463

_____________________________________________________________________

Chapter 15.
Tracking Moving Targets
15.1.Track While Scan (TWS)
The track of a target in 2D can be determined from a surveillance radar Plan Position
Indicator (PPI) display by plotting the target coordinates as they move when measured
from scan to scan.

Figure 15.1: Examples of plan position indicator display hardware

At its most simple, this tracking function can be performed by a radar operator
marking the face of the cathode ray tube with a pen. This is an inaccurate process and
limits the number of targets that can be handled at one time.

Figure 15.2: Plan position indicator showing a simulated and a real display

464
_____________________________________________________________________
Automatic trackers operate as follows (for a single target)
1. Target is detected as the received echo exceeds a threshold. There is no
information about its velocity. The software constrains the uncertainty to a
reasonable value for an aircraft target (large circle)
2. Target is detected again displaced in range and angles, but within the
uncertainty boundary. A crude velocity estimate is made and the position
where the target will appear next is predicted. The uncertainty is still large as
the position and velocity estimates are not good.
3. The target appears within this uncertainty boundary. Tracking filters estimates
of position and velocity improve, and the next sample prediction is made with
a smaller position uncertainty
4. As with (3)
5. The actual target position falls outside the position uncertainty boundary
because it has accelerated, and the prediction algorithm only used position and
velocity. Track is lost
6. A new target is detected with unknown velocity.

Figure 15.3: PPI Display sequence to illustrate the target tracking process

One method of performing the filtering and prediction function is to use filters
(see Chapter 13) for the polar co-ordinates (R,). An alternative would be to convert
the measured positions from polar to Cartesian (x,y) before filtering as shown in the
figure.

Figure 15.4: Track while scan processing

465
_____________________________________________________________________

15.2.The Coherent Pulsed Tracking Radar

Pulsed tracking radar systems, such as the BAE Systems unit shown in the figure
below, are mostly used for military applications such as fire control. These radars
track fast moving aircraft or missiles with high accuracy, and then use the estimates of
the target position and velocity to direct missiles or anti-aircraft guns.

Figure 15.5: Cluster of fire control tracking radars

Such coherent radar systems extract both amplitude and phase information from the
signal reflected by a target. This is required because the length of a single pulse is too
short to resolve typical target Doppler frequencies as shown in (c) of the figure below.

(a)

(b)

(c)

Figure 15.6: Doppler modulation of a pulse train

To extract Doppler shift, the returns from many pulses over an observation time T
must be analysed so that the spectrum can be resolved down to a bandwidth 1/T.
For this process to work, a deterministic phase relationship must be maintained over
the observation time T.

466
_____________________________________________________________________

15.2.1. Single Channel Detection

The block diagram of a single channel coherent radar is shown in the figure below.
The transmitter configuration is Master Oscillator Power Amplifier (MOPA)
The transmitted signal is obtained by mixing an RF signal from a Stable Local
Oscillator (STALO) with a frequency frf with a coherent oscillator (also very
stable) with a frequency fif.
The resulting frequency (after filtering) id fo is pulse modulated to form a
pulse train, amplified and transmitted.
The received signal is down converted to IF (typically 30-60MHz) by mixing with the
STALO signal. After amplification and filtering it is down converted further to
baseband (video) by mixing with the COHO signal
A consequence of this single channel down conversion is that there is no direction
information in any Doppler modulation since a target receding would produce exactly
the same Doppler signature as one approaching and the response is
V1 = k sin (2f d t + o ) ,
where V1 Video output voltage,
k Amplitude of the video signal,
fd Doppler frequency (Hz),
o Phase shift (rad).

Figure 15.7: Single channel coherent pulsed Doppler radar

(18.1)

467
_____________________________________________________________________

15.2.2. I/Q Detection

In an I/Q detector, the IF signal is split into two channels with the quadrature (Q)
channel being phase shifted by 90 with respect to the in-phase (I) channel
Though the two Doppler signals output by the I and Q channels will have identical
frequency whether the target is approaching or receding, their phase relationship with
each other will reverse, and so direction information can be obtained as discussed in
Chapter 13.

Figure 15.8: Coherent pulsed radar with a synchronous (i/q) detector

By sampling the I and Q outputs and using a complex fast Fourier Transform (FFT),
the magnitude and phase of the combined Doppler spectrum can be obtained.
Another benefit of I/Q detection over single channel detection is that a 3dB gain in
SNR is obtained at the output of the Doppler spectral analysis function.
If the actual Doppler frequency is not important, but only the fact that the target is
moving at important, then a process called Moving Target Indication (MTI) can be
applied.

15.2.3. Moving Target Indicator (MTI)

For a moving target. if the video output of the I or Q channel is examined, the
amplitude will vary on a pulse to pulse basis, as shown earlier, due to the changing
phase between the transmitted and received signals. For a static target, the phase will
remain unchanged, and the amplitude will remain constant.
An MTI based on a delay-line canceller operates by taking the difference of the
amplitudes of successive pulses as shown below.

468
_____________________________________________________________________
Difference
Bipolar Video
Delay Line
T - 1/prf

Unipolar Video

Full-wave
Rectifier
-

Figure 15.9: Delay line canceller used to suppress the returns from static targets

This is in effect a Finite Impulse Response (FIR) filter with a high-pass characteristic
that rejects signals that are unchanging or that are changing very slowly. However,
because it becomes in effect a sampled data system, the frequency response is
repeated with a period of 1/T as shown in the figure below.

Figure 15.10: Transfer function of single and double delay line cancellers for MTI

Standard FIR filter techniques can be used to adjust the weightings of a canonical
filter implementation, as shown below, to alter the filter response and narrow the stop
band and flatten the pass band.
Delay
T1

Input

W1

Delay
T2

W2

Delay
T3

W3

Delay
TN-1

W4

WN

Summer

Output

Figure 15.11: Canonical MTI filter structure

One of the disadvantages of classical MTI techniques is that they cannot discriminate
between a moving target and moving clutter, nor do they cope very well if the radar is
also moving.

469
_____________________________________________________________________
Blind Speeds
If the target Doppler frequency lies in the region where fd 1/T, it can be seen, from
the transfer function above, that it is attenuated. Zeros also occur at 2/T, 3/T n/T

fd =

n
= nf p ,
T

(18.2)

where fd Doppler frequency (Hz),

fp Pulse repetition frequency (PRF) (Hz).
The actual blind speeds are given by the following

vn =

n nf p
=
m/s,
2T
2

(18.3)

where vn Blind speed (m/s),

- Transmitter wavelength (m),
n 1,2,3,
To obtain a high first blind speed, it is necessary to either operate with a long
wavelength (often not practical), or to operate with a high PRF. However, a high PRF
becomes ambiguous at a short range which is also not ideal for surveillance or long
range tracking radar applications.
Staggered PRF and Blind Speed
The effect of blind speeds can be reduced by operating at more than one PRF as
shown below.

Figure 15.12: Effect of staggered PRF on MTI transfer function

470
_____________________________________________________________________
As the ratio of the pulse repetition interval, PRI, T1/T2 approaches unity, the greater
will be the value of the first blind speed. However, the first null also gets deeper and
so the rejection of slowly moving clutter will be compromised.
To cater for the moving clutter problem, a bank of Doppler filters can be implemented
instead of a delay line canceller. In modern radars, this process is generally
implemented digitally using the Complex FFT. The outputs (excluding sidelobes) of
such a filter implementation are shown below.

Amplitude
Response

To obtain sufficient rejection of unwanted signals in adjacent bins, the filter sidelobes
must be made as low as possible.

10 11 12

Frequency Bin

Figure 15.13: Filter bank implemented using the FFT

A tradeoff exists between the width of the mainlobe and the sidelobe level with
different windowing (weighting) functions as discussed in Chapter 11.
Table 15.1: Properties of weighting functions
Weighting Function
Rectangular
Hanning
Hamming
Taylor n=5
Taylor n=6
Dolph Chebyshev

Peak
Sidelobe
Level (dB)
-13.26
-31.5
-42.5
-34
-40
-40

MainLobe
Width
0.886
1.42
1.32
1.19
1.25
1.2

15.3.Limitations to MTI Performance

Sub-clutter visibility is the ratio by which the moving-target power may be lower than
the clutter in the same range bin, and still be detected with a specified Pd and Pfa.
It is usually limited by internal instabilities in the amplitude and phase of the various
waveforms generated by the radar, by the motion of a scanning antenna and the finite
time on target which has the effect of widening the clutter spectrum and also by the
bandwidth characteristics of real clutter. Rain is blown by the wind, the sea moves, as
do leaves and grass.

15.4.Range Gated Pulsed Doppler Tracking

The Doppler tracking loop and the range tracking loop are interrelated. The Doppler
tracking loop operates on video sampled at the range indicated by the range tracker
while range tracking is usually accomplished by means of the split (early/late) gate
technique.

471
_____________________________________________________________________
Before the signals in the early and late gates are compared to derive the range tracking
error, each is passed through a Doppler tracking filter to reject returns from stationary
objects at the same range.
The range tracking loop can be updated at the PRF unless an FFT that processes Np
pulses is used to perform the Doppler analysis in which case the update frequency is
reduced to PRF/Np.

Figure 15.14: Schematic diagram of a range gated doppler tracker

Initiation of tracking requires the near simultaneous initialisation of all four tracking
coordinates: range, two angles and Doppler frequency, this is achieved as follows.
The operator who identifies a target during the search phase designates the
selected target on the display using a cursor.
On the following scan, if the target is still present, the range, azimuth and
elevation angles and Doppler frequency are recorded.
The scan sequence is interrupted and the radar antenna returns automatically
to the designated area.
The range gate pair are moved to the designated range where they begin a
small search to compensate for uncertainties in the designation.
The sum channel video received by the combined early and late gates, or by a
third target gate that straddles the two is processed by a Doppler analyser (eg
FFT)
If a moving target is found at the designated range and sufficiently close to the
Doppler of the detection that triggered the acquisition sequence, then the
Doppler tracking gate is placed over the target.
The Doppler loop is closed by using a frequency discriminator to maintain the
target in the centre of the tracking filter.
Video signals in both the early and late gates are analysed by Doppler filters
slaved to the frequency of the Doppler tracker.
After allowing the Doppler loop time to settle, the range tracking loop is
closed by using the difference in the relative amplitudes of the early and late
gates to control the range position

472
_____________________________________________________________________

The video signals in bot the azimuth and elevation monopulse angle error
channels are sampled by a range gate slaved to the target gate of the range
tracker.
The sampled video for each of the two error signals is filtered by a Doppler
filter slaved to the frequency of the Doppler tracker. Thus angle errors are
derived only from the same range-Doppler cell that contains the target.
After a brief period to allow the range tracking loop to settle, the angle
tracking loop is closed.

The transition from search to track is not a trivial problem, and it is made even more
difficult in the military scenario, if the target either starts to manoeuvre or tries to
disrupt the process by deploying chaff or by some electronic means

15.5.Co-ordinate Frames
15.5.1. Measurement Frame
Radar measurements are made in polar space (R,,) as the radar can only measure
range, elevation and bearing (azimuth).

15.5.2. Tracking and Estimation Frame

The equations of motion that govern the profile of a target operate in Cartesian space,
(x,y,z), so it is advantageous to transform the co-ordinate system from polar to
Cartesian space.

Figure 15.15: Rectangular <> Polar (spherical) transfer function

Generally this frame of reference will remain centred at the radar, however, some fire
control systems translate and rotate the frame to make it target centred as the aircraft
dynamics can be better modelled in this frame.
If more than one sensor may be involved in the tracking function, and these sensors
are not co-located (they may be on different platforms that move relative to each
other), then an earth centred Cartesian frame is generally used.

473
_____________________________________________________________________

15.6.Antenna Mounts and Servo Systems

Figure 15.16: Antenna mounts

The pencil beam of a tracking radar must be pointed at the target for tracking to occur.
This is quite a challenge as a typical tracking radar has a 3dB beamwidth between 1
and 2, or smaller in the case of the radar for a close-in weapon system onboard ship.
A servo system is used to drive the antenna in the direction that minimises the
tracking errors. Most servo systems are Type II, or zero velocity error systems since,
in theory, no steady-state error exists for a constant velocity (angular rate) input. With
Type II systems, dynamic lags proportional to the magnitude of the target acceleration
do occur. To accommodate this, the tracking bandwidth is adjusted to minimise the
tracking error which is due to a combination of measurement noise and dynamic lag.

474
_____________________________________________________________________
Transmitter

Range
Tracker

AGC

Mixer
Duplexer
Sum Channel

Elev
Difference
Channel

Mixer

Elev Motor
& Feedback

Azim Motor
& Feedback

Amplitude
Det.

IF Amp

Phase
Sensitive
Det.

IF Amp

Phase
Sensitive
Det.

Mixer

Monopulse
Hybrid
Monopulse
Antenna

IF Amp

Local
Osc.
Posn & Rate
Current

Servo Amp

Posn & Rate

Current

Servo Amp

Video
Amp

Elev Angle Error

Secant
Gain

Azim Angle
Error

Figure 15.17: Monopulse tracking radar

At long range where the angular motion of the target is small, a very small tracking
bandwidth can be tolerated. However, at short range where target angular rates and
accelerations are large, a wider bandwidth becomes acceptable. Secant correction
increases the azimuth error signal gain as a function of the elevation angle.
Another restriction on tracking bandwidth is that it must be small (10%) compared to
the lowest natural resonant frequency of the antenna and mounting structure to reduce
the risk of instabilities occurring.

Figure 15.18: Lowest resonant frequency as a function of antenna diameter

475
_____________________________________________________________________
The pointing system, generally known as a pedestal must be able to accommodate
both target motion as discussed and its own base motion (if mounted on a moving
vehicle)
There are many different types of antenna mounts that can be used depending on the
tracking and stabilisation requirements.

15.6.1. On-Axis Tracking

The best tracking occurs using null-steering when the antenna is pointed towards the
target with an accuracy of only a few milliradians. This is known as on-axis tracking.
It reduces cross coupling between the axes by minimising cross polar levels and
reducing the effects of system nonlinearities.
It requires the following:
the removal by prior calibration of biases
a filter than can perform one sample ahead prediction
the selection of the appropriate co-ordinate system for tracking.
Target dynamics that dictate the real and apparent acceleration and tracking loop
bandwidth determines the tracking accuracy.

15.6.2. Crossing Targets and Apparent Acceleration

Figure 15.19: Geometry of a crossing target (a) in plan and (b) in perspective

476
_____________________________________________________________________
When viewed in radar polar co-ordinates, the target velocities will be

v
v
R& max = vt , A& max = t , E& max = t .
R
R

(18.4)

&& = at , E&& = at .
&& = a , A
R
max
max
max
t
R
R

(18.5)

The real accelerations are

The geometric accelerations and other derivatives are shown below. They are
normalised in the angular co-ordinates to A& max = m = vt / Rc . Where Rc is the ground
range at crossover.

Figure 15.20: Angular derivatives for crossing targets

In range, the derivatives are normalised to Ra = Rmin, at the point of closest approach.

477
_____________________________________________________________________

Figure 15.21: Range derivatives for crossing targets

Conventional servomechanism theory can be used to determine the lag errors in radar
tracking loops if conventional loops are implemented.
The lag error can be written as follows

a =

a
Kv

& a
Ka

&&a
K3

+ ...

(18.6)

Where the coefficients Kv, Ka and K3 are the servo error coefficients, the values of
which increase with increasing loop gain and bandwidth.
Servos are classified according to the first coefficient that is finite in the loop design.
A type 1 servo has a finite Kv (but infinite position error constant Ko), a type 2 has
finite Ka but infinite Kv, etc.
The acceleration error coefficient is intimately connected to the closed loop
bandwidth of the servo. If the bandwidth is expressed in terms of the equivalent noise
bandwidth Bn then
Ka =2.5Bn2 = 0.63/to2.

(18.7)

Where to is the equivalent averaging time for the tracking loop. The lag will be.

a =

& a
Ka

& a
2 .5 B

2
n

= 1.6& a t o2 .

(18.8)

This formula is applicable to azimuth, elevation and tracking lags with their
corresponding acceleration components.

478
_____________________________________________________________________
If thermal noise and dynamic lag are the primary sources of error, then an optimum
bandwidth can be obtained to minimise the total error variance. According to Barton,
this is

2 =

32 Bn
at2
,
+
k m2 f r B ( S / N ) 6.3R 2 Bn2

a 2 k 2 f B ( S / N )
Bn = t m r 2 2

1.6 3 R

(18.9)

1/ 5

(18.10)

where: at acceleration (real or geometric) (m/s2),

km Monopulse Gain Constant (typ 1.6),
fr Pulse repetition frequency (Hz),
B - IF bandwidth and pulse width (see matched filter),
S/N Single pulse signal to noise ratio,
3 Antenna beamwidth (rad),
R Crossing Range (m).

Figure 15.22: Typical noise and lag optimisation

15.6.3. Tracking in Cartesian Space

One method of maintaining the noise performance of the system while minimising the
dynamic lag is to operate with a wide angle-servo bandwidth, and to perform the
tracking and smoothing in Cartesian space.
As there are no geometric accelerations in Cartesian space, it is possible to reduce the
filter bandwidth to less than 1Hz. This determines the noise performance of the
tracker.

479
_____________________________________________________________________
Because the angle servo bandwidth is wide (typically 10 to 100Hz for a real system),
the dynamic lag for geometric accelerations is limited.
A simplified block diagram showing the loop configuration is shown below.

Radar

R
dAz
dEl

x
Az Polar
y
to
El Cartesian z

Tracking
Filter
Bank

El Meas
Encoders Az Meas

El Command
Angle
Servos

Az Command

Figure 15.23: Tracking in Cartesian space

x
y
z

Cartesian
to
Polar

480
_____________________________________________________________________

15.7.Fire Control Radar Design

15.7.1. Requirements
Ships Motion

Radar on the deck of a ship 8m above the water

Roll +/-25 Pitch +/- 10 Yaw +/-5
Period about 5s

Designation

From a surveillance radar (one time per second)

Elevation accuracy +/-5, Azimuth accuracy +/-2
Range accuracy +/-25m
No velocity information

Environment

Up to sea state 5 (very rough, wave height 2.4 to 3.6m)

Rainfall up to 25mm/hr

Target Types

Fixed Wing Aircraft (frontal RCS 1m2 independent of frequency)

Aircraft Height > 60m
Sea Skimming Missiles (frontal RCS 0.1m2 independent of frequency)
Sea Skimmer height 3m

Detection Performance

Probability of detection Pd = 0.95

Probability of false alarm Pfa = 10-6
Detection time from receipt of designation (excluding slew) 0.5s
Detection Range
Table 15.2: Tracking performance
Weather
Clear Air
Rain 12.5mm/hr
Rain 25mm/hr

Aircraft
15km
10km
5km

Sea Skimmer
6km
4km
2km

Tracking Performance

Minimum tracking range Rmin = 50m

Range tracking accuracy (<1m RMS)
Angle tracking accuracy (<1mrad RMS)
Tracking aircraft directly overhead at h>100m
Aircraft and missile velocity <280m/s

Safety Constraints

Average Transmit Power <100W

481
_____________________________________________________________________

15.7.2. Selection of Polarisation

For low flying aircraft over the sea we want to minimise the sea clutter. In calm seas
at low grazing angles the difference between the reflectivity for horizontal
polarisation and vertical polarisation exceeds 12dB. As the roughness increases the
difference decreases.

Figure 15.24: Sea Clutter reflectivity as a function of grazing angle

The radar will operate using horizontal polarisation, and it will be assumed that the
surface reflectivity is 35dB.

15.7.3. Pedestal Specifications

The maximum combined roll and pitch angle is

max < 252 + 52 .

This does not exceed 30

482
_____________________________________________________________________
The antenna can be stabilised with respect to this
tilt adequately without resorting to a 3rd axis,
however, it can result in a significant rotation of
the polarisation
The pedestal will be of the type elevation,, over
azimuth, .

defined as +ve up from the horizontal

defined as +ve anticlockwise from the x-axis
For a mounting height 8m above the sea with the minimum tracking angle for a sea
skimmer at R = 50m ( = -6) and a combined roll and pitch angle of 30 requires a
minimum angle of -36 (min = -40).
To allow the antenna to track over the vertical, and to have time to slew around in
azimuth without losing lock at the maximum combined roll and pitch angle requires
max = 90+30 = 120 (use 125)

15.7.4. Radar Horizon

For hr = 8m and ht = 60m, the radar horizon is given by
d = 130( hr (km) + ht (km) = 43km

The radar horizon is not a consideration for this design

15.7.5. Selection of Frequency

A reasonable maximum diameter for the antenna on a shipboard radar is 1.5m.
The beamwidth at various frequencies will be as follows:
Table 15.3: Beamwidth for different frequencies
Frequency
(GHz)
10
35
94

Band
X
Ka
W

Beamwidth
(deg)
1.4
0.4
0.13

A narrow beam decreases the effect of multipath and limits the area of clutter within
the tracking gate, while attenuation increases with frequency (particularly in the rain)
The minimum elevation angle when tracking at target at h = 60m and R = 15km
= 0.19 and the minimum tracking angle when tracking a sea skimmer at h = 3m
and R = 50m = -6
Though it may be possible to minimise the effects of multipath, it is not possible to
eliminate them when the radar is looking down at the target.

483
_____________________________________________________________________
Direct Path

B
Target
hr
hr

A
Radar
ha

Indirect Path
M

Reflecting Surface

Image
Figure 15.25: Multipath propagation

Multipath has a major effect on tracking accuracy as shown by the following

measured data for two systems tracking the same target.

Figure 15.26: Effect of beamwidth on low angle tracking accuracy

The best compromise is to use the narrowest beamwidth possible and to use multipath
reduction techniques. According to the literature application of these techniques can
maintain an RMS elevation tracking accuracy of between 0.05 and 0.1 beamwidths
Assuming that we can only manage 0.1 beamwidths, then the tracking accuracy will
be:
Table 15.4: Tracking accuracy with multipath
Band
X
Ka
W

Beamwidth
(deg)
1.4
0.4
0.13

RMS Tracking
(mrad)
2.4
0.69
0.22

This excludes the X-band option as the tracking accuracy does not meet the accuracy
criteria defined earlier.

484
_____________________________________________________________________

15.7.6. Adverse Weather Effects

Typical attenuation as a function of frequency under different weather conditions is
shown below:
Table 15.5: Atmospheric attenuation
Band

Clear
(dB/km)

X
Ka
W

0.02
0.15
0.3

Rain
12.5mm/h
(dB/km)
0.25
3
7

Rain
25mm/h
(dB/km)
1
7
12

15.7.7. Required Single Pulse Signal to Noise Ratio

The required S/N to achieve the specified Pd and Pfa is a function of the target
distribution and its fluctuation characteristics. For a non-fluctuating target this is
13.6dB.

Figure 15.27: Detection probability curves

485
_____________________________________________________________________
Swerling 1: Many independent scatters of similar RCS. This results in slow
fluctuations with time.
Swerling 2: One major scatterer and many smaller scatterers. This is typical of an
aircraft nose on. This results in fast fluctuations with time.

Figure 15.28: Fluctuating target effects

The additional SNR required to achieve these probabilities of the target is fluctuating
is determined from this graph. It will be 10.4dB. The single pulse SNR required is
thus 13.6+10.4 = 24.0dB.

15.7.8. Tracking Gate Size

Ideally, the tracking gate size is matched to the target length
A typical Jet fighter would be 15m long, so a gate size between 15 and 20m would be
ideal. We will use 20m.
This minimises the amount of clutter received, it does however complicate the target
acquisition process as the designation accuracy is only +/-25m, so at least 3 gates
would be required to span the uncertainty

15.7.9. Signal to Clutter

We assume that the reflectivity of the sea is the same at X, Ka and W bands
We assume that a similar SCR is required as SNR
For a gate size of 20m, the illuminated area will be a function of the antenna
beamwidth

486
_____________________________________________________________________
Table 15.6: Illuminated sea area as a function of frequency
Band

Beamwidth
(deg)
1.4
0.4
0.13

X
Ka
W

Area (6km)
(m2)
2928
840
264

Area (15km)
(m2)
7320
2100
660

For an average reflectivity = -35dB in all cases, the clutter RCS will be
Table 15.7: Clutter RCS as a function of frequency
Band
X
Ka
W

RCS (6km)
(dBm2)
-0.33
-5.8
-10.8

RCS (15km)
(dBm2)
3.6
-1.8
-6.8

For an aircraft RCS of 1m2 (0dBm2) or a sea skimmer with an RCS of 0.1m2
(-10dBm2), we are looking for an extra 20 to 30dB of signal to achieve the required
SCR of 24dB (for the required single pulse detection probability).
Integration improvement cannot be used for improving the SCR because, unlike the
white thermal noise used for target detection, clutter is correlated, so integration will
not be as effective.
The primary difference between the targets and the clutter is that generally the target
has a significant radial velocity and the clutter is static (or slow moving in high seas).

15.7.10. Moving Target Indicator

Some form of MTI will have to be implemented that can achieve a sub-clutter
visibility as follows
Table 15.8: Subclutter visibility requirements sea skimmer
Band
X
Ka
W

Clutter RCS
At 6km (dBm2)
-0.33
-5.8
-10.8

Sea skimmer RCS

(dBm2)
-10
-10
-10

Subclutter Visibility
For SCR=24dB
33.67
28.2
23.2

Table 15.9: Subclutter visibility requirements aircraft

Band
X
Ka
W

Clutter RCS
At 15km (dBm2)
3.6
-1.8
-6.8

Aircraft RCS
(dBm2)
0
0
0

Subclutter Visibility
For SCR=24dB
27.6
22.2
17.2

This can be achieved using a delay-line canceller, however because the ship will be
moving, the clutter will also have an effective velocity, so this is not a good technique
to use.

487
_____________________________________________________________________
A better alternative is to take more samples, to window them and to use an FFT to
isolate moving from static targets.
With a Hamming window, a static return rejection of 42dB can be obtained. However,
it is unlikely that if the measurement is unambiguous in range, that it will be
unambiguous in velocity.

15.7.11. The Pulse Repetition Frequency

To be unambiguous in range out to 15km, the maximum allowed PRF is calculated as
PRF = c/2Rmax = 10kHz
Using the Nyquist criterion, this is unambiguous in velocity up to a Doppler
frequency of 5kHz
The maximum unambiguous radial velocity is given by the following formula:
vr =

f d PRF .
=
2
4
Table 15.10: Unambiguous velocity as a function of frequency
Band
X
Ka
W

Unambiguous
Velocity vr (m/s)
75
21.5
8

All velocities will be folded down into this range, so it is possible that the folded
target velocity will fall directly into the bin (or bins) containing the clutter.
This can be ameliorated by changing the PRF on a block by block basis, which will
shift the relative positions of the clutter and the folded target Doppler frequencies
until an one is found in which the target is not in the same bin as the clutter.
It is quite likely that the clutter velocity spread will exceed 8m/s, so there may not be
an acceptable PRF at W-Band.
Because both the clutter and the ship may be moving, the actual gate in which the
former will be found must be tracked.

Figure 15.29: Separating target and clutter returns in a Doppler filter bank

488
_____________________________________________________________________

15.7.12. Search Requirement

Need to search a volume 10 4 50m in less than 0.5s to meet the specification.
Any of the following patterns can be adopted

(a)

(b)

(c)

Figure 15.30: Search pattern options

Option (a) requires that the pedestal changes direction more often than option (b), and
there is more of an overlap between scans with option (c), so adopt option (b)
For a 50% overlap between scans, the total distance that must be travelled to search
each area is determined from the number of vertical scans required to cover the 4 in
azimuth.
Table 15.11: Total search scan distance as a function of frequency
Band

Vertical Scans (n)

X
Ka
W

4/1.4x1.5=4.28 [5]
4/0.4x1.5=15 [15]
4/0.13x1.5=46 [46]

Total Distance
(deg)
50
150
460

Scan Speed
(deg/s)
100
300
920

Hits per scan (the number of pulses that illuminate the target as the beam passes over
it) is determined from the required scan rate and the antenna beamwidth.
Table 15.12: Hits per scan
Band
X
Ka
W

Hits Per Scan

(N)
140
13.3
1.4

There are not sufficient hits at either Ka band or W band to generate an FFT with
sufficient bins
Decreasing the overlap to 0%

489
_____________________________________________________________________
Table 15.13: Scan distance and hits per scan for 0.5s detection time
Band
X
Ka
W

Vertical
Scans
(n)
3
10
31

Total Distance
(deg)

Scan Speed
(deg/s)

Hits per Scan

(N)

30
100
310

60
200
620

233
20
2.1

There is the added burden that the target may have moved up to 140m during the
search, so a large bank of gates spanning the original 50m uncertainty plus 140m on
each side (330m) must be examined.
At least 17 gates will be required to span this range.
It is now feasible to use a 16 point FFT at Ka band or a 128 point FFT at X-band.

15.7.13. Integration Gain

The FFT process is the equivalent of a coherent integrator, which will produce gains
of 10log10(N). However, because the target is fluctuating, this effect will vary with the
observation time compared to the fluctuation rate and the probability of detection.
We use the following graph (which is not quite correct as it assumes post detection
integration).
For a Pd 0.95 (we use 0.9) and a Swerling 2 target, the integration gain is about
17dB for the 16 point FFT at Ka band.
For the 128 point FFT at X band, the integration gain is about 24dB.
Though these are higher than the theoretical maxima of 12 and 21dB respectively, the
extra is taken care of in the 10.4dB required for detection described earlier.

Figure 15.31: Integration improvement factor

490
_____________________________________________________________________

15.7.14. Matched Filter

We assume that the pulse is rectangular and that the filter is made up of 5 cascaded
tuned bandpass sections
For a pulse width of 20m (133ns), the optimum bandwidth = 5MHz
Table 15.14: Matched filter characteristics

Input Signal

Filter

Optimum .

Rectangular Pulse
Rectangular Pulse

Single tuned circuit

Two cascaded tuned
circuits
Five cascaded tuned
circuits

0.4
0.613

Loss in SNR compared

to Matched Filter (dB)
0.88
0.56

0.672

0.5

Rectangular Pulse

15.7.15. Transmitter Power

The average power allowed is 100W, the pulse width is 133ns and the PRF is 10kHz
making the duty cycle 0.133% and so the peak transmitted power Pt = 75kW.
This can be achieved using a Magnetron or Travelling Wave Tube at Ka or X band

15.7.16. System Configuration

A Magnetron can be used in a (pseudo) coherent radar if the LO is primed by the
random start phase of the magnetron on every pulse

Figure 15.32: Schematic diagram of a pseudo coherent pulsed radar

491
_____________________________________________________________________

15.7.17. Free Space Detection Range

Applying the radar range equation when there is significant attenuation is best
achieved using MATLAB
Various losses are listed in the M-file
% Fire Control Radar
r=(10:10:75000);
% Range (m)
sigmat = 1;
% Aircraft (sqm)
sigmas = 0.1; % Sea skimmer (sqm)
dant = 1.5;
% Antenna Diameter (m)
pt = 75e03;
% Trasmit power (W)
lamx = 30e-03;
% Wavelength (m)
lamk = 8.6e-03;
nfxdb = 3;
% Noise fig (dB)
nfkdb = 4.5;
tau = 133e-09;
% Pulse width (s)
k = 1.38e-23; % Boltzmann (Js)
t = 290;
nintxdb = 24; % Integ i mprove(dB)
nintkdb = 17;
ltxx = 3;
% Tx plumbing loss (dB)
ltxk = 4;
lscanx = 3.2; % 2D Scan loss (dB)
lscank = 3.2;
lmiscx = 3;
% Misc loss (dB)
lmisck = 3;
alphax = 1;
% Atmos attn (dB/km) 0.02, 0.25, 1
alphak = 7;
% cREAR, 12.5, 25mm/hr 0.15, 3, 7
SNx = 13.6+10.4;
SNk = 13.6+10.4;

% S/N for Pd=0.95, Pfa=1.0e-06

% Calculate the antenna gain

gxdb = 10*log10(4*pi*0.7*pi*dant*dant/(4*lamx*lamx));
gkdb = 10*log10(4*pi*0.7*pi*dant*dant/(4*lamk*lamk));
% Propagation factor
propxdb = 10*log10(lamx*lamx/(4*pi).^3);
propkdb = 10*log10(lamk*lamk/(4*pi).^3);
% Transmit power
ptdb = 10*log10(pt);
% Target RCS
sigmatdb = 10*log10(sigmat);
sigmasdb = 10*log10(sigmas);

% Aircraft
% Sea skimmer

prxdb
=
ptdb+2*gxdb+propxdb+sigmasdb-ltxx-lscanx-lmiscx-40*log10(r)2*alphax*r/1000;
prkdb
=
ptdb+2*gkdb+propkdb+sigmasdb-ltxk-lscank-lmisck-40*log10(r)2*alphak*r/1000;
% Threshold for detection
thermxdb = 10*log10(k*t*0.672/tau);
thermkdb = 10*log10(k*t*0.672/tau);
threshxdb = (SNx + thermxdb + nfxdb -nintxdb)*ones(size(r);
threshkdb = (SNk + thermkdb + nfkdb -nintkdb)*ones(size(r);
semilogx(r,prxdb-threshxdb,r,prkdb-threshkdb);
grid
title('FIRE CONTROL RADAR: MAGNETRON: SKIMMER: Clear Air')
xlabel('RANGE (m)')
ylabel('SIGNAL LEVEL ABOVE Smin (dB)')

492
_____________________________________________________________________
axis([0,100000,-50,200]);
lx=find((prxdb-threshxdb)<0);
lk=find((prkdb-threshkdb)<0);
min(r(lx))
min(r(lk))

Figure 15.33: Radar performance, clear air

Figure 15.34: Radar performance, 12.5mm/h rainfall

493
_____________________________________________________________________
Table 15.15: System performance
Target

Condition
Clear
12.5mm/h

X-Band Range
(km)
58
28.9

Ka-Band Range
(km)
37.5
6.8

Aircraft
RCS=1sqm
Sea Skimmer
RCS=0.1sqm

25mm/h
Clear
12.5mm/h
25mm/h

13.7
34.4
20.6
10.8

3.6
25.8
5.7
3.1

For aircraft detection, only the X-Band radar meets the specified range criteria of 15,
10 and 5km
Both the X-Band and the Ka-Band radars meet the sea skimmer detection
requirements of 6, 4 and 2km

15.7.18. Effects of Multipath on Aircraft Detection

The following graphs show that the constructive and destructive interference caused
by multipath can reduce the signal level Smin to below 0dB under some circumstances.
The fades are more pronounced at X-Band where detection may not occur between 13
and 14km, and again at just below 10km.
None of the fades at Ka-Band dip so low.

Figure 15.35: Multipath effects on aircraft tracking at X-band

494
_____________________________________________________________________

Figure 15.36: Multipath effects on aircraft tracking at Ka-band

Figure 15.37: Multipath effects on sea skimmer tracking at X-band

495
_____________________________________________________________________

Figure 15.38: Multipath effects on sea skimmer tracking at Ka-band

15.7.19. Detection Threshold and CFAR

The false alarm rate is very sensitive to the setting of the detection threshold voltage.
Because no velocity information is available, the radar must transmit a block of pulses
and look in all the Doppler gates in all of the range gates over the designated range of
the target.
Probability Density

0.8

0.6

Rayleigh
Noise
Detection
Threshold

Ricean
Signal + Noise
SNR = 15dB

0.4

0.2

12

14

Amplitude (V)

Figure 15.39: Selection of detection threshold

Changes in radar characteristics with time (ageing) and changes in the target
background characteristics mean that a fixed detection threshold is not practical. A
Constant False Alarm Rate Processor (CFAR) is required

496
_____________________________________________________________________
We propose a cell-averaging CFAR processor that averages the returns from a
particular Doppler bin across all of the range gates that span the designated range as
shown in the highlighted row.
This averaging process excludes the cell under test which proceeds sequentially from
gate 1 to gate N (it is shown in gate 3 in the Figure)
Range gates
D
o
p
p
le
r

B
i
n
s

1
1
2
3

125
126
127
128
Figure 15.40: Range/Doppler gate matrix showing cell containing target

15.7.20. Transition to Track

The mechanically operated FCR antenna moving at 60/s (X-Band) or 200/s (KaBand) during the search phase is unable to stop while the antenna is still pointing at
the target.
The angles, range and Doppler bin at which detection takes place is recorded as the
antenna sweeps past and the antenna returns to that designation more slowly.
If the target is not detected, then a slow search is conducted in angles while the range
gates broaden their search until a detection occurs or the system times out. The
angular rates of this new search are such that the antenna can stop while still
illuminating the target, and a transition to track mode is made.
The process of detection involves closing the Doppler loop, followed by the range
loop and finally the angle loop.

15.7.21. Target Tracking

A split tracking gate straddles the sum-channel range-gate in which the detection took
place and early and late gate measurements are made only in the Doppler bin in which
detection took place.
As the Doppler is generally ambiguous, it cannot be used to prime the range filter, so
a long (typically 50ms or so) period is used to obtain a good range rate estimate to
prime the range tracking filter.

497
_____________________________________________________________________
The range tracking loop closes using early and late gate signals from the appropriate
Doppler bin, and automatic tracking in range occurs using the difference between the
early and late gate signals to keep the gates centred on the target
Doppler Bank
Sample
&
Hold

ADC

2N

Range

L-E
D R=K-------L+E

Trigger
early
Sample
&
Hold

D R

late
Radar
Video

1
2
3

ADC

2N

1
2
3

Range
Tracker

Estimated
Range
Estimated
Rate

Normalisation
Function

Doppler Bank

Figure 15.41: Schematic diagram of range tracking configuration

The angle tracking loops are closed.

The azimuth and elevation error signals sampled simultaneously with the sum channel
signal drive the angle servos to keep the antenna pointed at the target being tracked.
Target dynamics that dictate the real and apparent acceleration and tracking loop
bandwidth determines the tracking accuracy.

Figure 15.42: Crossing target geometry

498
_____________________________________________________________________
When viewed in radar polar co-ordinates, the target velocities will be
v
v
R& max = vt , A& max = t , E& max = t .
R
R

The real accelerations are

&& = at , E&& = at .
&& = a , A
R
max
max
max
t
R
R

The geometric accelerations and other derivatives are shown below. They are
normalised in the angular co-ordinates to A& max = m = vt / Rc . Where Rc is the ground
range at crossover.
In range, the derivatives are normalised to Ra = Rmin, at the point of closest approach.
Assuming that the target is travelling at 280m/s and flies past the radar at a height of
60m and a ground range of Rc = 60m, then the peak velocities and accelerations
experienced (in polar co-ordinates)

m = 4.66rad/s & a = 13.75 rad/s2 (from the graph)

e = 1.6 rad/s & e = 10.42 rad/s2
&& = 923 m/s2
R& max = 280m/s R
max

Assuming that thermal noise and dynamic lag are the primary sources of error, then
an optimum bandwidth is calculated as follows:
at2 k m2 f r B ( S / N )
Bn =

1.6 32 R 2

1/ 5

where: at acceleration (real or geometric) (m/s2)

km Monopulse Gain Constant (typ 1.6)
fr Pulse repetition frequency (Hz)
B - IF bandwidth and pulse width
S/N pulse signal to noise ratio
3 Antenna beamwidth (rad)
R Crossing Range (m)

[923]
[1.6]
[10kHz]
[0.672]
[10]
[0.024 or 0.007]
[85]

The optimum angle servo bandwidth is 116Hz for the X-Band tracker and will be
significantly higher for the Ka-Band unit. This is very high as the specified crossing
range is extremely short.

499
_____________________________________________________________________
If the bandwidth is expressed in terms of the equivalent noise bandwidth Bn then the
acceleration coefficient
Ka =2.5Bn2 = 0.63/to2

Where to is the equivalent averaging time for the tracking loop. The lag will be.

a =

& a
Ka

& a
2 .5 B

2
n

13.75
= 0.409 10 3 rad
2
2.5 116

3
k m 2( S / N ) f r / Bn

0.024
1.6 2 10 10 / 116
4

= 0.36 10 3 rad

The RMS tracking error caused by the thermal noise will be = 0.36mrad
The RMS tracking error caused by dynamic lag will be a = 0.41mrad.
The alternative is to use a wide angle-servo bandwidth (200Hz) that will cope with
the geometric accelerations, and to perform the filtering in Cartesian space with a
bandwidth of between 1 and 4Hz that will cater for real target accelerations that
should not exceed 6g for the aircraft.

Radar

R
dAz
dEl

x
Az Polar
y
to
El Cartesian z

Tracking
Filter
Bank

x
y
z

Cartesian
to
Polar

El Meas
Encoders Az Meas

El Command
Angle
Servos

Az Command

Figure 15.43: Tracking in Cartesian space

15.8.References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]

M.Skolnik, Radar handbook, McGraw-Hill, 1970.

M.Skolnik, Introduction to Radar Systems, McGraw-Hill, 1980.
P.Bhartia, I.Bahl, Millimeter Wave Engineering and Applications, John Wiley & Sons, 1984
D.Barton, Radar Systems Analysis, Artech 1976.
C.Currie (ed), Principles and Applications of Millimeter Wave Radar, Artech, 1987.
G. Morris, Airborne Pulsed Doppler Radar, Artech, 1988.
N. Currie (ed), Radar Reflectivity Measurement: Techniques & Applications, Artech, 1989.
S. Sherman, Monopulse Principles and Techniques, Artech
M. Spiegel, Mathematical Handbook of Formulas and Tables, McGraw Hill, 1968.

500
_____________________________________________________________________