the occupancy
of both tracking and surveillance since they are both lower priority tasks. Figure 5-18(b)
shows the surveillance region occupancies decreasing in proportion with their desired
occupancies, such that they all reach 0% occupancy at the same time.
Oh .
Region 1
Rugion2
Regions
Region 4
Region 5
•.
Regions
. Region?
RegionS
oI . ..........__________
RegionS
K:. \
,j1••_
0 O •
Figure 5-18(a,b) Total achieved occupancies of each function vs time, and achieved
surveillance region occupancies vs time
Correspondingly, Figure 5-19 below shows the surveillance frame times increasing
throughout the simulation, until they become infinite when the surveillance occupancy
reaches 0%.
Region I
RegIon 2
on Regions
Region 4
RegonS
on Regions
Region 7
S • RegionS
• RegionS
I on
on
on
'rt!!! ,
Figure 5-19 Surveillance region frame times for MESAR, with track detections every 2s
108
5.3.7 Plot confirmation latency using the MESAR scheduler

The MESAR scheduling algorithm is not sophisticated in its method of trying to
schedule jobs very close to their desired schedule time. Many algorithms are more
complex, and attempt to shuffle jobs around to formulate an efficient time schedule in
advance [Orman & Potts, 1997], [Orman et al, 1996]. Although this strategy can result
in most of the jobs being scheduled very close to their desired times, there is a risk that
there will be empty periods in the schedule, such that the multi-function radar is not
utilised 100% of the time. To avoid this, MESAR selects the most appropriate job that is
ready at the instant the last job scheduled has finished. This means that the selected jobs
are always scheduled late by some amount- the maximum delay being the dwell time of
the previous task scheduled. This however, ensures that the multi-function radar is
100% utilised at all times.
The delay caused by the scheduler cannot be tolerated for functions such as plot
confirmation or track initiation in instances where long radar waveforms are used. To
get around this problem, large tasks are broken down and scheduled in smaller units.
The scheduling of one task may be interrupted by the scheduling of a higher priority
task. This is of particular significance to fast, crossing or low signal to noise ratio
targets, where rapid look back can be essential in maintaining the target in track, without
resorting to searching multiple beams.
Figure 5-20 plots the look number of each task scheduled as a function of time. Some
tasks are made up of in excess of 45 looks. It is difficult to see any interleaving of tasks
on Figure 5-20(a); however when the time axis is expanded, as in Figure 5-20(b), the
interleaving of plot confirmations with surveillance becomes clear.
SWcs
IS PIaqConIkma

I
I
HI. .•
4 .;Ls
W I'!
Figure 5-20 Look number scheduled as a function of time for plot confirmation tasks (green x)
and surveillance tasks (red +)
109
By performing a simulation over a large period of time, and inserting many targets, it is
possible to show the delay in scheduling of plot confirmation tasks. Figure 5-21
compares the delay in scheduling of plot confirmation tasks with the MESAR
scheduling algorithm (a), to the delay if the MESAR algorithm scheduled tasks as single
un-interruptible units (b). By scheduling tasks in smaller looks the delay in scheduling
of plot confirmation tasks is minimised; 97% of plot confirmations were scheduled
within lOms of their desired time. Conversely just 55% of plot confirmations were
scheduled within that period when tasks were not broken down, and more than 5% were
scheduled over 0.ls late.
flu 0S -. -. Ci,mul.bv. PS701.1t1S

- - - . —
flu OU
"7 07 '
Os . ..... - ........... .
05 O5
01 . .................................
7
02 02
:2 02
7,
2 fl 20
DI.,,.,
Figure 5-21(a,b) Histogram of the delay in scheduling plot confirmation tasks. (a) the MESAR
algorithm scheduling looks, (b) the MESAR algorithm scheduling un-interruptible tasks
5.4 Task scheduling in a rotating array multi-function radar
5.4.1 Background
In a rotating multi-function radar there are three significant complications to the task
scheduling problem.
i. blind arcs in the antenna coverage

ii. tasks must be scheduled close to array azimuth broadside where possible
iii. radar occupancy may be distributed non-un4formly with azimuth
Each of these is discussed in further detail below.
5.4.1.1 Blind arcs in the antenna coverage

Firstly, if full 360° illumination is not available from the multi-function radar's array
110
antenna, then the scheduler must deal with rotating blind arcs, where no tasks may be
scheduled. This also has the effect of limiting the maximum sustainable track update,
depending on the azimuth scan capability of the multi-function radar's phased array
faces and the rotation rate of the arrays.
5.4.1.2 Scheduling tasks close to broadside

Secondly, as discussed in Chapter 2, a heavy penalty may be paid for scheduling tasks
away from array azimuth broadside unnecessarily. Scheduling of tasks scanned away
from array azimuth broadside will result in broadening of the beam in azimuth and a
loss in signal to noise ratio. These in turn result in a degraded monopulse estimate
accuracy.
Using Barton's expression for estimating the monopulse accuracy that is achieved for a
dwell of n pulses [Barton, 1988];
[5-1]
63
cr
2j(snr).(n)
where is the monopulse accuracy (rads)

9 is the 3dB beamwidth (rads)
snr is the signal to noise ratio
we estimate the factor by which the monopulse estimate of a beam degrades as the beam
is scanned from broadside to be;
a 1 [5-2]
degradation factor = I
Lcos 2(ç) ]
where q is the scan angle off broadside(rads)

0 h.c is the monopulse acc. at array broadside
o is the monopulse accuracy at scan angle q
This is clearly of significance for both surveillance and tracking radar tasks, giving a
degradation of a factor of around 5.6 for a scan angle of 60 0 . Broadly speaking there are
three potential courses of action that may be taken as a result of the degradation
described above;
i. accept the degradation in both detection peiforinance and angular plot

accuracy
ii. try to regain the original broadside detection performance, but accept a
111
5 Resource Management in Static and Rotating Array Mu1tfunction Radar
degradation in angular plot accuracy due to a broader beam

iii. try to regain the detection performance, and moreover, try to regain the same
angular plot accuracy achieved on broadside (in radians/degrees)
Option (i) above obviously does not require any increase in radar dwell time, but will
result in poorer radar performance for tasks off array broadside.
Option (ii), assuming coherent integration, will require an increase in dwell time equal
to the loss caused by scanning; i.e. if a 3dB loss is experienced due to scanning the
beam, the dwell time must be doubled to recover the original signal to noise ratio. The
9dB loss caused by scanning at the extreme scan angle of 600 requires 8 times the dwell
time to compensate.
To regain the original broadside performance in both detection capability and angle
estimation, option (iii), requires a very large increase in the dwell time. From the
equations above we may go on to derive the factor by which the dwell time must be
increased to recover the broadside angular plot accuracy (assuming coherent integration
of pulses);
factor by which to increase dwell time=

\ COS (p)
This suggests a dwell time of a factor of 32 times longer than the broadside dwell time
is required to give the same monopulse accuracy when scanned to 60°.
In terms of surveillance, which may already take 80% of the multi-function radar's time-
budget when on broadside, options (ii) and (iii) may not be feasible without increasing
the surveillance frame times. Some degradation in signal to noise ratio or azimuth plot
accuracy may have to be accepted. It should be noted however, that, although the dwell
times increase, this is offset by the number of beams required to perform the search
decreasing, due to beam broadening. This means the total search function time increases
as l/cos4(p) to recover monopulse accuracy and l/cos2 (p) if only signal to noise ratio is
to be recovered.
Thus it is clear that the penalties for unnecessarily performing tasks away from array
broadside are large, manifesting themselves either in the form of degraded detection
performance and plot accuracy, or in terms of increased dwell times for a task, and that
112
the exact timing of track updates is crucial in maximising the efficiency of the tracker.
5.4.1.3 Distributing radar occupancy with a non uniform angular loading

The final complication in a rotating system is that, if we wish, radar-time may be
concentrated from all of the available array faces in a given direction. This can be
extremely useful if the multi-function radar loading is not uniform with azimuth, which
it is highly unlikely to be in an operational scenario. Nothing has been found in the
literature which addresses this topic, suggesting that current systems do not employ this
additional flexibility in their task schedulers.
Figure 5-22 below shows the difference between a static multi-function radar and a
rotating multi-function radar if the radar loading is not uniform with azimuth angle. In
both cases, the surveillance detection performances may be met if, for example, 80%
radar occupancy is dedicated to search. In this example, this leaves 20% for any other
functions such as tracking tasks or missile uplink. If any less than 20% is required for
these functions the surveillance detection performance will be in excess of the
requirement. Any more than 20% will cause a shortfall in surveillance detection
performance.
Static Antenna MFR System Rotating Antenna MFR System
I;
/
Figure 5-22 A comparison of static and rotating multi-function radars in the presence of a non-
uniform loading with azimuth
If a distinct azimuth angular sector exists with a higher radar loading than throughout
the rest of the coverage, then the static multi-function radar will have a non uniform
loading between its faces. Two or three of the faces (depending on whether it is a three
or four faced multi-function radar) will not be facing in the direction of the high load
angular sector. These faces will have less than the example 20% loading for functions
other than search, and will thus have a surveillance detection performance in excess of
the requirement. The one face that is pointing in the direction of the high load angular
113
sector will have more than the 20% loading allowed for other functions, and thus the
surveillance detection performance for this antenna face will be degraded below the
requirement. Unfortunately, this is likely to be just the direction where a good search
performance is desired.
In the rotating multi-function radar the unequal loading with azimuth angle does not
necessarily translate to an unequal loading between the multi-function radar array faces
as it does in the static case. Forward and back-scanning of the surveillance beams may
be used to dedicate higher amounts of radar time from each antenna face in a particular
direction.
In the directions where there is less than the example of a 20% tracking load above,
surveillance may expand to use this extra occupancy and search more beams. In these
directions the surveillance azimuth direction will move ahead of the antenna direction as
beams are being searched at a higher rate. This is termed forward scanning.
Similarly, in directions where more than the example 20% loading is present,
surveillance beams will be searched at a slower rate, and the surveillance direction will
fall behind the antenna broadside direction. This is termed back-scanning.
A rotating scheduler algorithm should be able to utilise forward and back-scanning

when necessary in order to cope with directions of higher than average radar loading.
However, forward and back-scanning should be avoided whereever possible due to the
penalties associated with performing tasks away from broadside. This suggests that the
rotating scheduler algorithm is likely to be required to monitor the multi-function radar
time loading and act appropriately for efficient task scheduling to be possible.
5.4.2 An algorithm for task scheduling in a rotating multi-function radar system

A scheduler algorithm for a single array faced rotating MFR has been described, based
upon the use of a queue to store delayed volumetric search tasks which cannot be
performed in a scan due to an overload of tracking tasks [Barbato & Giustiniani, 1992].
Here an algorithm which aims to avoid the occurrence of this situation is developed. A
rotating scheduler algorithm is proposed below which is designed to cope with the
complications of;
i. rotating blind arcs
114
ii. constraining tasks to array broadside where possible
iii.forward and back-scanning of surveillance when overloaded
The algorithm works using the same timebalance scheme to control the scheduling of
tasks as in the Modified MESAR static algorithm described above in section 5.3.3. The
timebalances are again calculated using the task look back interval (the time between
successive tasks - e.g. the track update interval, or the surveillance beam revisit time),
and the task time (the dwell time of the task).
Additionally, the rotating scheduler algorithm uses the concept of windows of

opportunity. These are angular limits, relative to array azimuth broadside, within which
the task must be inside, in order for it to be scheduled.
A surveillance window of opportunity is shown in Figure 5-23 below. In the example

shown, the window has a limit of — 10° ahead of array broadside and a limit behind array
broadside set to the extreme of antenna coverage. To understand the effect of the
window of opportunity consider a multi-function radar which starts its surveillance on
array azimuth broadside. As the multi-function radar rotates, it searches surveillance
beams at a rate which keeps the beam directions on array broadside, as long as
surveillance is receiving its desired occupancy. In areas where the multi-function radar
is under-loaded, surveillance beams will be searched at a higher rate and will move
ahead of array broadside. This will occur until the forward limit of the window of
opportunity is reached at 10°, beyond which surveillance is not allowed. In regions
where the multi-function radar is overloaded with tasks, surveillance beams will be
searched at lower data rates, freeing up occupancy for higher priority tasks. Surveillance
will start lagging behind array azimuth broadside until the overloaded region is passed.
In this way, forward and back scanning are used to take antenna time from angular
sectors which are under-loaded and given it to those which are overloaded.
D.p.s.d.M ci, r.d Io.dNig

Wlndcw q cppci1unSy
- _
Figure 5-23 A surveillance window of opportunity
115
The forward limit of the window of opportunity in the above example was set to 100
ahead of array azimuth broadside. In practice this would be varied depending upon the
loading of the multi-function radar expected over successive scans. If the multi-function
radar was under-loaded then there would be no reason that the window of opportunity
should extend ahead of array azimuth broadside, since this would incur off broadside
losses with no reason. If the multi-function radar was very heavily overloaded the
forward limit of the window should extend to the forward extreme of antenna coverage
to allow maximum forward scanning.
The backward limit of the window of opportunity need not be adaptive in the same way.
The surveillance beam directions will only lag behind array azimuth broadside if the
multi-function radar is overloaded.
This strategy allows the control of scheduling of tasks when surveillance is within its
window of opportunity. However, what happens when surveillance reaches the forward
or backward limits of the window? Currently, if the radar is under-loaded and
surveillance reaches the forward limit of the window, the simulation injects null tasks
for a duration of ims. i.e. the antenna is not used for lms. If the radar is overloaded, and
surveillance reaches the backward limit of the window, then surveillance beams will be
dropped. Clearly both these actions are not ideal and would not be desirable in a real
system. Possible ways to avoid these situations are taken up in section 5.4.5 below.
A flow diagram for the rotating scheduler algorithm proposed is shown in Figure 5-24.
The algorithm schedules in units of tasks rather than looks at this stage. A description is
given below.
116
HIGHEST PRIORITY LEVEL
Go down to the next Is this level surveillance

orion level J
NO YES
N there a job in it's window withi (ihere a job with a +ve TB, n Choose the most +ve job &
a Dositive TB? J in it's window (ie has a task find the next beam that is
j been missed)? inside it's window.
YES
Is there a job with a -ye TB, in Choose the job with the least
it's window (ie can a task be -ye TB
No tasks can be done. Insert
Schedule a look from the
Increment all
Updatepointing directions of _____________________

tasks. Recalculate surveillance Decrement this job's
windows based on predicted timebalance
loading
Figure 5-24 Flow diagram of the rotating scheduler algorithm
Starting from the highest priority level;
i. If this level is surveillance determine whether a surveillance region is behind

its window of opportunity - if so then drop this beam and find the next beam
inside the window. Choose this job.
ii. If this level is surveillance and a job has a positive timebalance then this
indicates a surveillance job can be done, which may be early. Choose this job.
iii. If this is surveillance and no jobs can be done from (i) or (ii) above, then the
forward limit of the window has been reached; insert a null task for ims.
iv. If the level is not surveillance (i.e. a higher level), and if there is a job with a
positive time balance, in its window of opportunity, then choose this job.
Otherwise, go down to the next level, and repeat from (i)
v. Schedule the chosen task or a null job.
vi. Increment all timebalances
vii.Decrement the chosen job's timebalance
117
5 Resource Management in Static and Rotating Array Muluflinction Radar
viii.Update the pointing directions of tasks, such as tracking tasks, to take

account of target movement.
ix. Recalculate surveillance window of opportunity based upon predicted radar

loading.
5.4.3 Beam search patterns with a rotating multi-function radar

Figure 5-25 shows how the beam directions are represented in space. (Note that the
circles on the figure are nominal, and in the simulation broadening of the beams in
elevation is taken account otT). As the elevation increases, the number of beams required
in each row decreases, until as we approach 90° elevation only a single beam is
necessary.
0 20 40 60 ao 100 120 140 160 100
Figure 5-25 Surveillance beam directions plotted in real space coordinates
Figure 5-26 shows the method used to number the beams in order to generate a search
sequence for each surveillance region.
0 0 0
7 S S S ... Toll IN*16 SI bsu. p.. * Saul.,
Figure 5-26 Search order of surveillance beam directions for a rotating system
The beams are numbered consecutively with increasing elevation in order that an
elevation column of beams is searched at each azimuth angle, from the bottom beam up.
Due to the irregular distribution of beams in space there will not be a uniform number of
beams per elevation column. In the figure the beams are numbered as belonging to one
large surveillance region from 0° elevation to 80° elevation. In practice, an elevation
118
sector this large would be split into several regions, or perhaps even 1 region per
elevation row, reducing this non-unifonnity.
The beam ordering scheme shown in Figure 5-26 will generate a uniform search pattern
of each region, in regular columns, from the bottom elevation of the region to the top
elevation of the region. If there is only one elevation row in a region, the scheme
produces simply a raster scan search pattern.
0 0 0
1 3 1 4 7 II 10121 ......I B.nmanbsu.Jb.J

Iti I I 14 12 1 I II Il I.I Sdidsr
Figure 5-27 Generating a pseudo-random search pattern with a rotating system
Figure 5-27, on the other hand, shows the simple method that is used within the
simulation to allow a pseudo random scan pattern to occur, for some preset angle away
from array broadside (or the forward or back-scan angle that the surveillance region is
currently at). The beams are still numbered in the same way as before. However, in this
case, the order in which they are searched is randomised over a small number of beams
ahead. The beam chosen for search is no longer the next beam in the sequence. The next
beam will instead be the next beam number plus some random integer. This random
integer should be chosen to represent one or more columns of beams. In the example
shown in the figure, if 18 were chosen to be the random integer, and no beams had been
searched (i.e. current beam number was 1) then the next beam to be searched may be
anywhere within the first three columns of beams. By keeping track of the beams that
have been searched a pseudo random scan around the array broadside direction is made.
5.4.4 Results with the rotating multi-function radar task scheduler algorithm
In principle surveillance may be scanned forward to the full azimuth scan limit of the
array in regions where the total occupancy is less than 100%. This would then mean that
the full angular coverage of the array may be dedicated down a single highly loaded
sector through forward and back scanning, if required, without reducing the surveillance
load. For an array with a scan capability of 600, this would allow distribution of 1/3 of
119
its resources (2*600/3600) in any given direction. Thus an occupancy of much more than
100% may be demanded from the multi-function radar within a small angular sector, as
long as a total occupancy of less than 100% is demanded in other regions. In the case of
scanning to the array limit of 60°, with a highly loaded sector of 30° in extent, an
occupancy within that sector equivalent to 400% C be sustained.
However, forward and back scanning of surveillance tasks to allow the multi-function
radar to distribute its load will degrade the performance of those tasks due to the
decreased gain, and increased beamwidth. Clearly, to obtain this maximum value of
forward and back scanning, the surveillance performance would be significantly
degraded. At 60° the surveillance beamwidth has doubled, and detection performance
reduced by 9dB.
In practice, as discussed in section 5.4.1.2, if we wish to maintain a uniform surveillance

detection performance, then as we scan forward or back from array azimuth broadside,
then we must increase the dwell time down each surveillance beam. Thus, even though
we are only achieving the same detection performance, the surveillance occupancy rises;
no further forward scanning will occur when the surveillance occupancy that reaches
100%. Therefore there is a limit on the angle of forward and back-scanning that can be
achieved, that is dependent on the array azimuth broadside surveillance occupancy
chosen [Butler et al, 1997]. This theoretical limit of forward or back scanning of
surveillance is shown in Figure 5-28, below, as a function of the surveillance occupancy
on array azimuth broadside.
Figure 5-28 Limit of forward/backward scanning, assuming a two-way gain loss with
cos3(ço)
For a system which is compensating surveillance waveforms with a factor of cos3(p),

the maximum radar time that may be dedicated towards tracking functions in an
120
overloaded angular sector through the use of forward and back scanning of surveillance,
tinax, befbre the onset of any surveillance degradation, is given by;

2. cos' [5-3]
i,., )
t
= Cs)
where Co is the antenna rotation rate

OCCu,yj!/,.,ch, is the surveillance occupancy on array azimuth broadside
For a multi-function radar which has a broadside surveillance occupancy of 80%, a

maximum scan limit of 21.8° is imposed. Thus the multi-function radar may dedicate
double this amount in a given direction, with no detriment to the surveillance detection
performance; equivalent to more than 12% of the total time-budget of each array face.
This theoretical value cannot be achieved in practice. It is limited by the need for
surveillance to 'catch up' within a rotation; after servicing a highly loaded sector,
surveillance must be able to go from fully back scanned, to fully forward scanned,
within one rotation period. This is scenario dependent - it depends on the angular
distribution of the tracking load. To determine the exact performance that may be
obtained through forward and back scanning, a simulation of the scenario is necessary.
Figure 5-29 and Figure 5-30 present the results of a simulation of a rotating multi-
function radar in the presence of a highly loaded sector. In this experiment a high
loading of tracking tasks were injected in an angular sector of 30° in azimuth. A
surveillance volume was constructed, consisting of a single region, which spanned 0°-
20.5° in elevation, covering all azimuth angles. The parameters of the regions specific to
the scheduler are shown in Table 5-5 below. An array leaning back at an angle of 20°
from vertical, and generating a 2.6° beamwidth was assumed.
Region Broadside Occupancy No of beams Frame time Function time Look Interval
Dwell time
I 0.589ms 80% 1359 Is 0.8s 0.736ms
Table 5-5 Scheduler specific parameters of the surveillance volume
Figure 5-29 demonstrates forward and back scanning of surveillance tasks during the
simulation. In the interval 0-Is, no tracking tasks were present. Forward scanning from
00, at the simulation start, to the limit of around 22° occurred in the first third of a
second . In the interval l-2s, a large number of long tracking tasks were introduced over
a 30° sector, causing a desired occupancy from the multi-function radar in that sector in
121
excess of 100%. Surveillance was temporarily suspended in this case, due to the
extremely high tracking load, and was reinstated some 18° behind array azimuth
broadside. It is also important to note that the tracking tasks were performed as close as
possible to azimuth array broadside with this method.
Figure 5-29 A comparison of the array pointing direction with the azimuth direction of
surveillance and tracking tasks
Figure 5-30(a) shows the desired occupancy from the multi-function radar, needed to
service the surveillance and tracking tasks as a function of azimuth. Since the tracking
tasks lay within a 30° sector, the desired occupancy over most of the coverage was
simply that required by surveillance (80%). In the 30° sector, however, the total
occupancy of tasks consisted of 80% for surveillance and 135% for tracking.
'.3..
•J-€ S.S.Itt,.

so • .
ill
US
I::
IN
.5
--I,.'
Figure 5-30 (a) Total desired occupancy of surveillance and tracking tasks as a function of
azimuth angle, (b) Surveillance and tracking occupancies used to service the desired
occupancies shown in (a)
Figure 5-30(b), shows the actual occupancies of surveillance and tracking that were
needed to service the desired occupancies of Figure 5-30(a). In regions where only an
80% occupancy was desired, 100% of the occupancy was devoted to surveillance to
allow forward scanning to occur.
122
Thus, this experiment showed a desired tracking occupancy of around 94% of the
theoretical value, could be serviced by the multi-function radar using forward and back
scanning. This experiment was repeated for several other values of sector size and
surveillance broadside occupancy. The results are presented in Table 5-6.
Surveillance Sector Width Simulated maximum Theoretical Percentage of

occupancy on (deg) track occupancy (%) maximum track theoretical value
broadside (%) ___________ _________________ occupancy (%) achievable (%)
80 10 410 437 93.93
80 20 205 218 93.93
80 30 135 146 92.78
80 40 100 109 91.63
70 10 535 548 97.67
70 20 265 274 96.75
70 30 175 183 95.84
Table 5-6 Theoretical and Practical (through simulation) maximum achievable track
occupancies
5.4.5 Other resource management issues for a rotating multi-function radar

Currently the algorithm performs forward scanning of surveillance in azimuth angular
sectors where the radar loading is less than 100%. It may be desirable to forward scan
the antenna even when the loading in an angular sector is 100% or more. Also, when
surveillance reaches the boundaries of its window of opportunity either null tasks must
be inserted (if the forward boundary is reached), or surveillance beams must be dropped
(if the backward boundary is reached). Neither of these are desirable for the multi-
function radar system.
The following are examples of action that may be taken in the radar control in the case
of under-loading or overloading of the radar;
i. increase or decrease surveillance frame times in some regions

This must be done in multiples of the rotation period of the antenna.
ii. increase or decrease the surveillance dwell times

Two methods exist for achieving this; firstly, the number of pulses in a given
direction may be increased or decreased to modify the dwell time. The
disadvantage of this is that the number of pulses added or subtracted may have to
be in units of bursts of pulses. Another method is to modify the PRFs of the
surveillance waveforms adaptively, extending or decreasing the unambiguous
range.
123
iii. increase or decrease the surveillance beam spacing

This is an effective alternative to changing the surveillance frame time
significantly, which will modify the surveillance detection performance. A
disadvantage is that if the beam spacing is increased too much, then there may be
regions in the surveillance coverage with large dips in detection performance.
This may be alleviated by utilising scan interlacing as suggested by Billam.
iv. increase tracking data rate or dwell times

This applies only in the case of under-loading, since our main aim is to avoid
decreasing higher priority functions in the case of overload.
These mechanisms may then be included in the resource management of the rotating
system allowing radar time to be freed from areas where it is not required and spent in
other areas.
5.5 Penalty functions and fuzzy logic for efficient scheduling
The approach of the rotating task scheduling algorithm implemented above is closely
related to the MESAR static task scheduler algorithm of section 5.3.2. All tasks are
scheduled at, or slightly after their desired time, depending upon the radar load, and
their relative priorities. It is not an optimal method of scheduling the tasks. What is not
reflected in this scheduling strategy is what the penalties of scheduling these tasks early
or late are. Different functions will be affected in different ways when their tasks are
scheduled early or late, or ahead of or behind azimuth array broadside. Surveillance
tasks for example may suffer no penalty if delayed by only a small amount of time
relative to the search frame time. However the same delay may cause a track to be lost,
or a plot confirmation to fail.
An improvement would be to reflect these penalties in the task constraints that are
passed from the resource manager to the scheduler. This is being considered in the
airborne case by Harvey, although no work has been published at this stage. In the static
case, the penalty functions need only reflect the temporal variation of the task
usefulness. In the rotating case, the penalty functions need also to include the effect of
the azimuth scan angle on the task.
124
This is proposed as an area for further work; however it is considered outside the scope
of this study.
5.5.1 The use of fuzzy logic

In the rotating case, the window of opportunity concept for regulating the scheduling of
tasks may be extended using fuzzy logic to represent the windows. Rather than the
forward and backward limits of the window representing hard and fast boundaries,
beyond which no tasks may be scheduled, but inside of which they may, fuzzy logic may
neatly represent the penalties of scheduling beyond, or just inside of, a window
boundary.
Figure 5-31 and Figure 5-32 show example windows of opportunity for surveillance and
for a tracking update. The figures show how fuzzy windows may be used to cope with
the fact that in reality there is no sharp cut-off, beyond which a surveillance update or a
track update should not be scheduled.
1FL
Forward scan dependent
on radar loading
—k Broadalde
•.......... Window ot opportunity
- Antenna coverage
Figure 5-31 Fuzzy windows of opportunity for a surveillance task
IIIL
—b Broadalde
•-...--* Track Updatee
- Ant.nna coverage
Figure 5-32 Fuzzy windows of opportunity for a tracking task
Simple membership functions, for example 'in_window', may be generated which

represent how appropriate it may be to schedule this task, at this time. The shape of the
membership function should be based on relevant penalty functions such as scheduling
125
away from the exact update time and scheduling away from array azimuth broadside.
Fuzzy logic has been successfully applied to the control of an airborne intercept radar by
Popoli [Popoli & Blackman, 1987]. A fuzzy logic control strategy has not been
attempted thus far, but it may provide a simple, effective and flexible solution to the
rotating scheduling problem.
5.6 Conclusions
A set of objectives which a scheduling algorithm should attempt to meet have been
proposed. It is immediately clear that not all objectives can be perfectly met in a
practical situation, thus the importance of each objective to the multi-function radar
system must be reflected in the scheduling algorithm.
A simulation model of a static multi-function radar, centered upon the task scheduler
has been implemented. This allows design of multi-function radar search coverage
volumes and analysis of the performance at a macroscopic and a microscopic level.
Three static task scheduler algorithms have been incorporated into the simulation and
tested. The first is a simulation of the MESAR look based scheduler. The other two
work on the same principle, but are modified to calculate and reset time balances in a
different way. Of these, one is a look based scheduler as per MESAR, the other is a
simplified version which assumes that tasks are not broken down into looks, having
some execution time advantage.
The simulation has been tested and validated using the MESAR search region
parameters as a test case. A detailed assessment of the MESAR scheduler has been
made. This was important since there was some question over whether this algorithm
was performing as expected in the MESAR radar [Godbehere, 1996].
The results show that the scheduler algorithms achieve 100% utilisation of the multi-
function radar antenna in all circumstances. This was the second of the three objectives
which applied to a static multi-function radar, and reflects the importance attached to
full utilisation of the multi-function radar antenna. This objective is achieved at the
expense of scheduling tasks as close as is possible to their desired time (the first
objective). Tasks are always scheduled after their desired update time to some extent.
126
This is minimised in the scheduling architecture, by splitting waveforms up into smaller

looks so that the scheduling of a long waveform may be interrupted by a higher priority
function. This entails, however, some extra complication in the signal processing since
data from the interrupted waveform must be buffered to allow the interruption.
The MESAR algorithm is not optimal, in that better algorithms could be devised which
give 100% radar utilisation and schedule tasks closer to their desired time by allowing
tasks to be scheduled early. However, according to the requirements set out for timely
scheduling, the MESAR algorithm is sufficient, and need not be more complicated.
The scheduler algorithms used in MESAR themselves are simple and flexible. This is
because MESAR uses what may be called a distributed resource management
architecture. Many of the computations which result in allocation of the multi-function
radar resources are made in other separate areas of the radar control software.
An analysis of the differences between static and rotating multi-function radars has
drawn out the complications that the scheduling algorithms must overcome for efficient
operation of the multi-function radar. These are summarised;

iii.radar occupancy may be distributed non -unformly with azimuth
It is concluded that a rotating scheduler algorithm should include some form of

prediction of the radar load over one or more successive scans in order to allocate
resources efficiently in situations of high radar loading. For resources to be allocated
correctly in this situation a rotating multi-function radar scheduler must allow the use of
forward and backward scanning of tasks to enable distribution of resources with angle.
The theoretical extent to which this can be achieved has been determined for a multi-
function radar which does not compensate the transmitted energy in surveillance with
scan angle, and for a multi-function radar which does.
The cost of unnecessarily scheduling tasks away from array broadside in rotating multi-
function radar systems has been determined in terms of the decrease in detection
performance and in terms of the decrease in angular estimation accuracy when using
monopulse. (and resolution capability). It was found that dwell times must be increased
127
by l/cos3 ((p) in order to maintain detection performance, but by l/cos 5 ((p) in order to
maintain plot accuracy. In surveillance the increased dwell times are offset somewhat by
the increase in beamwidth (so a reduced number of search beams), and therefore the
search function time needs only be increased by a factor of 1/cos 2(ç) in order to
maintain detection performance, and by 1/cos4 (ip) in order to maintain plot accuracy.
An algorithm has been presented for the scheduling of tasks in a rotating radar that
copes with the problems outlined in (i) to (iii) above. It is highly efficient in distributing
the radar occupancy to the angular sectors which require it most. It will automatically
adjust from a situation in which tasks are uniformly distributed throughout the MER
coverage to a situation in which the threat may come from a particular angular sector, so
that many of the MFR tasks are concentrated in a small azimuthal sector.
A simulation of a rotating multi-function radar centered upon the above algorithm has
been implemented. Results from this have validated that the algorithm performs well in
both under-loaded and overloaded situations. Simulation has shown that around 94% of
the theoretical maximum loading can be serviced through the use of forward and
backward scanning. This smaller value arises due to the time taken for surveillance to
'catch up', from scanned to the maximum angle behind broadside, to the maximum
angle ahead of broadside, in one rotation of the array antenna.
128
6 ADAPTIVE TRACKING IN MULTIFUNCTION
RADAR
6.1 Introduction
This chapter looks at the problem of tracking highly manoeuvrable targets efficiently
with a multi-function radar. It investigates strategies to prevent the loss of manoeuvring
targets or targets with high angular crossing rates. Importantly the case of a rotating
array multi-function radar is considered, for which little published literature has been
found.
The following section describes the problem of tracking manoeuvring targets with a
multi-function radar in more detail, and poses some of the significant questions that
arise. Section 6.3 then compares the approach of the MESAR tracking system with some
pertinent published work that was developed for the ELRA bistatic multi-function radar
built by FFM, Germany, on parameter optimisation for a static array multi-function
radar. It is shown that there are some important differences between the two, and that it
would be beneficial to apply the parameter optimisation method that has been used in
the ELRA system to a model of the MESAR tracking system.
Section 6.4 then presents the development and validation of a model of a MESAR like
multi-function radar tracker system. This section contains the development of
algorithms for the multiple beam updating of targets and for performing track updates
with a rotating array antenna. These models are then applied to perform a similar
parameter optimisation to that used for the ELRA system, using the benchmark tracking
problem scenarios, developed by Blair and Watson [Blair et al, 1994]. An extensive
analysis of the performance and radar time-budget requirements of using systems with a
narrow radar beamwidth and systems using a rotating array antenna system is made.
These results are given in section 6.5. Finally, the conclusions of the chapter are
contained in section 6.6.
129
6 Adaptive Tracking in Mu1tfunction Radar
6.2 Tracking of manoeuvring targets with a narrow beam adaptive

tracker
Chapter 4 showed that construction of a multi-function radar with a narrow beam is

possible without increasing the search time-budget and search frame times to excessive
lengths, through the use of simultaneous multiple beam processing. This would give
good tracking performance compared to multi-function radars that are forced to use a
broad beam from the inherent compromise between beamwidth requirements for the
search and tracking functions. In Chapter 5 it was seen that multi-function radar
resources are limited, and that radar time saved from one function may be used by
another within the scheduler if it is correctly designed. Thus there is a strong drive to
keep track update rates low where ever possible, to minimise the radar time-budget used
for tracking each target, and allow this radar time to be used elsewhere. Also in Chapter
5, it was seen that rotating arrays that do not provide instantaneous coverage of the
azimuthal volume (i.e. systems which utilise just one or two array faces), -
limit the maximum track update that can be supported.
Thus, although the inclusion of simultaneous multiple beam processing will allow
multi-function radars to utilise a narrower beamwidth for the tracking function, the
same, relatively low track update rates that are associated with a multi-function radar
will still be enforced. These low update rates, coupled with a narrow beam will increase
the probability that a manoeuvring target, or a target with high angular crossing rate,
will not be illuminated by the multi-function radar pencil beam at the next track update.
This will either result in an increased probability of losing the track, or in the necessity
to widen the search around the expected target position for track updates.
Several papers have been published looking at adaptive update rate target tracking with
a phased array radar ([Cohen, 1986], [Gardner & Mullen, 1988], [Wilkin et al, 1991],
[Munu et al, 1992]). These papers however have not quantified the efficiency of such
algorithms in terms of their use of multi-function radar resources. This has been
performed more recently by Shin et al. [Shin et al., 1995], who propose an adaptive
update rate ilvIM tracker for efficient use of multi-function radar resources. None of the
above mentioned papers, however, consider the additional question of what energy
should be transmitted on the target with a given update strategy. Neither do they
consider the effects of rotating array systems, or the use of multiple beams for track
130
6 Adaptive Tracking in Multifunction Radar
updating.
Some significant questions that arise when considering these problems are given below.
i. Is there an optimal signal to noise ratio at which to perform tracking?

ii. Is it possible to track the required range of manoeuvring targets with a given
heamwidth?
iii.Is this possible in the rotating array case when blind arcs limit the track
update rate?
iv. How much radar time is used in a rotating system compared to a static
system?
v. How much radar time is used in a narrow beamwidth system?
vi. To what extent do we require the use of multiple beams for track updates
against manoeuvring targets?
These questions are addressed throughout the remainder of this chapter.
6.2.1 Simplified analysis of tracking manoeuvring target with a narrow

beamwidth
As a precursor to the detail simulation results that follow later in this chapter, a
simplified analysis of the problem of maintaining track on manoeuvring targets is given.
The case of a linear predictor with no smoothing filter and noiseless measurements is
considered.
The antenna configurations that are studied as part of this investigation are shown in
Table 6-I below, along with the angular width of the blind sector that the antenna
configurations suffer, and also the period, TbId , that this leads to. This is the period for
which no updates may be made in a given direction due to the blind sector, and therefore
defines the minimum update interval which may be sustained. It is defined by the
interval between two updates in the target direction, the first of which is made at the
maximum back-scanned angle (600 behind of azimuth array broadside), and the next
possible update in the target direction, which must be made at the maximum forward
scanned angle (60° ahead of azimuth array broadside). It is this interval which is one of
the limiting factors in maintaining track on manoeuvring targets.
131
6 Adaptive Tracking in Multfunction Radar
Antenna
of Faces I Rotation Period I Blind Sector Width
Static - - - 0.1 s
Rotating I I S 240° 0.67 s
Rotating 1 2 S 240° 1.33 s
Rotating 2 1 s 60° 0.17 s
Rotating 2 2 S 600 0.33 s
Table 6-1 Multi-function radar antenna configurations considered
In the case of the static array multi-function radar, there is no blind sector (given 3 or
more faces), and thus TbIIUd should actually be zero. However, in Table 6-1 TbId is
represented by the minimum track update interval which can be achieved, which for this
investigation has been taken to be 0.is. The benefit for the tracking of manoeuvring
targets of having a second array face in a rotating system is apparent. The blind sector
width is reduced by a factor of 4 giving smaller values of TbIId. (Obviously in the case
of a third array face, TbIId would have no significance as in the static case).
A target trajectory was created, consisting of a period of constant velocity flight before
execution of a constant acceleration 180° turn in azimuth. Both the range at which the
manoeuvre occurred, and the strength of the manoeuvre were varied. The predictor used
simple linear extrapolation between the last two measurements to predict target position
at the time of next update. The maximum error between the predicted target position and
true target position is then the value plotted for the following figures.
Figure 6-1 shows the results for each of the values of TbId considered in the study. It
should be noted that these results are highly optimistic. The introduction of any
smoothing will introduce a lag which will worsen the maximum prediction error. Also,
the introduction of noise into the measurements and of a probability of detection of less
than I will increase the maximum prediction error. Thus it is difficult to relate these
graphs to the absolute tracking prediction errors that may be expected in a realistic
situation. Figure 6-1 does however given an indication of the relative performance that
may be expected of the different antenna configurations.
The results show that an increase in TbIjd implies a shift of the surface up the z axis.
Figure 6-1 (a) and (b) show indiscernible results for the minimum update rates of 0.ls
and 0. 17s, suggesting that the tracking performance against manoeuvring target for the
two faced array system, rotating at is should give nearly identical performance to that of
a static system.
132
6 Adaptive Tracking in Multftinction Radar
This is not however the case for the remainder of the results in Figure 6-1. Each shows a
significant increase in the maximum prediction errors that occur. In the last case of a
single array, rotating with a period of 2s, strongly manoeuvring targets at close range
give maximum prediction errors of several degrees. Given that the maximum prediction
error must be kept within the radar beamwidth, and that these results are likely to be
very optimistic, this would imply that tracking using multiple beam updates would be
essential to maintain track on close, manoeuvring targets with such a system.
0 '00
M ,- (M

(a) (b)
.;.... . . I
/ .. .1
IC
R )
(c) (d)
j
I
R
(e)
Figure 6-1(a,b,c4e) Maximum prediction error between a constant velocity prediction and a
manoeuvring target, with varying manoeuvre strength and range of manoeuvre start. (a)
T=O.1s, (b) T=O.17s, (c) T=O.33s, (d) Tu=O.67s, (e) T=1.33s
133
6 Adaptive Tracking in Multifuncrion Radar
Figure 6-2 extracts the special case of a 7g manoeuvre at a range of 40km from the
above results. This corresponds to the situation that occurs in benchmark trajectory 6.
Up. C.)
Figure 6-2 Maximum prediction error between a constant velocity prediction and a target
manoeuvring with a lateral acceleration of 7g at a range of 40km
A cubic spline (green line) has been fitted through the extracted data points (red
crosses). This gives an indication of the relative performance that the different systems
will give. The difference between the two rotating systems with the smallest values of
Tb11d (the twin faced systems) and the static system is minimal. However the two single
faced systems experience significantly increased maximum prediction errors, and we
may therefore expect a decreased track robustness from these systems against
benchmark trajectory 6.
These results have given a relative indication of the performance we may expect to
obtain from the different antenna configurations. However it is impossible to relate
these graphs to the absolute tracking performance of a given tracker due to the
complexity involved. This may only be obtained through a detailed simulation of the
track smoothing filters, plot measurement errors, track update strategy and use of
multiple beam updates.
6.3 Allocation of resources in tracking for a static array system
Little has been found in the open literature on tracking and resource management for a
rotating array system, thus we start by looking at the static case. The first two important
questions that must be addressed for an adaptive tracking system are;
i. when should track updates be scheduled?
134
ii. what energy should be used for these updates?
Two different strategies to address these points have been adopted in the MESAR multi-
function radar system and in the ELRA multi-function radar system, developed at FFM,
Germany. These two strategies are compared below, and the impact of using
significantly narrower beamwidths than those that are implemented in the MESAR and
ELRA systems are considered.
6.3.1 Allocation of resources for tracking in MESAR

As discussed in section 3.6.3, MESAR calculates the time of next update for a track
based on two criterion [Noyes, 1990];
i. there is a high probability that the target falls within the 3dB beam on the
next update. This value will depend on the azimuth and elevation smoothed
accuracy of the track, and on the amount that the track can possibly
manoeuvre in azimuth and elevation.
ii. there is a constant smoothed track accuracy in position or velocity.
Choosing an update interval, T, that is too large will mean that one of the above is not
fulfilled. Choosing an update interval that is too small will mean the tracking function
will have a track occupancy greater than its necessary value. Thus in MESAR, there are
two times that are used when requesting a track update;
a. latest update time. This is the earliest of the two possible outcomes of (i)
above, depending on whether the azimuth or elevation manoeuvre capability
is larger, and the track uncertainty in each coordinate. It represents the time by
which an update must be made before which there is a chance that the target
will fall outside of the 3dB beamwidth, and not be detected.
b. desired update time. This is the time which is given from (ii) above. It may be
greater or smaller than the latest update time.
The track updates will be scheduled using the minimum of these two intervals. The
track update interval is constrained to a maximum of 5s and a minimum of 0.1 s.
MESAR attempts to maintain the target signal to noise ratio between 26dB and 29dB
135
[Stafford, 1991]. A smoothed estimate of the integrated signal to noise ratio is

maintained, and the number of waveforms integrated for each track update is modified
in an attempt to keep this smoothed estimate constant. The figures of 26dB and 29dB
are thought to be chosen rather arbitrarily in MESAR, to give good plot accuracy from
the monopulse estimate. It is possible that there is room for improving the efficiency of
the tracker in this respect.
6.3.2 Parameter optimisation for allocation of resources in the ELRA adaptive

tracking system
The problem of efficient allocation of multi-function radar resources for the tracking
function has also been investigated by Van Keuk [Van Keuk & Blackman, 1991]. He
attempted to obtain an optimal strategy for tracking by looking at the update interval,
energy used for the updates, and the accuracy requirements as a coupled system.
As Van Keuk states, if AT is chosen to be small, then the probability density function of
the extrapolated target position is highly concentrated in angle space, which will lead to
a good signal to noise ratio at the next update. This has the consequence of a higher
detection probability and a higher plot accuracy. Also, since the beamwidth is
significantly larger than the variance in the probability density function of the target,
there is no need to search in adjacent beam directions to ensure detection. On the other
hand, a very small value of AT is expensive in terms of radar time.
Conversely, a large value of AT leads to a large variance of the probability density

function of the extrapolated target position, which in turn leads to a reduced signal to
noise ratio, and plot accuracy. Also, a large value of AT increases the probability that the
extrapolated target position will exceed the area illuminated by the multi-function radar
pencil beam, and will thus necessitate a search of more than one pencil beam.
There is clearly an optimal choice of AT which will minimise the radar time expended
on the target, and this AT will have a corresponding track accuracy.
Equally, a similar optimisation may be applied to the signal to noise ratio, SNRO , used
for each track update. Assuming coherent integration, the track update dwell time will
be proportional to the signal to noise ratio. Using a long dwell time will lead to good
signal to noise ratio and plot accuracy, but can lead to wasted resources. Conversely,
136
although a shorter dwell time saves resources, a lower signal to noise ratio and plot
accuracy is achieved, and the probability of wasting energy due to missed detections is
increased.
By making a considerable number of simplistic assumptions, an analytical solution may

be found that gives an optimal tracking strategy in terms of radar time. Assumptions are
made about beam positioning losses, target detection criterion, target dynamics, and the
search strategy employed which allow formulation of analytical expressions relating
track update rate, signal to noise ratio, accuracy requirement and radar time on target.
Van Keuk introduces a track sharpness parameter, V0 . This may be understood as the
point in time at which a track update must be scheduled to maintain a fixed angular
accuracy. Let the major axis of the target position error ellipsoid, defined by the
positional elements of the predicted target covariance matrix, be denoted by G. This
parameter will increase as a function of time. A threshold is defined as V0.(bw/2), as the
acceptable track inaccuracy. Thus AT, the period after which another track update must
be scheduled, may be calculated from the point at which G is greater than V0.(bw/2),
shown in equation [6-1] below.

[6-1]
GKtN+i N)
A useful approximation to AT, valid when the tracking filter converges, given by Van
Keuk [Van Keuk, 1978] is shown in equation { 6-2], below. Thus, given the parameters
of a Singer target manoeuvre model, we may calculate an update interval for a track as a
function of the tracking accuracy required.
0.4
[6-2]
u24
ATEO.41''
Ji+!
2
where r is the target range

a is the measurement accuracy
e is the Singer model acceleration correlation time
I is the Singer model acceleration
U = (½bw. V,,)Ia
Also, given a search strategy for a track update of one or more pencil beams, an
approximation to the number of beams that must be searched in order to detect a target
137
6 Adaptive Tracking in Multfunclion Radar
may be made. Van Keuk uses the expression given in equation [6-31, below.

i^(aV,2)2} 16-31
N
where a 1+14(IInPA/SNR0)½
_DII+SNl)
DD
The track update interval and number of beams required are plotted in Figure 6-3,
below, for a range of values of V0 and SNRO. The target parameters chosen were
r=6Okm, =10m/s2, O=60s, Pf=10 5, and bw=1°.
II
I
Ii) 5NR
Figure 6-3(a,b) Track update interval, and number of beams required for each track update as
a function of track accuracy requirement (V 0) and signal to noise ratio (SNR0)
From these parameters we may now make an estimate of the radar load. The average
number of track update beam directions required per unit time, Nbav, is given by the
average number of beams for each track update, divided by the average track update
interval used. Then the radar load of the track is Nbav multiplied by the dwell time of
each track update. Since the dwell time is proportional to the required signal to noise
ratio for tracking, SNRO, the radar load can be approximated by equation [6-4].
J0.4 (6-4]
Lcc
r (v0 , SNR0 , p1)
The form of equation [6-4] shows that Lr may be optimised in terms of V 0 and SNRO,
independently of target range and dynamics. Figure 6-4 shows Lr as a function of V 0 and
SNR D, for the same set of target parameters as used in Figure 6-3 above.
The analytical simplification of the problem results in a minimised value of radar load
for values of SNR0=l5dB, and V0=0.3. Thus the optimal accuracy requirement in terms
of minimising the radar load for tracking is 0.15 beamwidths, independent of the
138
beamwidth used.
Figure 6-4 Radar load as a function of track accuracy requirement (V0) and signal to noise
ratio (SNR0)
A more accurate estimation of the optimal values of SNR O and V0 can only be obtained
through a Monte Carlo simulation. By performing this, Van Keuk concluded that the
estimate for V0 was accurate. However the analytical estimate of the optimal signal to
noise ratio was an underestimate; a value of 18dB for SNRO was indicated from the
simulation.
6.3.3 Comparison of the two strategies

Both the MESAR and Van Keuk strategies arrive at similar values for the track update
intervals that must be used. A difference between the two strategies comes from the
difference in the tracking accuracy requirement that is demanded. The Van Keuk
strategy is to allow the accuracy demanded to be a parameter which may be optimised,
resulting in a value of 0.15 beamwidths. In the MESAR system, this is a user specified
is
value; an assumption thain keeping with an air defence tracking system.
Neither strategies reflect different priorities of target, in the requirements used. A

weapons system radar is likely to have to track a large number of targets, only a few of
which are of genuine interest to the weapons system. Thus different requirements need
to be specified for these different target types. Low priority targets such as airliners and
other commercial traffic need only be tracked with an accuracy that allows track
maintenance to occur with a low probability of track loss, meanwhile high priority
targets, such as incoming missiles, need to be tracked with fixed accuracy requirements,
specified by the weapons system demands.
139
The difference in the values of signal to noise ratio chosen for tracking is considerable
(up to 1 1dB). Therefore, it will be beneficial to apply the same optimisation as has been
used by Van Keuk on the ELRA system, to the MESAR system.
This will be achieved through development of a Monte Carlo simulation of the MESAR
tracker, described in the following sections of the chapter. Rather than investigate the
performance of this system with a random model, such as the Singer model, it is thought
to be more useful to use a range of realistic target trajectories, representative of those
that a multi-function radar may encounter in its operation. Therefore, the model has
been developed and validated using the target trajectories proposed in the Benchmark
Tracking Problem. This was first published in 1994 [Blair et al, 1994] to cope with the
lack of a common means of comparing different tracker systems. The benchmark
problem addressed beam pointing control of a phased array radar against highly
manoeuvring targets, and has been widely applied in the U.S. ([Rhatigan et al, 1994],
[Sastry et al, 1994], [Sworder & Boyd, 1994], [Daeipour et al, 1994], [Tsaknakis &
Athans, 1994], [Blair & Watson, 1994], [Kirubarajan et al, 1998]). It was considered
useful to utilise these trajectories since they are representative of a number of different
air targets, and will ultimately allow results to be more widely comparable. The
benchmark tracking problem and trajectories are described in Appendix A.
6.4 Development of an adaptive tracking system simulation model
A model has been developed that is based on the MESAR adaptive tracker. The
development of the model is described here. The validation and testing of the model,
and subsequent modifications that were necessary, are described in the following
section. Before the tracker itself, and the validation of it, are presented, the modelling of
the target measurement process is presented.
6.4.1 Modelling of the target measurement process

The beamwidth is assumed to broaden with the reciprocal of the cosine of the scan
angle, and is assumed to be separable in azimuth and elevation, as in equation [ 2-8 ].
The gain is assumed to decrease with the cube of the cosine of the scan angle, as in
equation [ 2-9 1 . The scan angle of the beam from array broadside is calculated using
equations [ 4-6 ], where the array is assumed to lean back from vertical at an angle of
140
200.
The target is modelled as having a signal to noise ratio, with a one way loss which varies
with the angle from boresight as in equation [6-5], [Skolnik, 1982]. Here ( is the angle
of the target from beam boresight, and 0 is the radar 3dB beamwidth at the current scan
angle.

2.78!y [6-5]
snr = snr0e
This gives rise to a probability of detection, PD, given by equation [6-6], where PF is
the probability of false alarm, taken to be 10.6.
- I/(1+5NR) [6-6]
D F
The accuracy of the monopulse estimate of the target's position within the beam is given
by equation [ 2-5 ] . Figure 6-5 shows the monopulse accuracy of the estimate as a
function of the signal to noise ratio. The standard deviation of the range measurements
used was 15m.
I
i
()
Figure 6-5(a) Signal to noise ratio as a function of target angle from beam boresight, (b)
Accuracy of monopulse estimate as a function of the target signal to noise ratio
6.4.2 Kalman filtering

In order that a Monte Carlo analysis/optimisation may be performed to find the most
efficient method of tracking targets, a Kalman filter model must be developed. For this
the same strategy has been used as has been adopted in the MESAR radar. A rotated line
of sight coordinate system is used ([Blackman, 1986], [Daum & Fitzgerald, 1983]). This
has the advantage that the filter may be de-coupled, since the positional and velocity
errors are maintained in the diagonal of the covariance matrix. Also, the Cartesian filter
141
terms may be simply related to the polar measurements.
A simple 2 state, constant velocity Kalman filter was adopted for the study. The stages
of the Kalman filter are shown below [Tunnicliff, 1980].
1. Find forecast position and velocity
X -KXK [6-7]
2. Find the uncertainty in the forecast

[6-8]
P;1 =KKk
3. Usc the uncertainty in the forecast and measurement to calculate the filter gain
[6-9]
K K+I = P 1 H 1 [H k+I P;+I H + I + Rk+J
4. Find the smoothed position and velocity

[6-10]
Xk+I = X +I - K 1 (Hk^I x;, - Zk^l)
5. Find the uncertainty in the smoothed position and velocity

[6-11]
k+I = - Kk+IHk^II'+I
This type of filter is optimal only for constant velocity target motion. In the instance of a
manoeuvring target, this type of filter must be augmented with some form of manoeuvre
detection and adjustment of the filter to avoid very poor performance. This is seen
clearly in Figure 6-7 - Figure 6-13 below, which show the output of such a Kalman filter
model using the benchmark trajectories 1 and 3. Figure 6-6 shows both a 3 dimensional
and a 2 dimensional view of the first trajectory, the modelled noisy measurements, and
the smoothed filter estimate of position. Figure 6-7 - Figure 6-9 then show the errors
between the smoothed position and velocity and the true position and velocity, and
compares those to the errors between the measurements of radar plots. Figure 6-10 -
Figure 6-13 then show this for benchmark trajectory 3. Tracks were updated with a
constant update rate of 1Hz for this example. A target which gave a 20dB signal to noise
ratio on beam boresight and at array broadside was used in this case.
The filter showed good performance in periods of constant velocity flight, however,
very poor performance during a manoeuvre; smoothed estimates of position were
considerably poorer than the raw measurements. After a manoeuvre, the error in the
smoothed estimate decayed slowly. Although the amount of smoothing that was applied
142

6 Adaptive Tracking in Mult(function Radar
was limited to prevent an ever increasing confidence in the filter's estimate, the filter,
unsurprisingly, still reacted poorly to a target manoeuvre.
I I
0' __ç__cU - S-
:: ' a'
Figure 6-6(a,b) 3 dimensional and 2 dimensional plot of Benchmark Trajectory 1, showing
position truth, position measured and snw othed filter output
S FOS
0 0 Mu
j •
L I
r 5': •........... : . ....
Figure 6-7(a,b) X-coordinate position and velocity errors for Benchmark Trajectory 1
S
...0 0
For.cl-Iruth
Muisw.d-Truli SW
SW
u. •
S
.5
• :
..........* I . - V
:
0 *
:.\.
I"
0 0 dSp -
0 •
0•
ro
Figure 6-8(a,b) Y-coordinate position and velocity errors for Benchmark Trajectory 1
143

6 Adaptive Tracking in Multi unction Radar
ForiciOt -
00 0 YUSUiSdTiUthI
0
hO 4 0 . 0 0
0
0 o0 0 0 ,.
0 00 0000*
10 0Ø •0
0
0
I- -
:;°."°' •*%%0
J :f
i0 • 40 0 0 0 0
0 o° 0
.120 0
0
0 . -
- . 12 12
20 40
Figure 6-9(a,b) Z-coordinate position and velocity errors for Benchmark Trajectory 1
2t' - --
II
Kçmj
Figure 6-1O(a,b) 3 dimensional and 2 dimensional plot of Benchmark Trajectory 3, showing

position truth, position measured and smoothed filter output
• Foracast - Tr,th
10
0 M.ursd_TruthI
I
00% 00 0 % * ,o0 0

4*0
20 40 & & •% 2 I.:
144

• Forqi,t - Truth
0 0 MI20ur.d-T,uVI
•
I
I V
•0•
0 •
%o .% 0 0. ::: J
20 40 h 0 00 2C r
Figure 6-12(a,b) Y-coordinate position and velocity errors/or Benchmark Trajectory 3
I. • Fcrici.t-Truth
0 00 10 0 Mstnd-ThjIh
0
0 o 00 0 0
Q •e°
0 0
:
b ° t• 0 0 b° 0IP
• 0 4 0 0 0 .1•.
0•0+*:0
-2X
2
•
0 •..
0
.
0
,•
0
0
00
o% 0
0 0 00
U.
-120 .....
4' 20 r 2 4 --
6.4.3 Augmenting the constant velocity Kalman filter for manoeuvring targets
The performance of the standard, constant velocity Kalman filter described above is too
poor to be used in this study. The errors indicated in the figures above are such that the
target would have manoeuvred out of the beam and been lost, purely due to the lag in
the filter response. A manoeuvre detection and response scheme was thus implemented
to reduce this problem. The strategy adopted was that first suggested by Quigley
[Quigley, 1975].
6.4.3.1 Manoeuvre detection

The onset of a target manoeuvre may be indicated by the normalised distance (i.e. the
number of variances or standard deviations) between the predicted target position and
the measured target position. This quantity, let us call it J, may be broken down into its
three components in X, Y and Z. J, J and J are simply the residuals (X, Y and Z)
between the target expected position and measured position, divided by the standard
deviation expected of the residual.
145
- Xre [6-12]
Jx -
Here, is the standard deviation of the prediction error, and c y , is the standard
deviation of the measurement error. These residuals are then smoothed from one update
to the next, so that the manoeuvre detection decision is based on more than one update.

E = (1 - + (A)J [6-13]
Where E is the smoothed, normalised residual in the x coordinate, and X is the

smoothing constant. Once this smoothed residual exceeds a certain threshold, in any of
the three coordinates, a manoeuvre may be indicated.
If the manoeuvre threshold is exceeded in any one of the three coordinates, a

confirmation update is requested 0.ls later. Only upon confirmation of the threshold
crossing is a manoeuvre flagged. The state covariance matrix would then be adjusted to
reduce the smoothing applied through the filter. This strategy was adopted to reduce the
number of false alarms due to manoeuvre.
6.4.3.2 Manoeuvre response

The manoeuvre response is achieved by adding a plant noise term, Q, to equation [6-8]
above. This results in a smaller confidence in the filter prediction, and thus more weight
is given to the radar measurements.
P11 - b K'K IT+Q [6-14]

K
The Q matrix is derived for a single dimension as, per equation [6-15] below [Castella,
1980]. The value of q that is used is proportional to the smoothed residual, so that a
larger manoeuvre will result in a higher manoeuvre response.
rT 2 I3 T/2 [6-15]
Q=T,2
The manoeuvre detection strategy has the effect of increasing the Kalman filter
performance in the instance of a manoeuvre, but with some reduction in the
performance against constant velocity targets due to declaration of false manoeuvres.
This is seen in the results shown in Figure 6-14 - Figure 6-23 below. The simulations
146

above were repeated for benchmark trajectories 1 and 3, with manoeuvre detection
implemented. A constant track update rate of 1Hz was again used, however updates
were scheduled alter O.ls when a manoeuvre was flagged.
-'—'I
II
Figure 6-14(a,b) 3 dimensional and 2 dimensional plot of Benchmark Trajectorj 1, showing

position truth, position measure4 and smoothed filter output
- . * • For.t-Tiuli
0 0 M..iumd-T,uti
0
00.. 0... * G
Q . 0. 0 j.O
I
I.
I
* * Fosct - TillIh
0 0 Muur.d-Truth
0 0
:....i..i..:::.;...............
00 0
0 4
0 0 %6
0
. 0.. .
I 0 0 0 000
00 b 0
0° + 0 U, •'%• •1
0 00 . 0
* I
• - I— •'I
+ 0* 0 000
°° 0
• bo I
q60 0 0 :
0
0
0
10 II 10 •10 III 110 III 110 0 II 10 10 10 •10 10 110 110 110
147
00
° • •
0 0 00
00 0
0
F ••.:
uo •c. or •.
Figure 6-1 7(a,b) Z-coordinate position and velocity errors for Benchmark Trajectory 1
00 .................................................
00
IT
II .................................................................................
01
II . . .
01 .
7 - • • • •7 •
Figure 6-18(a,b) Manoeuvre detection flag for Benchmark Trajectory 1
I 1•

position truth, position ,neasured and smoothed filter output
148

6 Adaptive Tracking in Mult (function Radar
0 Foriut - In..Ifl
0 0 M.uuw.d_InithI.
0 0
0 0
• jof 0 0 I;
0 O 0
I
0•
' eOi.o0
0*
:°
••
0 0
0 9
o 0 4 0
0
*
0
u • •r •..:
I. • Fac,csst-Tnili
0 a I 0
0
*
4* 0
o0 I;
• 0
10b0
I
I •Io
I
0 00+
0 • ,# 0 0 •.1, 0 000
00 0 0
0
0
Tflflo fl!
,
I a *
0 0
* For.csst-Tiui
Mu..ur.d-Thji
(III.
1111.
I. 0 0 0
0 0 0 0 0
fl_fl
00aeo
o
!1000000e4+
• • 0
0 04cp%coo *o
' o2.. go.. (t°' .1 I
.:4 •.0 uoo.o°0
0 0*0 4.•
0 0 0 0 o % 0 • 0 + 0 0 0
0 0
0 0.0
0 o: -- -
1 . __
Figure 6-23(a,b) Manoeuvre detection flag for Benchmark Trajectory 3
149
The results show a considerable improvement in performance during target manoeuvre,

with only a small impact on the constant velocity filter performance.
As stated by Watson [Watson & Blair, 1993], tracking of manoeuvring targets with
single model track filters supplemented by manoeuvre detection in this way, gives poor
performance in comparison to multiple model techniques. The common method of
increasing the filter state estimate error variances suffers from two problems. Firstly, the
filter response is delayed because the error variances are not increased until after a
manoeuvre is declared. Secondly, after a manoeuvre has finished, the filter has large
error variances, making a bias difficult to detect, and thus resulting in a delay in the
decision to reduce the covariance matrix errors. Watson proposes the use of the
Interacting Multiple Model (1MM) to improve performance in these circumstances.
However, the objective of this study is to determine an efficient method of multi-

function radar resource allocation for target tracking. Absolute tracking accuracy is not
the most important issue. Therefore, the approach taken here will be to develop an
efficient tracking strategy for static, rotating and simultaneous multiple beam multi-
function radars using a simple single model filter, supplemented with manoeuvre
detection. This may then be extended to use multiple model methods such as 1MM, for
implementation into a real system, when high accuracy tracking is crucial.
Thus the aim of the optimisation will therefore be to determine the most efficient
tracking strategy that will give the desired track accuracy, neglecting that during target
manoeuvre. It is accepted that using a constant velocity predictor during phases of
manoeuvre will give a poor smoothed tracking accuracy, and that this may be improved
upon in several ways.
6.4.4 Adaptive update rate strategy

The model uses the MESAR adaptive update strategy described in section 6.3.1 above.
The time of next update is bounded between 0. ls and 5s. An example of this is shown in
Figure 6-24 - Figure 6-26, which show results from tracking benchmark target 1. Figure
6-24 shows the three dimensional and plan views of the simulation output. The adaptive
update rate is clearly visible; during the early phase of the track, the update rate is high,
before becoming lower as the confidence in position and velocity grow. At the points of
manoeuvre, the update rate can be seen to increase again, reflecting the increased
150
6 Adaptive Tracking in Mulufunction Radar
uncertainties in the covariance matrix, causing a decreased desired update interval. This
is more clearly visible in Figure 6-25, which piots the latest, desired and chosen update
times over the course of the simulation. The results presented are those for a target
which gives a 20dB signal to noise ratio on beam boresight and array broadside, and for
a radar beamwidth of 2.5°.
'00

position truth (blue crosses), position measured (red circles), and smoothed filter output (green
plus signs)
• • MESAR Laleut Up Irtlurvut

MESARDI*SdUpthIIaIIWI
1,.
1, c1IOSSnUISInIVII
1pT;' 7f.r
I,)
Figure 6-25(a,b) Track update interval used for tracking of Benchmark Trajectory 1
• • Forucuat - Truth • •. • Forec.st-Tru

0 •
0 Mui,ed-Trut • • Med-Tr
Smoothud-TrM • 0 •• • Sinocthsd-Tr
0• 00 -
0D 0 000
v.
1, • •#' •:r •. .;L. 0181,
III 0 '4' 0
& 'to
00
U •, 0 0
0 •%o ;54 . •.
0
I•.:ç •°p
'e
p0 00 : 0
I %0
0 0
0 0 •
-03 0 • 0• ji•
0
.o,J.•0 • 0
I • 0
Figure 6-26(a,b) X and Z coordinate (azimuth and elevation) position and velocity errors for
Benchmark Trajectory 3
151
6.4.5 Adaptive transmitted waveform energy

The modelling of MESAR's adaptive transmitted waveform energy is performed in a
simplistic fashion. A nominal waveform of Ims is assumed to give a signal to noise
ratio of 18dB. This waveform length is then increased or decreased in proportion,
assuming perfect coherent integration, to achieve the desired signal to noise ratio. Each
time a track update is made, this waveform length is added to the total radar time that
has been used to track the target. The waveform length is adjusted on a continuous scale,
and takes no account of the actual PRF of the waveform, or whether it was some kind of
Doppler burst which would inevitably quantise the units in which the waveform may be
adjusted.
6.4.6 Missed detections and declaration of lost track
if a track update failed to detect the target, updates are scheduled at the minimum update
interval, 0. 1 s, until the target is re-detected. The same energy is used in the transmitted
waveform as for the previous update.
If the target is not re-detected within 20 updates then the target is declared lost. No data
from such a target will be included in the Monte Carlo results, except the fact that the
target was lost.
6.4.7 Development of a multiple beam track update strategy

A strategy devised for multiple beam track updates, to allow robust tracking of
manoeuvring targets, is described below.
In the instance that the target has not been detected for a set interval of time, multiple
sequential beams are searched at each track update rather than a single beam. The period
of time after which this occurs is based upon the latest update interval calculated within
the MESAR tracker, described in section 6.3. 1.
The latest update time gives the period after which a target which starts executing a
manoeuvre of the maximum assumed strength will be outside of the radar 3dB
beamwidth. In the event that multiple missed detections occur, it is assumed that the
start of the manoeuvre is not detected, and thus the target could manoeuvre out of the
radar beamwidth in a smaller interval than the latest update interval. Thus, if the target
has not been detected for more than half of the latest update interval, a multiple beam
152
update is triggered.
The use of multiple beam track updates is shown in Figure 6-27. Initiaily, a four beam
cluster is searched around the predicted target position. Each of the beams is searched in
turn, until either the target has been detected, or all of the four beams have been
searched. The four beam cluster is continually used until the target is re-detected. Also,
the latest update interval is recalculated using an effective beamwidth of double the radar
beamwidth originally used in the calculation, If the target remains undetected for longer
than this second latest update interval, then the use of a seven beam cluster is initiated.
- . Similarly the latest update time is again recalculated
using an effective beamwidth of three times the original value, and if the target remains
undetected still, then the tracker activates the use of a 19 beam cluster for track updates.
(rn hi ici di
Figure 6-27(a,b,c,d) Beam update strategies, and their periods of use
Figure 6-28 and Figure 6-29, below, show results from using the multiple beam update
tracking strategy.
S
I
•
a I
S
20,
'S IS
5I
•,S__ - -.
• •
'4
Thl.. Ii
Figure 6-28(a,b) Multiple beam update tracking of Benchmark target trajectory I using a signal
to noise ratio of 25dB and a desired accuracy of 1/4O of a beamwidth. (a) Plan view of track
showing position truth (blue crosses), position measured (red circles), and smoothed filter
output (green plus signs), (b) Number of beams as a function of time
Figure 6-28 shows the situation for a signal to noise ratio of 25dB, and a desired track
accuracy of 1140th of a beamwidth. In this instance, with the 2.5° beamwidth used,
multiple beam updates were at no point requested. However, in Figure 6-29, where the
153
signal to noise ratio has been reduced to just 12dB, and the desired accuracy reduced to
1120th of a beamwidth their use is clear. (Note that the target would certainly have been
lost in this instance if multiple beam updates had not been used).
-
•1
I
j
0
40 40 00 100 100 110 100 100

T(l
Figure 6-29(a,b) Multiple beam update tracking of Benchmark target trajectory 1 using a signal
to noise ratio of 12dB and a desired accuracy of 1120" of a beamwidth. (a) Plan view of track
showing position truth (blue crosses), position measured (red circles), and smoothed filter
output (green plus signs), (b) Number of beams as a function of time
6.4.8 Development of a rotating array track update strategy

No literature has been found on adaptive update rate tracking for a rotating array multi-
function radar. It is thought that the majority of current systems simply use a fixed
update interval, related to the rotation period of the array.
This strategy could have been adopted here, however such a system suffers a poor
efficiency in terms of radar time. Therefore, the adaptive track update strategy used in
the MESAR static array system has been adapted for use with a rotating array. This also
enables a comparison to be made between the performance and efficiency of a static
array and rotating array system, based upon similar tracking strategies.
The desired and latest update intervals are calculated in the same manner, and are then
adjusted to cope with the additional constraints imposed by the rotating array/s. The
strategy adopted updates tracks on array azimuth broadside wherever possible to ensure
that maximum efficiency and accuracy is attained in the majority of track updates. Track
update intervals are shortened in most cases to ensure the desired update and latest
update intervals are always met. This results in a slightly higher average update rate in
the rotating array tracker to that in the static array. This in turn may result in a higher
total radar time required for target tracking, although this is offset by the lower average
track waveform dwell times that are required, since the majority of updates are made on
array azimuth broadside and do not require extra integration gain.
154
Once in a situation of a stable track, update intervals will extend over one or more
sweeps of the array faces past the target direction. In this situation the update interval
may be shortened so that it occurs at the closest passing of array azimuth broadside. This
is shown in Figure 6-30, which plots the track update directions relative to the array
coverage.
dupdale
L
° of rotatton
Within array coverage

+60° —0. Current update
0 Desired next update
Quantised next update
Figure 6-30 Quantisation of track update directions to array azimuth broadside, shown for a
twin faced array multi-function radar
Figure 6-31 shows the different conditions of track update interval which require
different action to be taken. An explanation of these different conditions and the actions'
arising from them is given below.
If the current track update has been made ahead of array azimuth broadside, then there
are just two conditions for the calculation of the next update interval. These are shown
in Figure 6-31(a).
° )nirection
r LFj0fl
of rotation /01 rotation
/

+60° 460°
Figure 6-31(a,b) Different conditions for the quantisation of track updates, shown for a twin
faced array multi-function radar. (a) Conditions for a current track update ahead of array
azimuth broadside, (b) Conditions for a current track update behind of array azimuth broadside
In the first condition, the interval to the next track update is less than the time before the
track direction crosses array azimuth broadside. In this instance, the desired update
155
interval remains unchanged to ensure the track accuracy is met, or to ensure the target
remains within the radar beam. In the second condition, the desired update interval is
long enough for the track direction to pass in front of the array azimuth broadside
direction. In this case the update interval is shortened by the smallest amount that
ensures the track update occurs on array azimuth broadside. Both conditions ensure that
the next track update will occur within the array coverage limits.
Figure 6-31(b) shows the different conditions that arise if the current track update was
made behind of array azimuth broadside. In this case there are four different conditions
requiring different forms of action. If the next update is within the same array coverage
as the current update, condition 1, then the update interval remains unchanged. If
however the next track update falls outside of this array coverage, as in condition 2, then
the update interval is shortened so that the next update is made 600 behind azimuth array
broadside. This shortening gives rise to the possibility that the resultant track update
interval will be less than the minimum allowed, assumed to be 0.ls here. Then, the next
track update interval has to be extended to allow the blind arc to pass, and the update
will be made at 60° ahead of array azimuth broadside. If this occurs then the desired
track update interval is not met.
Condition 3 shows an update interval that is long enough to have exceeded the period of
time in which the track direction falls within the blind arc, but still not long enough for
the track to have passed in front of array azimuth broadside. In this instance, the track
update interval again remains unchanged.
Finally, condition 4 shows the case of update intervals long enough for the track
direction to have passed in front of array azimuth broadside. Again, in this instance, the
track update interval is quantised down to allow the update to be made on azimuth array
broadside.
The results from the rotating array adaptive update strategy are shown in Figure 6-32 -
Figure 6-34 below. These are for a 1140th of a beamwidth desired track accuracy with a
beamwidth of 2.5°, and for a signal to noise ratio of 20dB. The case of a multi-function
radar with a twin faced array rotating with a period of 2s has been chosen.
Figure 6-32(b) shows the desired, latest and chosen update intervals for the rotating
array update strategy. The quantisation of the update interval is clear. In periods of
156

6 Adaptive Tracking in Multfiinction Radar
constant velocity flight the update rate can be seen to be quantised to the interval
between array azimuth broadside crossing the track direction (is). This is clearer still in
Figure 6-33, which shows the azimuth angle of the track updates away from array
broadside. In periods of constant velocity flight the track updates are almost always
made very close to 00, resulting in the maximum azimuth plot accuracy and minimum
possible scanning losses.
MESAR Latmst Upd.Iu This

5 . . MESARD..ITedUpdt.ThI.
0 0 UpdstsThisU.sd
........-4......
14
' r
2 1 J0 ;.-•
14.. Is)
Figure 6-32(a,b) Rotating array update strategy tracking of Benchmark target trajectory 1
using a signal to noise ratio of 25dB and a desired accuracy of 1/40" of a beamwidth. (a) Plan
view of track showing position truth (blue crosses), position measured (red circles), and
smoothed filter output (green plus signs), (b) Track update interval as afunction of time
– _t n . —
• _•
•WflhI 5.4. IT! .5
" .r
•I = • 4fl

20 10 50 l);0 ,0!10 10 10 200
I... II!
Figure 6-33 Angle of track updates away from array azimuth broadside for the rotating array
update strategy (red squares). Overlaid is the manoeuvre flag (blue circles), showing where the
tracker perceived a manoeuvre
Also shown on Figure 6-33 is the manoeuvre flag of the tracker. In periods of
manoeuvre when the desired update interval becomes small, track updates are forced to
be made away from azimuth broadside. This is also visible during track initiation for the
first few seconds of the track life.
Figure 6-34 shows the track accuracy that was achieved against benchmark target
trajectory 1. The adaptive update rate rotating tracker was successful in achieving an
accuracy in excess of that desired, excluding periods of manoeuvre.
157

6 Adaptive Tracking in Muirjfunction Radar
01 -
Forecast - Truth I 0 • Foracast-Tr
0.4 0 0 0
I. 0 Msusurud-TiuIhI 03 0 0 I • Mua,wsd-Tu
Smoothsd_Trral • 0 • Smoothed-Ti
S
coj •0
0.4
o
0 ' o;0 ••
• •_ 0 :.
f • It.ad,
I*:0 •:i0.:,
0, 0
Icc 0 p000 0
0 00' •100
005 0
00.10.0
.40
-Ci
00 0
0 -Ca ' 0
-0l
0 -04
0
. a, ID ID 100 140 140 I 140 10 40 10 00 100 Ia, 140 lID 140
Figure 6-34(a,b) X and Z coordinate (azimuth and elevation) position and velocity errors for
Benchmark Trajectory I using the rotating array track update strategy
6.4.9 Validation and testing of the adaptive tracking system simulation model
For the purposes of parameter optimisation the models were to be used over a very wide
range of parameters. It was essential to ensure that the tracker performed sufficiently
well over the parameter range for successful conclusions to be drawn. In practice, this
became an iterative process, and improvements to the tracking algorithms were made
when inconsistencies arose. These are outlined below.
The modifications that were made to the initial model are listed below;
i. Manoeuvre response; it was found that the response to a manoeuvre was

often not large enough; i.e. the uncertainty terms in the covariance matrix
terms were not increased enough to ensure the filter gave much greater weight
to the target position measurements. This was due to a very short update
interval being used at the point of manoeuvre detection. In some
circumstances, particularly when using a narrow beam, this was significant
enough to cause track loss. The solution was to use the maximum of either the
last update interval, or the average of all of the update intervals in the track
life, in the manoeuvre response calculation.
ii. Termination of a manoeuvre declaration; it was found that the flagging of
manoeuvres was often turned off prematurely. The solution was to enforce a
strategy for termination of manoeuvre similar to that of declaration. Two
consecutive updates were required with residuals below a threshold for the
manoeuvre declaration to be terminated.
iii. Calculation of the latest update time; it was found that at very high signal to
noise ratios, and with narrow beamwidths, targets were lost unexpectedly
- 158
during a manoeuvre. This was due to the fact that in these circumstances, the
target would be tracked throughout the manoeuvre using the latest update
time, instead of the desired update time (because the tracking accuracy
requirements were met purely by the plot accuracy). However, the latest
update time is calculated assuming a manoeuvre has only just begun, and was
too long to retain the target in the beam if the manoeuvre had been underway
for some time.
iv. Calculation of the desired update time; it was found that at very low signal to
noise ratios, the desired track accuracy was not being successfully met. This
was due to several missed detections in a row causing the desired update time
to be exceeded. The solution was to scale down the desired update interval by
a factor proportional to the target probability of detection.
This process was considered valuable, and ultimately, resulted in a robust tracker
capable of performing over a wide range of signal to noise ratios, desired accuracies and
radar beamwidths. It is thought a worthwhile process for any such system, since it is
generally unknown in advance what parameter range the system will be operating in,
and how that will change. For example, at the point of target detection, the tracker may
be operating with a very low signal to noise ratio. Conversely there may be times where
the signal to noise ratio is high, for example if the target approaches close in range.
Also, the desired accuracy on the target may change, depending on whether it is a target
of genuine interest or not.
6.5 Results
6.5.1 A summary of the results collected

Since the tracking systems proved to have a complicated performance with the variation
of the parameters, it was decided to perform the Monte Carlo simulations over all of the
parameters, rather than a select few. This would ensure that any anomalies in the tracker
performance against each of the bench mark targets with a given parameter set were
observed. Also, it would allow any deviation in the optimal signal to noise ratio or
desired track accuracy to be observed.
159
Table 6-2 shows the variation in the parameters that were performed. A total of 10
different signal to noise ratios and 20 different desired smoothed track accuracies were
calculated for each data set; averaged results from tracking 25 targets were used at each
data point. This was performed for each of 7 target trajectories (the 6 benchmark
trajectories, and I constant velocity test target), and for each of the 4 different radar
beamwidths. Since this was repeated for each of the final total of 10 different static and
rotating adaptive tracker configurations the total number of targets tracked in the data
set was 1.4 million.
Parameter Minimum Maximum Step

Signal to Noise Ratio 12 dB 30 dB 2 dB
Desired Smoothed Track Accuracy N = 2 N 40 N=2
(Fraction of a beamwidth - 1/N BW)
Radar Beamwidth 0.5° 3.5° 1°
Target Trajectory Number 1 7 1
Table 6-2 Variation in parameters for the Monte Carlo simulations
A total of 10 different tracker configurations were examined, shown in Table 6-3. The
majority of these were for the different configurations of arrays for a rotating multi-
function radar system. The most likely candidates for a system performing air defence
were considered, namely either a single or double faced system with a rotation period of
either 1 or 2 seconds.
AntennaConfiguration _____________________
Static/Rotating Number of Rotation Single/Multiple Beam
_____________ Faces Period Updates
Static - - Single
Static - - Multiple
Rotating 1 1 s Single
Rotating 1 1 s Multiple
Rotating 2 I s Multiple
Table 6-3 Antenna configurations considered for the Monte Carlo simulations
Clearly, it is not possible, or worthwhile, to display all of the results of such an analysis
in this thesis. Therefore it has been decided to present the results for benchmark target 6
only here. This proved to be one of the most stressful cases, containing strong angular
manoeuvres in both azimuth and elevation (up to 7g), and also longitudinal manoeuvres
(up to 1 g). The track filtering performance against other targets was generally superior
to the results obtained against this target, unless otherwise stated.
160
A complete set of results against all 7 target trajectories is given for the cases of the
static array multiple beam update system, and the rotating array multiple beam update
system. These are contained in Appendices B and C respectively. Results for the other
rotating array configurations considered are given for benchmark target trajectory 6
only, for comparison. These are also contained in Appendix C.
6.5.2 Single beam static array multi-function radar tracker

The first results obtained were for a static array multi-function radar using a single beam
for each update. This tracker is equivalent to that currently implemented in the MESAR
system. A relatively broad radar beamwidth of 2.5° was used, representative of that used
or envisaged for current multi-function radar systems. The track update rate was limited
to between O.ls and 5s. The target signal to noise ratio was varied from 12dB up to a
maximum 30dB, and the desired smoothed tracking accuracy was varied from ½ a
beamwidth to 1/40t of a beamwidth.
Figure 6-35 - Figure 6-37 show the variation of the significant outputs of the Monte
Carlo analysis; the total radar time required to maintain the track, the percentage of
successful tracks, the average track update interval used, and the smoothed tracking
accuracy that was achieved. Each data point on the graphs, here as throughout this
chapter, has been calculated by taking an average of the results of tracking 25 successive
targets.
Figure 6-35 shows clearly that there are large savings to be made by reducing the signal
to noise ratio at which tracking is performed from that used in MESAR (26dB-29dB).
However, this may only be safely undertaken if the robustness of the tracker is not
decreased (i.e. the chance of losing a track), and the smoothed track accuracy
requirements are met.
These two parameters are shown in Figure 6-35(b) and Figure 6-36(b). Although the
smoothed tracking accuracy is met at low signal to noise ratios (the most stressing case
1140th of a beamwidth is met above a signal to noise ratio of around 15dB), the track
of
robustness decreases at low signal to noise ratios. This reflects the difficulty in tracking
highly manoeuvrable targets with a single pencil beam.
Figure 6-35 also shows that there are savings to be made from using a low desired track
accuracy wherever possible, though the savings made are less pronounced. In an air
161
defence application this parameter is likely to be set by the requirements of the weapons
system for high priority targets, and thus will not be a variable parameter. However for
low priority targets, that must be tracked, but are not of genuine interest to the system, it
shows the savings that may be made.
These results show that not only is the inclusion of multiple beam track updating
important from the aspect of maintaining track on highly manoeuvrable targets, but it is
essential to enable target tracking to be efficient in its use of the multi-function radar
time resources. If multiple beam track updates were utilised to increase the track
robustness, it would be safe to use a considerably lower signal to noise ratio than that
used currently in the MESAR system. A conservative reduction is suggested, down to a
figure of 20dB. This is considerably higher than the minimum signal to noise ratio
required to meet even the most stringent tracking accuracy requirement, but is low
enough that the majority of the saving in radar time-budget is met.
Figure 6-37 shows the potential savings in radar time that could be made for this track
trajectory as a function of the desired smoothed track accuracy. It shows the percentage
of the radar time required if tracking were performed at 20dB instead of the upper value
of the MESAR tracker of 29dB. If only a poor desired tracking accuracy is required, (the
case for the majority of low priority targets), then less than 20% of the radar time is
needed. This would effectively allow the multi-function radar to track as many as 5
times the number of such targets as it currently does. The saving that is made decreases
with increasing tracking accuracy requirement, to a value of around 40% of that when
tracking with a 29dB signal to noise ratio, for the most accurate tracking accuracy
requirement of i/4O of a beamwidth.
162
6 Adaptive Tracking in Mult(function Radar
I
I
D.w.dT, M.5.5 q.NoI ILl () : i 5_ mNo...R,
Figure 6-35(a,b) Total radar time required to maintain track on Benchmark target trajectory 6,
and (b) the percentage of successful tracks, with a beamwidth of 2.5°
'Is I
II
I
.5
IL
I.:
- Inl.u. ,_ I._,
trajectory 6, (b) Achieved smoothed track accuracy for Benchmark target trajectory 6,
excluding periods of manoeuvre, with a beamwidth of 2.5°
Figure 6-37 Percentage of radar time required for tracking Benchmark target trajectory 6 using
only a 20dB signal to twise ratio instead of a 29dB signal to noise ratio, with a beamwidth of
2.5°
163
Figure 6-38 - Figure 6-40 show the results from the same situation as above with the
exception that a radar beamwidth of 0.5° instead of 2.5° has been used. The results show
the increased difficulty in maintaining track on a manoeuvring target with a narrower
beamwidth. Only at the very highest signal to noise ratios considered here is robust
target tracking achieved, i.e. successful tracking throughout the target trajectory life of
close to 100% of the targets. This effectively limits this single beam tracking strategy to
these high signal to noise ratio values, and thus high uses of the multi-function radar
time resources.
The variation of radar time with signal to noise ratio and desired track accuracy is
similar to that of the 2.5° case, the decrease in beamwidth increasing the radar time
required for maintaining track. Thus these results show even more vividly the need for
multiple beam updating for the purposes of robust and efficient tracking. The potential
savings from tracking at a signal to noise ratio of 20dB instead of 29dB, shown in
Figure 6-40, are seen to be slightly greater than for the 2.5° system. A value of around
30% of that when tracking with a 29dB signal to noise ratio, for the most accurate
tracking accuracy requirement of l/4dh of a beamwidth is seen.
164
'1
I
a-
I:
C..1 Y• Sfl A &h'. - •1._ aW
D.. .d '. a a,
.1-i
::::
D.'.d TrnaA -
59I T NIa
excluding periods of manoeuvre, with a beamwidth of 0.5°
I
I
1
I
I
only a 20dB signal to noise ratio instead of a 29dB signal to noise ratio, with a beamwidth of
0.5°
Figure 6-4 1, below, compares the radar time required to maintain track on benchmark
165
6 Adaptive Tracking in MuItfunction Radar
target trajectory 6 for the 2.5° and 0.5° single beam track update cases. The increase in
radar time that is needed to maintain track on the manoeuvring target varies with both
signal to noise ratio and desired accuracy, and at worst it is seen that the radar time
required to track with the narrower beamwidth doubles.
D....d T.1 I. .,... 5flNo..R()
Figure 6-41 Ratio of radar time required to track Benchmark trajectory 6 with a 0.50
beamwidth to that with a 2.5 °beamwidth, for a system employing single beam track updates
Figure 6-42 then takes the case of a target tracking signal to noise ratio of 20dB and a
desired tracking accuracy of 1140th of a beainwidth. These values are chosen because
they represent a case of both efficient and accurate tracking of a target. They show the
variation in the radar time required to track benchmark target 6 and the percentage of
successful tracks as a function of radar beainwidth. Although the increase in radar time
required to track the target at a lower beamwidth does not increase dramatically over the
ranges considered, (comparing a 2.5° beamwidth to a 0.5° beamwidth system gives
roughly a 40% increase in radar time), the track robustness does decrease. This would
prohibit considering narrower beamwidths for weapons system applications where track
robustness is required to be close to 100%.
I
I
-
Figure 6-42(a,b) Total radar time required to track Benchmark target trajectory 6, and (b)
Percentage of successfisl tracks, as afunction of radar beamwidrh
166
6.5.3 Multiple beam static array multi-function radar tracker

The results from the previous section show that to enable efficient as well as robust
target tracking of a manoeuvring target, multiple beam updates must be employed.
Results from the same example cases as in the previous section of a 2.5° and 0.5°
beamwidth system are presented below.
Figure 6-43 - Figure 6-48 show the results for the 2.5° and 0.5° multiple beam update
systems (adhering to the multiple beam update strategy developed in section 6.4.7). The
format of the Monte Carlo results presented is the same as in the previous section,
although in this case the average number beams that were required over the track life is
also presented (Figure 6-44(b) and Figure 6-47(b)).
Again we see a similar variation in the total radar time required to maintain track on
benchmark target trajectory 6 as in the single beam cases of section 6.5.2. However we
see a considerable increase in the track robustness, illustrated by the increase in the
percentage of targets tracked (particularly in the 0.5° case). The use of multiple beam
updates allows a lower signal to noise ratio to be used for tracking whilst still retaining
the desired track robustness and tracking accuracy requirement.
The average number of beams required for each track update is seen to be very close to
one over large ranges of signal to noise ratios and desired track accuracies. This shows
that it was only very occasionally necessary to employ multiple beam updates in most
cases. The exception to this occurs when the desired track accuracy approaches the half
beamwidth limit, and the signal to noise ratio used for tracking is low. In the worst case
over three beams were required on average per update, causing a threefold increase in
radar time.
167
6 Adaptive Tracking in Mult(/iinction Radar
.l
j
'In -
D..'..J T'.
Sq,iiNoe. tI R.p,
and (b) the percentage of successful tracks, with a beamwidth 012.50
.
,
... -1 A I.
F4.IA,t,
A &_ I.N_.
2.5°
I
j: UI
I, I.
No. R(IP)
Figure 6-45(a,b) Achieved smoothed track accuracy for Benchmark target trajectory 6,
excluding periods of manoeuvre, and (b) Percentage of radar time required for tracking
Benchmark target trajectory 6 using only a 20dB signal to noise ratio instead of a 29dB signal
to noise ratio, with a beamwidth of 2.5°
168
6 Adaptive Tracking in Multfiinction Radar
!.1
ut. . -
D...d Ac-, D..o.d 'e. Aotoo, •2 t
to N'. 0.o SU.011N,., P..t. t)
0I.l
102
DNhsd Tao0 ko., 02 10 D....d T *2

SIo Not..() OIUnUON*.. tt)
0.5°
1m F
I::
1
1
I.
TIIy 09t*N. D....d T.o0 *00Y
to noise ratio, with a beamwidth of 0.5°
169
Figure 6-49, below, shows the ratio of radar time required for multiple beam update
tracking at 0.5° to that at 2.5°. It shows an extremely similar result to that in Figure 6-
41, showing roughly a doubling of radar time is required to track the same target with
the narrower beamwidth. Figure 6-50 then takes the case of a 20dB signal to noise ratio
and 1140th of a beamwidth desired tracking accuracy, and shows the variation in radar
time and successful tracks as a function of radar beaznwidth.
Figure 6-51 compares the radar time required for tracking benchmark target trajectory 6
using the multiple beam system to that required for the single beam system, for the same
values of signal to noise ratio and desired track accuracy as above. A very small
difference is seen between using multiple beam track updates and single beam track
updates. In the case of the smallest beamwidth considered, a difference of less than 10%
was observed.
These results show that accurate tracking of manoeuvring targets is most efficiently
performed at a lower signal to noise ratio than that used in the MESAR system
currently. The use of multiple beam updates in the tracker provides track robustness at a
small cost in terms of radar time. For a 2.5° beamwidth, the multiple beam tracker
operating at a signal to noise ratio of 20dB and a desired accuracy of 1140th of a
beamwidth requires just 38.9% of the radar time of the single beam tracker operating at
a 29dB signal to noise ratio, representative of the MESAR system. It suffers no
degradation in track robustness. For the 0.5° system the saving in radar time increases,
the multiple beam system requiring just 29.5% of the radar time of the higher signal to
noise ratio, single beam system.
":•
Figure 6-49 Ratio of radar time required to track Benchmark trajectory 6 with a 0.5
beamwidth to that with a 2.5 °beamwidth, for a system employing multiple beam track updates
170
I
I
U (
Figure 6-50(a,b) Total radar time required to track Benchmark target trajectory 6, and (b)
Percentage of successful tracks, as a function of radar beamwidth, using multiple beam track
updates
Figure 6-51 A comparison of the total radar time required to track Benchmark trajectory 6 with
the static and multiple beam track update systems
6.5.4 Single beam rotating array multi-function radar tracker

The results for the two rotating array antenna configurations considered (single array
system rotating with a period of is, and a twin faced array system rotating with a period
of 2s), are presented in the current and the following section. The results for the two
other antenna configurations discussed in section 6.5.1 (single array system rotating with
a period of 2s, and a twin faced array system rotating with a period of is), are contained
in Appendix C.
Figure 6-52 - Figure 6-57 show the results for the antenna with a single array, rotating
with a period of is, using single beam track updates. The radar time surfaces, shown in
Figure 6-52 and Figure 6-55, show similar trends to the static array case, although the
surface is not exactly the same shape, reflecting the difference in the track update
intervals used. Importantly they still suggest that using a lower signal to noise ratio
would be beneficial in terms of radar time.
171
6 Adaptive Tracking in Mu/tifunction Radar
The desired accuracy, shown in Figure 6-53(b) and Figure 6-56(b), in the most strenuous
1140th of a beamwidth, is not achieved at a signal to noise ratios as low as that in
case of
the static case. This is due to the limit in update rate imposed by the blind sectors in
antenna coverage. The desired accuracy is met in all cases above a value of around
19dB. Therefore the value of around 20dB for tracking of targets suggested from the
static array results would remain appropriate in this case, since the desired accuracy
could still be met. Comparing the radar time that could be saved by reducing the signal
to noise ratio from 29dB to 20dB, shown in Figure 6-54 and Figure 6-57, we see that the
rotating system gives similar reductions to the static array system, of between around
15% to 30% (depending on the desired accuracy).
Although the results again confirm the need for multiple beam track updates, the track
robustness, is in fact slightly increased above that in the static array case. This is due to
the necessity to make track updates at large scan angles away from azimuth array
broadside to support the desired update rate. This results in a broadening of the azimuth
beamwidth during periods of manoeuvre, reducing the chance of failing to illuminate the
target at the next update. At 60° from azimuth array broadside, the azimuth beamwidth
has doubled in size.
It is important to note that the necessity to reduce the update interval, and hence increase
the number of track updates that are made off of array azimuth broadside during periods
of manoeuvre, has a significant effect on the total radar time required to track the target.
The updates made away from array azimuth broadside not only have a broadened
beamwidth, but also require a large increase in dwell time to recover the loss in gain.
Thus, the total radar time required to track the target in the case of the rotating array
tracker system is highly dependent on whether the target manoeuvres, and period of this
manoeuvre.
Figure 6-58 - Figure 6-63 show the results for the twin faced system, rotating with a
period of 2s. The results show only modest differences to the single array case. The
track robustness is increased at very low signal to noise ratios, and the signal to noise
1140th of a beamwidth is met
ratio at which the most strenuous desired track accuracy of
is decreased to 16dB. This reflects the reduction in the value of TbIId between the two
systems. Broadly, however, the performance is essentially the same as the static array
system rotating twice as fast.
172
6 Adaptive Tracking in Multfiuzction Radar
rn,
121
10,
I
0.
1::
O...d " ('s)

D..,.dT,. A0rn.y

SNRI)
and (b) the percentage of successful tracks. Single faced rotating system with a beamwidth of
2.5°
101
II
1
0
J
In
010
— 0?
EN.S.Rrn0() I0N0I.. RrnntJ
excluding periods of manoeuvre. Single faced rotating system with a beamwid:h of 2.5°
0.
I:
I
:.,c..A ....
only a 20dB signal to noise ratio instead of a 29dB signal to noise ratio. Single faced rotating
system with a beamwidth of 2.5°
173
6 Adaptive Tracking in Mulr(function Radar
Do ..d7,.A,n,..,0 ' I,, Sq.iI. N M. (,

D....d To .:
I Aol, i
0.50
oh
1041 1
•1
I
0
1•0)
.
DsdT Nd I
excluding periods of manoeuvre. Single faced rotating system with a beamwidth of 0.5°
I...,
only a 20dB signal to noise ratio instead of a 29dB signal to noise ratio. Single faced rotating
174
12 I.,,,
II
I
I
j
0
S.
..,. - .
Figure 6-58(a,b) Total radar time required to maintain track on Benchmark target trajectoly 6,
and (b) the percentage of successful tracks. Twin faced rotating system with a beamwidth of
2.50
I, •
1 'i
Ii
I! j :::
J ?o No.,. Ro() SNoSR.()
excluding periods of manoeuvre. Twin faced rotating system with a beamwidth of 2.5°
I
I
I
1
I
I
Mi
only a 20dB signal to noise ratio instead of a 29dB signal to noise ratio. Twin faced rotating
175
6 Adaptive Tracking in Multçfunction Radar
I,
1'
1:
#1
D....d T.0 4n.rSnn '2 I.. . 4I' S9'I'T NW
Figure 6-61(a,b) Total radar tune required to maintain track on Benchmark target trajectory 6,
and (b) the percentage of successful tracks. Twin faced rotating system with a beamwidth of 0.5°
01
j06
I:
J0! j"dl.
I0
In 0,dlI,N'...R.bn.' :,s!k0......... N.n. hdl,. ..
excluding periods of manoeuvre. Twin faced rotating system with a beamwidth of 0.5°
Fii
I
I
only a 20dB signal to noise ratio instead of a 29dB signal to noise ratio. Twin faced rotating
176
6.5.5 Multiple beam rotating array multi-function radar tracker

Finally, Figure 6-64 -Figure 6-73, show the results for the two rotating systems using
multiple beam track updates.
The results show the track robustness to be greatly improved, as would be expected.
They show that very satisfactory tracking performance in terms of achieved accuracy,
track robustness and efficiency can be achieved with the rotating systems as was
achieved with the static array system.
The value of 20dB chosen in the static case for efficient, accurate and robust tracking at
1140th of a beamwidth remains appropriate in this instance.
The final figure, Figure 6-76, compares the radar time required for the different single
beam and multiple beam rotating systems. As in the static case, the difference in radar
time required for the multiple beam updating of tracks was less than 10% in all cases. In
fact, in all but one case (the single faced array with a 0.5° beamwidth), the difference
was less than 5%.
177
6 Adaptive Tracking in Mul:function Radar
12
7i
I0
e
J'.
J1..
D....dT..,
() D.P..d T,.. ,,
2.50
:... ....

123
Io " . lb
I.
2 I.. '.2 I.
Sq..hIcNew.Pa, S..IIo Ne.. llo I)
Figure 6-65(a,b) Average track update interval required to maintain track on Bench,nark target
I 1u01
4-
/
il:
I
1
DUIT I 0
to noise ratio. Single faced rotating system with a beamwidth of 2.5°
178
Is
10
1,
0
a.
:.. ' •.: ,., ,,,, -: i

!&., R.br ., I.. Ro0,dB,
0.50
,,,
0I!J
15

I0 01.d T*a. 2 Ia
dNd ()
al.
I
'a.
h1.
1
a20
1
I
50
00d T.a 0:
E N Pa.
to noise ratio. Single faced rotating system with a beamwidzh of 0.5°
179
6 Adaptive Tracking in Mulr(function Radar
12
I0
0
I., -.
•I.• -
loo.ov •2 in
!gmNo.R (I R()
and (b) the percentage of successful tracks. Twin faced system with a bewnwidth of 2.5°
j in
.,.._i T .... A ..? I:

&A t flj r.a .*
trajectory 6, (b) Average number of beams required for each track update. Twin faced system
with a beamwidth of 2.5°
I-
I
1
I
I
02
to noise ratio. Twin faced system with a beamwidth of 2.5°
180
o<.
'0
•1' -
D....d t -: I..
0q.0IN.
Figure 6-73 Total radar time required to maintain track on Benchmark target trajectory 6, and
(b) the percentage of successful tracks. Twin faced system with a beamwidth of 0.5
1'. I.,
SJt,N.'0IRo(
r,....i T. 6q.aI.. N. (I
trajectory 6, (b) Average number of beams required for each track update. Twin faced system
I:: F
F
j:
I
F
I'
to noise ratio. Twin faced system with a beamwidth of 0.5°
181
Figure 6-76(a,b) Comparison of the radar time required to track Benchmark target trajectory 6.
(a) Single faced array system, (b) Twin faced array system
6.5.6 Comparison of static and rotating array tracker performance

The results presented in sections 6.5.2 to 6.5.5 have shown that very satisfactory
performance can be obtained from both the static and rotating systems operating at a low
signal to noise ratio, coupled with the efficient use of multiple beam track updates. Both
systems can achieve a track robustness close to 100%, even in the case of the narrowest
beamwidth considered of 0.50. Finally, a comparison of the radar time the different
rotating and static systems required to perform tracking of benchmark target trajectory 6
is presented.
It is important to note however in this comparison, that the radar time required by both
the static and rotating systems is dependent on the trajectory characteristics. In the static
case, a target trajectory which is close to the edges of the array coverage when compared
to a similar trajectory close to array azimuth broadside will require a considerably
greater amount of radar time. Conversely in the rotating case, although the general
azimuthal direction of the trajectory does not affect the total radar time requirements, a
target which is manoeuvring for much of its trajectory will require a greater amount of
radar time than one which follows mostly a constant velocity trajectory.
An important difference however, is that the rotating array tracker can overcome this
problem to some extent, where the static array tracker cannot. By inclusion of another
smoothing filter model of the target manoeuvre, track updates may continue to be made
close to array azimuth broadside as in the case of level flight. It is likely that updates
away from array azimuth broadside would be required only at a point close to switching
between the two models. This would allow the rotating array tracker to benefit from the
minimum azimuth beamwidth for much of the track life, and give the potential for
182
superior accuracy to the static array system.
Figure 6-77 and Figure 6-78 compare the radar time required to track benchmark target
trajectory 6 for the static and rotating array systems using a 20dB signal to noise ratio
and a desired accuracy of 1140th of a beamwidth. The single array rotating system
requires between 2.1 and 2.5 times radar time of the static array system to track
benchmark trajectory 6, depending upon the beamwidth being used. This figure
increases slightly for the twin faced rotating system.
Figure 6-77(a,b) Comparison of the radar time required to track Benchmark target trajectory 6
with a single faced rotating array multiple beam system and a single faced static array multiple
beam system (a) Absolute radar time required by each system, (b) ratio of the radar time used
by rotating system to the static system
• 8.Asr. SMcS stem
:JJT'
0! 1 15
Figure 6-78(a,b) Comparison of the radar time required to track Benchmark target trajectory 6
with a twin faced rotating array multiple beam system and a single faced static array multiple
beam system (a) Absolute radar time required by each system, (b) ratio of the radar time used
by to rotating system to the static system
This situation may be contrasted to the case of the constant velocity target, approaching
at an azimuth angle of 0° (array azimuth broadside for the static system). This can be
seen from the constant velocity results shown in Appendices B and C. In this the static
system requires roughly the same amount of radar time as the single faced rotating
system. If this trajectory was rotated such that it lay 60° away from azimuth array
broadside of the static array system, then the static system would require a greater
183
amount of radar time than the rotating array system.
6.6 Conclusions
Most importantly these results have shown the viability of a narrow beam multi-function
radar tracker for the tracking of manoeuvring targets, for both static and rotating array
systems. They have indicated the most efficient values of signal to noise ratio to aim for
in the updating of tracks, and have shown that the robustness of track can be maintained
through the use of multiple beam track updates. The extra radar time for such updates is
not prohibitive, requiring less than a 10% increase in radar time in most cases.
The study has quantified the variation in both radar time and track robustness when the
radar beamwidth is varied. At narrow beamwidths the use of multiple beam track
updates is imperative to allow robust tracking. An efficient strategy for the use of
multiple beam track updates has been proposed, and the increase in track robustness
made using this update strategy has been quantified.
The results have shown that accurate tracking of manoeuvring targets is most efficiently
performed at a lower signal to noise ratio than that used in the MESAR system
currently. The signal to noise ratio may be reduced considerably (by at least 6dB),
without decreasing the track robustness below a reasonable level, or the tracking
1140th
accuracy below that desired. In the case of a 2.5° beamwidth, and a of a
beamwidth desired tracking accuracy, the results indicate that the savings in radar time
from utilising a multiple beam tracker operating at 20dB, rather than 29dB as in
MESAR, would allow more than double the number of targets be tracked. In the
case of a 0.5° radar beamwidth, the savings in radar time would allow around three
times the number of targets to be tracked using this system.
A rotating array adaptive tracker has been constructed based upon the same track
updating strategy used in the MESAR static array tracker. The rotating array tracker
uses roughly the same amount of radar time as the static array tracker for constant
velocity targets. However, it requires considerably more radar time than the static array
system during periods when the target executes a strong manoeuvre. This is due to the
necessity for track updates to be made at large scan angles from array azimuth
broadside. In the case of the most stressful target trajectory considered, benchmark
184
target trajectory 6, the rotating array system required roughly 2½ times the radar time of
the static array system.
However, the increase in radar time is heavily dependent on the exact target trajectory;
on the period of the manoeuvres, and on the azimuth angle of the target. This is due to
the amount of track updates which must be made at large scan angles from array
azimuth broadside, necessitating long dwell times to offset the loss in antenna gain. For
example, in the case of a manoeuvring target trajectory which was mainly close to array
azimuth broadside of the static array, the radar time required for tracking in the rotating
array case would almost certainly be more than the static case, due to the relative
number of track updates that must be made at large scan angles between the two
systems. Conversely, if the same target trajectory was now rotated such that it was close
to 600 away from azimuth broadside of the static array, the situation is likely to be
reversed; the rotating array tracker would require less radar time than the static array
tracker.
It is thought that the implementation of a multiple model tracker which more accurately
represents the target dynamics during manoeuvre would reduce the necessity to make
track updates at great angles from array azimuth broadside. This would reduce the radar
time required for tracking with a rotating array multi-function radar. This is proposed as
an area of further work.
185
7 A ROTATING, NARROW BEAMWIDTH MULTI-
FUNCTION RADAR SYSTEM
7.1 Introduction
This chapter takes the main findings from chapters 4,5 and 6, and applies them to
demonstrating the viability and performance of a narrow beamwidth rotating array
multi-function radar with the capability of simultaneous multiple beam processing. The
methods for analysis of the search function time developed in chapter 4 are used to
compare this system to an equivalent static array system which does not have the ability
to process simultaneous multiple beams.
Chapter 4 has shown that multi-function radars utilising rotating arrays are more
efficient at performing the search function than those utilising static array faces. It also
showed that the use of simultaneous multiple beam processing allowed the possibility of
a narrow beamwidth multi-function radar system which did not have excessive search
function times. This is important since it shows the feasibility of constructing a narrow
beamwidth system which can undertake the surveillance and tracking functions
successfully, within the stringent time requirements placed on the multi-function radar.
Chapter 5 then demonstrated the fact that a rotating array multi-function radar system
can distribute the radar time resources across the angular coverage in a way that static
array systems cannot, and can therefore better support a high loading of radar tasks in a
confined angular sector. This gives the rotating array system an advantage in the
operationally likely scenario that targets of interest are clustered around a certain
direction, rather than equally distributed throughout the radar coverage volume.
Finally chapter 6 went on to show that manoeuvring targets can be successfully tracked
with a narrow beamwidth multi-function radar utilising either a static or rotating array
antenna. Track robustness can be attained through the use of multiple beam track
updates at little cost in terms of radar time.
186
7 A Rotating, Narrow Beamwidth Multi-Function Radar System
Also this region requires the highest search rate, since targets may pop up over the
horizon at short range. Region 2 extends from the top of region 1 to an elevation of 100,
and requires MTI Doppler processing for cancellation of rain clutter. Region 3 requires
no Doppler processing. Table 7-I shows the surveillance waveforms and frame times of
the example requirement.
Region Elevation Frame Waveform Type Azimuth Broadside

______ Extent Time ___________________________ Waveform Duration
1 Bottom beam 2 s Pulsed Doppler - 2 bursts of pulses 2(3+8)/2KHz = I lms
________ ________________ _________ (8 pulses_plus_3_guard_pulses) _________________________
2 Second beam to 4s MTI - 3 bursts of pulses 3(2+3)12KHz = 7.5ms
_______ 100 ________ (3 pulses plus 2 guard pulses) _______________________
3 10° - 90° 4 s Non Doppler Processed 1/2KHz = 0.5ms
_______ ______________ ________ (I_pulse) ______________________
Table 7-I Surveillance frame time and waveform requirements for the example search volume
These waveforms are increased in length to account for the decrease in gain with
azimuth scan off that occurs in the static systems, assuming coherent integration is used
to offset the cos3 (0) scan off loss. It is assumed that the rotating system performs all
surveillance on azimuth array broadside, and thus no increase in the waveform length is
required. This will be the case for a multi-function radar with a total occupancy of 100%
or less as described in section 5.4.
A peak total of 30 high priority targets must be tracked using dedicated waveforms. It is
assumed for simplicity that all low priority targets will be tracked using track while
scan, and thus have no impact on the radar resources. The 30 high priority targets will be
spread evenly between regions 1 and 2.
7.3 Antenna configurations considered
Chapter 5 showed that a rotating array multi-function radar performed the search
function more efficiently than the equivalent static array systems. Also, Billam [Billam,
1997] has shown that a single faced array antenna is the most efficient arrangement of a
fixed number of elements in an array antenna. Thus a single face, rotating array system
has been chosen here, and compared to both three and four faced static array systems.
Each multi-function radar system will contain a total of 10,000 elements to be

distributed equally between the array faces. This has been arbitrarily chosen, to allow a
relative comparison of the systems. Obviously a greater number of total elements would
188

give greater performance and conversely a smaller number would give worse
peformance, but it is not absolute performance against the requirement that is of interest.
The static systems have the advantage of having independent steerable beams from each
array face, giving 3 or 4 simultaneous beams compared to just one beam for the single
array rotating system. Conversely the ERP of the static systems is considerably less than
the rotating system due to the distribution of the elements amongst the faces. The
antenna characteristics of the systems are given in Table 7-2. A frequency of 3GHz is
assumed, and a power feeding each element of lOW.
System Independent Elements Beamwidtli Gain Relative

_______________________ Beams per face (deg) (dB) ER?
Four Faced Static Array 4 2,500 2 39.0 1
Three Faced Static Array 3 3,333 1.76 40.2 1.78
Single Faced Rotating Array 1 10,000 1 45.0 16
Table 7-2 Antenna characteristics of the different array configurations
7.4 Analysis of time budget and free space detection performance
7.4.1 Surveillance
This section presents the results of applying the UV space and rotating UV space
methods for estimation of the surveillance time-budget requirements. It also gives a
comparison of the likely detection performance of the different systems using the radar
equation.
We start by considering the radar time-budget and performance aspects for the
surveillance function. Table 7-3 below shows the results from applying the UV space
estimation technique developed in chapter 4 to the four faced static array multi-function
radar system.
gion Number of Beams Function Time Frame Time Occupar

I 192 15.7ms 3.0229 2 152%

2 771 1 0.6ms 8.1697 4 204%

3 4356 0.616ms 2.6854 4 67%
Total Occu pancy 423%
Table 7-3 Search time-budget requirements of the 4 faced static array multi-function radar
The 2° beamwidth of the static arrays gives a total of 53 19 beams to search, requiring a
total radar time of 13.88s. Given the frame time requirements of the three regions, this
translates into a total required surveillance occupancy of 423%. Thus an occupancy of
189
106% is required from each of the multi-function radar array faces to meet the
surveillance requirement.
Clearly this is not possible, and in practice the requirement would be alleviated in some
way to bring the surveillance occupancy below 100%, allowing enough time for other
functions such as tracking. This could be achieved for example by increasing the frame
times of the regions somewhat. However this figure is close enough to 100% to allow
the requirement to be used for a comparison of the different antenna configurations.
Using the radar equation expressed in [7-1] below we can calculate the broadside signal
to noise ratio, SNR0, against a target in free space. If we assume a duty cycle of 30% for
the radar then for a target at a range of 50km, with a radar cross section of 0.1 m 2, a
single pulse signal to noise ratio of 1 1.5dB is achieved. Assuming Swerling 3 target
characteristics this equates to a single pulse probability of detection of 0.4.

P GG,t2ci [7-1]
SNR0 - (4ir)3R4kTB
where Pa. is the average transmitted power

G, is the gain of the antenna on transmit
Gr is the gain of the antenna on receive
A is the wavelength of the transmitted signal
o is the radar cross section of the target
R is the target slant range
k is the Boltzmann constant, 1.38 x 10-23 J/deg
T is the temperature in degrees Kelvin
B is the receiver bandwidth in Hertz
For comparison, the occupancy requirements of an equivalent four faced rotating system
are shown in Table 7-4.
Time

1 213 lims 2.35 2 117%

2 853 7.5ms 6.40 4 160%

3 4721 O.Sms 2.36 4 59%
ccunancv 336%
Table 7-4 Search time-budget requirements of the 4 faced rotating array multi-function radar
As was shown in chapter 4, the rotating system requires significantly less radar time to
complete the search since it can perform surveillance on azimuth array broadside. For
this particular requirement, the rotating system requires just 79% of the radar time of the
equivalent static system, whilst achieving the same detection performance.
190
This system is therefore more efficient than the static 10,000 element system, and can
meet the requirement within the time-budget constraints, requiring a surveillance
occupancy of just 84% per face. However it has the same 2° beamwidth as the static
system, which we wish to reduce.
Table 7-5 shows the situation if all 10,000 elements are configured into a single, rotating
array. The decrease in beamwidth gives rise to a vast total number of surveillance beams
of 23149. This total surveillance occupancy required for the single array face to perfonn
the volume search rises to 1190%, which is clearly totally impossible. Table 7-5 shows
in a striking way the reason the beamwidth of a multi-function radar is compromised
between the surveillance and tracking functions.
The large increase in gain, and the increased total power output, however give a
significantly increased detection performance. The single pulse signal to noise ratio
against the 0. 1m2 target at 50km rises to 29.6dB, some 18dB greater than that of the four
faced array.
1 425 urns 4.68 2 234%

2 3840 7.5ms 28.80 4 720%
3 18884 0.Sms 9.44 4 236%
Table 7-5 Search time-budget requirements of the single faced rotating array multi-function
radar, with no capability for simultaneous multiple beam processing
Table 7-5 shows the radar time constraints that must be overcome through the use of
simultaneous multiple beam processing. To enable the timely surveillance of regions 2
and 3, we consider a quadrupling of the transmit beamwidth, coupled with narrow
receive beams, shown in Figure 7-2(a), below.
(a) (b)
Figure 7-2(a,b) Simultaneous multiple beam configurations for (a) Regions 2 and 3, (b) Region
1. Dashed line shows transmit beam pattern, filled circles show possible receive beam positions
This allows the search of these regions to be undertaken with a 16th of the number of
beams. It will give a comparable search occupancy to that of the four faced systems,
191
since those systems have a ¼ of the total occupancy per face, and 1/4 of number of search
16th
beams, thus a of the occupancy of the single array.
For the same reasons, a broadening of a factor of 8 is applied to the search of the single
beam in elevation of region 1, shown in Figure 7-2(b). This again will give a
comparable surveillance occupancy to the four faced system, but will also give the
narrowest beamwidth in elevation which is important to minimise the effects of
multipath in the bottom beam.
Table 7-6 shows the time-budget requirements for the 10,000 element array applying
this scheme of simultaneous multiple beam processing. The use of the broadened
transmit beam patterns enables a reduction in the total required surveillance occupancy
to 89%.
The free space detection performance of the surveillance regions is now different since
the transmit beam used for region 1 has a higher gain than that used for regions 2 and 3.
A signal to noise ratio of 20.6dB is achieved against the 0.1m2 target at 50km in region
1, and 17.6dB in regions 2 and 3.
Thus the single array system utilising multiple beam processing can successfully
conduct the surveillance of the volume within the multi-function radar time resource
constraints, and achieve a satisfactory free space detection performance. This system
requires just 84% of the radar time that the four faced static system requires, whilst
achieving a signal to noise ratio 9dB higher in region 1 and 6 dB higher in regions 2 &
3.
Number of rell Time Function Time Frame Time Occupa:

53 urns 0.5855 2 29%

2 240 7.5ms 1.8008 4 45%

3 1181 0.5ms 0.5903 4 15%
Total Occuuancv 89%
Table 7-6 Search time-budget requirements of the single faced rotating array multi-function
radar, with the capability for simultaneous multiple beam processing
7.4.2 Tracking
In this section we compare the tracking performance of the single array rotating system
utilising simultaneous multiple beam processing to the four faced static system which
does not have the capability to perform simultaneous multiple beam processing.
192
The comparison is made difficult by the variation in the radar time requirements of both
the static and rotating array tracker systems with the trajectory characteristics. As
discussed in section 6.5.6, a target trajectory which is close to the edges of the array
coverage when compared to a similar trajectory close to array azimuth broadside will
require a considerably greater amount of radar time in the static array system.
Conversely in the rotating case, although the general azimuthal direction of the
trajectory does not affect the total radar time requirements, a target which is
manoeuvring for much of its trajectory will require a greater amount of radar time than
one which follows mostly a constant velocity trajectory.
This is reflected in the fact that the constant velocity target required less radar time for
tracking in the rotating case than the static case, whereas for benchmark trajectory 6 the
rotating array system required more than double that of the static array system (see
results in Appendices B & C). Taking an average of the radar time required for tracking
all of the 7 targets considered, the static system required 60% of the radar time used by
the single array system rotating with a period of ls. This figure will be used as a
comparison here, but the dependency on the likely target trajectories must be noted. It is
also important to note that the use of multiple Kalman filtering techniques such as 1MM
may reduce the radar time requirements of the rotating array tracker against
manoeuvring targets.
Tracking benefits from knowledge of the target direction, and so broadened beams need
only occasionally be used. The rotating array system can therefore obtain maximum gain
in both transmit and receive for most track updates. This will give a free space detection
performance some 18dB higher than the static system if no beam broadening is used.
Since the minimum length of the transmitted waveform that must be used for tracking
will be defined by the clutter conditions, the difference in the radar time that will be
required by the two systems for tracking will be dependent on the clutter environment in
which the system operates. The case of benign clutter and severe clutter are discussed
below.
7.4.2.1 Benign clutter environment

In the absence of severe clutter the waveforms may be reduced in length, since Doppler
discrimination of the target and the clutter is not required. If the tracking waveforms are
to aim to achieve a signal to noise ratio of around 20dB against the target, as
193
7 A Rotating, Narrow Bea,nwidth Multi-Function Radar System
recommended in chapter 6 for efficient tracking, then the difference in radar time the
two systems will require for tracking will depend on the following factors;
i. increase in track update rates for narrow beam tracking

ii. increase in radar time to track manoeuvring targets with a rotating array
tracker
iii. decrease in track waveform length due to higher free space detection
performance
Chapter 6 quantified the first of these factors. The increase in radar time necessary to
track using a narrower beamwidth is shown in Figure 6-77. An increase of around 15%
in radar time is required to track using a 10 beamwidth rather than a 2° beamwidth.
The second factor, has been discussed in section 7.4.2 above, and is dependent on the
specific target trajectory assumed. The single array system rotating with a period of is
required a factor of 1.67 times the radar time of the static array system, based upon an
average of the 7 target trajectories considered in this study.
These two factors suggest a radar time requirement for the rotating array system of
almost double that of the static system. This combined with the fact that the static
system has four times the total occupancy available for tracking from its four faces,
suggests the static system should be able to track almost 8 times as many targets.
However, the waveform length requirements of the rotating array system will be
considerably less than the static system, since the free space detection performance of
the rotating system is 18dB greater. This potentially allows a reduction in the
transmitted waveform energy by a factor of 64 to allow tracking to occur at the same
signal to noise ratio as in the static case. In practise the savings made in radar time by
reducing the transmitted waveform length will be limited by the need to transmit and
receive at least one pulse, and will therefore depend upon the target range (assuming
range unambiguous operation), but they should certainly outweigh the increase in radar
time required for factors discussed above.
It is concluded that in a benign clutter environment, the rotating array system is likely to
be able to track at least as many, if not more, high priority targets than the static system,
and with a better resolution and accuracy.
194

7A Rotating, Narrow Beamwidth Multi-Function Radar System
7.4.2.2 Severe clutter environment

In a severe clutter environment the length of the waveforms used for tracking is likely to
be dominated by the requirement for Doppler resolution. In this situation the waveform
length will not be able to be reduced in the rotating array tracker, resulting in tracking
taking place at a signal to noise ratio higher than that necessary, and a greater radar time
requirement for the rotating system. This is the more stressful of the two clutter
situations in terms of radar time, and thus we consider the tracking requirement outlined
in section 7.2 for this case.
Waveforms assumed for tracking the 30 targets are based on those for surveillance of
regions I and 2. In the case of tracking however, where the target velocity is known,
only a single Doppler pulse burst need to be transmitted to cover the required velocities
of interest. The tracking dwell time for region 1 is thus half of that for surveillance, and
for region 2 one third of the for surveillance.
The tracking occupancy requirements for the static and rotating array systems are shown
in Table 7-7 and Table 7-8 below. An average update rate of 1Hz has been assumed for
the rotating system, which is decreased by 15% for the static system to account for
tracking with a broader beamwidth. The update rate is reflected in the frame time in
each table.
The rotating array system requires an occupancy of 12% to cope with the peak loading
of tracks given by the requirement. The static array system requires nearly 15%,
however this is shared amongst the four faces, and thus the tracking occupancy required
per face is just 3.7%.

Dwell Time Number of Function time Frame Time Occupar

I 5.Sms 15 82.Sms is 8.25%

2 2.5ms 15 37.5ms Is 3.75%
Total Occuuancv 12%
Table 7-7 Tracking time-budget requirements of the single faced rotating arra)' multi-function
radar
of targets Function time Frame Time Occupan
gion Dwell Tii

1 7.85ms 15 117.75ms 1.15s 10.23%

2 3.53ms IS 52.95ms i.i5s 4.6%
Total Occunancv 14.8%
Table 7-8 Tracking time-budget requirements ofthefourfaced array multi-function radar
In this severe clutter environment, for a given total radar tracking occupancy, the static
system will be able to track at least four times the number of targets as the rotating array
195
system. The signal to noise ratio at which tracking would occur would be 18dB higher in
the rotating array system.
The rotating array system could reduce the occupancy required for tracking through the
use of cluster tracking techniques. Simultaneous multiple beam processing may be used
for the updating of multiple tracks closely spaced in angle. This is beyond the scope of
this study and is proposed as an area of further investigation.
7.4.3 Summary
The rotating array requires a greater occupancy for tracking of high priority targets
against the target trajectories considered within this study. However, against the
example requirement, this was more than offset by the efficiency with which this
antenna configuration is able to perform the search function.
The rotating array system, utilising multiple beam processing, required an occupancy of
89% to meet the search requirement, and 12% to meet the tracking requirement set.
Thus within the limits of the accuracy of the simple analysis performed, the total
occupancy of 101% indicates that the rotating array system was able to meet the
example requirement.
The static array system required an occupancy of 106% per face to meet the search
requirement and 3.7% per face to meet the tracking requirement. Thus in total the four
faced static system required an occupancy around 9% greater than the rotating system in
order to meet the requirement set.
The rotating system achieved a higher signal to noise ratio in all cases due to the
increased antenna gain. In surveillance, a free space signal to noise ratio of 9dB and 6dB
greater was achieved in regions 1 and 2 respectively. In tracking, since the maximum
antenna gain is achieved in both transmit and receive, an 18dB better free space signal to
noise ratio was achieved. However, to achieve timely surveillance, a considerable
broadening of the transmit beam was considered, resulting in illumination of a greater
area of clutter. The impact of this must be considered carefully before such a strategy
may be adopted.
196
7.5 Conclusions
This chapter has compared two different multi-function radar systems against a simple,
but realistic requirement. A fixed number of active array modules have been considered,
since to a large degree the cost of an active array multi-function radar is dependent on
this factor. These have been distributed differently in each of the two systems.
The first system is a multi-function radar utilising four fixed antenna faces. This has
traditionally been regarded as the best system if it can be afforded. The second system
considered takes on the findings of chapters 4, 5 and 6, and utilises a single rotating
array with the capability of performing simultaneous multiple receive beam processing.
This system has a narrower beamwidth than the static array system, and would
ordinarily have difficulty in meeting the time-budget constraints placed on the multi-
function radar.
The analysis shows that the single rotating array multi-function radar is able to meet the
example requirement for surveillance and tracking within the radar time-budget
constraints, through the use of simultaneous multiple beam processing. A total
occupancy of 101% is required, which to within the accuracy of the analysis made, is
achievable. The static system requires a total occupancy of 110% to meet the same
requirement, which is not achievable, and would thus require a small reduction in the
requirement specified.
The rotating array system requires more radar time resources for tracking than the static
array system. However in the case of the example requirement, the extra time needed
was more than compensated for by the efficiency with which the rotating array system
performed the search.
The benefits and drawbacks of the narrow beamwidth rotating array system compared to
the static array system considered are summarised below.
197

Advantages Disadvantages
. Increased free space detection performance • Increased radar time requirement

in surveillance for tracking in some cases
• Increased free space detection performance • Increased illumination of clutter

in tracking • Increased receiver or digital
• Better plot resolution in tracking and beamformer complexity
surveillance
• Better smoothed track accuracy

• Better performance in multipath conditions
• Increased resistance to main lobe jamming
• Greater ability to cope with an uneven

loading of tasks, or tasks confined to a
narrow angular sector
Essentially, the use of simultaneous multiple beam processing may be regarded as

another degree of freedom in the control of the multi-function radar. It allows the time
and energy resources to be more closely fitted to the requirement. A multi-function radar
with a narrow beam searching in clutter may have a free space detection performance in
excess of that required, but clutter cancellation below that required within the time
constraints. The use of multiple beams allows the adjustment of time and energy
resources allocated in a given direction independently, and therefore to trade off the free
space detection performance with the Doppler cancellation.
198
8 CONCLUSIONS AND FURTHER WORK
8.1 Summary
A phased array antenna can electronically steer the direction of the antenna beam near
instantaneously. In a multifunction radar this capability is used to allow the system to
multiplex its time between many different functions, the primary functions usually being
volume surveillance and multiple target tracking. Potentially the use of automated
digital control of the multi-function radar system allows the tasks it is performing to be
dynamically adjusted. Savings in radar time made in meeting the requirement of one
function allow that time to be devoted beneficially to other functions. It is important
therefore to ensure that the multi-function radar's resources are efficiently allocated
between the functions, and that each function is performed with the minimum use of
radar time and energy possible. The aim of the work presented in this thesis has been to
investigate and improve techniques for multi-function radar control and tracking. The
research has been focused particularly on coping with the effects of utilising rotating
arrays and on the use of techniques such as beam broadening and simultaneous multiple
receive beam processing, that will be available in future multi-function radars with
advanced digital beamforming capabilities.
A fundamental disadvantage of the multi-function radar is the compromise that must be

made on the radar beamwidth. To allow surveillance of the volume to occur at a useful
rate whilst supporting the required levels of clutter suppression through Doppler
processing, a broad radar beamwidth must be utilised. However, the requirements for
good plot accuracy and resolution for the tracking function conflict with this, desiring a
narrow radar beamwidth. The result is that many current multi-function radar systems
have a radar beamwidth larger than is ideal for the tracking function to allow timely
volume surveillance, and as a result have poor performance when compared to dedicated
tracking radars. By applying novel methods for estimation of the search time
requirements of different multi-function radar configurations, it has been concluded that
this compromise on radar beamwidth may be avoided through the use of simultaneous
multiple beam processing. This allows the surveillance of the volume to be achieved
199
8 Conclusions and Further Work
inside the surveillance frame time requirements with a narrow radar beamwidth. This in
turn allows the benefits of a narrow radar beamwidth to be accrued by the tracking
function.
A second important conclusion from the research is that the traditional view of four
fixed array faces as the 'ideal' antenna configuration for a multi-function radar is not
supported. It has been shown that multi-function radars which utilise rotating array
antennas are considerably more efficient at performing the surveillance function in their
use of radar resources than their static array counterparts.
There are also benefits in the resource management and scheduling of a rotating array
system. A multi-function radar utilising a rotating array antenna has the ability to
distribute radar time across the angular coverage in a way that a static array antenna
system cannot. This allows better support high loading of surveillance and tracking
tasks in a confined angular sector, which is a likely operating environment for the multi-
function radar. A novel task scheduling strategy for a rotating array multi-function radar
which achieves this has been presented, and its performance assessed through computer
simulation.
The utilisation of a narrow beamwidth in the multi-function radar brings disadvantages

as well as advantages. The coupling of a narrow radar beamwidth with a low track
update rate, enforced by the radar time-budget constraints, increases the difficulty in
tracking of manoeuvring targets. This is particularly the case for the rotating array which
has blind arcs in the antenna coverage.
Efficient tracking strategies have been developed which address this problem for both
static array and rotating array multi-function radars. Successful tracking of highly
manoeuvrable targets with a narrow beam multi-function radar has been demonstrated
though simulation. High levels of track robustness can be attained through the use of
multiple beam track updating strategies which cost little radar time.
These findings have been summarised by presenting a simple assessment of a narrow

beamwidth, rotating array multi-function radar utilising simultaneous multiple beam
processing against a realistic, fictional, requirement. This system has been compared to
a four faced static array system using the same total number of array elements/active
array modules. The rotating system has been shown to be able to meet the requirement
200
within the radar time-budget constraints successfully, at the cost of increasing the clutter
illuminated by the transmit beam. When compared to the four faced static system, the
rotating system required less radar time resources to meet the requirement, and achieved
a higher free space detection performance in both surveillance and tracking.
Thus the possibility of a multi-function utilising a narrow radar beamwidth has been
demonstrated within the radar time and energy resources available. Coupled with the use
of a rotating array antenna, such a system would be both efficient, and give good
performance for the surveillance and tracking functions. However, if significant beam
broadening is required, then the implications of illuminating more clutter than is
illuminated by current systems must be assessed. The advantages and disadvantages of
such a system are summarised below.
Advantages Disadvantages
• Increased free space detection performance • Increased radar time requirement
in surveillance and tracking for tracking in some cases
• Increased efficiency in use of radar • Increased illumination of clutter
time/energy resources • Increased receiver or digital
• Better plot resolution in tracking and beamformer complexity
surveillance
• Better smoothed track accuracy
• Better performance in multipath conditions
• Increased resistance to main lobe jamming
• Greater ability to cope with an uneven

loading of tasks, or tasks confined to a
narrow angular sector
8.2 Evaluation of static array, rotating array and variable beam

pattern multi-function radar
A novel method of calculating the search function time for a static radar has been
201
developed. This gives analytical expressions which may be quickly evaluated for a range
of beamwidths and search sectors, without explicitly calculating each beam direction
and dwell time. This will allow surveillance configurations of a multi-function radar to
be evaluated rapidly using simple equations. The method is based upon estimating the
number of beams and the dwell times required in each beam direction in UV space, in
which the beam area remains invariant.
By creating a similar projection to the UV space projection, involving a transformation

onto the surface of a 3D cylinder, another spatial system may be created in which the
beam area of a rotating multi-function radar system remains invariant. This projection is
somewhat akin to the Sanson-Flamsteed sinusoidal and Mollweid projections used in
the field of cartography. The above methodology for estimating the beam directions and
dwell times, and thus the search function time, may be applied in this spatial system for
the rotating system.
The above methods for static and rotating multi-function radars have been compared
with the process of calculating beam directions and dwell times exactly, and have found
answers within a few percent. The accuracy decreased somewhat as the beamwidth
considered increased (to around 8% for a 100 beamwidth), due to the exact number of
beams required to exactly cover the edges of a given region.
A comparison of the search function for different configurations of static and rotating
array multi-function radars has been made. It was found that the static
systems require 20% (4 faced) and nearly 50% (3 faced) more radar time to execute a
complete search of the same volume. When comparing the 3 faced static system with a 2
faced rotating system, it was found that nearly the same amount of radar time must be
dedicated from each of the 3 faces as is required from just 2 faces in the rotating case.
Thus this comparison suggests one of the 3 faces could be eliminated if rotation was
employed with very little detriment to the search. These figures will vary to some
degree, depending upon the exact volume that is to be surveyed, and the tilt angle of the
arrays.
The impact of using a broadened beam both in transmit and receive on the multi-
function radar system has been made. For any significant broadening of the beam there
is a heavy penalty to be paid both in the tracking and surveillance functions.
202
The impact of using a broadened beam in transmit, coupled with narrow receive beams
has also been made. The variation of surveillance search time has been determined as a
function of transmit beamwidth both in thermal noise and in clutter conditions,
assuming multiple narrow receive beams. When target detection is limited by thermal
noise the result is as predicted by the familiar search radar equation; the search time is
invariant with transmit beamwidth, since a wider beam requires a longer dwell time to
regain the signal to noise ratio.
When target detection is limited by clutter, the search time may be decreased by
broadening of the transmit beam. This is because the dwell time in that instance is
determined by the resolution required in the Doppler domain, governed by the length, of
the transmit waveform. The greater quantity of clutter illuminated is far outweighed by
the Doppler discrimination possible.
The implication from these arguments is that the compromise on beamwidth usually
associated with air defence multi-function radar systems does not necessarily hold if the
multi-function radar has the capability of processing multiple receive beams. Systems
may be built with narrower receive beamwidths than those currently, whilst still
retaining the ability to cancel clutter through Doppler processing within the time-budget
constraints of the multi-function radar. From the tracking perspective all the benefits of
a dedicated tracking radar may be realised, such as excellent monopulse accuracy,
resolution of closely spaced targets, and resistance to main beam jamming.
8.3 Resource management of static and rotating array multi-function

radar
A set of objectives which a scheduling algorithm should attempt to meet have been
proposed. It is immediately clear that not all objectives can be perfectly met in a
practical situation, thus the importance of each objective to the multi-function radar
system must be reflected in the scheduling algorithm.
A simulation model of a static multi-function radar, centred upon the task scheduler has
been implemented. This allows design of multi-function radar search coverage volumes
and analysis of the performance at a macroscopic and a microscopic level. Three static
task scheduler algorithms have been incorporated into the simulation and tested. The
203
first is a simulation of the MESAR look based scheduler. The other two work on the
same principle, but are modified to calculate and reset time balances in a different way.
Of these, one is a look based scheduler as per MESAR, the other is a simplified version
which assumes that tasks are not broken down into looks, having some execution time
advantage.
The simulation has been tested and validated using the MESAR search region
parameters as a test case. A detailed assessment of the MESAR scheduler has been
made. This was important since there was some question over whether this algorithm
was performing as expected in the MESAR radar.
The results show that the scheduler algorithms achieve 100% utilisation of the multi-
function radar antenna in all circumstances. This was the second of the three objectives
which applied to a static multi-function radar, and reflects the importance attached to
full utilisation of the multi-function radar antenna. This objective is achieved at the
expense of scheduling tasks as close as is possible to their desired time (the first
objective). Tasks are always scheduled after their desired update time to some extent.
This is minimised in the scheduling architecture, by splitting waveforms up into smaller
looks so that the scheduling of a long waveform may be interrupted by a higher priority
function. This entails however some extra complication in the signal processing since
data from the interrupted waveform must be buffered to allow the interruption.
The MESAR algorithm is not optimal, in that better algorithms could be devised which
give 100% radar utilisation and schedule tasks closer to their desired time by allowing
tasks to be scheduled early. However, although relatively simple, the current algorithm
meets the requirements for timely scheduling adequately. The algorithm has therefore
been validated against the proposed objectives as far as is possible within this study.
An analysis of the differences between static and rotating multi-function radars has
drawn out the complications that the scheduling algorithms must overcome for efficient
operation of the multi-function radar. These are summarised;

iii. radar occupancy may be distributed non-uniformly with azimuth
204
It is concluded that a rotating scheduler algorithm should include some form of

prediction of the radar load over one or more successive scans in order to allocate
resources efficiently in situations of high radar loading. For resources to be allocated
correctly in this situation a rotating multi-function radar scheduler must allow the use of
forward and backward scanning of tasks to enable distribution of resources with angle.
The theoretical extent to which this can be achieved has been determined for a multi-
function radar which does not compensate the transmitted energy in surveillance with
scan angle, and for a multi-function radar which does.
An algorithm has been presented for the scheduling of tasks in a rotating radar that
copes with the problems outlined in (i) to (iii) above. It is highly efficient in distributing
the radar occupancy to the angular sectors which require it most. It will automatically
adjust from a situation in which tasks are uniformly distributed throughout the multi-
function radar coverage to a situation in which the threat may come from a particular
angular sector, so that many of the tasks are concentrated in a small azimuthal sector.
The cost of unnecessarily scheduling tasks away from array broadside in rotating multi-
function radar systems has been determined in terms of the decrease in detection
performance and in terms of the decrease in angular estimation accuracy when using
monopulse. It was found that dwell times must be increased by l/cos 3 (p) in order to
maintain detection performance, but by l/cos5 (ç) in order to maintain plot accuracy. The
total increase in the radar time for surveillance will therefore increase, however this is
offset somewhat by the increase in beamwidth (so a reduced number of search beams).
Therefore the search function time needs only be increased by a factor of 1/cos 2(p) in
order to maintain detection performance, and by l/cos 4(p) in order to maintain plot
accuracy.
The use of fuzzy logic and penalty functions has been suggested as a simple and
effective technique for the resource management of a multi-function radar. Such
techniques would allow a flexibility in the time at which tasks are scheduled, resulting
in a more efficient allocation of the multi-function radar resources. Also, these
techniques have the capability of allowing simple, adaptable and comprehensible control
rules to be used in the radar resource management. Fuzzy logic has been successfully
applied to the control of an airborne intercept radar by Popoli [Popoli & Blackman,
1987]. This is considered a fruitful area for further work.
205
8.4 Adaptive tracking in multi-function radar
The problems of tracking manoeuvring targets with a narrow beamwidth multi-function

radar system and a rotating array multi-function radar system have been addressed
through computer simulation and Monte Carlo analysis. Computer simulations of both
static array and rotating array multi-function radars have been constructed, and have
been applied to the target trajectories given in the benchmark tracking problem [Blair et
al, 1994].
Most importantly these results have shown the viability of a narrow beam multi-function
radar tracker for the tracking of manoeuvring targets, for both static and rotating array
systems. They have indicated the most efficient values of signal to noise ratio to aim for
in the updating of tracks, and have shown that the robustness of track can be maintained
through the use of multiple beam track updates. The extra radar time for such updates is
not prohibitive, requiring less than a 10% increase in radar time in most cases.
The results have shown that the signal to noise ratio that is aimed for use in the MESAR
radar for tracking is unnecessarily high. The signal to noise ratio may be reduced
considerably (by at least 6dB), without decreasing the track robustness below a
reasonable level, or the tracking accuracy below that desired.
The study has quantified the variation in both radar time and track robustness when the
radar beamwidth is varied. At narrow beamwidths the use of multiple beam track
updates is imperative to allow robust tracking. An efficient strategy for the use of
multiple beam track updates has been proposed, and the increase in track robustness
made using this update strategy has been quantified.
A rotating array adaptive tracker has been constructed based upon the same track
updating strategy as that used in the MESAR static array tracker. The rotating array
tracker uses roughly the same amount of radar time as the static array tracker for
constant velocity targets, close to array azimuth broadside. However, it requires
considerably more radar time than the static array system in periods when the target
executes a strong manoeuvre. This is due to the necessity for track updates to made at
large scan angles from array azimuth broadside. In the case of the most stressful target
trajectory considered, benchmark target trajectory 6, the rotating array system required
roughly 2½ times the radar time of the static array system.
206
However, the increase in radar time is heavily dependent on the exact target trajectory;
on the period of the manoeuvres, and on the azimuth angle of the target. This is due to
the amount of track updates which must be made at large scan angles from array
azimuth broadside, necessitating long dwell times to offset the loss in antenna gain. A
manoeuvring target trajectory, remaining mainly close to array azimuth broadside of the
static array, would require a greater radar time for tracking in rotating array case.
Conversely, for a target trajectory close to 600 away from azimuth broadside of the static
array, the situation is likely to be reversed, the rotating array system requiring less radar
time for tracking.
It is thought that the implementation of a multiple model tracker which more accurately
represents the target dynamics during manoeuvre would reduce the necessity to make
track updates at great angles from array azimuth broadside. This would reduce the radar
time required for tracking with a rotating array multi-function radar, as well as increase
the smoothed track accuracy during target manoeuvre. This is proposed as an area of
further work.
Another method by which the radar time required for the tracking function may be
reduced in a system which utilised simultaneous multiple beam processing would be to
utilise group tracking methods, i.e. synchronising track updates on closely spaced
targets. This may also be worth further study, although the benefit that will be accrued
will depend upon the likely target density in relation to the radar beamwidth.
207
9 REFERENCES
Adams, M.: 'The Little Book of Map Projections', G. Phillip and Son, 1914
Applebaum, S.P.: 'Adaptive Arrays', IEEE Trans. AP-24, l976.ff 6?62
Barbato,A. & Giustiniani, P.: 'An Improved Scheduling Algorithm for a Naval Phased
Array Radar', Proc. of Radar 92, lEE Conf. Pubi., No. 365, 1992 ff Z -S'
Barton, D.K.: 'Modern Radar System Analysis', Artech House, 1988
Bayliss, E.T.: 'Design of Monopulse Antenna Difference Patterns with Low Sidelobes',
Bell System Technical Journal 47, No.5, 1968
Best, J.H.: 'Phased Array Co-ordinate Transformations', The Microwave Journal, July
1962
Billam, E.R.: 'Design and Performance Considerations in Modern Phased Array Radar',
Proc. of Radar 82, lEE Conf. PubI. No. 216, 1982 / )S—Zj
ff
Billam E.R.: 'New Techniques in Phased Array Radar', Proc. of European
Microwave Conference, 1985 / if
Billam, ER.: 'The Optimisation of Beam Position Separation in Phased Array Radar',
Proc. GIE International Conference on Radar, 1986 / WI
Billam, E.R.: 'Phased Array Radar and the Royal Navy', Journal of Naval Science, Vol.
12, No. 2, 1986
Billam, ER. & Harvey, D.H.: 'MESAR - An Advanced Experimental Phased Array
Radar', Proc. Radar 87, lEE Conf. PubI. No. 281, 1987 , rf
Billam, E.R.: 'Phased Array Radar and the Detection of Low Cross Section Targets',
Proc. of European Military Microwave Conference, 1989a,
Billam, E.R.: 'MESAR - The Application of Modern Technology to Phased Array
208
9 References
Radar', lEE Tutorial Meeting on Phased Array Radar, September 1989b
Billam, E.R.: 'Parameter Optimisation in Phased Array Radar', Proc. of Radar 92, lEE
Conf. Pubi., No. 365, 1992
Billam, E.R.: 'Parameter Optimisation in Phased Array Radar', Ph.D. Thesis, University
of Birmingham, 1994
Billam, E.R.: 'The Problem of Time in Phased Array Radar', Proc. of Radar 97, WE
Conf. PubI., No. 449, 1997a ^
Billam, E.R.: 'Rotating versus Fixed Active Arrays for Multifunction Radar', Proc. of
Radar 97, lEE Conf. Pubi., No. 449, 1997b, -
Billeter, D.R.: 'Multifunction Array Radar', Artech House, 1989
Birch, P.: 'MESAR Task Scheduler', MESAR Technical Design Note No. 240, Siemens
Plessey Radar Technical Note, 1990
Blackman, S. S.: 'Multiple Target Tracking with Radar Applications', Artech House,
1986
Blair, W.D. & Watson, G.A. 'Benchmark Problem for Radar Resource Allocation and
Tracking Manoeuvring Targets in the Presence of False Alarms and ECM', Naval
Surface Warfare Center, Dahlgren Division, Technical Report No. NSWCDD/TR-
96/10, 1996
Blair, W.D. & Watson, G.A.: '1MM Algorithm Solution to Benchmark Problem for
Tracking Manoeuvring Targets', Proc. SPIE 2221, 1994 /7'
Blair, W.D., Watson, G.A. & Hoffman, S.A.: 'Benchmark Problem for Beam Pointing
Control of Phased Array Radar Against Manoeuvring Targets', Proc. of American
Control Conference, 1994 / p 2oI - zor
Blair, W.D., Watson, G.A., Gentry, G.L. & Hoffman, S.A.: 'Benchmark Problem for
Beam Pointing Control of Phased Array Radar Against Manoeuvring Targets in the
Presence of False Alarms and ECM', Proc. of American Control conference, l995 ZO 1 - 14S
Butler, J.M.: 'UK/US MESAR Trials Report', Defence Research Agency Technical
209
9 References
Report, No. DRA/AW/AWL/AWIRJCR/Cr95306/ 1.0, 1995
Butler, J.: 'Beam Pointing Errors in Tracking Manoeuvring Targets', Defence Research
Agency Technical Report, No. DRA/AW/AWL'AWIS/WPIWp9421O/1 .0, 1994
Butler, J.M., Moore, A.R., & Griffiths, H.D., 'Resource Management for a Rotating
Array Multi-Function Radar', Proc. of Radar 97, lEE Conf. Pubi., No. 449, 1997
Castella, F.R.: 'An Adaptive 2-Dimensional Kalman Tracking Filter', IEEE Trans. AES,
Vol. AES-16, No. 6, Nov. 1980 , ' P Zi.-8?
Chamberlin, W.: 'The Round Earth on Flat Paper: Map Projections used by
Cartographers', National Geographic Society Press, 1947
Chestnutt, J.: 'Overview of the MESAR Task Scheduler Subsystem', MESAR

Technical Design Note No. 131, Siemens Plessey Radar Technical Note, 1984
Cohen, S.A.: 'Adaptive Variable Update Rate Algorithm for Tracking Targets with a
Phased Array Radar', lEE Proc. F, Vol. 133, No. 3, 1986 -
Daeipour, E., Bar-Shalom, Y., Li, X.: 'Adaptive Beam Pointing Control of a Phased
Array Radar using an 1MM Estimator', Proc. of American Control Conference, 1994
' 2oS- 2p9
Daum, F.E & Fitzgerald, R.J.: 'Decoupled Kalman Filters for Phased Array Radar
Tracking', IEEE Trans. on Automatic Control, Vol. 28, No. 3, 1983 4' .2.t
Dolph, C.L.: 'A Current Distribution for Broadside Arrays which Optimises the
relationship between Beamwidth and Sidelobe Level', Proc. iRE, Vol. 34, 1946
Gardner, L.A. & Mullen, R.J.: 'Constant Gain Tracker with Variable Frame Time',
IEEE Trans. on Aerospace and Electronic Systems, Vol. 24, No. 4, 1988 3
Godbehere, D.B.: 'The MESAR Scheduler Investigation', Private Communication

between Siemens Plessey Systems, Cowes and Defence Research Agency, Portsdown,
1996
Harris, F.J.: 'On The Use Of Window Functions For Harmonic Analysis With The
Discrete Fourier Transform', Proc. IEEE, Vol. 66, No.!, 1978
210
9 References
Kahrilas, P.J.: 'Electronic Scanning Radar Systems Design Handbook', Artech House,
1976
Kingsley S. & Quegan,c .: 'Understanding Radar Systems', McGraw-Hill Book

Company, 1992
Kirubarajan, T., Bar-Shalom, Y., Blair & W.D., Watson: 'IMMPDAF Solution to
Benchmark for Radar Resource Allocation and Tracking Targets in the Presence of
ECM', IEEE Trans. on Aerospace and Electronic Systems, Scheduled to appear in
October 1998
Levanon, N.: 'Radar Principles', John Wiley & Sons, 1988
Milne, K.: 'Synthesis of Power Radiation Patterns for Linear Array Antennas', lEE
Proc., Vol. 134, Pt. H, No.3, June 1987
Money, D.G.: 'Naval Environment Clutter, Attenuation and Propagation Specifications',

Defence Research Agency Technical Report, No. DRA/SS/SSWI/WP950 16/1.0, 1995
Moore, A.R. & Butler, J.M.: 'Digital Adaptive Beamforming Progress', Defence
Research Agency Technical Report, No. DRA/AW/AWJJAWJS/CR94304/1 .0, 1994
Moore, A.R., Levett, W.A., & Butler, J.M.: 'Active Multifunction Radar in Theatre
Missile Defence', Proc. 8th tnt. Conference on Theatre Missile Defence, 1995
Moore, A.R., Salter, D.M.: 'Multifunction Radar for Naval Weapons System
Applications', Proc. International Maritime Defence Exhibition 95, 1995
Munu, M., Harrison, I., Wilkin, D. & Woolfson, M.S.: 'Comparison of Adaptive Target
Tracking Algorithms for Phased Array Radar', lEE Proc. F, Vol. 139, No.5, 1992
Noyes, S.P.: 'Finding the Time of Next Update', MESAR Technical Design Note No.
243, Siemens Plessey Radar Technical Note, 1990
Noyes, S.P.: 'Plot to Track Association', MESAR Technical Design Note No. 246,
Siemens Plessey Radar Technical Note, 1990
Noyes, S.P.: 'Track Smoothing Algorithms', MESAR Technical Design Note No. 242,
Siemens Plessey Radar Technical Note, 1990
211
9 References
Orman A.J. & Potts C.N.: 'On the Complexity of Coupled-Task Scheduling', Discrete
Applied Mathematics, No. 72, 1997
Orman A.J., Potts C.N., Shahani A.K. & Moore A.R.: 'Scheduling for a Multifunction
Phased Array Radar System', European Journal of Operational Research, No. 90, 1996
Popoli, R. & Blackman, S.S.: 'Expert System Allocation for the Electronically Scanned
Antenna Radar', Proc. of American Control Conference, 1987jf'f ,2,—/fg
Quigley, A.L.C. & Holmes, J.E.: 'The Development of Algorithms for the Formation
and Updating of Tracks', Admiralty Signals and Weapons Establishment Technical
Report (Classification Unlimited), 1975
Rhatigan, B.R., Kalata, P.R., Chmielewski, T.A.: 'An a— Target Tracking Approach to
the Benchmark Tracking Problem', Proc. of American Control Conference, l994i ia7(jSo
Rhodes, D.R., An introduction to monopulse, McGraw-Hill Book Company, 1959
Sabatini, S. and Tarantino, T.: 'Multifunction Array Radar, System Design and
Analysis', Artech House, 1994
Salter, D.M.: 'MESAR - The Plessey/Admiralty Research Establishment Adaptive

Phased Array Demonstrator', Proc. Military Microwaves 88, 1988 if
Salter, D.M.: 'IS&T Final Report', Roke Manor Research Technical Report, No.
72/96/R/63C, 1992
Sastry, C.R., Slocumb, B.J., West, P.D., Kamen, E.W., Stalford, H.L.: 'Tracking a
Manoeuvring Target Using Jump Filters', Proc. of American Control Conference, 1994
Shin, H.J., Hong, S.M., Hong, D.H.: 'Adaptive Update Rate Target Tracking for Phased
Array Radar', fEE Proc. Radar, Sonar and Navigation, Vol. 142, No. 3, June 1995f//?-i.3'
Skolnik, M. I.: 'Radar Handbook', McGraw-Hill, 1970
Skolnik, M.I.: 'An Introduction to Radar Systems', McGraw-Hill, 1981
Stafford, W.K.: 'Real-time Control of a Multifunction Electronically Scanned Adaptive

Radar (MESAR)', lEE Colloquium on Real Time Management of Adaptive Radar
Systems, June 1990
212
9 References
Stafford, W.K.: 'Tracking Waveform Selection', MESAR Technical Design Note No.
248, Siemens Plessey Radar Technical Note, 1991
Sworder, D.D. & Boyd, J.E.: 'Dual-Path Algorithms for the Bench Mark Tracking
Problem', Proc. of American Control Conference, 1994 IOO -
Taylor, T.T.: 'Design Of Line-Source Antennas for Narrow Beamwidth and Low
Sidelobes', IRE Trans. Antennas and Propagation, Vol. AP7, 1955
Tsaknakis, H. & Athans, M.: 'Tracking Manoeuvring Targets Using H Filters', Proc.
of American Control Conference, 1994 ,y I -
Tunnicliff, J.: 'The MESAR Tracking Algorithm', MESAR Technical Design Note, No.
23, Admiralty Signals and Weapons Establishment Technical Note, 1980
Van Keuk, G. & Blackman, S.S.: 'On Phased Tracking and Array Parameter Control',
IEEE Trans. on Aerospace and Electronic Systems, Vol. 29, No. 1, 1993 ,, -I'7
Van Keuk, G.: 'Software Structure and Sampling Strategy for Automatic Tracking with
a Phased Array Radar', AGARD Conf Proceedings, No. 252, 1978
Villeneuve, A.T.: 'Taylor Patterns for Discrete Arrays', IEEE Trans. Antennas and
Propagation, Vol. Ap-32, 1984 '
Von Aulock, W.: 'Properties of Phased Arrays', Proc. IRE, Vol. 48, 1960
Watson, GA. & Blair, W.D.: 'Multiple Model Estimation for Control of a Phased Array
Radar', SPIE, Vol. 1954, 1993 C/
Wilkin, D.J., Harrison, I. & Woolfson, M.S.: 'Target Tracking Algorithms for Phased
Array Radar', lEE Proc. F, Vol. 138, No.3, 1991
Woodward, P.M.: 'Method of Calculating the Field Over a Plane Aperture Required to
Produce a Given Polar Diagram', J. lEE, Vol. 93, Pt. lIla, 1947
Wray, M.: 'Software Architecture for Real Time Control of the Radar Beam within
MESAR', Proc. of Radar 92, lEE Con!. Pub!., No. 365, 1992
213
APPENDIX A
THE BENCHMARK TRACKING PROBLEM
The benchmark tracking problem was proposed in 1994 [Blair et al, 1994] to cope with
the lack of a common means of comparing different tracker systems. The benchmark
problem addressed beam pointing control of a phased array radar against highly
manoeuvring targets, and has been widely applied in the U.S. ([Rhatigan et al, 1994],
[Sastry et al, 1994], [Sworder & Boyd, 1994], [Daeipour et al, 1994], [Tsaknakis &
Athans, 1994], [Blair & Watson, 1994], [Kirubarajan et a!, 1998]). The bench mark
problem specified;
i. a set of target trajectories against which to assess the tracker

ii. a model of the radar measurement process
iii. a set of metrics by which the tracker may be assessed
It has since been used, especially in the U.S., as a standard against which different
phased array tracker systems may be compared, and has been revised subsequently in
1995 [Blair et al, 1995] and 1996 [Blair & Watson, 1996].
It was intended to employ part of the benchmark for further work in this area. However,
it was necessary to extend the benchmark model extensively to study this problem, since
it applies only to a static face phased array multi-function radar, and also, does not allow
for the generation of multiple receive beams. Also, the model is provided in MATLAB,
which would have proved to slow to allow the parameter investigation that was required
to be undertaken. The entire model would thus have needed to be translated into a
compilable language such as C or Fortran to enable the Montecarlo analysis undertaken.
Therefore, employing the entire Benchmark model is proposed as an area of further

work. In this study only the Benchmark target trajectories are included within the
investigation.
214
Appendices
Target Scenarios
Six target scenarios are specified within the benchmark problem. The benchmark
scenarios have been reproduced from [Blair & Watson, 1996], and are shown in Figure
A- I to Figure A-3 below. The start point of each trajectory is marked with a circle. The
targets exhibit cross sectional fluctuations according to the Swerling Case 3 model and
perform up to 7g lateral acceleration, and 2g longitudinal acceleration. Target ranges
vary between 20 - 120 km, and target elevations vary between 00 to 80°. Since the model
considers only a static single faced multi-function radar, the target azimuth varied
between ±60°. The six target scenarios each represent a different target type.
The first trajectory is representative of a large aircraft, such as a military cargo aircraft.
The aircraft flies with a constant speed of 290 mIs and performs a 2g turn, and then a 3g
turn.
The second trajectory is representative of a smaller, more manoeuverable, high

performance commercial aircraft. The aircraft flies with a speed of 305m1s and performs
a 2.5g turn through 90°, before descending and performing a second 4g manoeuvre.
The third and fourth trajectories are both representative of medium, high speed
bombers, with good manoeuverability. The third trajectory aircraft flies at a constant
height, and initial speed of 457m1s. A 45° 4g turn is made, followed by a second 4g turn
of 90°. In the subsequent period of straight and level flight the aircraft decelerates to a
speed of 274m1s.
The fourth trajectory has an initial speed of 251m1s and height of 2.29km. The first 4g,
450 manoeuvre is followed a period of straight and level, constant velocity flight. A 6g
turn is then made, and the aircraft accelerates to a height of 4.57km, where the aircraft
levels out and flies at constant velocity.
The final two trajectories are representative of fighter/attack aircraft. Trajectory five
begins with a period of longitudinal maximum thrust acceleration (<ig) at a height of
1 .5km. A 5g turn is made, whilst the throttle is maintained. 20s later a 7g turn is made,
after which straight and level, constant velocity flight occurs. A 6g turn is then made,
whilst the aircraft pitches up concurrently, and climbs to an altitude of 4.45km, where
the aircraft levels out and flies at constant velocity.
215
Appendices
The sixth trajectory begins with straight and level, constant velocity flight at a speed of
426m/s and height of 1.55km. A 7g turn is made, followed later by a 6g turn, in which
the aircraft is turned nose down to decrease its altitude to a height of 0.79km. A period
of constant height, low longitudinal acceleration flight follows, before a 6g turn is made.
A period of full thrust acceleration (<ig) is followed by a 7g turn, and constant velocity
flight.
cTTIi,
-. -.
Figure A .l(a,b) Benchmark tracking problem trajectories 1 (a) and 2 (b)
...
..
L.
Figure A.2(a,b) Benchmark tracking problem trajectories 3(a) and 4 (b)
'S
'S
'a
I'
L'
• - i . - .•
Figure A.3(a,b) Benchmark tracking problem trajectories 5(a) and 6(b)
216

Appendices
Modeling of the Target Measurement Process
The radar that is specified in the Benchmark Problem is a 4 GHz phased array using
amplitude comparison monopulse. The array consists of 3025 elements, and is tilted
back 15° from vertical. Each track update consists of one phase/frequency coded pulse,
from a selection of eight specified waveforms. The range gate is approximately 1.6km
for a track dwell and 10km for a search dwell. The radar 3dB beamwidth is 2.4° on
broadside, and broadens to 4.5° at maximum scan. The radar has a maximum scan angle
of 60°, and a single pencil beam.
The model uses the radar equation to model target detection. The output signal to noise
ratio for the compressed pulse, SNRC , is given by equation [ A-I], and the radar
parameters in Table A-i.
FG i G r 2L2 Fi 2 Fr2 Q, ( Nr YG{R}" [A-I]

SNR= )3
(4ir L1(,, k, T F R4 J
L
where Fr is the transmit propagation factor

Fr is the receive propagation factor
o is the target rcs
Lt(,, are the total losses for the radar system
Nc is the number of discrete sub-pulses in the transmitted pulse
. is the sub-pulse width, assumed to be equal to the compressed pulse width
k(, is the Boltzmann constant
T0 is the reference temperature
F is the receiver noise figure
G 1 (R1 is the sensitivity time control gain with range
1MW
Pt
36.8 dB
Gt,Gr
7.5 cm
0 dB
Ft , F1
21.6dB
290 K
T0
Table A-i Radar equation parameters for the Benchmark Problem
Eight sets of waveform parameters are specified, as is the sensitivity time control
function with range. Functions are specified for the calculation of antenna gain patterns,
sum and difference channel voltages and monopulse angular estimation.
The RCS of the target is modeled as Swerling 3, where the average RCS, varies
between target scenarios. It may be computed by solving equation [A-2] below, where
217
Appendices
x is a random number, uniformly distributed between 0 and 1.
[A-2]
l—x
Tate )
Metrics
The primary measure of tracking performance in the Benchmark Problem is a weighted

sum of the average radar energy per second and the average radar time per second. The
average energy per second, Eave, is the sum of the energy of the track dwells scheduled
by the tracking algorithm divided by the number of seconds of the trajectory. The
average radar time per second, Tave, is the sum of the radar time of the track dwells
scheduled by the tracking algorithm, divided by the number of seconds of the trajectory.
Two cost functions are then defined;
C1 = E jy +1 Ta,e
[A-3]
C2 = E + l05T
C 1 corresponds to a period of operation when radar energy is critical, and C2

corresponds to a period of operation when radar time is critical. The objective is to
minimise one of the cost functions on request. The best algorithm is determined as the
one that minimised the appropriate cost function, while satisfying a constraint of 4% on
the maximum number of tracks lost during the Monte Carlo simulation.
A secondary metric is the computer resources that are required for implementation of
each algorithm. The number of floating point operations per second that the algorithm
requires in its final implementation is to be calculated.
218
APPENDIX B
STATIC ARRAY TRACKER MONTECARLO
SIMULATION RESULTS: BENCHMARK TARGETS
1 TO 6 & CONSTANT VELOCITY TRAJECTORY
A complete set of Montecarlo simulation results is presented here of the multiple beam,
static array tracker that was developed. A beamwidth of 2.5° has been used.
219
Appendices
Static Array Multi-function Radar
3,
q' N5 R 1S .. L,. A.w ..
Figure B-1(a,b) Total radar time required to maintain track on Benchmark target trajectory 1,
and (b) the percentage of successful tracks. Static array system with a beamwidth of 2.5°
:::::::::
2 IS
Figure B-2 (a,b) Average track update interval required to maintain track on Benchmark target
trajectory 1, (b) Average number of bewns required for each track update. Static array system
UIIS
I SIS
"55
1
'10
I
I
S2O
I
I
0
I fl 5' 510 010 00 0!' S I" St

t.tw1'.:I S.
Figure B-3(a,b) Achieved smoothed track accuracy for Benchmark target trajectory 1,
Benchmark target trajectory I using only a 20dB signal to noise ratio instead of a 29dB signal
to noise ratio. Static array system with a beamwidth of 2.5°
220
Appendices
14
dT.S ?4c.,y q..I ' N R.I (1 .dT.. P2 In

P9.4 P N,.,. R.4n
plo Plo
e
on •.__
plo
•' Dlo,.d '.4.4 Ac,.p
., I•
d In N. R.. ..JO. Np.4P Na.. Nnb
Figure B-5(a,b) Average track update interval required to maintain track on Benchmark target
trajectory 2, (b) Average number of beams required for each track update. Static array system
'I,.
I
1.0o
I
1 P10
I
I
J l.20 I
I
0.
SqIp In Np
.. R.on. d T. .uiy
excluding periods of manoeuvre, and (b) Percentage of radar rime required for tracking
221
Appendices
25
'1 D.sdT *2 Ia
Sigl No.. () !qb N,... Rn()
a?' a
...d T,ad, c,..W 12 In D....d Ta.d. A,n,.a.c Ia2 I.•

Ia Na... Rain tdi a. Na... Rain dBj
*1W
I
I125
1
LII'
II2 In I
$9.11 a. Rala
222
Appendices
010
Dn.dT Duuwdtu10 — 0! 10
..,y I' Iq.Il lo 010w 0010 R (Fj
Figure B-1O(a,b) Total radar time required to maintain track on Benchmark target trajectory 4,
and (b) the percentage of successful tracks. Static array system with a beamwidrh of 2.5°
T).o,10 '.wk 10o,wy fl

SqToI To No,. fob dO, t10obd Tb 0ooy 0 In
To011i To No.,. R.10 (dOj
0.101
I::
010
i.F
i
-
I.:
0bj RModO, Aw
223

Appendices
V VP
•,,
•,

O...d Tm VV.,.VV dY oy 10
'3
I'
D.i.sdTl A00.. 1 I..

ip* N RD
I
Is'.
I I
is's'
I'
I
1
0 I
sIn
3q.II RD. Rs'op
224
Appendices
I. Do.,.d T.o,, .:
No... R. ,oa, Plc,.. RoIo;
and (b) the percentage of successful tracks. Static array system with a beamwidrh of 2.5°
! 10
Iq'd R (I
02 10
Sy. Ni. RMoll
JO,'0
1
iu.
S2D
S
000. R()
225
Appendices
22
ii
Os
D..dT.. *,...o, ' i
Figure B-19(a,b) Total radar time required to maintain track on the constant velocity
trajectoty, and (b) the percentage of successful tracks. Static array system with a beamwidth of
2.50
j1
.
-- — —I-I DdT. - "
Figure B-20(a,b) Average track update interval required to maintain track on the constant
velocity trajectory, (b) Average number of beams required for each track update. Static array
I
I
1
I
'L . I
r....s L•' -,
Sb,s (
Figure B-21(a,b) Achieved snwothed track accuracy for the constant velocity trajectory,
excluding periods of manoeuvre, and (b) Percentage of radar time required for tracking the
constant velocity trajectory using only a 20dB signal to noise ratio instead of a 29dB signal to
noise ratio. Static array system with a beamwidth of 2.5°
226
APPENDIX C
ROTATING ARRAY TRACKER MONTECARLO
SIMULATION RESULTS: BENCHMARK TARGETS
1 TO 6 & CONSTANT VELOCITY TRAJECTORY
Montecarlo simulation results are presented here of the multiple beam, rotating array
tracker developed. A beamwidth of 2.5° was used. A complete set of results are
presented for the case of single array multi-function radar rotating with periods of is.
For the remaining three systems considered, only results for benchmark target trajectory
6 are given for comparison.
227
Appendices
Single Array Multi-function Radar, Period is
-- in

m m Nn.s. P.m Ii
Figure C-1(a,b) Total radar time required to maintain track on Benchmark target trajectory 1,
2.50
.:i:11j..

D..i..d Ti..m m.mc, s in Ds..sdlmth c.n..cy n in
Figure C-2(a,b) Average track update interval required to maintain track on Benchmark target
ts
Jim.
1
-..m
I
I
I
in I
ii.m N.. I) DuUWTm.E
Figure C-3(a,b) Achieved snwothed track accuracy for Benchmark target trajectory 1,
Benchmark target trajectory I using only a 20dB signal to noise ratio instead of a 29dB signal
228
Appendices
'2
:
'..o T..0 I
.Il 0000 08, J&,
2.50
6I0
•0 lip

'LI 3 10 NUI.(
Figure C-5(a,b) Average track update interval required to maintain track on Benchmark target
I
is60
I
520
I
1
0 I
p0 I

D..00d 00I000 ' I 6' •I •1' • 0' 0l
Figure C-6(a,b) Achieved smoothed track accuracy for Benchmark target trajectory 2,
229
Appendices
12
'I,. -
r.6..d T 0 D..a,d Ta, Aaaa,. ..: i
6.il So la,.. R.Ua dU, Oq. ?, Na,o
2.50
ii:
:
S'2 50
2 50 — r Nla. ()
Figure C-8(a,b) Average track update inten'aI required to maintain track on Benchmark target
Sb
JUdo
I
- 3*
I
*2.
I
010
I
30 1
0.Sd T,.do Acc,.c Sop. 3.5.08
3 11° 555 506 025 00 305 6*'
6, No... o.0 A,,t*v
230
Appendices
I.
JO
0su.WT,. Ao'.op 0 I n •1 10
Ns 0.00(00)
Figure C-1O(a,b) Total radar time required to maintain track on Benchmark target trajectory 4,
2.50

Du00udT, •2 to 0uIdT 0 II
0.10 (00(
Figure C-11(a,b) Average track update interval required to maintain track on Benchmark
target trajectory 4, (b) Average number of beams required for each track update. Single faced
rotating system with a beamwidth of 2.5°
'Ito
I
t o00
1.00
1
p40
.20
'40
I
-1,
20
02 tO 0 WO 000 003 UPi fl Wfl UC

0.00(00)
231
Appendices
I.
J O
:i
OIfl
--
0..d T'. A'i 2 10
Sq,d 0o No.. 0.,, ()
2.50
120 020
p. 1..
lIo__.____
12 IN lO II
0, I 0.b . * 0p.U0 No,.. NoNo (J
rotating system with a beamwidrh of 2.5°
0101
I
010
I
1
JO00
1
1
I
20 I
row 7. o... - 0U0 0000 000.IO,
232
Appendices
3
I
I
I
Id T I I.. ___ ___ (4I Tfl.* 1 II ___
2.50
'S
I
I
Io
4P'
- II

2 is
1 550
1 I
I
1 .20
1
I
20 I
. a.
- I., 0 05 01 SYI • Ill *1' ftc,
233

Appendices
Os
3
I
0
1010
10
-.
.
107 10
10 0' Ne.. D...d T.10 10Ilo 0,0. ONe i10
Figure C-19(a,b) Total radar time required to maintain track on the constant velocity
trajectory, and (b) the percentage of successful tracks. Single faced rotating system with a
beamwidth of 2.5°
:1
0' Nmis0,Io()
Figure C-20(a,b) Average track update interval required to maintain track on the constant
velocity trajectory, (b) Average number of beams required for each track update. Single faced
0,110,
"Si
Li i
I
01 I
00 I
02 00 0
0' No.. 110,
Figure C-21(a,b) Achieved smoothed track accuracy for the constant velocity trajectory,
excluding periods of manoeuvre, and (b) Percentage of radar time required for tracking the
constant velocity trajectory using only a 20dB signal to noise ratio instead of a 29dB signal to
noise ratio. Single faced rotating system with a beamwidth of 2.5°
234
Appendices
Single Array Multi-function Radar, Period 2s
Ii
is

D.Nd Trnth Sq. Ds..d Tmth AAC...c, 2 IA
2.50
IL. ....-..
I.
D...d T,.d, kAA,.v I2 In

Sq. In Nn.,. R.sn (AD) D.. .d T,id, Aoc,.Icv n 2 1.1
Sq. In Noin. NoOn (AD)
I
I
1
0)
N) IA
SqSIn Ns. NIOn (AD,
235
Appendices
Twin Array Multi-function Radar, Period is
12
I.
I..
I' I
- - - .---. -

D..i,.d 1,.d, ..: i ___ D...,.d T,. 02 10
2.50
ri..•: .1........I....
003 10
•
DUdT,I 2 Ia DUI.UdT ..a0 0! In

NO. 0U1 0ql 10 P P0o (Is)
rotating system with a beamwidih of 2.5°
0I
t O.
I I
I
I
Sq m. P.r .0. DsT'
236
Appendices
Twin Array Multi-function Radar, Period 2s
12
I.
I,
l0
S10
D.d Tm • I' ___r.. OuI..d T A0$..0V $2 In
,, o
2.50
:...
$2 IS DN.,dTI. $2 10
$ Noin. R.No() b N.s No
UII
1
SI)
I
J $20 1
S I
'I0 j
D.o.d T.. A$) 0: '• !r Nn. 0I)n :0.., •. ..N A.•*•b
Figure C-30(a,b) Achieved snwothed track accuracy for Benchmark target trajectory 6,
excluding periods of manoeuvre, and (b) Percentage of radar time required for tracking LOtDON
Benchmark target trajectory 6 using only a 20dB signal to noise ratio instead of a 29dB signal UNW
237

Tracking and Control in Multi-Function Radar: by Joseph Mackay Butler

Uploaded by

Copyright:

Available Formats

Tracking and Control in Multi-Function Radar: by Joseph Mackay Butler

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Tracking and Control in Multi-Function Radar: by Joseph Mackay Butler

Uploaded by

Copyright:

Available Formats

TRACKING AND CONTROL IN

Joseph Mackay Butler

A thesis submitted to the University of London for the Degree of

Doctor of Philosophy in Electronic Engineering

Department of Electronic and Electrical Engineering

University College London

4 THE IMPACT OF VARIABLE BEAM WIDTH AND

5 RESOURCE MANAGEMENT IN STATIC AND ROTATING ARRAY

8.1 Summary 199

APPENDIX A: THE BENCHMARK TRACKING PROBLEM 214

APPENDIX B: STATIC ARRAY TRACKER MONTECARLO

APPENDIX C: ROTATING ARRAY TRACKER MONTECARLO

Symbol Definition Units

A I,eamuv area of beam in UV space m2

DERA Defence Evaluation and Research Agency

E(ø) antenna gain in direction 0 dB

ix normalised distance between the target predicted and measured position

Jz normalised distance between the target predicted and measured position

n number of pulses integrated

N number of elements, or subarrays in the array

n_beams number of surveillance or tracking beams

0 occupancy of the task or function %

Q plant noise term

R covariance of measured target state

r, R slant range of the target m

S received signal power at the radar W

S mm minimum detectable signal W

snr signal to noise ratio at a scan angle of ( dB

tfr frame time of the task or function S

Ty time taken by the pulse to travel to the target and return S

TWs Track While Scan

Ur coordinate of rotating TJV space antenna coordinate system

(1) antenna rotation rate radians/s

target transition matrix

abeam solid angle of beam steradians

sector solid angle of surveillance sector steradians

standard deviation of the tracker prediction error radians

standard deviation of the measurement error radians

scan angle away from array normal radians

smoothing constant for tracker residual estimate

03 3dB beamwidth radians

oc, beam width at scan angle of p radians

monopulse accuracy radians

monopulse accuracy at scan angle p radians

1.1 Historical background

performed adequately. This is also addressed within the study.

1.3 Thesis Layout

Chapter 2 gives an introduction to some of the background topics pertinent to Multi-

Chapter 3 then gives a detailed introduction to the MESAR experimental multi-function

Chapter 6 presents a detailed investigation of a static and rotating array adaptive

1.4 Novel Aspects

Chapter 7 demonstrates the viability of a narrow beamwidth multi-function radar

This chapter gives an introduction to some of the topics pertinent to Multi-function

The multi-function radar is equivalent to a suite of radars, employed for some

At its simplest level a multi-function radar consists of the following:

i. a phased array antenna, (including the associated beamforming and

It is the coupling of an electronically steerable beam, adaptive waveform generation and

2.2 Multi-function radar functions

Figure 2-1 Functions of a multi-function radar Jhr naval air defence

Physicalaperture