Qinetiq Report On Geolocation

q
CONDITIONS OF SUPPLY
This document is supplied by QinetiQ for Ofcom under Contract No. C31400/008
Ofcom AMS
Final report
QinetiQ; PR Edmonds , B Ashforth, DI Elliner & S Madsen

TRL; D Harvey, R Allan, S Clover
ARUP; B Laidlaw, K Kilfedder
QINETIQ/D&TS/SS/CR0600316
Publication number: QINETIQ/06/00039
July 2006
'This report was commissioned by Ofcom to provide an independent view on issues relevant to
its duties as regulator for the UK communication industry, for example issues of future
technology or efficient use of the radio spectrum in the United Kingdom. The assumptions,
conclusions and recommendations expressed in this report are entirely those of the contractors
and should not be attributed to Ofcom.'
q ARUP
Copyright © QinetiQ ltd 2006
Administration page
Customer Information
Customer reference number 830000/083
Project title OFCOM AMS System
Customer Organisation OFCOM
Customer contact Dr Christos Politis
Contract number C31400/008
Authors
QinetiQ Ltd: Paul Edmonds, David Elliner, Sue Madsen, Ben Ashforth
TRL: Donald Harvey, Roger Allen Steve Clover,
ARUP Communications: Bruce Laidlaw, Ken Kilfedder
Release Authority
Name Giles Bond
Post Research Group leader
Date of issue 07/07/2006
Record of changes
Issue Date Detail of Changes
0.1 17/01/06 First draft delivered as pdf
0.2 18/01/06 Second draft delivered as .doc
1.0 12/5/06 First issue
2.0 15/06/06 Second issue
3.0 07/07/06 Issue for publication on Ofcom website
QINETIQ/06/00039 Page 2
Executive Summary
This report has been prepared for Ofcom under contract C31400/008 by QinetiQ with
contributions from its partners TRL Technology Ltd and Arup Communications
through the 2005-2006 Spectrum Efficiency Scheme (SES) initiative.
Introduction
The radio spectrum is an increasingly important national asset, and so efficient

management of it is a high priority. Good spectrum management has direct benefits
in terms of improved communications, but also in indirect terms as it supports a vital
part of the country’s infrastructure. To understand spectrum usage, Ofcom
commissioned a study into the design of an Unattended Monitoring Station (AMS)
that could be deployed in large numbers to automatically detect, identify and locate
the source of interfering radio signals over a large part of the UK.
Ofcom currently have two programmes which utilise unattended equipment for
spectrum management and interference resolution purposes. These are a remote
monitoring and direction finding system and an unattended monitoring system. Both
are deployed at a limited number of sites and cover just a small proportion of the UK.
A fully comprehensive network will need far more monitoring stations (original
estimates were 2000 but work carried out during this study suggest that as few as
1200 could provide nationwide detection and location of handheld PMRS or 982 for
location of 3G base stations). Such stations would need to be automatic in
operation, have means of downloading appropriate reports and be relatively cheap to
purchase and maintain.
It is expected that an AMS network would have the following benefits:
• Enhanced scope of spectrum monitoring, both geographical and transmission

types,
• Greater precision in identifying and locating sources of interference,
• Improved responsiveness to interference problems,
• More detailed information on spectrum usage,
• More detailed information on interference patterns.
Project Scope
The approach to the study was to begin with a literature search that built on the
consortium’s existing knowledge of the state-of-the-art in spectrum monitoring. In
parallel with this work, Arup Communications carried out a business case study by
assessing the economic aspects of a deployed AMS system and they completed a
cost-benefit analysis. Following this, the system approach was analysed using
recognised propagation models and geolocation prediction tools to establish the
likely performance of the system. Information from the modelling was fed back to
the business case. Requirements for a final AMS system were assessed and a
design specification created.
Concurrently with all these activities, a ‘Proof of Principle’ prototype system was built
and demonstrated. TRL led on the hardware build, and the software was developed
by QinetiQ based upon previously fielded proprietary products. The purpose of the
demonstrator was to de-risk the final system design, demonstrate the concept and to
allow trials to take place to evaluate the efficacy of the approach for geolocation in
particular.
This final report details the findings of the various work packages associated with
this programme. The structure of the document is as follows:
Section 1. Introduction.
Section 2. A review of state-of the-art interference monitoring products.
Section 3. The business case study, conducted by Arup Communications. (This
has also been supplied to Ofcom as a stand alone document).
Section 4. The results from applying QinetiQ modelling tools to predict the
geolocation performance of the system using time-difference-of-arrival (TDOA).
Section 5. A description of the design considerations for a final deployed system.
Section 6. An overview of the design of the demonstrator units.
Section 7. Results of the trials carried out with the demonstrator units.
Section 8. Conclusions, including further requirements to improve performance for a
fully deployed network.
Review of current state-of-the-art automatic interference monitoring

equipment.
Work Package 1 required a literature review of automatic interference monitoring

equipment with the aim of determining the current state of the art.
The main search took place between February and March 2005 and a full report [1]
was delivered to Ofcom. The consortium reviewed the collated evidence against the
requirements of an automatic monitoring system that would meet Ofcom’s needs
both technically and in terms of cost. A watching brief has been maintained in the
interim.
The AMS would be expected to identify unauthorised radio transmissions without
operator intervention with a minimum of false alarms. This will require the system to
determine the form of modulation in use and to access a database of licenced users.
The requirement implicitly calls for an automatic monitoring system that is equipped
with a wideband, fast tuning receiver, both to accommodate the widest bandwidth
signals and to achieve the spectrum revisit time. This receiver will also ensure a high
probability of intercept against fleeting or intermittent interference. The specified
frequency band is 20MHz to 3GHz and there are potentially a very high number of
signals simultaneously present.
Conventionally, low-cost receivers with acceptable intermodulation performance are
narrowband superheterodyne types. However, both scan rate and instantaneous
bandwidth could be well below what is required for this application. There are
several manufacturers of wideband tuners of varying size, cost, and fidelity.
Unsurprisingly the higher performance models are generally more expensive.
All of the automatic monitoring systems that were reviewed use signal location
entirely based on direction finding (DF) techniques. These rely upon multi-element
antenna arrays which are generally large and their profiles may not meet planning
authority requirements for Ofcom’s application. No systems were found which use
the Time Difference of Arrival (TDOA) location technique.
The monitoring capability of the commercially available automatic monitoring
systems is generally provided by off-the-shelf spectrum analysers, rather than
digitisers. While these instruments offer wide operational bandwidths, the ability of
the monitoring system to intercept, analyse and classify transitory or sporadic
interference is limited. These off-the-shelf instruments are also designed for
laboratory use and are generally bulky and expensive assets.
The survey also revealed that the commercially available automatic monitoring
systems:
• have a limited capability against modern signal types such as time and code
division multiple access (TDMA and CDMA) where the signals can effectively be
stacked on top of each other;
• exhibit particularly poor performance in multipath environments which limits their

use in urban or metropolitan areas;
• cannot cope with co-channel signals and have limited classification capabilities in
dense signal environments;
• will not support fast scan rates;
• require high signal levels (meaning limited receiver dynamic range), which limits
operational range of the system.
• are based on large, calibrated antenna arrays, laboratory spectrum analysers and
add-on signal classifiers that makes the systems high cost, high value assets.
An examination of fielded systems in Australia and New Zealand was carried out as
these countries have had spectrum trading arrangements in place since 1997 and
1989 respectively. However trades are not taking place regularly enough for them to
have had any impact on the spectrum monitoring role performed by the regulators
and many of the trades which have taken place have been as a result of mergers
and acquisitions, which have not complicated monitoring requirements at all. New
Zealand completed their network of Spectrum Monitoring equipment, including fixed
stations and vehicles in early 2003, but the system has not yet been used to support
or monitor spectrum trades.
Finally, very few of the systems that were reviewed offered any integration with
existing databases of license holders and license types. This would be a useful
feature as it will help operators to determine quickly whether particular transmissions
are licensed or not.
The conclusion of the study was that there is a wide range of automatic monitoring
systems currently available or in use on a global basis. However, there are currently
no systems that are capable of fully meeting the Ofcom requirements.
Identification and Analysis of System Benefits
A study was carried out by Arup Communications [2] to assess the costs of
introducing new equipment and an assessment of the likely benefits gained. In
performing this study, Arup Communications received input from the other
consortium parties and also had discussions with Ofcom to get an understanding of
the current cost model.
The AMS could be deployed to cover a greater geographic area than the current
spectrum monitoring system and is fully effective up to 3GHz. Thus it encompasses
high-value transmissions such as GSM and UMTS mobile telephony, broadcast
television, and potentially the IMS bands.
The most obvious beneficiary of the AMS system will be Ofcom itself. Additionally
however, if Ofcom’s spectrum management duties can be carried out more
effectively, all spectrum users will benefit. Agencies such as the Police and
Coastguard will have better information with which to pursue a prosecution and
some spectrum users such as broadcasters and mobile telephony operators will also
benefit from certain summary information that could be made available from the
system, either in the form of regular static reports, or by providing direct access to
the system itself. A further tangible benefit will be help in establishing a spectrum
trading mechanism, by which spectrum holders may sell or rent rights to particular
bands to other organisations. With access to good information about current
spectrum use the value of the spectrum could be more accurately determined; and
underused spectrum could be easily identified by purchasers.
The costs of commissioning and operating an effective and inclusive AMS system
are significant and it would represent a clear step-change from the current UMS and
DF operations undertaken from Baldock. Recommendations have been put forward
as to how these costs could be financed, and where potential savings can be made
from partnerships with site or infrastructure owners.
However, there are huge, less easily quantified benefits of an operational and fully-
featured AMS system. In principle, if a very modest efficiency gain of 1% in the use
of the spectrum could be attained, that would be worth some £250 million every year.
If benefits of this scale could be captured, whether by Ofcom itself or by a
commercial enterprise, then a national AMS system could be deployed on a self-
sustaining basis.
Modelling Study
A modelling study of the proposed AMS formed Work Package 3. The aim was to
determine the location accuracy of the interfering signal and the detection range of
the AMS allowing a more refined estimate of the likely number of stations. Two
location techniques were considered, direction finding (DF) based on multi-element
arrays and Time Difference of Arrival (TDOA) – the latter method formed the greater
part of the study as it was quickly shown that it provides greater accuracy and
sensitivity than DF, and importantly met Ofcom’s requirement for low profile
antennas.
TDOA is based on the principle that the distance a signal has travelled is related by
its velocity to the time taken. If the signal is simultaneously intercepted by two
sensors, the difference between the time-of-arrival of the signal at the two sensor
positions can be used to estimate the relative range of the emission source to the
two sensors. A line-of-position can then be generated to represent the locus of
geographical points that can satisfy the measured time difference.
A model has been developed to determine the range at which a signal could be
detected and the accuracy to which it could be located using a TDOA based system.
The principle outputs of the modelling are a plot of geolocation error for a regular grid
of points within a defined rectangular region, and an indication as to the range over
which the signal can be detected.
The model was used to estimate the number of monitoring stations which would be
required to provide monitoring coverage of the whole United Kingdom. The country
was divided into types from ‘remote rural’ to ‘dense urban’ according to data
provided by Arup and a number of exemplar transmitter types were used as test
cases for calculating detection ranges. Once predicted propagation losses had been
calculated it was possible to estimate the required station density for each
transmitter type and area. The results of the exercise showed a significant diversity
in the number of required monitoring stations to cover the country for different
transmitter types ranging from 13 for pager base stations to 26,727 for GSM mobiles.
A sensible compromise would be to use between 950 and 1200 based on the
estimates for 3G base stations and personal mobile radio results.
The model was also used to show the relationship between the amount of data used
in the estimation of TDOA and the achievable geolocation precision. These results
were to assist in determining the requirement for communications bandwidth
between the stations in order to transfer sufficient data for TDOA calculation.
Finally a comparison with DF accuracy was undertaken. Provided the three sensors
used for TDOA surround an emitter, the geolocation error does not vary significantly
with the distance from the emitter to the sensors. For direction finding however, there
is a constant angular error so the calculated position error scales with distance from
the sensor.
AMS System Design
The purpose of Work Package 4 was to design an unattended monitoring system

whose primary purpose is to automatically detect, identify and located the source of
interfering radio signals over a large part of the UK.
The requirements called for a solution comprising of a large quantity of monitoring
stations, which are networked for control, data collection, and reporting purposes.
The system should be capable of fully automatic operation or manual control. It
should perform routine monitoring of the spectrum to establish occupancy and locate
the source of emissions potentially providing Ofcom with a means of validating a
database of licensed user activity.
Based on the findings of the modelling study the system will use TDOA as the
location technique.
All functionality will be accommodated in a single environmentally sealed unit that
will be mounted close to the antennas in an external location. The power supply will
be in a separate unit located inside a building or protective structure.
The design considers options for the various functional elements including receivers,
digitisers, timing, processors and antennas and environmental constraints such as
cooling and heat dissipation.
Software for spectrum monitoring and signal location has been developed by QinetiQ
over the last ten years, and it is recommended that the software for the final system
is based on this capability. The following functions have been identified as essential
to the system:
• Determine the presence of emissions in the environment and then relate them to
emitters using matched filtering techniques. This allows the production on
network structure for a frequency and very advanced automation of location and
modulation classification assets.
• Display sensor results,
• Log sensor results to hard disk at regular intervals,
• Manage the transfer of the information between sensors to optimally achieve
location of all emitters,
• Stream data between sensors to locate all emissions (subject to communications
bandwidth available between the units),
• Calculate position fixes based on TDOA techniques based on information from
multiple sensors,
• Show position fixes on a digital map,
• Report results dependent upon pre-set filters.
Communication between sensors will be achieved through use of the internet. The
system will be capable of providing full control of remote sensors from either regional
or national control centres. A network hierarchy for data management is essential to
prevent the maximum available bandwidth from being exceeded for the control
centres.
Hardware and Software development of the demonstrator system.
A key part of AMS project was the design and build of three demonstration units to
de-risk key technologies that will be required for the final deployed system and to
help in identifying the benefits of novel techniques to address the AMS problem. The
system is designed as a software based architecture. That is, the hardware is
generic and will support a number of different functions, and the intelligence is in the
software. This gives the flexibility to upgrade or enhance the capability simply by
downloading a new file rather than having to change hardware.
As per the final design, a one unit design was employed. For portability and
convenience the systems have been built into transit case type 19” rack units.
Commercial-off-the-shelf (COTS) components have been used where possible to
reduce development costs. The system consists of two main units: a rack mounted
PC, and a custom “RF Tray” designed and built by TRL.
The software installed on the demonstrator system is a prototype of that proposed
for the final system and has been developed at QinetiQ over approximately 6 years.
Variants of this are used by a number of customers. The challenge was to create a
system that would integrate with relatively low cost hardware. The basic capabilities
of the application are:
• Real-Time Display – an animated real-time window displays when a connection
is made to a remote sensor. This shows a real-time spectrum display along with
detected emissions.
• Historical Display - Within this view each emission is represented as a line on a
time versus frequency plot. The colour of the line indicates the received signal
amplitude and the length of the line indicates the duration of the emission. A real-
time and historical view can be displayed for every sensor the system is
connected to.
• Maps - Maps are used both for sensor management (identifying locations of
sensors, and as a means of connecting to sensors) and for displaying TDOA
position fix results.
• TDOA position fixing - The AMS system supports position fixing by TDOA. When
requested to locate an emission the requesting sensor and its two nearest
neighbours are identified and tasked to simultaneously capture the signal. This is
transferred back to a central point, correlated and mapped onto the ground to
give a location.
• Filtering – By using filters it is possible for the operator to focus on signals of
interest. Selection can be made on a number of criteria including signal
parameters such as signal name or frequency, or geographically to select only
signals coming from a selected area.
• Modulation classification – a generic blind classification algorithm is used to
classify the modulation type of detected signals.
Sensor performance and trials of the AMS
Initial trials were carried out at Malvern and subsequently at Ofcom’s HQ in London
to measure the sensitivity of the sensors, in particular to assess their performance in
a densely populated signal environment compared with that in more rural areas.
Further trials have been carried out to test the communication across a networked
demonstrator system and to assess the effectiveness of the AMS in being able to
pinpoint and identify RF transmission sources in real time by TDOA.
The London trials showed that signal levels in urban environments can greatly
exceed the expected -40dBm at the antenna. Measurements showed a number of
signals at the -20dBm level with maximum levels around -10dBm. Receivers used
in the system must be able to handle this power whilst maintaining sufficient
sensitivity to detect low power signals.
Communication between the sensors is essential. Without a communication link the
sensors can still perform spectrum monitoring individually, but they are unable to
work co-operatively. To calculate locations a sensor will task its neighbours via the
communications link and data will be returned back to the tasking sensor for location
processing. The results from all sensors will be complied through a network
hierarchy for access from the central control station. The communication link also
allows access to real time or historical data from any of the sensors in the network.
The demonstrator was designed to operate with 3G mobile phone data cards, but the
service at the time in the Malvern area proved to be too unreliable. Instead the
system was further developed to operate with a broadband connection. Whilst this
introduced time delays into the project the result has been a much more flexible
solution which should work with almost any type of internet connection.
The disadvantage of using the internet to network was that it limited the number of
available sensor sites available for trials. Initially, limited testing was carried out
using employees’ homes but these were often not in optimal locations and the
sensors were re-located to more realistic separations in Malvern, Tewkesbury and
Pershore which forms a triangular configuration spaced about 17 km apart.
This enabled full testing of the TDOA capability. Multiple captures were taken against
a number of signals to allow a good statistical analysis of the results. Most of the
targets were FM broadcast radio stations but also pager transmitters, personal
mobile radios and mobile phone base stations were used.
Results from the trials generally showed good consistency with the predictions of the
modelling exercise. The choice of available targets within the “good” area enclosed
within the triangle of the sensors was very limited, so many of the test emitters were
in areas where an accurate location would not be expected. The emitters that were
close to the triangle of sensors were located to within 150-300m and whilst the
accuracy was less further away from the triangle and the geometry of the sensors in
relation to the target was poor, in most cases it would still be possible to calculate an
accurate direction fix.
In a full nationwide networked system it should always be possible to surround a
target by three sensors so the level of accuracy will be more consistent and it should
be possible to position fix most emitters of the type studied to within 100 – 200m
using a bandwidth of 100kHz.
The trials were limited by the availability of transmitters in suitable locations. It is
recommended that further trials be performed using transmitters under Ofcom
control. Testing in alternative environments and with different sensor spacing would
provide comparative data and will further enhance our understanding of the
capability of the system.
Conclusions
The project therefore has concluded that whilst there are a number of automatic
monitoring systems currently available they would not fulfil the full requirements of
the system that Ofcom wish to implement. Modelling of the TDOA method of
geolocation has confirmed that this is an appropriate approach and that it has many
advantages over the more traditional DF method based on multi-element arrays. A
demonstrator system has been built and trialled and the results were consistent with
expectations from the modelling. This has de-risked elements of a final design and
provided confidence that the design would meet the requirements that have currently
been identified. Finally the cost/benefit analysis carried out by Arup has shown that
whilst the cost of implementing a nationwide system are high, so too are the benefits
and there are a number of options for financing both the installation and running
costs of an AMS system.
List of contents
1 Introduction 17
1.1 Background 17
1.2 An Automatic Monitoring System 18
1.3 The Consortium 19
1.4 Approach 20
1.5 Report Overview 21
2 Review of current state-of-the-art automatic interference monitoring equipment
– (Work Package 1) 22
2.1 Introduction 22
2.1.1 Scope 22
2.1.2 Review of current state of the art automatic interference monitoring
systems 23
2.2 Commercially available systems 23
2.2.1 Literature search summary 24
2.2.2 Equipment summary 24
2.2.3 Internet links 29
2.3 Press Releases 29
2.3.1 Conatel Signs $10Mil TCI Spectrum Monitoring Contract 07/02/99 30
2.3.2 Dielectric receives $11 million contract 30
2.3.3 TCI International, Inc. Announces Final System Acceptance of
Spectrum Monitoring System Contract with the Government of
Colombia 31
2.3.4 CTS announce systems for India 31
2.3.5 Spectrum to spend US$1-m on airwaves equipment 32
2.3.6 Vodafone Makes the First Data Call on HSDPA (3.5G Wireless) 32
2.4 Conclusions 33
3 Identification and Analysis of System Benefits – (Work Package 2) 35
3.1 Introduction 35
3.2 Benefits 38
3.2.1 More Geographically Inclusive 38
3.2.2 More Inclusive of Transmission Types 38
3.2.3 Precision in Source Identification 39
3.2.4 Responsiveness 39
3.2.5 Information On Current Use of Spectrum 39
3.3 AMS Beneficiaries 39
3.3.1 Ofcom 39
3.3.2 Operators 40
3.3.3 End Users 41
3.3.4 Law Enforcement 41
3.3.5 Emergency Services 41
3.3.6 Research Organisations 41
3.3.7 Spectrum Traders 41
3.4 Benefits Quantification 42
3.4.1 Benefits to radio regulation 43
3.4.2 Benefits to operators 44
3.4.3 Benefits to end users 44
3.4.4 Benefits to R&D 45
3.4.5 Benefits to spectrum trading 45
3.4.6 Summary of Realisable Annual Benefits 46
3.5 Costs Quantification 47
3.5.1 Modelling Assumptions 47
3.5.2 Source Data 48
3.5.3 Build-out 52
3.5.4 Results 53
3.5.5 Sensitivity Analysis 53
3.5.6 Alternative Deployment Scenarios 56
3.6 Financing and Procurement Options 57
3.7 Conclusion 59
4 Modelling study of the proposed AMS – (Work Package 3) 60
4.1 Introduction 60
4.2 Geolocation techniques 60
4.2.1 Direction finding 60
4.2.2 Time Difference of Arrival (TDOA) 62
4.3 TDOA modelling 63
4.3.1 Overview 63
4.3.2 Calculation of propagation loss 64
4.3.3 Theory of operation 65
4.3.4 Calculation of the rms error associated with each TDOA measurement 66
4.3.5 Mapping of TDOA measurement errors to position error 66
4.3.6 Model output 68
4.3.7 Limitations of the model 69
4.4 Example applications 69
4.4.1 Required number of monitoring stations 69
4.4.2 Practical sensor layouts 74
4.4.3 Relation between correlation sample length and geolocation precision 77
4.4.4 Results 78
4.4.5 Scaling laws 80
4.5 Graphical output 80
4.5.1 Co-channel signals 83
4.6 Comparison with DF accuracy 84
5 AMS System design – (Work Package 4) 88
5.1 Introduction 88
5.2 Requirements for the AMS 88
5.3 The proposed solution 97
5.4 Solution Architecture 97
5.4.1 ‘Two-Unit’ Architecture 98
5.4.2 ‘One-Unit’ Architecture 100
5.4.3 Outdoor Unit 102
5.4.4 Thermal Analysis: Power Dissipation 102
5.4.5 Thermal Analysis: Solar Generated Heat 104
5.4.6 Form Factor 104
5.4.7 Mounting Options 105
5.4.8 Connectors 105
5.4.9 Indicators 105
5.4.10 Materials & Construction 105
5.4.11 Environmental Specification 105
5.4.12 EMC Specification 106
5.4.13 Power Supply Unit 107
5.4.14 Summary 107
5.5 Antenna 107
5.5.1 Testing 108
5.5.2 Antenna choices 110
5.5.3 Switching 111
5.6 Functional modules 111
5.6.1 Receiver 111
5.6.2 Digitiser 114
5.6.3 Processor 115
5.6.4 GPS 116
5.6.5 ADC Clock 116
5.7 TDOA architectures 117
5.7.1 Introduction 117
5.7.2 Sensor Selection 117
5.7.3 Timing errors 118
5.7.4 Multipath 118
5.8 Emitter Location techniques 119
5.8.1 Calculation of TDOA position loci 119
5.8.2 The extended three sensor algorithm 120
5.8.3 Maximum likelihood search 121
5.8.4 Simplex Search 122
5.9 Communication between sensors 122
5.9.1 Network control 123
5.9.2 Data storage 124
5.9.3 Data Transfer Messages 126
5.10 Software 127
5.10.1 Current QinetiQ capability 127
5.10.2 Key functionality 128
6 Hardware and software development of the demonstrator system – (Work
Package 5) 130
6.1 Overview 130
6.2 Component Layout 131
6.3 Component parts 132
6.3.1 PC 132
6.3.2 Receivers 134
6.3.3 Digitiser 135
6.3.4 GPS 136
6.3.5 ADC Clock 137
6.3.6 Device Control 137
6.3.7 RS232 137
6.3.8 TTL 138
6.4 The RF tray overview 139
6.5 Demonstrator network connectivity 142
6.5.1 Wireless solution 142
6.5.2 Wired solution 143
6.6 Software 145
6.6.1 Introduction 145
6.6.2 AMS GUI 145
6.6.3 AMS GUI Displays 146
7 Sensor performance and trials of the AMS – (Work Package 6). 155
7.1 Aim 155
7.2 MRX Receiver sensitivity 155
7.2.1 Central London trials 155
7.2.2 Possible Receiver Design enhancements 159
7.3 Communications link between sensors 160
7.3.1 Local testing 161
7.4 TDOA location testing 161
7.4.1 Sensor locations 161
7.4.2 Target frequencies 164
7.5 TDOA performance results 164
7.5.1 100kHz Bandwidth results 164
7.5.2 20 kHz Bandwidth results 169
7.5.3 Impact of sample length on accuracy 171
7.6 Comparison with simulation 173
7.7 Conclusion 179
8 Conclusions 181
9 Recommendations 188
10 References 189
11 Abbreviations 191
A Appendix A Literature study – system details 196
A.1 Codem Systems 194
A.2 TCI Spectrum monitoring and management 196
A.3 Communications Research Centre, Canada 199
A.4 SAT SigMon 208
A.5 InterConnect Communications 216
A.6 Rhode and Schwarz – Argus 218
A.7 Transportable Regional Remote Station 224
A.8 Tadiran Electronic Systems 227
A.9 Thales 232
B Appendix B Communications options available within the UK 237
1 Introduction
1.1 Background
QinetiQ have led a consortium, with TRL Technology and Arup Communications, to
carry out a research study which included the design of an unattended monitoring
system that meets Ofcom’s needs, and the construction of a unit which would
demonstrate elements of the whole system. This final report details the findings of
the programme and is broken down into sections that cover the various work
packages.
The radio spectrum is an increasingly important national asset, and so efficient
management of this asset is a high priority. Spectrum management has direct
benefits in terms of improved communications, but also in indirect terms as it
supports a vital part of the country’s infrastructure. In the same manner that it is
difficult to manage a company without management information, it is difficult to
manage the spectrum well without information on its usage and quality. The design
of an unattended monitoring system whose primary purpose is to automatically
detect, identify and locate the source of interfering radio signals over a large part of
the UK, is a key component of the Ofcom Spectrum Efficiency Scheme.
The overall net economic benefit from the use of the radio spectrum in the UK was
last estimated in 2000 at about £20 billion a year. Allowing for inflation, even a 1%
increase in usage brought about by more effective and efficient management would
be worth £250 million a year.
Furthermore, as the spectrum, and in particular the unlicensed bands become more
congested it is only a matter of time before overuse causes degradation of service.
Many services operating within the unlicensed band, for example wireless LAN rely
on a low occupancy within their allocated frequency band to give a reasonable
chance of getting messages through. As usage increases the number of message
failures increases and messages have to be resent leading to even more traffic
which can lead to the whole network grinding to a halt. The only answer is to have a
better monitoring capability, and to be able to act upon the results of that capability to
reactively allocate bandwidth to those users most in need.
Routine monitoring of the radio spectrum by unattended sensors will provide Ofcom
with the means of:
• Validating a database of licensed user activity and identifying unauthorised or
“pirate” operators or other sources of interference.
• Modifying the rules governing unwanted emissions or modifying the parameters of

ultra-wideband transmitters if the noise floor rises rapidly.
• Monitoring if the licence-exempt spectrum is becoming congested or whether the

basis on which we have made assignments is not in line with actual usage.
• Providing important information to make a judgement about the introduction of
new technologies such as cognitive radio.
1.2 An Automatic Monitoring System
To understand spectrum usage, Ofcom commissioned a study into the design of an

automatic monitoring system (AMS) that could be deployed in large numbers to
monitor the radio spectrum and detect interfering radio signals with sufficient
accuracy to localise and identify the source of the interference.
Ofcom currently have two programmes which utilise unattended equipment for
spectrum management and interference resolution purposes.
• The remote monitoring and direction finding (RMDF) system currently consisting
of 24 sites with the capabilities to detect signals and their direction of arrival from
20MHz to 3GHz.
• The Unattended Monitoring System (UMS) currently with 44 sites capable of
providing occupancy information from 20MHz to 3GHz.
These systems have been deployed separately because the RMDF systems are
optimally sited on high ground so that they have the highest chance of receiving an
interfering signal, whereas the UMS systems are optimally mounted in city centres
where they only receive local signals and so can accurately map the usage in that
city.
However, a fully comprehensive network, capable of monitoring the radio spectrum
and detecting interfering sources over a large part of the country will need far more
monitoring stations than are currently in the network. An original estimate of the
number of stations required for covering much of the urban, suburban and some of
the rural area was 2000, but this has since been reduced after a business study was
conducted, (see Section 3). These stations will have to be automatic in operation,
have means of downloading reports of interfering sources automatically or on
request and be relatively cheap to purchase and maintain to ensure the network
costs are not prohibitive.
Additionally, it is expected that the AMS network would have the following benefits:
• Enhanced scope of spectrum monitoring, both geographical and transmission
types,
1.3 The Consortium
The consortium, which has already established a track record of providing

consultancy to Ofcom, comprised of three parties: QinetiQ Ltd (Prime), TRL
Technology Ltd and Arup Communications. Each member brought a wide range of
“world class” skills to the consortium.
The consortium members have worked successfully together in the past, and
brought a proven track record of research and achievement in all aspects of
electronic surveillance and technology development and production.
QinetiQ Ltd is Europe’s largest science and technology organisation. Formerly an
agency of the MOD, the company has a distinguished heritage as a leading provider
of technology solutions and a supplier of impartial and trusted advice.
QinetiQ has over 9,000 staff, and can supply the creative thought-power of Britain’s
biggest independent team of scientists, engineers and internationally acclaimed
experts.
QinetiQ’s core business is providing scientific research, test and evaluation for
military and civil customers to give them leading edge competitive advantage. In
many areas the problems posed by the defence environment require the creation of
intellectual property well beyond the state of the art in commercial markets and a
reduction in risk.
QinetiQ’s move into the commercial sector provides technology to customers from
fields as diverse as transport, health, energy and telecommunications.
The Electronic Surveillance & Counter Measures Group (ES&CM Group) within
QinetiQ ltd was the Prime Contractor providing the working interface with Ofcom and
overall responsibility for programme delivery.
The ES&CM Group employs more than 120 scientists and engineers who work
either in the research of new techniques for exploitation of the RF spectrum or in the
continued development of electronic surveillance products for military and
commercial customers. The expertise in the design, development and optimisation of
hardware and software systems has resulted in our products being used in major
procurements for the Ministry of Defence, NATO and the United States Department
of Defence. The ES&CM Group has produced software and hardware units that
have been sold to Thales, General Dynamics, Lockheed Martin and others, building
up a track record of successful delivery.
By teaming with recognised leaders in the manufacture of state of the art technology
the Group is establishing a position as one of the leading suppliers of spectrum
monitoring products.
TRL Technology is a specialist supplier of surveillance, monitoring and counter
measure solutions for satellite and radio communications to defence and
government organisations worldwide. TRL has been delivering innovative solutions
to these markets for over 20 years since its formation in 1983.
TRL’s customer base is wide ranging and includes the UK Government including
Defence Procurement Agency (DPA), Defence Science and Technology Laboratory
[dstl] and Government Communications Headquarters (GCHQ). It also supplies to
foreign governments of over 20 approved countries and commercial companies
including QinetiQ, BAe Systems, Thales, Inmarsat and BT.
Arup Communications has over 20 years experience as IT and business
consultants offering fully integrated information and communication technology
solutions for the construction, government, transportation, telecoms, education,
finance, research, healthcare and corporate sectors.
The core range of services offered by Arup Communications includes strategic
advice, technology and business planning, and IT design and implementation. As
technology specialists Arup Communications have a wide range of skills including
information systems networking, wireless, audio-visual, security and control systems.
The company strives to add value by looking at the best ways to use the latest
technology to help the public and private sector clients reach their business
objectives.
With QinetiQ’s technical and management approach, the expertise from a leading
RF systems production company and the in-depth knowledge from a leading
business consultancy organisation, the consortium provided Ofcom with the means
to carry out an in-depth study into the design and development of an Automatic
Monitoring System.
1.4 Approach
The QinetiQ approach began with a literature search that built on the team's existing
knowledge of the state of the art in AMS. In parallel with this work, Arup assessed
the economic aspects of a deployed AMS system. This was achieved by comparing
the predicted costs of equipment options and providing first-pass cost estimates for
deploying and running a real-world monitoring network. Following this, the system
approach was analysed using recognised propagation models and geolocation
prediction tools to establish likely performance. A prototype system was built and
demonstrated, with TRL leading on the build, using their knowledge of developing
and manufacturing high-technology systems, particularly RF receiver products. This
lead to a final design specification for a deployable AMS system. These performance
figures and costs were then fed back to further refine the business case.
QinetiQ's design objective was for an AMS unit that is capable of hosting a variety of
geolocation and analysis techniques (including hybrid approaches). This enables it to
adapt to evolving requirements post network deployment. The key aim has been to
design and build a system that is flexible, reconfigurable and upgradeable. This has
been achieved by a software based approach – by use of readily available, low cost
generic hardware with the intelligence built into the software, the functionality of the
system can be changed or upgraded by simply installing a new version of the
software.
The design can process emissions from a large number of emitters, encompassing
analogue and digital waveforms, and can accurately locate these emitters.
1.5 Report Overview
This Final Report details the findings of the various work packages associated with
this programme. Additionally a Design File has been supplied which provides a
specification for the design of the final system. It summarises the outcome of the
design issues considered in this report, and details the user requirements.
The structure of the document is as follows:

A review of state-of the-art interference monitoring products is given in Section 2.
The findings of the business case study, conducted by Arup Communications and
considering the economics of various technical approaches, are presented in Section
3. The cost of deploying, maintaining and operating such a network in the United
Kingdom are also discussed.
Section 4 presents the results from applying QinetiQ modelling tools to predict the
geolocation performance of the system using Time Difference of Arrival (TDOA).
A description of the proposed solution for a final deployed system can be found in
Section 5. To demonstrate the concept, three “Proof of Principle” prototype units
have been built. An overview of the design of these demonstrator units can be found
in Section 6.
These prototypes have since demonstrated the advantages of the TDOA location
techniques as well as fast, wideband scanning and networked monitoring. Details
about the trials can be found in Section 7.
Finally the conclusions, including further recommendations to improve performance
for a fully deployed network, are in Section 8.
2 Review of current state-of-the-art automatic
interference monitoring equipment – (Work
Package 1)
2.1 Introduction
Work Package 1 required a literature review of automatic interference monitoring

equipment with the aim of determining the current state of the art.
The main search took place between February and March 2005 and a full report [1]
was delivered to Ofcom.
In addition, the consortium team reviewed the collated evidence on how the current
state of the art automatic monitoring systems can be used to support spectrum
regulation, enforce licensing and spectrum monitoring for spectrum trading to
support the Spectrum Efficiency Scheme.
A watching brief has been maintained through the remainder of the project and,
although this has not identified many developments in the interim, the first data call
on 3.5G, also known as HSDPA (High Speed Downlink Packet Access) that enables
up to 1.5Mbps transmission speeds has been made by Vodafone and has been
noted and added to section 2.3.
2.1.1 Scope
The design of an unattended monitoring system whose primary purpose is to

automatically detect, identify and locate the source of interfering radio signals over a
large part of the UK, is a key component of the Ofcom Spectrum Efficiency Scheme.
Routine monitoring of the radio spectrum by unattended sensors will provide Ofcom
with the means of validating a database of licensed user activity and identifying
unauthorised or “pirate” operators or other sources of interference.
In automatic mode, the monitoring system would be expected to identify
unauthorised radio transmissions without operator intervention. Since a key
requirement is to minimise false alarms, the system must be able to determine with
high confidence when a transmission is valid. As a minimum, this will require the
system to determine the form of modulation in use, and to access a database of
licensed users. However, this may be insufficient, and the system may need to
progressively build up its own database of dynamic traffic activity/behaviour to
identify unusual events.
The Ofcom requirement implicitly calls for an automatic monitoring system that is
equipped with a wideband, fast tuning receiver, both to accommodate the widest
bandwidth signals and to achieve the spectrum revisit time. This will also ensure a
high probability of intercept against fleeting or intermittent interference.
Over the specified frequency band (20MHz to 3GHz) with potentially a very high
number of signals simultaneously present, maintaining high receiver fidelity will be
essential for correct signal characterisation and avoidance of false alarms.
Conventionally, low-cost receivers with acceptable intermodulation performance are
narrowband superheterodyne types. However, both scan rate and instantaneous
bandwidth are well below the requirement for this project.
There are several manufacturers of wideband tuners of varying size, cost, and
fidelity. Unsurprisingly the higher performance models are generally more expensive.
An important part of this study was to focus on the affordable automatic monitoring
systems that have the potential to meet (or partially meet) the Ofcom AMS
requirement.
2.1.2 Review of current state of the art automatic interference monitoring systems
The consortium team has a wealth of experience in systems that provide emitter
detection, location, and signal analysis and have an extensive range of contacts in
relevant fields. This meant that the consortium team was ideally placed to review the
state of the art in interference monitoring products, using sources and means that
may not easily be available in a standard open literature search.
Multiple routes have been taken to compile information for this review. They include:
• Internet searches for relevant technical information; the ITU members web page
have been reviewed for relevant information and links.
• QinetiQ ltd using its own sources of information.
• Liaison with Ofcom to compile a list of specific companies that it would wish
particular attention to be paid.
The collated information has been tabulated so that Ofcom can compare the current
state of the art capability of commercially available automatic spectrum monitoring
systems.
In addition, the consortium team has reviewed the collated evidence on how the
current state of the art automatic monitoring systems can be used to support
spectrum regulation, enforce licensing and spectrum monitoring for spectrum trading
to support the Spectrum Efficiency Scheme.
2.2 Commercially available systems
The consortium carried out an extensive search of current commercial and research
literature of automatic monitoring systems between February and March 2005. This
section describes the commercially available “off the shelf” spectrum monitoring
systems that have the potential to meet (or partially meet) the current Ofcom
requirement.
A description and a specification of the automatic monitoring systems that were
reviewed are given in Appendix A of this report.
2.2.1 Literature search summary
A summary of the search of current commercial and research literature of automatic

monitoring systems is tabulated in Table 2.1.
The information given was acquired from numerous sources available to the QinetiQ
led consortium and provides a good overview of the current capabilities of
commercially available automatic monitoring systems used for spectrum
management applications. The information is based on manufacturer’s data sheets
and is correct to the best of the consortium’s knowledge, but has not been
independently tested so the accuracy cannot be guaranteed.
The summarised data provides all of the known capabilities of all of the systems
detailed in the Appendices of this report. If a box is empty then it has not been
possible to find the information during the literature search or from follow-up
enquiries to the companies concerned.
2.2.2 Equipment summary
The systems found during the literature search meet the requirement for large
manual or automatic direction finding systems against single channel interferers
such as hoaxers or pirate radio stations.
In addition they all locate signals using angle of arrival direction finding techniques.
These have a constant angular error so the position error scales with distance
resulting in very large probability ellipses when attempting to locate an emitter at
distance.
Against modern multi-channel emitters such as GSM mobile phones or TETRA the
systems may possibly work against the base channel but would not distinguish
multiple mobile handsets using a single frequency as the DF technique cannot
distinguish between multiple co-channel signals.
In addition the systems would not be able to locate an interfering emitter in the
presence of the licensed emitter.
Commercially available spectrum monitoring systems
Rohde &
WinRadio Sat Corp Codem Aselsan TCI Thales
Capabilities Schwarz
MS-8108SR SigMon 1000 Eaglenet TRRS TCI 715 Esmeralda
TMS 100 & 200
Summary
Mode of A narrowband Wideband and Narrowband Wideband 10 Narrowband up to Wideband

Operation receiver Narrowband MHz 500 kHz
monitors a
single channel
Spectrum Scan Yes Yes Yes Yes Yes as per ITU Yes Up to 750
recommendation channels/s (HF),
6 to 6000 per 1000 channels/s
minute (VHF/UHF)
dependent upon
band.
Antenna Up to 6 One Multiple Options Various options Multiple options for

for mobile DF, dependent upon fixed or mobile HF
fixed DF or intercept or and VHF/UHF
monitoring only direction finding
system purchased
Rohde &
TMS 100 & 200
Frequency 0.15 to 1500 Any Agilent 20 to 2500 MHz 9 kHz to 3000 R&S EB200 9kHz – 30MHz
Range (MHz) default 0.14 to Spectrum MHz 10 KHz to 3000 (HF), 20 MHz –
4000 extended. Analyser MHz 3000 MHz (VHF/
UHF)
Tuning 100 10 Hz to 25 kHz 1 Hz 1Hz (HF), 10Hz
resolution (Hz) (VHF/UHF)
Noise Figure 9dB (HF) 15 dB 12 dB Typical 14dB typical (HF),

(VHF) 9dB typical
(VHF/UHF)
Sensitivity Typically 1 uV -33 to -8 dBuv/m -124 to -100dBm
(HF),
-128 to -108dBm
(VHF/UHF)
Dynamic Range 85 dB
No. Narrowband 6 to 8 Unspecified, cost

Channels dependent
Modulation FM, AM, Phase / ITU-R SM 1268-1

Classification ITU (appendix 2), SM
recommendations. 328, BS412-6,
chap. 4.6 of ITU-R,
SMH, ed. 2002
Rohde &
TMS 100 & 200
Modes AM, SSB/CW, AM/FM/SSB AM/FM/LSB/USB/ AM:A2A, A2B,

FM-N, FM-W CW A3E,
CW:NON, A1A,
A1B
USB/LSB
(SSB):J2A, J2B,
J7B,
H3E, J3E, R3E
ISB:B8E (2
reception
channels)
FM:F3E
ϕ M:G3E
FSK:F1A, F1B
(HF), F1B
(VHF/UHF)
TCPIP Data Yes Yes Yes Yes Yes
Communications
Location Method Direction Finding Direction Finding Direction Finding Direction Finding Direction Finding Triangulation (HF,
(AOA) (AOA) (AOA) (AOA) (AOA) UHF/VHF), option
of Single Station
Location (SSL)
(HF)
Supports > 1
emitter on No No No No No
Channel
Rohde &
TMS 100 & 200
Manual Tasking Yes Yes Yes Yes Yes Yes
No No No Licence Violation Licence Violation Can integrate with No

Detection – via Detection – via an existing
Licence
Monitoring user-defined user-defined national licence
presets presets DB to track
violations
Scheduled Runs Yes Yes
Client Server Yes Yes Yes Yes

Operation
Table 2.1, Commercially available spectrum monitoring systems
2.2.3 Internet links
The following table gives the Internet URL’s of the commercially available
systems covered by this report.
The links to the listed websites were valid to the date of publication of the
literature study report.
System Internet URL
WinRadio http://www.winradio.com/home/ms-systems.htm
Sat Corp http://www.sat.com/products/terrestrial/sigmon_options/sigmon_1000/
Codem http://www.codem.com/content/datasheets/sigint/eaglenet.pdf
Aselsan http://www.aselsan.com.tr/msting/trrs_eng.htm
TCI http://www.tcibr.com/PDFs/715webs.pdf
Rohde & http://www.spectrummonitoring.rohde-schwarz.com

Schwarz
Tadiran http://www.tadsys.com
Electronic
Systems
Thales http://www.thalesgroup.com/land-joint/portfolio/pdf/esmeralda_ag.pdf
Table 2.2, Internet URL’s of the commercially available systems covered by this
report
2.3 Press Releases
In addition to a review of the commercial and research literature for automatic

monitoring systems, the consortium members have conducted a search of press
releases dating back to 1999. This avenue of research has provided a good
indication of the current use and deployment of automatic monitoring systems
across the globe.
The information presented in this section has been collated from searches of
international technical, business and scientific journals and publications. The
reports presented in this section are informative and cover a number of systems
or manufactures/suppliers listed in this report.
Other press releases were gathered, but the quality of these reports in relation to
the worldwide application, technical detail and the deployment of automatic
monitoring systems were judged to be less informative to Ofcom than the
following examples.
2.3.1 Conatel Signs $10Mil TCI Spectrum Monitoring Contract 07/02/99
CARACAS, VENEZUELA 1999 JUL 2 (NB) -- By Staff, GNCS Business News.

Venezuela's telecommunications regulator Conatel has signed a $10 million
contract with TCI International (TCII) to buy a country-wide Automated Spectrum
Management and Monitoring System for Venezuela.
The system consists of one national control centre, five remote monitoring
stations and 10 mobile units, and is scheduled to be completed in 14 months.
Reported by Newsbytes.com, http://www.newsbytes.com
2.3.2 Dielectric receives $11 million contract
February 20, 2004 – SPX Corporation (NYSE:SPW) today announced that its
Dielectric business has been awarded an $11 million contract to supply the
Botswana Telecommunications Authority. The 2-phase, 13-month contract will
include 18 of the company’s TCI brand Spectrum Monitoring Systems, 16 fixed
and 2 mobile, along with custom RF Spectrum Management software. The
Botswana Telecommunications Authority (BTA) is one of the pre-eminent
agencies of its type in Africa and is at the forefront of establishing many of the
spectrum monitoring techniques and standards adopted on the African continent.
The BTA will use these systems to establish a nationwide network of Spectrum
Monitoring Systems to monitor, control and license their wireless spectrum.
Dielectric’s systems utilize the latest radio and signal processing technology and
will allow the BTA systems to be controlled from a central location in the capital
city of Gaborone and cover the 20-3000 MHz frequency range. Many of the
Spectrum Monitoring Systems will be installed in unmanned, remote locations.
John Capasso, President of Dielectric said, “BTA is leading the development of
spectrum monitoring technology in Africa. As we expand our systems globally,
we appreciate this vote of confidence from such an innovative partner.” Dielectric
Communications is the nation's largest manufacturer of broadcast antenna
systems (TV/FM/HF/MF), communication towers, lighting, and signal processing
equipment. Based in Raymond, Maine, Dielectric offers complete system
monitoring, maintenance, and service to the broadcast and wireless markets.
SPX Corporation is a global provider of technical products and systems,
industrial products and services, flow technology and service solutions. The
Internet address for SPX Corporation’s home page is www.spx.com.
http://www.dielectric.com/broadcast/news_story.asp?ID=63
2.3.3 TCI International, Inc. Announces Final System Acceptance of Spectrum

Monitoring System Contract with the Government of Colombia
Jan. 31, 2000 TCI International, Inc. announced today that on 28 January 2000,
the Ministry of Communications of the Republic of Colombia formally and
unconditionally accepted TCI's $18M Nationwide Automatic Spectrum
Management and Monitoring System in Colombia.
The TCI Model 710 System is believed to be the first such state-of-the-art
Telecommunications Infrastructure Modernization Project to be purchased and
operated by an indigenous Telecommunications Authority in Latin or South
America.
The Colombian Ministry of Communications and TCI are currently discussing the
terms of a follow-on contract to provide technical support during the year 2000.
TCI looks forward to cooperating closely with the Ministry in future years to
supply System spares, maintenance, technical upgrades, and training.
2.3.4 CTS announce systems for India
PARIS, France, 6thMay, 2003 Cril Telecom Software (CTS), the leading
provider of spectrum management systems and services, today announced a
groundbreaking deal to provide its ELLIPSE spectrum management and
monitoring solution to India’s Ministry of Communication. Selected as part of a
consortium led by Thales Communications and its partner Himachal Futuristic
Communications Ltd (HFCL), this multi-million Euro deal reinforces CTS’s
presence in Asia. The CTS spectrum management software, ELLIPSE Spectrum,
will allow the Ministry’s Wireless Planning and Coordination (WPC) Wing to
securely and efficiently manage its spectrum resources in accordance with
International Telecommunications Union (ITU) regulations. The project will be
deployed over the next 18 months and enable the Indian Ministry of
Communications to realise the following benefits: -Enhanced process automation
for improved service quality to network operators; -Optimized network
deployment for enhanced end-user services; -Reinforced security measures for
aviation and military deployments; -Increased revenues from superior spectrum
usage management; -Guaranteed data consistency and increased productivity; -
Adoption and use of new telecommunications technology in India. Mr. Y.L.
Agarwal, managing director at HFCL International said: “CTS’s Ellipse Spectrum
software is a very scalable solution. We were confident we could rely on CTS for
our nationwide deployment. The company’s customer base, wide-ranging
coverage and compelling commitment to quality were significant factors in our
decision to work with them.”
http://www.criltelecom.com/news_documents/023026_WPC_FINAL_gb.pdf
2.3.5 Spectrum to spend US$1-m on airwaves equipment
Wednesday, February 09, 2005 - The technology ministry through one of its agencies,
the Spectrum Management Authority (SMA), will invest upwards of US$1 million in
equipment and software to improve the monitoring of Jamaica's airwaves for
telecommunication firms and radio stations.
The equipment is an automated spectrum management system which can
assess, co-ordinate and plan the actual assignment of frequencies.
Another instrument being secured - a remote spectrum monitoring system - will
monitor the signals of users anywhere in the island.
"It will provide us with the tools we need to efficiently manage the spectrum," said
Ernest Smith, the managing director of SMA.
"The remote monitoring system would be the first in the English-speaking
Caribbean, to the best of my knowledge."
The project will be funded from a portion of the US$6 million that the US telecom
giant AT&T Wireless, paid the Jamaican government for a cell licence last year.
This means, according to Smith, that the spectrum users would not be charged
for the equipment. The SMAs budget is approximately J$60 million.
The SMA is now evaluating four firms to supply the equipment and software the
agency having closed the tenders on January 17.
"We are now in the process of evaluating those tenders," he said, adding that the
successful candidate would supply the equipment within a year.
"Within 12 months of placement, we expect to have those two operating systems
installed and commissioned and functioning."
http://www.jamaicaobserver.com/magazines/Business/html
2.3.6 Vodafone Makes the First Data Call on HSDPA (3.5G Wireless)
14 November , 2005, Europe Portugal :Vodafone today made the first data call
on 3.5G, also known as HSDPA (High Speed Downlink Packet Access) as part of
the testing program of this important technology.
HSDPA is a technological evolution of the 3G network that permits a
considerable increase in transmission speeds. Using this new technology, it is
already possible in this initial phase to achieve higher transmission speeds of up
to 1.5Mbps, approximately 4 times the speed currently available on 3G
(384kbps).
With the introduction of HSDPA, it will be possible to offer new mobile data
services and boost Vodafone’s current mobile broadband services such as
Internet and e-mail access – the Vodafone Mobile Connect Card 3G and the
Vodafone live! 3G mobile portal, along with many of the currently available
information and entertainment services, in particular Mobile TV and video
streaming and download on demand.
At the beginning of next year, Vodafone will begin to provide this technology to a
limited number of users, including certain partners, customers and employees,
with a view to rolling out HSDPA (3.5G) in the second half of 2006.
Through this evolution of the 3G network, Vodafone aims to maintain its
leadership position in innovation and data services by ensuring that its customers
enjoy high download speeds and an improved mobile experience when
accessing various data services.
Vodafone has always been in the forefront of the introduction of new
technologies and the development of mobile data services, having pioneered
tests in Portugal of both the 2.5G (GPRS) and 3G networks, and having been the
first operator to offer 3G services in Portugal with the launch of the Vodafone
Mobile Connect Card 3G.
http://www.3g.co.uk/PR/Nov2005/2201.htm
2.4 Conclusions
An extensive study of commercial and research literature has been carried out.
There is a wide range of automatic monitoring systems currently available or in
use on a global basis. However, a closer inspection of the literature suggests that
there are fewer than ten systems that come close to meeting the Ofcom
requirements, and none that will fully meet them.
All of the automatic monitoring systems that were reviewed use signal location
entirely based on direction finding (DF) techniques, which rely entirely upon multi-
element antenna arrays. These antenna arrays are generally quite large and
bulky, particularly for operation at HF and VHF and their profiles are not
necessarily inconspicuous or unobtrusive.
The monitoring capability of the commercially available automatic monitoring
systems is generally provided by off-the-shelf spectrum analysers, rather than
digitisers. While these instruments offer wide operational bandwidths, the
instantaneous bandwidth has to be reduced if signals close to thermal noise
levels need to be analysed. This affects the revisit time and reduces the ability of
the monitoring system to intercept, analyse and classify transitory or sporadic
interference. These off-the-shelf instruments are also designed for laboratory use
and are generally bulky and expensive assets.
The survey also revealed that the commercially available automatic monitoring
systems:
• have a limited capability against modern signal types such as TDMA and
CDMA where the signals can effectively be stacked on top of each other;
• have limited (if any) signal analysis capabilities particularly against TDMA and
CDMA or other complex methods of signal modulation;
• exhibit particularly poor performance in multipath environments which limits

their use in urban or metropolitan areas;
• cannot cope with co-channel signals and have limited classification

capabilities in dense signal environments;
• will not support fast scan (sample) rates;
• require high signal levels (meaning limited receiver dynamic range), which
limits operational range of the system;
• are based on large, calibrated antenna arrays, laboratory spectrum analysers

and add-on signal classifiers that makes the systems high cost, high value
assets.
In addition, the search has found that in Australia and New Zealand, which have
had spectrum trading arrangements in place since 1997 and 1989 respectively,
trades are not taking place regularly enough for them to have had any impact on
the spectrum monitoring role performed by the regulators. Consequently in these
countries the amount of trading is sufficiently low that they do not have a
requirement for sophisticated monitoring systems.
Many of the trades which have taken place have been as a result of mergers and
acquisitions, which have not complicated monitoring requirements at all. New
Zealand completed their network of Spectrum Monitoring equipment, including
fixed stations and vehicles in early 2003, but the system has not yet been used to
support or monitor spectrum trades.
Very few of the systems that were reviewed offered any integration with existing
databases of license holders and license types. This is a useful feature as it will
help operators to determine quickly whether particular transmissions are licensed
or not. In a dynamic spectrum trading environment, it could be otherwise difficult
for operators to determine the validity of unusual transmissions.
Ove Arup & Partners Ltd
3 Identification and Analysis of System

Benefits – (Work Package 2)
3.1 Introduction
In this chapter, an outline evaluation of the potential benefits and costs from the
deployment of a national network of automatic monitoring stations (AMS network)
is developed. It describes the types of benefit that could be expected to arise,
the likely beneficiaries and the expected scale of benefits. The capital costs of
producing and installing a national AMS network have been developed, and the
annual running costs estimated.
Both the benefits and costs of the AMS network have been assessed by
comparison with the current methods of monitoring radio frequency use and
interference in the relevant bands. Ofcom currently has deployed two systems of
unattended equipment for monitoring spectrum use and resolving interference
problems in the range 20MHz – 3GHz:
• 24 remote monitoring and direction finding (RMDF) sites located on hill top
sites;
• 44 sites of the unattended monitoring system (UMS) located in city centres for
the purpose of recording occupancy of the radio spectrum.
The number of RMDF sites may be increased. However, both systems will
remain significantly less than national in scope.
The anticipated benefits of the AMS network are:
• Enhanced scope of spectrum monitoring: geographical,
• Enhanced scope of spectrum monitoring: transmission types,
These benefits may be summarised as better information on spectrum
occupancy or use, including unauthorised use, and better information to support
the resolution of interference problems.
Improved information is not a benefit that can be valued directly. Information
enables activities to be carried out more effectively or, by reducing uncertainty,
enables other activities to be carried out that, in the absence of the information,
would not be considered worthwhile.
The potential beneficiaries are:

• Ofcom,
• End Users (citizen consumers),
• Operators – current,
• Operators – future,
• Law Enforcement,
• Emergency Services,
• Research Organisations,
• Spectrum Traders.
Note that the armed forces are excluded from this list, which concerns only civil
use of the radio spectrum. The assignment of potential benefits to beneficiaries is
described in Figure 3.1.
Transmission Responsive Interference

Geographical Precision Usage info
Types ness Info
Ofcom
Operators
New
Operators
Users
Law
Emergency
Services
Research
Traders
Figure 3.1, Benefits and Beneficiaries: Dark Green - Significant, Pale Green -
Less Significant, White - Not Significant
Economic benefit is not the same as revenue. It is not uncommon for large
infrastructure deployments to be undertaken without realisable financial proceeds
in excess of total costs being anticipated. The classic example is the Victoria
Line on the London Underground network, the construction of which was justified
on the basis of improved journey times by users, rather than increased revenues
for the service provider or any other party. More recently, the construction of the
Channel Tunnel Rail Link, which will replace the current rail link to Waterloo with
a new high-speed line from the Tunnel to St. Pancras, has been justified because
it will make a 20 minute improvement in journey times from the continent to the
centre of London.
Cost-benefit calculations estimate the net effect on economic activity. So, in
addition to building up the benefits assessment by looking at a series of discrete
benefits and beneficiaries, it is worthwhile also making an overall assessment of
the net impact. This can provide a check on the scope and scale of the assessed
benefits.
Benefits can be assessed using the standard techniques of cost-benefit analysis.
The standard method involves estimating how willingness-to-pay (and
willingness-to-produce) varies with price, and taking into account the benefits
received by users willing to pay more than the market price, and by suppliers
willing to produce for less than the market price. This technique is suited to
evaluating the effect of incremental change in a product or service which is
already traded in a competitive market. In the case of AMS, however, existing
data on spectrum usage is not charged for, so there is not a reference market
price on which to base the cost-benefit calculation.
Moreover, while we have developed good information on the costs of supply, this
information is not sufficient to support an analysis of potential benefits to
producers (“producers’ surplus” in cost-benefit language). We are therefore
limited to comparing willingness-to-pay, on the one hand, with estimated costs of
supply.
The starting point for the assessment of costs and benefits of the AMS system is
the assumption that Ofcom will be responsible for it and the principal beneficiary
of the data it produces. We assume that AMS will replace other methods of
generating data about frequency use and interference, generating some cost
savings. Then, to the extent that the deployment of an AMS network makes radio
regulation more efficient or the use of a particular frequency band less
susceptible to interference, then the value of these changes can be valued fairly
robustly – though not necessarily with precision.
On the other hand, if it is supposed that an AMS network would promote
innovation or facilitate the development of a service which does not currently
exist– such as spectrum trading, then the valuation of potential benefits become
more speculative.
A second major issue is the limited degree to which potentially realisable benefits
can be ascribed to the information provided by an AMS network as distinct from
other relevant factors that may contribute to greater efficiency in spectrum use or
the reduction in problems of interference.
In general, information that enables economic activity to be more efficient, or the
value of an asset to be determined more completely or more precisely, does itself
have a value (someone is willing to pay to obtain the information), but this value
is limited. For example, the effort that purchasers of assets such as property or a
company are willing to expend to confirm the value of their purchase (“due
diligence”) may amount to 1% of the value of the asset, but is often much less
and rarely more. This is because the alternative to having available good quality
information is to make do with less good information. Information primarily
reduces uncertainty about asset values; its worth reflects the benefit that can be
obtained from the reduction in uncertainty, rather than from the intrinsic value of
the asset.
3.2 Benefits
3.2.1 More Geographically Inclusive

The proposed AMS system could be deployed to cover a greater geographic
area than the current spectrum monitoring system. It is assumed for the
purposes of the benefits analysis that all but the most sparsely-populated areas
have full coverage. Although it is possible to deploy AMS units in a more limited
fashion, or to deploy them area by area fully benefits will only be realised with a
near-complete deployment. We have assumed that some 83% of the UK would
be covered by the system, with only regions described by the Office of National
Statistics as “Remote Rural” omitted.
The currently-deployed systems are focussed on the busiest areas of the
country, with other areas much harder to reach. Each of the beneficiaries listed
in section 3.3 has a particular collection of benefits associated with greater
geographic coverage.
3.2.2 More Inclusive of Transmission Types

The focus of the existing UMS and RMDF equipment is on frequencies below the
2GHz range. The existing systems are optimal for services such as broadcast
radio and PMR. The systems are horizontally polarized, which makes detection
of television and some fixed-links problematic. The AMS system is fully effective
up to 3GHz, and therefore encompasses high-value transmissions such as GSM
and UMTS mobile telephony and broadcast television; and potentially the IMS
bands used for wireless LANs over relatively short ranges.
3.2.3 Precision in Source Identification

The currently-deployed DF equipment can identify the location of a high-power
transmitter provided there is limited co-channel interference with it, and that there
are limited reflections. Even when these two criteria are met, the location
precision is within about 600m or so. The AMS system is designed to operate
effectively, even for lower-powered transmissions with severe co-channel and
reflection effects present. The anticipated location accuracy is in the range of
100m or so.
3.2.4 Responsiveness
The AMS system should be configurable to provide pre-emptive alarms in the
event that certain conditions are met. For example, if licensed spectrum were
used by a source in an unexpected location, indicating a possible violation. This
would enable Ofcom's officers to act in advance of receiving a compliant.
By providing greater location accuracy, the time taken to find infringing
transmitters will also be reduced.
3.2.5 Information On Current Use of Spectrum

The AMS system will be able to make a record for every emission it detects and
store this in a database. This emissions database will form a valuable resource,
with potential applications in academia, industry and to inform Ofcom’s activities.
3.3 AMS Beneficiaries

The most obvious benefits from the AMS system will be in radio regulation. If
Ofcom’s spectrum management duties can be carried out more effectively, all
spectrum users will benefit.
Some spectrum users will also benefit from certain summary information that
could be made available from the system, either in the form of regular static
reports, or through direct access to the system itself.
3.3.1 Ofcom
Ofcom will be the key beneficiary of the system. The Communications Act 2003
outlines some specific statutory duties and regulatory principles that would be
supported by the AMS network. One of Ofcom’s specific duties is to ensure the
optimal use of the electromagnetic spectrum. Good information on the current
usage of the spectrum is vital to discovering any inefficiencies.
Another duty is to ensure that a wide range of electronic communication services
- including high speed data services - is available throughout the UK. The extent
and quality of radio communication services available across the UK could be
monitored by a geographically widespread AMS system.
Although Ofcom operates with a presumption against intervention in the market,

it must ensure that all such interventions are “evidence-based”. The AMS system
could extend the scope of evidence collection for radio communications systems
in terms of service type and geographic spread.
Ofcom is required to research markets and remain at the forefront of
technological understanding. Knowledge of spectrum occupancy and other
market factors and use, especially under the increasingly deregulated, spectrum
trading regime could be difficult to obtain without an effective spectrum
monitoring capability.
Finally, Ofcom is expected to facilitate law enforcement actions against those
who misuse the spectrum, such as pirate radio broadcasters. The greater
transmission location accuracy will save time and effort in tracking down
misuses.
The AMS system has the potential to improve the efficiency and effectiveness of
their work on these all these activities. However, significant additional resources
will be required to deploy and operate the AMS system, over and above the
currently-deployed systems.
3.3.2 Operators
By operators we mean organisations that operate radiocommunication services,
such as broadcasters and cellular mobile telephony operators. A distinction can
be made between national operators, with national network obligations and
hence an interest in information that is national in scope, and operators whose
infrastructure is geographically limited. Incumbent operators are likely to place a
higher value on the information than new entrants, since they will be more certain
to be able to realise the benefits.
Many national operators are currently working to expand and improve their
services. The cellular mobile phone operators continue to install new base
station sites to improve the coverage and capacity of their 2G networks, and to
roll out 3G systems. Broadcasters are introducing new digital services, both for
radio and television.
In both cases, improving geographic coverage is a key driver for the expansion.
An AMS system covering the whole of the UK could help them to identify
proactively potential coverage and capacity problems without waiting for
consumer feedback or deploying survey personnel to the field.
Current operators will also be interested to discover interference sources
proactively.
New operators are increasingly deploying new services and installing networks
using wireless technologies. For example, there is much current interest in
mobile TV, WiMAX and TD-CDMA data networks and in extending WiFi hotspot
coverage in metropolitan areas. This trend is likely to accelerate in the near
future, as additional frequency bands are auctioned. Spectrum trading may
make it possible for frequencies already assigned to be utilised by others, for
example frequencies used by cellular mobile operators. Incumbent and new

operators would find it useful to survey the spectrum they intend to sell or acquire
before engaging in spectrum transactions.
3.3.3 End Users

End users are also called citizen consumers - the users of radiocommunication
services. They will benefit from any improvements to the services offered by the
operators. They will also benefit from the potential of the AMS system to
streamline Ofcom’s handling of complaints about interference and reception
problems.
3.3.4 Law Enforcement

Law enforcement agencies can benefit from improved information on illegal uses
of the spectrum. However there is also a potential benefit from the ability of the
AMS system to report radio transmissions by criminals. The greater transmitter
location accuracy and geographic coverage offered by the AMS system will be
key to the benefits for law enforcement.
3.3.5 Emergency Services

Ofcom’s current DF equipment is occasionally used to support HM Coastguard
by locating distress beacons and mayday transmissions. Improvements to the
accuracy of these location services, the greater geographic coverage offered, or
the ability of Coastguard Officers to use the AMS system directly could be very
valuable to them on such occasions.
3.3.6 Research Organisations

Academic and commercial researchers are at work developing new
radiocommunication techniques and equipment. More complete information
about the actual current use of the spectrum could be made available to them to
assist in their work.
For example, the database of historical transmissions could be used to inform a
realistic simulation of the spectrum usage for the development of cognitive radio
technology.
For these groups, the more complete and accurate the dataset, the better.
3.3.7 Spectrum Traders

A key component of the current drive towards spectrum liberalisation is the
establishment of a spectrum trading mechanism, by which spectrum holders may
sell or rent rights to particular bands to other organisations, without recourse to a
central authority. With access to good information about current spectrum use,
and how susceptible it is to interference, the value of the spectrum could be more
accurately determined; and underused spectrum could be easily identified by
purchasers. Any specialised spectrum trading agents who emerge would have a
direct interest in access to accurate and comprehensive information on spectrum
utilisation, in order to identify and value spectrum trading opportunities.
3.4 Benefits Quantification

It was noted in the introduction that a realistic valuation of the benefits form
deployment of the AMS system would be based on an assessment of the
willingness to pay of relevant parties for its output. The scope of this research
has not included a rigorous investigation of commercial interest in AMS; rather
benefits have been assessed analytically.
Willingness-to-pay is an important criterion because it is, in principle, an objective
measure of value and because it underpins calculations of potential future
revenues. The total benefits from deployment of AMS will, however, exceed
potential revenues. A distinction can be made between the worth of specific
information about the value of an asset and the effect of knowing that specific
information is now available that was not available before. For example, if
precise information can be obtained regarding the performance characteristics of
second hand cars, then some purchasers will pay to obtain it, while others may
not. But the fact that such information can be obtained at reasonable cost will
have a generally positive effect on the value of all second hand cars. Similarly,
the knowledge that frequency usage and interference can be readily measured
will have a generally positive effect on the value of all frequencies.
The overall net economic benefit from the use of the radio spectrum in the UK
was last estimated in 2000 at about £20 billion a year (ref: The Economic Impact
of Radio, Radiocommunications Agency 2001). Updating for inflation and the
overall growth of the economy would imply that the current value of spectrum use
to the country is about £25 billion a year. Almost all this annual value is
associated with frequency bands that would be monitored by the AMS system.
It then follows that, if a 1% improvement could be made in the value derived from
use of the spectrum, this would be worth £250 million a year. The net present
value of this stream of annual benefits would be of the order of £3 billion1. The
AMS network may or may not be able to achieve a 1% improvement in efficiency
with which frequencies are assigned and used. However, this calculation
illustrates the immense potential worth of even minor improvements in spectrum
usage.
Whatever their scale, it would probably be impossible to relate the improvement
in frequency utilisation to a specific change in economic activity or to a specific
impact on asset values. However, assuming an AMS network were able to
generate good quality information regarding usage and interference across a
broad range of frequencies, then the knowledge that such information was
available should raise the value of rights to spectrum by at least 1%, and so will
in principle produce a one-off benefit of at least this scale.
1
Based on NPV calculation with a 5% discount rate and a 20-year system lifetime.
Public mobile communications and broadcasting activities accounted for over

75% of the overall net benefits in the 2000 report, suggesting that these two
sectors should be the principal focus of attention concerning the potential overall
effect of deploying an AMS network.
In public mobile communications, there would appear to be great potential for the
enhanced usage of radio spectrum, which could be achieved through a variety of
means:
• The assignment of additional frequencies;
• Facilitating secondary assignment;
• Spectrum trading;
• Improved engineering of networks.
The critical question is the role that the AMS network might play in enabling
enhanced use of radio frequencies for public mobile communications. In general,
it would appear that deployment of the AMS network, though not a necessary or
sufficient condition of enhanced use of frequencies being realised, could play a
major role in facilitating secondary assignment and spectrum trading (see section
3.4.5 below).
In broadcasting, the methods by which frequencies are planned and allocated
means that the potential benefits are likely to be largely limited to the improved
resolution of interference issues and more effective deterrence of transmissions
from pirate radio stations.
At a high level, therefore, the benefits from successful deployment of a national
AMS system are potentially very large in relation to the cost of the investment. In
addition to this high-level approach, we have built up estimates for specific types
of user of their willingness-to-pay for use of the AMS network, and to identify
specific savings that the AMS system might make possible. These figures are
possible sources of revenue for the an AMS operator as well as realisable
components of the overall economic benefit described above.
3.4.1 Benefits to radio regulation

The first step is to consider the impact on the costs of current regulatory
activities. The operation of the AMS network should enable Ofcom to reduce the
numbers of its staff engaged in monitoring spectrum usage and in tracking
specific cases of interference. A reasonable assumption is that staff numbers
might be reduced by about 50, producing savings of the order of £4 million a
year. Note that the costs, including staff costs, of operating the AMS system are
accounted for in section 5 below.
Improved information should also enable improvements to be made in the
conduct of radio regulation more generally. Better information would facilitate
faster and more effective decision making. What might be the annual value of
this effect as regards Ofcom? Quantitatively the most significant element is likely
to be in more rapid detection of unauthorised use of radio frequencies, where the
likely impact of utilising data from a national AMS system would be to produce
improvements in the efficiency of detection of the order of 5-10%. Such
efficiency gains might be realisable in the form of staff cost savings. The annual
cost to Ofcom of detecting unauthorised frequency use is not known precisely but
is estimated at about £50 million at current prices. We therefore estimate this
benefit at a further £2.5 million a year.
The benefits arising from the increased value of frequencies have been
estimated separately above. One consequence of such increased value would be
an increase in the proceeds of spectrum auctions, which is a realisable cash
benefit. The scale of future spectrum auctions can only be estimated within very
wide parameters: based on recent trends, and Ofcom’s plans, we estimate
average proceeds for HM Treasury from spectrum auctions of £200 million a
year. Applying the 1% value improvement attributable to the AMS system
described above produces an annual benefit of £2 million.
The sum of the annual estimated benefits to radio regulation of a national AMS
system are therefore £8.5 million. This estimate is subject to a wide margin of
error, but is likely to err on the low side. It excludes for example any estimate of
the efficiency gains from improved data on spectrum use, other than in detection
of unauthorised use.
3.4.2 Benefits to operators

Operators require good information of spectrum usage in order to optimise the
deployment of their networks. Cellular operators, for example, can already
monitor the throughput of traffic at individual base stations, but this provides
limited information on the location of mobile handsets. More precise location
information would enable better coverage – and hence higher revenues - or a
requirement for fewer installations (lower costs).
Because of the availability of alternative sources of data, whether using
alternative techniques as described above, or relying on expert opinion, or a
combination of the two, operators’ willingness-to-pay for better information will be
only a small fraction of the potential gains to be made from its use. One measure
of the benefit of the information to operators would then be the saving in the cost
of the additional expertise required to interpret and extrapolate from inferior data.
The potential saving in the requirement for expertise in network planning by a
national operator would be about 2-3 man years. At a total employment cost of
£100,000 per expert, the worth of the improved information is about £250,000 per
national operator per year. Assuming 6 national operators, (five mobile operators,
plus Airwave), and allowing for a smaller benefit (saving of one man-year of
expertise) to each of at least an equal number of regional and specialised
operators, suggests a total annual benefit to operators of about £2 million.
3.4.3 Benefits to end users

The benefits of using radio frequencies accrue disproportionately to end users, if
benefits are measured by willingness-to-pay. In relation to interference, however,
end users appear to be relatively tolerant; Ofcom investigates just a few

thousand cases each year. Even if the benefit to each user of resolving an
interference problem is quite high (figures used to estimate the rectification costs
of TV reception problems caused by planned tall buildings would suggest about
£1-200), the total value to end users of improved efficiency in the resolution of
interference problems is unlikely to exceed £1 million.
Benefits to end users arising from greater utilisation of frequency bands and
improvements in the services offered by operators will be more substantial; these
are considered in the relevant section.
3.4.4 Benefits to R&D

The improved flow of information relating to spectrum usage and interference that
an AMS network would produce will be of interest both to academic researchers
and to commercial enterprises engaged in R&D activities. While AMS information
might well be made available free of charge for these purposes, in principle
researchers are willing to pay for information used in their research projects.
We can distinguish between the value of the initial improvement in knowledge to
which deployment of an AMS network would contribute, and the continuing flow
of data that the network would generate. Valuing in advance the initial gain in
knowledge is exceptionally difficult; academic researchers would probably not be
willing to pay a significant amount, although an approach to funding R&D that
took account of the commercial opportunities that might (or might not) arise
would suggest a value of the information in the low millions of pounds.
Once there is an understanding of the type of information that can be produced
on a regular basis, researchers should be able more easily to assess its worth.
Research undertaken by commercial interests is likely to be the primary source of
potential revenues; academic researchers may be reluctant to pay for access to
data, even though they would derive significant benefit from doing so – for
example, by being able to devote available research budgets to other activities.
The value of the data to researchers can be measured as the saving that would
be made in having to generate equivalent information experimentally. We
assume that the cost of generating equivalent information would be to employ
one research assistant for one year, at a total employment cost of £50,000,
including laboratory space and equipment. By surveying ongoing research
projects in the radio field, we identified at least five where the type of data the
AMS system will produce would be valuable. So we consider a reasonable
estimate of the willingness of researchers in the UK to pay for AMS data would
be about £250,000 a year.
3.4.5 Benefits to spectrum trading

Spectrum trading is a form of secondary market in rights to use radio
frequencies. It should achieve a more efficient allocation of spectrum than would
be possible through auction or administrative allocation methods alone. The
precise scope and method of spectrum trading to be adopted in the UK is not yet
clear.
The principal effects of an AMS network on spectrum trading are likely to be:
a) improved information regarding the usage made of specific frequency bands
in specific areas, thereby enabling more frequencies to be traded;
b) improved ability to enforce property rights in regard to specific radio
frequencies, thereby raising the value of tradable frequencies;
c) reducing the cost of spectrum trading transactions
The scale, timing and precise form of spectrum trading in the UK has yet to be
determined. Once it is established, the annual trading turnover is likely to be
significantly less than the value of frequency rights acquired at auction. It is
assumed of the volume of spectrum trading will be 50% of all auction proceeds in
any year2, that is about £100 million3 a year – a small fraction of the annual value
of the frequencies that are tradable. The willingness of traders - buyers or sellers
of spectrum - to pay for the information that an AMS network would produce to
underpin their transactions is unlikely to exceed £1million (based on usual level
of expense on due diligence).
This underestimates the contribution that an AMS network could make to
spectrum trading. The availability of better information about usage and
interference will also have a one-off positive effect on the value of spectrum
property rights, increasing willingness to pay for frequencies at auction, or to pay
licence fees set by an administrative process.
3.4.6 Summary of Realisable Annual Benefits

The table below summarises the annual economic benefits that have been
identified and estimated in foregoing sections.
2
That is to say the annual value of spectrum trades is estimated to be about half the total
value of auctioned spectrum.
3
Currently planned auctions account for 290Mhz of spectrum in the next 12 months or so.
The most recent award (the 1781 Award in May 2006), in a perceived lower-value region,
brought in around £0.6m per Mhz. Future awards in the UHF band could be very highly
valued by the market; therefore an assumption of £200m auction income was made.
Beneficiary Annual benefit
Ofcom £8.5m
Operators £2m
End users interference £1m
Researchers £0.25m
Spectrum trading £1m
Total £12.75m
Table 3.1, Estimated realisable annual benefits
While there is almost certainly a wide margin of error around these estimates,
each estimate is likely to be towards the low end of the possible range of
estimates. That is, we consider that in each case there are additional benefits
that have not been quantified.
3.5 Costs Quantification

An economic model has been constructed which captures the chief costs of
procuring, building and operating a network of AMS units to cover all but the most
sparsely-populated areas of the UK.
In practise, many of these costs would not be incurred by the same organisation;
the model covers the costs that be born by all involved parties. A discussion of
financing and procurement options is included at section 3.6.
3.5.1 Modelling Assumptions

The UK can be modelled as regions of Remote Rural, Rural, Sub-urban, Urban
and Dense Urban population. We are not interested in covering those areas with
very limited transmissions, and therefore omit Remote Rural areas. The Office of
National Statistics defines these regions as having fewer than 20 people per
square kilometre. We have made no effort to determine where in particular any
of these regions are.
The AMS units will be deployed in a set of regular grids of varying densities, with
denser grids being required in more populous areas. The combination of known
areas and grid densities gave a total population of 982 installed AMS units. (An
alternative method, which began with a fixed number of AMS units and worked
back to a set of grid densities, was used until detection range information
became available).
Each of the covered regions is modelled as having a different type of AMS

installation, to account for the cost differences in installing and maintaining AMS
equipment in each region. Costs that vary from region to region were site
acquisition costs, mounting purchase and installation, and power and network
provision. A profile of capital and operational costs was built up for each region
type.
Having determined the quantity and location types of the AMS units, an outline
build-out schedule was devised, based on a limited build capacity, and resulting
in a four-year programme.
Maintenance costs were modelled as a programme of replacement, with a fixed
percentage of AMS units being replaced in every year. Separately, a certain
percentage of AMS units in each region are modelled as requiring re-location, to
account for changes in the surrounding environment. This is in line with Ofcom
experience in running the existing RMDF and UMS sites.
3.5.2 Source Data
3.5.2.1 Area Classification

In order to estimate the number of stations required, the anticipated densities in
the various urban and rural landscapes has been developed.
The following area classifications have been defined to suit:
• Remote Rural – Remote, largely inaccessible areas of the UK where a
density of zero stations will be required.
• General Rural – Sparsely populated countryside and coastal areas of the UK
containing villages and towns with a population of up to 10,000.
• Suburban – Suburban areas of the UK, which make up the largely residential
outlying areas of settlements with a population of 10,000 or more. Typically
consists of 1-2 storey houses.
• Urban – Built-up areas of the UK consisting of settlements of 10,000 of more.
Largely town and city centre areas with medium rise commercial and
residential premises.
• Dense Urban – A highly developed landscape with many close-knit and high-
rise, mainly commercial buildings.
It is recognised that a number of different urban / rural definitions exist and no
single classification meets the needs of all users.
The Office of the Deputy Prime Minister (ODPM) recommends using a definition of
Urban Settlements being those with a population of 10,000 or more [Source 1
below]. This results in an urban / rural split (by area) of 7% to 93%.
Population density statistics from the Office of National Statistics [Source 2 below]
are used to separate out Remote Rural from Rural. Remote Rural areas are
defined at the local authority level as those with a population density of 20 people
per square km. This largely includes the Highlands of Scotland and the Orkney
and Shetland isles.
Further data from the Office of National Statistics [Source 3 below] is used to
separate out Suburban, Urban and Dense Urban areas at statistical ward level.
This results in the following profile of the UK:
Rural Urban
93% 7%
Dense
Remote Rural Rural Suburban Urban
Urban
17% 76% 4% 2.5% 0.5%
Table 3.2, Percentage of UK in each Region
Sources:
1. Office of the Deputy Prime Minister
Urban and Rural Definitions: A User Guide
http://www.odpm.gov.uk/stellent/groups/odpm_control/documents/contentserve
rtemplate/odpm_index.hcst?n=3331&l=2
Accessed August 2005.
2. Office of National Statistics
Population Density, 2002
http://www.statistics.gov.uk/StatBase/ssdataset.asp?vlnk=7662&Pos=1&ColRa
nk=1&Rank=272
3. Office of National Statistics
Area Classification for Statistical Wards, 2001
http://www.statistics.gov.uk/about/methodology_by_theme/area_classification/
wards/cluster_summaries.asp
3.5.2.2 Device Detection Range- Number of Units Required
A simulation exercise was performed to determine the likely effective range of

detection of the AMS devices. The simulation was based on a regular grid of
AMS units at the points of equilateral triangles, placing each one in the centre of
a regular hexagon. Each of the four regions was determined to require a
different-sized grid of AMS units. Taken together with the area of each region,
this leads to a total number of AMS units of 9824.
Note that the large range shown against Remote Rural is arbitrarily chosen to
prevent the allocation of any units to those regions.
Range / km Area per device/ km2 (hex) Number of Devices
Remote Rural 10005 2598076.2 0
Rural 15.5 624.2 295
Sub-urban 4.3 48.0 201
Urban 2.4 15.0 405
Dense Urban 2.4 15.0 81
Total 982
Table 3.3, Range, Area and Number of Devices
3.5.2.3 Site Acquisition
Time and effort will be required to locate and acquire each site required for the
AMS system. This is modelled as a number of man-days effort with a cost of
employment for each officer involved. A different number of days was assumed
for each region-type, to include location of a suitable site, negotiations, and other
administrative works. A cost of employment of £186 per day was assumed for
the officers. It is assumed that sites in urban regions are more straightforward to
acquire.
In each region a flat procurement cost was also assumed, to cover costs of legal
fees, planning permission, etc.
Leading to the following capital costs (capex) in each area:
Rural Sub-urban Urban Dense Urban
Officer Days 20 20 14 14
Site Acquisition
(admin, etc) £ 3,720.9 £ 3,720.9 £ 2,604.7 £ 2,604.7
Site Procurement £ 1,000.0 £ 1,000.0 £ 1,000.0 £ 1,000.0
Table 3.4, Costs of Site Acquisition (per site)
3.5.2.4 Mounting Installation
Three different designs for mountings were considered - a steel lattice tower, a
monopole tower, and installation on existing roofs. These were each modelled
as the sole mounting type in the Rural, Suburban and in the two Urban regions
respectively- a simplifying assumption. Tower capital costs were taken from
estimates supplied by cellular operators; rooftop mounting capital costs were
supplied by Ofcom.
4
This figure is based on detection of 3G base stations.
5
1000 was chosen to be sufficiently large so that the model would not allocate any AMS
devices to this region.

Mounting purchase £ 15,000.0 £ 10,000.0 £ 1,500.0 £ 1,500.0
Mounting installation £ 15,000.0 £ 10,000.0 £ 1,000.0 £ 1,000.0
Table 3.5, Mounting Costs.
3.5.2.5 Site Rental
A different annual site rental cost was assumed in each region, based on
discussions with Ofcom about the rental costs paid for placing the existing DF
and UMS equipment, as follows:
• Rural - £2,000.0
• Suburban - £3,000.0
• Urban - £5,000.0
• Dense Urban - £10,000.0
These were treated as Operational Costs (Opex).
3.5.2.6 Churn
Experience of operating the existing DF and UMS equipment has shown that
there is a significant level of Churn. Owing to various factors, new sites have to
be acquired from time to time. For example, changes to building ownership or
use, demolitions, proximity of new radio-sources, etc, could all lead to the need to
decommission a site, and acquire a new one.
We have assumed that each re-location calls for a new expenditure of Site
Acquisition Costs and Mounting costs as described in sections 3.5.2.3 and
3.5.2.4, as opex. In line with Ofcom experience, we have assumed the following
levels of re-location are required annually:
• Rural - 1%
• Suburban - 5%
• Urban - 10%
• Dense Urban - 10%
3.5.2.7 Power and Network Access
Each AMS site requires a power supply and access to a reliable data network. In
each region, a Capital Cost was assumed for the supply of the required
infrastructure, and an Operations Cost for the services were assumed, as follows:

Power Provision £ 10,000.0 £ 6,000.0 £ 3,000.0 £ 3,000.0
Network Provision £ 2,000.0 £ 1,500.0 £ 1,000.0 £ 1,000.0
Power (annual) £ 250.0 £ 250.0 £ 250.0 £ 250.0
Network (annual) £ 500.0 £ 500.0 £ 500.0 £ 500.0
Table 3.6, Power and Network Access Costs
Power provisioning and service charges are in line with Ofcom experience from
the UMS equipment. Annual network costs are based on monthly charges of
£40, the norm for low-rate wired access. However, this cost is likely to vary with
the market trends in telecommunications costs, and according to the particular
communications network architecture chosen for deployment.
3.5.2.8 Maintenance
Although it seems clear that the AMS units have few risk points with the potential
for hardware failure, a regular program of maintenance has been modelled.
Maintenance costs are based on the assumption that, in each region, one in
twenty devices, and one in twenty mountings fail every year, and must be
replaced. This is not to imply any particular lifetime for the production-spec AMS
units.
3.5.2.9 Other Costs
Several costs for planning and operating the network have been included in the
model, these are:
• Network Planning - £50,000
• Core HQ Equipment - £100,000
• HQ Staff Costs - £200,000
• Core Network Maintenance - £50,000
• AMS Units Hardware & Antenna (each)- £10,500
• AMS Unit Software (each)- £1,018
The total software cost is allocated amongst the individual AMS units, so that
each unit has a software cost of just over £1000. It is clear that the software cost
per unit does not scale linearly with the number of units. The figure quoted
above is based on an enterprise-wide license of £1,000,000 divided between the
982 units required. If fewer than, 250 units were procured; a per-unit cost would
probably be introduced.
3.5.3 Build-out
It is assumed that in each year of construction, 75 rural, 75 suburban, 100 urban
and 25 dense urban AMS units are installed. The core network and HQ is
established in the first year.
Year 0 Year 1 Year 2 Year 3 Year 4 Year 5

Core Sites 1 1 1 1 1 1
Rural Sites 75 150 225 295 295 295
Sub-urban Sites 75 150 201 201 201 201
Urban Sites 100 200 300 400 405 405
Dense Urban Sites 25 50 75 81 81 81
Total AMS Sites 275 550 801 977 982 982
Table 3.7, Installed Base of AMS Sites
3.5.4 Results
This gives a cash flow model as follows, in which for every year subsequent to
Year 5, a steady state is reached, requiring no further Capital Expenditure and
£6.5m in Operational Expenditure:
Capex Year 0 Year 1 Year 2 Year 3 Year 4 Year 5

Core £150,000 £- £- £- £- £-
Rural £4,367,945 £4,367,945 £4,367,945 £4,076,748 £- £-
Sub-urban £3,280,445 £3,280,445 £2,230,702 £- £- £-
Urban £2,162,298 £2,162,298 £2,162,298 £2,162,298 £108,115 £-
Dense Urban £540,575 £540,575 £540,575 £129,738 £- £-
Opex
Core £250,000 £250,000 £250,000 £250,000 £250,000
Rural £655,831 £983,747 £1,289,802 £1,289,802 £1,289,802
Sub-urban £901,657 £1,208,220 £1,208,220 £1,208,220 £1,208,220
Urban £1,392,093 £2,088,140 £2,784,186 £2,818,988 £2,818,988
Dense Urban £598,023 £897,035 £968,798 £968,798 £968,798
Total £10,501,262 £14,148,866 £14,728,661 £12,869,790 £6,643,923 £6,535,808
Table 3.8, Cash Flow Model
If the scheme extended to Year 24, the whole project would have a project net
present value (NPV) of these costs, given a 5% discount rate, of about £115m.
This can be considered the cost of building and operating the system as
described. The total Capital Cost (the sum of the first 5 rows of this table) is
£36.6 million. Once construction is complete, from Year 6 on, the assumptions
above contribute to an annual operational cost of £6.5 million.
3.5.5 Sensitivity Analysis

Some charts follow indicating the relative values of each component of the
overall system. They have been constructed by considering the relative
contribution of each component to the NPV, total capex and opex above.
The following chart shows the relative cost of each of the modelled regions. Note
that the cost levels are disproportionate to the size of the region.
3%
12%
27%
Rural
Sub-urban
Urban
Dense Urban
Other
37%
21%
Figure 3.2, Contribution of Each Region to the Project NPV

The following chart shows how each cost element contributes to the system cost.
The mountings and site rental costs are the highest elements in the model. AMS
units themselves form a comparatively small fraction of the total cost.
7% 3%
11% Site Aquisition
10%
Mounting
7% Pow er
Netw ork
6% AMS Units
Other (2)
9% Site Rental
Maintenance
44% 3%
Churn
Figure 3.3, Contribution of Each Cost to NPV
The following chart shows the contribution of each cost element to the Capital
Cost. Here, the mounting costs dominate.
0% 11%
31% Site Aquisition

Mounting
Pow er
Netw ork
39% AMS Units

4% Other (2)
15%
Figure 3.4, Contribution of Each Cost to the Capital Cost

As shown in the following chart, site rental costs are the highest component of
the operational costs of the model.
10% 4%
8%
4%
13% Pow er
Netw ork
Other (2)
Site Rental
Maintenance
Churn
61%
Figure 3.5, Contribution of Each Cost to the Operational Cost
3.5.6 Alternative Deployment Scenarios

The cost model seeks to account for all potential costs incurred in constructing
and operating the AMS system on a national scale. It assumes that every site
must be surveyed and commissioned individually as part of the deployment of the
system.
It should be possible to reduce the capital or operational expenditure of an AMS
system in a number of ways. For example, deployment of the system could be
focussed more precisely on more populous areas, effectively taking a longer
period to achieve national coverage than the four years assumed. However this
is likely to reduce both the tangible and intangible benefits to many of the
potential beneficiaries for reasons outlined elsewhere in this paper.
Costs could also be reduced by partnering with other organisations to provide
some of the required facilities. By partnering with a major landlord, site
acquisition and rental costs could be reduced. Site rental is a major component
of the operational costs of the modelled AMS system at over 60%. Potential
landlords would have to have a nationwide presence in rural, suburban and
urban regions - examples include the supermarket chains, the Highways Agency,
the Post Office or Network Rail.
Partnering with an existing operator of telecommunications infrastructure might
additionally reduce the costs of provisioning mountings, power, and network
infrastructure. As indicated, these costs make up about 55% of the capital costs
of the AMS scheme. Churn and maintenance costs might also be reduced in this
case. The UK’s cellular network operators each utilise thousands of cell sites for
their equipment, which have similar infrastructure requirements. The television
and radio broadcasters also have significant networks, covering the country with
over one thousand transmitters.
Cellular operators typically utilise specialist agents to identify suitable sites for
their equipment, negotiate with site owners and manage the installation. In
addition, many transmission sites are operated by independent organisations like
Arqiva and National Grid Wireless. However, a thorough assessment of the
potential for cellular or broadcast equipment to interfere with the AMS units’
effectiveness would have to be undertaken before a site-sharing approach could
be adopted - it is quite possible that it is technically impossible to site AMS
equipment close to major spectrum users without reducing the sensitivity and
therefore the usefulness of the AMS network.
Assuming that it is possible to reduce the overall site acquisition and rental costs
by 50% through a partnership with a major landlord, the capital cost of the AMS
network would be reduced from £39 million to £37 million. Assuming it is
possible to reduce the mounting, power provision and network provision costs by
50%, through partnership with an existing infrastructure owner, a further
reduction to in Capex to £27 million would be possible.
Scenario Potential Effect Capex Opex
Self-built N/A £36,630,942 £6,535,808
Landlord 50% reduction in site £34,584,221 £4,521,808

Partnership acquisition an d
rental
Infrastructure As above, plus 50% £24,045,971 £4,521,808

Provider reduction in power
Partnership and network
provision and
equipment mounting
Table 3.9 - Effect of partnership arrangements on Capex and Opex
3.6 Financing and Procurement Options

The analysis presented above suggests that the potential benefits of a
nationwide AMS system exceed the prospective costs. The benefits are,
however, distributed across a range of parties, including the public at large. The
direct benefits to Ofcom itself are not likely to exceed the direct costs that Ofcom
would incur in establishing and running the AMS system. That is, there would be
a net cost to Ofcom of deploying an AMS system. In this section, we consider
the options for financing this net cost.
First, as a regulatory authority, Ofcom does have the option of recovering the net
costs of deploying an AMS system via charges to users of the radio spectrum.
On reasonable assumptions, the increase in charges to spectrum users
necessary to cover the net cost to Ofcom of an AMS system would be about
10%6.
It should, however, be possible to take a more commercial approach to the
financing and procurement of the AMS system. We consider three options:
• A wholly commercial venture;
• A PFI-type approach to procurement;
• Outsourcing.
6
Assuming a four year deployment programme and an average life of assets of 10 years,
the annualised capital cost would be £2.4m; the annual operating cost is estimated at
£6.5m; the savings on existing staff costs are estimated at £4m. So the annual net cost
to Ofcom would be about £5m, which is 10% of Ofcom’s licence fee income, as reported
in the Annual Report 2004-5.
Under the commercial venture option, a private organisation establishes the AMS
network with a view to selling access to it, analyses and reports generated by it.
Commercial interest is likely to be stimulated by several factors:
a) Net revenue potential: the annual benefits quantified in section 4 that could in
principle be captured by a commercial venture exceed the projected annual
costs of the AMS system7;
b) Strategic positioning: the information generated would give the operator of the
AMS system potential first-mover advantages in developing new services,
including spectrum trading;
c) Export potential: successful establishment of a national AMS system might
lead on to similar opportunities in other countries.
These factors suggest operation of an AMS system might be commercially
attractive to equipment manufacturers, infrastructure operators and
communication service providers.
It would be necessary of course to ensure that the operator of the AMS system
did not secure an unfair commercial advantage through its ability to control
access to the information generated.
Ofcom would act as the project sponsor and would be a principal customer,
entering into a long term contract but not providing any guarantees. The transfer
of commercial risk to the operator would have the consequence that the charges
to Ofcom and to other users would have to exceed costs though obviously not
their willingness-to-pay. If the estimates of willingness-to-pay in Section 3.4 are
robust, then commercial interest would largely depend on the prospect of
securing significantly lower capital or operating costs than outlined in Section 3.5.
Under a PFI-type approach, a private organisation would establish the AMS
network, with a guaranteed annual revenue stream from Ofcom over its expected
lifetime (15-25 years). The scale of the investment is above the minimum level at
which project financing can be considered economic.
In return for the guarantee, Ofcom would expect to have a greater say in the
design and deployment of the system, and in the terms on which information
generated by the system was made available to third parties. While this would
have advantages in maximising efficiency gains in radio regulation and non-
discriminatory access to information, the result might be a solution too closely
tailored to UK requirements to be successfully exported.
Under an outsourcing approach, Ofcom would own the assets comprised in the
AMS system but it would be operated by a commercial organisation acting on
behalf of Ofcom. Ofcom’s outsourcing partner could also be responsible for
handling sales of information and other services to third party users of the
system.
7
The estimated annual benefit of £1m to end users from reduction of interference
probably could not be captured directly.
The merits of outsourcing in this case are that Ofcom would be able to secure
reductions in staff costs and efficiencies in operating the system while avoiding
giving long-term revenue guarantees to the operator. Ofcom would, however, be
required to bear the capital cost and the commercial risks of deploying the
system.
Which, if any, of these financing and procurement options can be taken up will
depend on market-testing to establish the degree of commercial interest, and on
the extent to which the conditions under which an operator would accept all or
part of the commercial risk of deploying an AMS system are compatible with
Ofcom’s requirements.
3.7 Conclusion
The costs of commissioning and operating an effective and inclusive AMS
system are significant. With a likely capital cost of tens of millions of pounds, and
an ongoing operational expenditure of millions of pounds, it would represent a
clear step-change from the current UMS and DF operations undertaken from
Baldock. We have shown how these costs could be financed in several ways,
and outlined potential savings from partnerships with site or infrastructure
owners.
However, the benefits of an operational and fully-featured AMS system are
considerable, and potentially dwarf the costs. Specific benefits have been
quantified above by assessing the willingness of the relevant parties to pay for
the information that an AMS system would produce; these benefits amount to
£12.75 million every year. The capital cost of the system is estimated at £36.6
million over four years, with an annual running cost thereafter of £6.5 million at
current prices. Based on these estimates, the project has a positive net present
value in excess of £42 million.
If benefits of the scale estimated could be captured as revenue, whether by
Ofcom or by a commercial enterprise, then a national AMS system could be
deployed on a self-sustaining basis. This may not be possible. Accordingly, a
range of ways in which the AMS system could be funded and operated have
been explored.
4 Modelling study of the proposed AMS –
(Work Package 3)
4.1 Introduction
A modelling study of the proposed AMS formed Work Package 3. The aim was to
determine the location accuracy of the interfering signal and the detection range
of the AMS allowing a more refined estimate of the likely number of stations.
Two location techniques (described briefly in Section 4.2) were considered:
• Direction finding (DF) based on multi-element arrays,
• Time Difference of Arrival (TDOA).
Most of this section concentrates on the Time Difference of Arrival technique.
Direction finding was discounted at an early stage as it can be shown that both
the accuracy and, more importantly the sensitivity is far greater for the TDOA
system. Additionally, a DF system would require large calibrated multi element
antenna arrays which do not meet Ofcom’s requirement for small inconspicuous
installations
4.2 Geolocation techniques
4.2.1 Direction finding
Obtaining an emitter position estimate from a DF (Direction Find/Fix) is

conceptually one of the simplest methods of geolocation (Figure 4.1). DFs can be
combined most simply by using triangulation in order to estimate the position of a
target. The same method can be used by multiple sensors each obtaining a DF
estimate of the same targets, or a single moving sensor obtaining multiple DFs of
the same targets at different positions as it moves.
Figure 4.1, Combination of DF to give a position
This method of geolocation is versatile, in the sense that it can be used with
either single or multiple-platform systems, and can be used even when there are
distinct differences between the sensors involved, as long as the errors
associated with each DF measurement are known.
Combination of DFs
The algorithms which combine the DF measurements have a significant influence
on the performance of the technique. The Stansfield algorithm is probably the
most common choice. This provides a method for statistically determining the
most likely location of an emitter given at least three lines of bearing (LoB) and
the associated error (if there are only two LoB, the most likely emitter location is
where the two lines cross) (Figure 4.2). Stansfield (1947) provides a statistical
means to determine the most likely location of an emission source, given LoB to
that source and the error associated with each bearing measurement. Using this
method, Stansfield shows that if there are three intersecting bearings, the most
likely emitter position, may not necessarily lie in the centre of where those
bearings cross. Stansfield also provides a method to calculate an area of
uncertainty/probability associated with the estimated emitter location. This
typically has an elliptical shape and is associated with a specified level of
probability. For example a 95% probability ellipse indicates that in 95/100 times
the actual emitter location will lie within that ellipse.
An alternative method of estimating an emitters location is to overlay lines of
bearing on a map to determine a probability image.
Figure 4.2, Combination of DF by Stansfield
4.2.2 Time Difference of Arrival (TDOA)
TDOA is based on the principle that distance/range is related to velocity and time
(Figure 4.3). The distance between an emission/signal source and the position of
a sensor can be calculated if the velocity of the signal is known (RF emissions
travel at c, the speed of light), and the time taken for the signal to travel that
distance
Figure 4.3, Time delay on a transmitted signal
The sensors do not have any a-priori signal information, so, although the sensors
can report the time at which a signal was received, a single sensor on its own
cannot determine the emission range, since it does not know at what time the
signal was transmitted.
Alternatively, if the same signal is intercepted by two sensors, the difference
between the time of arrival of the signal at the two sensor positions can be used
to estimate the relative range of the emission source to the two sensors (Figure
4.4). The Time Difference of Arrival of a signal between two sensors can be used
to generate an isochron. This isochron represents the locus of geographical
points that can satisfy the measured time difference. (a) With two static sensors,
only a single isochron can be generated. (b) Using TDOA alone, a position fix
can be generated if a third sensor is available, so that TDOA curves can be
generated for two additional baselines, and then combined to determine a PF.
Figure 4.4, TDOA with 2 and 3 sensors
4.3 TDOA modelling
4.3.1 Overview
A model has been developed to determine the range at which a signal could be
detected, and the accuracy to which it could be located using a TDOA based
system. For a comparison with a DF based system see Section 4.6.
The modelling was restricted to the use of TDOA by three monitoring stations to
estimate emitter position. One station was regarded as the ‘master’, with the
TDOA with respect to the other two stations being used to determine emitter
position.
The modelling assumes that TDOA values are estimated by finding the position
of the “peak” in the cross-correlation of the two signals.
The principle output of the modelling is a plot of geolocation error for a regular
grid of points within a defined rectangular region (Figure 4.13). The metric used
for geolocation error is Circular Error Probable (CEP) which is defined as the
radius of the circle centred on the true emitter position within which there is 50%
probability of the estimated emitter position falling. This metric is equal to the
median error distance of the position estimate.
Modelling of a particular scenario is in two stages. The first is the calculation of

propagation loss as a function of distance. The second is the calculation of
geolocation error based on that propagation loss profile, using a MATLAB model.
4.3.2 Calculation of propagation loss
Propagation loss as a function of range was calculated using the Microcomputer

Spectrum Analysis Models Land Mobile Services application (MSAM LMS)
produced by the US Dept of Commerce National Telecommunications &
Information Administration.
The following is an extract from the application’s Help file, regarding the pedigree
of the application:
Most of the programs were developed at National Telecommunications and

Information Administration (NTIA), both at the Office of Spectrum Management
(OSM) and Institute of Telecommunications Sciences (ITS). Some of the models
were developed at the Joint Spectrum Center (JSC), formerly the
Electromagnetic Compatibility Analysis Center (ECAC) of the Department of
Defense (DOD), Annapolis, MD, and were modified to meet NTIA's requirements.
The programs have been verified to be correct for many scenarios run over a
number of years. NTIA assumes no responsibility for the results generated by
these programs except to guarantee that the programs will run on Microsoft
Windows.
Various parameters can be set including frequency (MHz), environment (‘open’,

‘suburban’, or ‘urban’), city size (if applicable), height of base station antenna (m),
height of mobile antenna (m) and the percentage of time and locations for which
the calculated propagation loss is not exceeded.
The ‘Base Station’ is taken to be the monitoring station, and the ‘Mobile Antenna’
to be the transmitting source. ‘Height of Base Station Antenna’ is the HAAT
(Height Above Average Terrain) value, This is a formally defined metric, and is
essentially the height with respect to the average height of the terrain between 3
and 16 km from the antenna.
There are three propagation loss models available; Okumura-Hata ITU 529 [3][4],
the Modified Cost 231 [5] and the Okumura-Hata Davidson [6]. The ranges of
parameter values for which these are valid are shown in Table 4.1.
Model Okumura-Hata Modified Cost 231 Okumura-Hata

ITU 529 Davidson
Frequency 150 – 1500 1500 - 2000 30 - 1500
range (MHz)
Distance (km) 1 - 100 1 - 100 100 - 300
Base station 30 - 200 30 - 200 200 – 2500
antenna height
(m)
Mobile antenna 1 - 10 1 - 10 1 - 10
height (m)
Table 4.1, Parameter ranges for the available propagation loss models
The maximum distance for which the propagation loss is calculated needs to be
large enough to suit the scenarios that will be subsequently modelled. For larger
distances than this, the MATLAB model sets the propagation loss to an arbitrary
high value (300dB). The minimum distance for which the calculated propagation
loss is valid is 1 km. However, the ‘Report for Distance Range’ mode allows a
minimum distance of 0.1 km to be used. For distances less than the minimum
available, the MATLAB model sets the propagation loss to that for the minimum
available distance.
4.3.3 Theory of operation
Calculation of the geolocation error for each point of the grid has two steps. The
first is the calculation of the rms error associated with each of the two TDOA
measurements. The second is conversion of the TDOA measurement errors into
the corresponding error in position.
There are two conditions which lead to a valid error estimate not being obtainable
for a particular point. The first is poor geometry, as detailed in 2.4.4 below. The
second is if the cross-correlation of the signals used to estimate TDOA does not
have sufficient SNR for the correlation peak to be reliably detected. Note that the
modelling does not require the signal to be detected by any of the stations prior
to correlation.
4.3.4 Calculation of the rms error associated with each TDOA measurement
The rms error associated with each TDOA measurement is modelled as having
the following independent components:
1. The fundamental error in determining the position of the correlation peak due
to the finite SNR of the signals being correlated. This is the Cramer-Rao
lower bound (CRLB) value given by reference [7]. In computing the SNR, the
presence of a co-channel interferer is accounted for by adding its received
power to the computed noise power.
2. The error in synchronisation between the two stations.
3. A component to account for the shift in the signal correlation peak caused by
a co-channel interferer having its correlation peak very close to that of the
signal of interest. Although this is actually a deterministic bias in the
estimated TDOA value, it has to be modelled as a random component.
4. A component to account for the shift in the signal correlation peak due to
multi-path propagation effects.
The errors in the two TDOA measurements are also assumed to be independent.
4.3.5 Mapping of TDOA measurement errors to position error
To allow an analytic solution, geolocation using TDOA is assumed to make use

of the signals from only 3 monitoring stations. It is also assumed that the 2 TDOA
measurements which are used to compute the emitter position are between
station 2 and station 1, and station 3 and station 1 respectively.
Let the bearing and range from the emitter to the ith station be φi and Ri
respectively. Here bearing is defined as the angle anti-clockwise from the x-axis.
If the emitter were to be displaced from its true position by (dx dy), as shown in
Figure 4.5, then
dRi = -dx cos φi - dy sin φi

Equation 1
Φ2
R2 + dR2
R2
dx
dy
R3 + dR3
R1 + dR1
R1
R3
Φ3
Φ1
Figure 4.5, Relation between TDOA errors and position error
The effect on the TDOA measurements would be:
d TDOA21 = [ dx ( cos φ1 - cos φ2 ) + dy (sin φ1 - sin φ2 ) ]/c

Equation 2
d TDOA31 = [ dx ( cos φ1 - cos φ3 ) + dy (sin φ1 - sin φ3 ) ]/c
Equation 3
This can be expressed in matrix form
 dTDOA21   dx 
  = M  
 dTDOA31   dy 
Equation 4
Suppose now that the dTDOA terms are errors in measuring TDOA. (dx dy)
would be the resulting errors in the estimated position according to:
 dx   dTDOA21 
  = M −1  
 
dy  dTDOA31 
Equation 5
If the errors in the TDOA values are independent with variance σ2TDOA , then it
can be shown that the covariance matrix for the errors (dx dy) is given by:
σ2TDOA M-1 (M-1)T

Equation 6
If M is not invertible, it indicates that the geometry for geolocation is very poor, in
that one or both of the TDOA values is insensitive to the emitter position. This
occurs if the two stations making the TDOA measurement have the same relative
bearing with respect to the emitter.
If the covariance matrix has eigenvalues e21 and e22 , then the median error
distance CEP is approximated by 0.59 (e1 + e2).
Note that the covariance matrix is sensitive to the choice of station 1.
4.3.6 Model output
The model generates five outputs:
1. The plot of the geolocation error, over the defined rectangular region.
2. A thresholded plot of the geolocation error. The colour map of this plot has
been set to clearly indicate the points within the region for which a valid
geolocation estimate could not be obtained.
3. A contour plot of the geolocation error.
4. A text file detailing various statistics of the geolocation error, distinguishing

between the whole rectangular region and the triangular region defined by the
3 monitoring stations.
5. A plot showing the number of sensors which could detect the signal over a
given area.
4.3.7 Limitations of the model
The main limitations are associated with the calculation of the propagation loss.
For example, the effect of multi-path on geolocation accuracy is not explicitly
considered. More generally, the model only predicts geolocation accuracy for a
typical area of the defined type (‘Open’, ‘Suburban’ etc.). Much more detailed
modelling would be required in order to predict accuracy for a specific scenario.
4.4 Example applications
4.4.1 Required number of monitoring stations
The aim of this exercise was to determine the number of monitoring stations
which would be required to provide monitoring coverage of the whole United
Kingdom. The exercise consisted of the following steps:
1. Divide the total area of the UK into specified percentages of different area
types, corresponding to different propagation conditions.
2. Define a set of exemplar transmitter types.
3. For each combination of area and transmitter type, calculate the maximum
detection range by the monitoring station.
4. For each transmitter type, determine from the maximum detection ranges the
required density of monitoring stations.
The UK was divided into area types as shown in Table 4.2 below. This data was
supplied by Arup as part of their investigations into usage of the system.
Total area of the UK in km2 242,514
Area Type Proportion Area (km2)
Remote Rual 17.00% 41,227.4
Rural 76.00% 184,310.6
Sub-urban 4.00% 9,700.6
Urban 2.50% 6,062.9
Dense Urban 0.5% 1,212.6
Table 4.2, Division of the UK into area types
The area types were mapped into the Okumura Hata-Davidson (OHD) area types
according to Table 4.3 below.
Area type (Arup) Area type (OHD)
Remote rural Not required
Rural Open
Sub-urban Suburban
Urban Urban-Small/medium city
Dense urban Urban – Large city
Table 4.3, Mapping of area types
The following exemplar transmitter types (Table 4.4) were used as test cases for
calculating detection ranges. (The parameters were chosen in conjunction with
Ofcom).
Type Frequency Tx power Antenna Required receiver
(MHz) (W) height (m) bandwidth
Pirate radio station 100 10 10 150 kHz
Pager base station 153 100 10 25 kHz
Personal Mobile 450 5 10 12.5 kHz

Radio (PMR) base
Personal Mobile 450 0.5 2 12.5 kHz

Radio (PMR) mobile
GSM base station 950 1500 10 200 kHz (single

(Vodafone & O2) channel)
GSM mobile 900 2 2 200 kHz (single

(Vodafone & O2) channel)
GSM base station 1840 1500 10 200 kHz (single

(T-Mobile & Orange) channel)
GSM mobile 1750 2 2 200 kHz (single

(T-Mobile & Orange) channel)
3G base station 2000 20 10 5 MHz
Table 4.4, Exemplar transmitter types
Receive antenna height 20 m, 50 m for ‘Dense Urban’ area

type
Receive antenna gain 0 dBi
Transmit antenna gain (to monitoring 0 dBi

station)
Received SNR required for detection 10 dB
Receiver noise figure 14 dB
Table 4.5, Other modelling parameter values
The propagation loss observed at a given range for a given area type varies with
both time and location. For this exercise, the values of propagation loss used
were those which would not be exceeded for 50% of times and 50% of locations.
The criterion for detection is that the ratio of signal power to noise in the signal
bandwidth to exceeds 10dB. This is slightly different to how detection is carried
out in the demonstration system. In this, detection is with respect to a fixed
channel bandwidth of ~3kHz and there is some averaging of the noise prior to
detection.
Figure 4.6 and Figure 4.7 show some sample received SNR profiles. The
detection range is taken as the point at which the received SNR drops below the
10dB line.
Received SNR for several transmitter types

Rural environment GSM Base Station
80
Pirate Radio Station
GSM Mobile
60
3G Base Station
Personal Mobile Radio
40
10 dB detection threshold
20
SNR (dB)
0
1 10 100
-20
-40
-60
Range (km)
Figure 4.6, Received SNR for several transmitter types: Rural environment
Received SNR for several transmitter types
Urban environment GSM Base Station
80
Pirate Radio Station
GSM Mobile
60
3G Base Station
Personal Mobile Radio
40
10 dB detection threshold
20
SNR (dB)
0
1 10 100
-20
-40
-60
Range (km)
Figure 4.7, Received SNR for several transmitter types: Urban environment
The maximum detection ranges for all combinations of transmitter and area type
are given in Table 4.6 below.
Rural Sub-Urban Urban Dense

Urban
Pirate radio station 57 26 20 23
Pager base station 100 50.5 37.5 47.5
PMR base station 52.5 25 16.8 14.5
PMR mobile 12.5 4 2.3 3.7
GSM base station (Vodafone/O2) 81 37.5 24 19.5
GSM mobile (Vodafone/O2) 6 1.8 1 1.4
GSM base station (T-Mobile/Orange) 61 30 16.5 23.2
GSM mobile (T-Mobile/Orange) 3.4 0.95 0.4 0.5
3G base station 15.5 4.3 2 2.4
Table 4.6, Maximum detection ranges (km) for transmitter/area type

combinations
The table shows that in some cases the model predicts the detection range for
“dense urban” as being greater than for “urban”. A possible explanation is that
the model becomes less reliable as the distances become small, and it is starting
to exaggerate the advantage of antenna elevation.
For all practical purposes the detection range can be considered to be the same
for these two cases.
An approximation to the required monitoring station density for each transmitter
type and area can be calculated by assuming that the percentage of the UK
assigned to that area type consists of a single region, and ignoring edge effects.
The monitoring stations can be assumed to be disposed on a triangular grid, the
spacing of this grid being such that there is no point with a distance from at least
one station exceeding the maximum detection range. Each station therefore had
a notional hexagonal cell within which signals will be detected.
No coverage is required for the “Remote rural” areas of the UK.
The results of the exercise are given in Table 4.7 below.
Transmitter type Total number of

monitoring stations
Pirate radio station 35
Pager base station 13
PMR base station 44
Personal mobile radio (PMR) 1,217
GSM Base station – Vodafone/O2 21
GSM Mobile – Vodafone/O2 5,923
GSM Base station – T-Mobile/Orange 35
GSM Mobile – T-Mobile/Orange 27,776
3G Base station 982
Table 4.7, Total number of monitoring stations vs transmitter type
4.4.2 Practical sensor layouts
The above approach gives sensible values for those emitter types for which the
urban detection range is smaller than the size of a city. For instance, Table 4.7
states that 1166 sensors would be required in order to detect PMR mobiles
across the whole of the UK and Figure 4.8 backs this up by showing that 851
would be sufficient to cover England and Wales.
Figure 4.8, 851 sensors providing the ability to detect PRM mobiles across
England and Wales
The assumption that the urban land forms one contiguous region is however not
valid for those emitter types which have significantly longer detection ranges. In a
theoretical unbounded urban region, a triangular grid of sensors separated by
34.6 km, no point would be more than 20 km from the nearest sensor, enabling
pirate radio stations to be detected anywhere. In such a grid, each sensor would
provide urban coverage over a hexagonal cell with an area of 1039 km2, which is
roughly five times the combined area of the urban regions of Norwich, Ipswich
and Cambridge. In reality however, it is not possible for a single sensor to provide
coverage over even these three cities because they are separated from each
other by about 65 km. The quoted figure of 35 sensors may therefore not be
sufficient to provide a nationwide capability against pirate radio stations.
The model used above for determining detection ranges made the assumption
that a propagation path would be entirely within one land type. It was not able to
predict the range over which an emitter in a rural area could be detected by an
urban sensor. Initially, assume that the detection range depends on the location
of the emitter so that any rural pirate radio station can be detected from a range
of 57 km regardless of whether the sensor is in a city or not.
Figure 4.9 has three panes, which all use the same colour scale. The colour
represents distance between a sensor and an emitter, as shown in the
logarithmic scale at the bottom of the figure. In all three panes, the position on
the map refers to the position of the emitter. The left pane shows the maximum
possible range over which detection will be possible. For instance, an emitter in
an urban region can only be detected if the distance to the nearest sensor is less
than 20 km so the urban areas are shown in the dark blue colour which
represents a range of 26 km. The rural areas appear in the orange colour which
represents 57 km.
The middle pane shows what the actual distance to the nearest sensor would be
from each point on the map if sensors were deployed in 51 specific locations.
The colour of each point in the middle pane is at least as far to the blue end of
the spectrum as the corresponding point in the left pane, showing that for any
potential emitter location, the nearest sensor is closer than the maximum
detection range.
If 51 sensors are able to cover England and Wales then the number required for
the whole of the UK will probably be about 70 to 80.
.
Figure 4.9, Pirate radio coverage of England and Wales provided by 51 sensors
The right pane in Figure 4.9 shows, for the same arrangement of 51 sensors, the
distance from each point to its third closest sensor. It can be seen that very few
places are within 40 km of three sensors but that should not matter since the
received SNR is allowed to be much lower at the second and third receiving
station than at the first. The worst case is the city of Middlesbrough, which is
roughly 100 km away from its third closest sensor. This is about five times the
predicted detection range. Under an inverse fourth power law of propagation,
increasing range by a factor of five reduces power by 28dB so it is possible that a
signal originating in the city could arrive at the third sensor with a SNR of -18dB.
This should still be sufficient to produce a measurable peak when cross-
correlated with the signal recorded with +10dB SNR at the detecting sensor.
The assumption that any rural pirate radio emitter can be detected from 57km
away regardless of the sensor location is unlikely to cause a problem since most
of the cities shown in Figure 4.9 are small enough that the sensors which monitor
them could themselves be situated in a rural area just outside the city. For
instance, a sensor just to the east of Bristol could have the whole of the city
within its 20km urban range while providing rural coverage as far as
Marlborough.
4.4.3 Relation between correlation sample length and geolocation precision

The aim of this exercise was to generate modelling results which showed the
relationship between the amount of data used in the estimation of TDOA and the
achievable geolocation precision. These results were to assist in determining the
requirement for communications bandwidth between the stations in order to
cross-correlate signals and hence estimate TDOA.
The following parameters were used:
Area type Open
Transmit antenna height 2m
Receive antenna height 20 m (minimum

allowed by
propagation model)
Transmitter power 1W
Transmitter gain 0dBi
Receiver/signal bandwidth 20kHz
Receiver gain 0dBi
Receiver noise figure 14dB
Propagation loss not exceeded for 50% of the time and for
50% of locations.
Table 4.8, Parameters for modelling exercise
An ideal sampling rate is assumed at twice the receiver bandwidth. In practice, a
slightly higher sampling rate would be required (and hence slightly more
samples).
The receiver and emitter geometry is close to the optimum for geolocation. The
receivers are positioned at three corners of a square, with the emitter at the
centre of that square.
The only source of error is that in estimating the position of the correlation peak
due to the finite SNR of the signal. This is computed as the Cramer-Rao lower
bound (CRLB) value given by reference [7]. In practice there would be additional
errors due to (for example) the presence of co-channel signals, multipath effects
and less than perfect synchronisation between the receivers. However, these
additional errors are not sensitive to the number of data samples used.
The error metric used is ‘circular error probable’ (CEP). There is a 50%
probability that the estimated position will be closer to (or further away from) the
true position than the CEP.
4.4.4 Results
Results of simulations based on the conditions stated in Section 3.2.1, for three
receivers – emitter ranges and at 5 emitter frequencies, are given in Figure 4.10
to Figure 4.12. The number of samples refers to the number of data points
captured at each sensor for correlation to calculate the TDOA.
Geolocation error vs number of samples
Range 50 km
10000
1000
Error (CEP) metres
30 MHz
100 MHz
100 300 MHz
1 GHz
2 GHz
10
1
100 1000 10000 100000 1000000
Number of samples
Figure 4.10, Geolocation error vs sample length. Range 50 km

Range 25 km
10000
1000
Error (CEP) metres
30 MHz
100 MHz
100 300 MHz
1 GHz
2 GHz
10
1
100 1000 10000 100000 1000000
Number of samples
Figure 4.11, Geolocation error vs sample length. Range 25 km
Range 10 km
10000
1000
Error (CEP) metres
30 MHz
100 MHz
100 300 MHz
1 GHz
2 GHz
10
1
100 1000 10000 100000 1000000
Number of samples
Figure 4.12, Geolocation error vs number of samples. Range 10 km
4.4.5 Scaling laws
The geolocation error will be inversely proportional to the square root of the
signal power, so will halve for a 6dB increase in signal power.
If the receiver bandwidth (and hence the sampling rate) only increases, then for
the same number of samples the geolocation error will be proportional to the
square root of the receiver bandwidth.
If the receiver bandwidth matches the signal bandwidth, then (for the same
number of samples) the geolocation error will be inversely proportional to the
square root of the receiver bandwidth.
If the receiver bandwidth is less than the signal bandwidth, then (for the same
number of samples) the geolocation error will be inversely proportional to the
receiver bandwidth.
4.5 Graphical output
The figures included in this sub-section were generated using parameter values
which illustrate the utility of the figures rather than values which correspond to a
particular scenario.
The most important graphical output is a plot of geolocation error over the
defined rectangular region of interest as shown by the scales on the two axes. An
example of such a plot is shown in Figure 4.13. For this example, an area of
15km by 15km is being assessed and the monitoring stations are positioned just
outside the top-left, bottom-left and top-right corners and are represented by
black crosses. The colour at a given point represents the median error position
error which there would be on a geolocation result for an emitter located at that
point.
Figure 4.13, Plot of geolocation error as a function of target emitter location
Figure 4.14 refers to the same hypothetical scenario as Figure 4.13. The colour
at each point represents the number of sensors which would have sufficient SNR
to report that a signal was present. It is only necessary for one sensor to report
the signal in order for an automatic geolocation to be tasked.
Figure 4.13 showed that geolocation would still give a result even if the emitter
was in the region at the bottom right of the plot where it could not be reported by
any of the sensors. However, to geolocate an emitter in this position, the sensors
would have to be tasked manually.
In a nationwide system with a large number of sensors, the sensors should be
close enough together that every possible emitter position will result in a
detection at at least one sensor.
Figure 4.14, Number of sensors with SNR above detection threshold
The height of the correlation peak relative to the noise in the correlation surface
must be above a threshold value to ensure that the peak can be found reliably
without excessive false alarms. Figure 4.15 is a plot showing how many reliable
TDOA measurements would be given as a function of emitter position in the
same scenario. For an emitter in the blue or green regions, it will not necessarily
be possible to identify a clear peak in both correlation surfaces. However, if both
peaks are identified, the error in the resulting geolocation will be that given in
Figure 1.3. The geolocation errors are small even for correlation peaks which are
barely detectable because the large bandwidth used in this scenario means that
the correlation peaks are narrow.
Figure 4.15, Number of reliable TDOA measurements
4.5.1 Co-channel signals
Figure 4.16 shows the potential effect of a co-channel signal on geolocation

capability. For this example, the co-channel emitter is situated at the bottom-right
corner. For each of the two sensor pairs, there is a green band within which the
TDOA peak for the target emitter would overlap that of the co-channel emitter,
resulting in an error in the determined TDOA. The width of the bands is
determined by the bandwidth of the emitters. The actual geolocation error due to
the presence of the co-channel emitter will depend on its power. A co-channel
signal still has an effect even in the red region because it acts as noise, reducing
the SNR for the target signal.
Figure 4.16, Potential effect of a co-channel signal
4.6 Comparison with DF accuracy
Table 4.9 gives the maximum detection distances allowed from an emitter to its
closest sensor for different types of signals in different environments. Provided
the three sensors used for TDOA surround an emitter, the geolocation error does
not vary significantly with the distance from the emitter to the sensors, only on the
received SNRs and the number of points captured.
For direction finding however, there is a constant angular error so the transverse
geolocation error scales with distance from the sensor. With a modern ground-
based DF sensor and 10dB SNR in a clean rural environment, the DF error is
typically about 2 degrees. In an urban environment, errors of about 20 degrees
can be expected. The DF error does not depend on the signal bandwidth.
Provided that a close to optimum antenna array size can be chosen for each
signal of interest, the error is not frequency dependent.
If an emitter is geolocated by DF from two sensors at roughly equal distances
from it then the expected geolocation error is roughly the product of the distance
to one sensor and the sine of the angular DF error. The geolocation errors which
result for emitters at the maximum detectable range for each of the scenarios in
Table 4.6 are given in Table 4.9.
Dense
Emitter type Rural Suburban Urban
urban
Pirate radio station 2000 910 7000 8000
Pager base station 3500 1800 13000 17000
PMR base station 1800 870 5900 5100
PMR mobile 440 140 800 1300
GSM 900 base

2800 1300 8400 6800
station
GSM 900 mobile 210 63 350 490
GSM 1800 base

2100 1000 5800 8100
station
GSM 1800 mobile 120 33 140 170
3G base station 540 150 700 840
Table 4.9, Geolocation errors, in metres, expected from DF systems at the

maximum detection ranges
The yellow line in Figure 4.6 shows that a 2W 1GHz emitter at a height of 2m
would be received with a SNR of +1dB at a sensor which was 10km away and
had a height of 20m. A 1W emitter at the same frequency would therefore be
received with a SNR of roughly -2dB. Figure 4.12 shows that this 1W emitter
would be geolocated with an error of about 9m if captures of 105 points were
used with a receiver bandwidth of 20kHz.
Therefore, if TDOA sensors are perfectly synchronised and spaced so that they
receive signals with about 10dB SNR then captures of 105 points can locate
emitters with 20kHz bandwidth to about 9m. The narrowest of the example
emitter types used in this report has a bandwidth of 12.5kHz so it would
theoretically give a position error of about 15m. The signals with greater
bandwidths should give smaller errors. However, the most likely way of
synchronising the sensors is using GPS, which introduces further errors of about
50ns, resulting in 20m geolocation errors.
In previous trials, mobile phones have been located to about 40m and broadcast
FM radio stations to about 100m. A TDOA system in which the distances from
the emitter to the sensors was roughly the maximum detection range should be
able to give similar accuracy whereas Table 4.9 predicts that a DF system with
the same sensor spacing would give larger errors.
In a recent trial, two groups of emitters with bandwidths of the order of 10kHz
were used as targets for geolocation by a pair of TDOA sensors and by a
network of three DF sensors. The DF sensors were fixed and formed a 10km
baseline. The errors with which they geolocated the targets depended on the
distances from the sensor baseline to the targets. It was found that targets 10km
from the DF baseline generally gave positional errors of about 5km while targets
at 30km range gave errors of about 20km. The TDOA sensors in the trial were
constantly moving and were typically each about 30km away from the emitters.
The RMS TDOA errors were about 700ns against one type of emitter and 9000ns
against the other. If the TDOA sensors had been stationary and surrounding the
emitters, these figures would translate into positional errors of the order of 250m
and 3km. In this situation, TDOA was shown to give greater accuracy than DF by
roughly an order of magnitude
It has been suggested that more sensors will be required in order to provide a
nationwide geolocation capability using TDOA than would be necessary using DF
since for TDOA location, the signal needs to be received at three sensors
whereas for DF location, it only needs to be received at two. However, there are
two reasons why this is not true.
Firstly, in a regular square grid of sensors where adjacent sensors are separated
by one unit of distance, the distance from an emitter to its second closest sensor
is a maximum when the emitter is located close to one of the sensors, as shown
in Figure 4.17. The distance to the second closest sensor is then just under one
unit. The distances to the third, fourth and fifth are each also close to one unit.
Similarly, in a triangular grid, the distance to the second closest sensor is again a
maximum when the emitter is close to one sensor but then the seventh closest
sensor is hardly any further away from it than the second. In general, any regular
two-dimensional arrangement of sensors which ensures that there will always be
two sensors within a given distance of an emitter regardless of the emitter’s
position will inevitably provide a third sensor within the same distance.
Figure 4.17, The maximum possible range to the third closest sensor is not
significantly greater than the maximum possible range to the second closest
sensor
Secondly, for a TDOA system, only one sensor needs to be sufficiently close to
the emitter to have the 10dB SNR necessary for signal detection and
identification. The second and third sensors are tasked by the first and it will be
possible to produce TDOA measurements by cross-correlation even if the SNRs
at the second and third sensors are negative. With a DF-based system, it is still
only necessary for one sensor to do the detection and identification but DF from
the second sensor will not be possible unless the SNR of the signal received
there is positive.
5 AMS System design – (Work Package 4)
5.1 Introduction
The Ofcom ITT requested the design of an unattended monitoring system whose
primary purpose is to automatically detect, identify and locate the source of
interfering radio signals over a large part of the UK. This was carried out under
Work Package 4 of the project. A requirements analysis was performed which
took into account the results of Work Packages 1 and 3 and also discussions with
Ofcom. This section of the report describes the outline design of the final
operating system.
The solution should comprise a large quantity of monitoring stations, which are to
be networked for control, data correlation, and reporting purposes and the
system should be capable of fully automatic operation or manual control.
Additional requirements are routine monitoring of the spectrum to establish
occupancy, potentially providing Ofcom with a means of validating a database of
licensed user activity.
5.2 Requirements for the AMS

The following points summarise the requirements of the system as described in
the ITT. These were expanded upon in the QinetiQ proposal, and it is mainly
these requirements that are listed. Some further requirements are also included
which have emerged during subsequent discussions with Ofcom.
1. The units should be capable of measuring interference from 20MHz to 3GHz

and should have a low noise floor, close to thermal noise, for detection.
2. Due to the numbers involved these units should be low cost and, to minimise
planning permission issues, should be small and use small antennas.
3. The frequency channels and bands to be monitored should be set
automatically over a communications link. It is expected that the units will
have a controllable bandwidth, scan rate and dwell time.
4. They should use a reliable method of detecting the interfering signal with
sufficient receiving sensitivity achieved to be usable across the entire
frequency range.
5. To maximise the use of the AMS units, they need to be capable of carrying
out routine spectrum monitoring tasks in measuring channel or band
occupancy, as well as automated and manual interference checks.
6. The system can be used in places without dedicated communications links,
hence they will be able to use existing structures such as radio or cellular
networks to download recorded data and send new instructions.
More precise details of the exact requirements are beyond the scope of this
document but can be found in the Design File [8].
The following two tables summarise how these requirements have been met in
the demonstrator system, and also what additional features would be included,
and what changes could be made to reduce costs and improve performance for a
fully deployed network of sensors. Table 5.1 details the physical characteristics
and Table 5.2 the features.
Component Requirement Demonstrator Final System Comments
Enclosure Small, light and easy to 4u 19" ruggedised rack Custom built sealed For cost reasons on the demonstrator (as
mount. / transit case from outdoor unit. Small enough there are only three) COTS units are being
Dragon Cases, for wall mounting, or on an used wherever possible. With economy of
suitable for trials. antenna tower. Approx scale a production run could customise
COTS 19" rack mount size of a desktop PC. components allowing a more robust, compact
PC and RF trays. design. There may still be benefits to a 2 unit
design.
Power Mains power. Low voltage dc with a For safety reasons it is more sensible to
separate indoor mains operate the final unit at a low voltage. This
power supply unit. simplifies installation and maintenance
procedures, as well as reducing the
environmental requirements on the enclosure.
PC DSP processing Intel dual 3.2 GHz Probably AMD option for PC choice is a compromise between
techniques. XEON, COTS 19", 2u lower power dissipation. performance, cost and power consumption.
rack unit. AMD processors use less power which is
The philosophy is to use important for heat generation in an enclosed
a software-based This provides the high outdoor unit. Even so, cooling must be
architecture on low cost level of processing considered carefully.
hardware.This make the power required for the
system easily intensive modulation
upgradeable with recognition and
software updates. location processing.
Digitiser Reasonable probability Echotek GC314-PCI- Custom made digitiser with This is a high specification digitiser offering an
of intercept requires a FS. 105 Msample per on board signal processing increase in performance in both dynamic range
wide stare bandwidth. second digitiser / or FPGA to reduce and staring bandwidth over our previous
Following discussions tuner. workload on PC. systems.
with Ofcom it has
become apparent that a
high dynamic range will
be required to work in a
“noisy” environment
such as London.
Receiver 20MHz to 3GHz TRL MRX 3500 TRL MRX 3500, possibly This is a high specification receiver which
frequency range. Low with a low bandwidth option provides a wide stare bandwidth and high scan
noise figure and good to give improved sensitivity rates with a compact form and low cost
sensitivity. in a high signal strength
environment.
GPS TDOA location Trimble Thunderbolt. Either GPS chips bought in For size and production cost a customised
calculations require direct from Motorola, or option may prove to be more suitable than a
accurate position of the This has a 20ns timing existing TRL GPS module. COTS item. QinetiQ will work with TRL to
sensor, and precision accuracy, to meet the further develop an in-house designed board
timing to sub 100ns precision timing of that they have.
accuracy. TDOA
Demodulation Discussion with Ofcom Software based Software demod on IQ Hardware demodulation has also been
has highlighted the architecture provides data, with ability to store considered, but software is considered to offer
importance of being able upgrade path for and access across the a more flexible approach
to demodulate an audio features such as network
signal. demodulation with no
change to the
hardware.
Switching / 20MHz to 3GHz TRL in-house circuit TRL printed circuit board. To cover this frequency range at least two
distribution frequency range. board - milled. antennas are required. Switching is needed to
select the relevant one for the frequency band.
Sampling Low phase noise clock TRL in-house clock TRL in-house clock This is the clock source that drives the ADC.
clock to drive ADC (phase module. 102.4MHz. module. 102.4MHz.. Good phase noise is essential to make best
noise will reduce the use of the high quality ADC cards.
sensitivity of the ADC).
Networking Sensors to be remotely 3G phones / Broadband. Wi-fi. 3G to The internet will form the backbone of the
configurable. broadband to connect connect to the internet. network. It is anticipated that the final system
to the internet. would use whatever network medium is
Co-operative TDOA available to connect to the internet. This is
locations. likely to be broadband wherever possible. Real
time spectrum displays may not be available,
Data to be transferred and location of weak signals could have
from remote sensors for latency across a narrow band link.
analysis.
Antennas Small antenna. HIGH BAND: AOR- HIGH BAND: AOR-DA5000 Size of antennas is important on the final
DA5000. . system for planning permission reasons. This
Discussion at meetings makes low frequency performance more
suggests that an challenging. Antennas for the final system
antenna should fit into must be robust to survive outdoor conditions
1m3 for planning without degradation.
permission reasons.
LOW BAND: Scanking LOW BAND: Robust

Royal Discone base version of demonstrator
antenna - RN40T. antenna.
Cables Low noise figure and For both Antennas Cable types and length The demonstrator cables need to be flexible /
good sensitivity. CFD400 Low loss 50 chosen for specific low bulk for packing and transport. The final
Ω coaxial cable with N installations. system would be fixed, so cables could be
type connections. Max more rigid to reduce cost, and give better
2.8dB loss. performance.
Table 5.1, Physical characteristics
Human Remote operation. Interactive software As for demonstrator but The software is based on QinetiQ software
Machine interface allowing with additional control for specifically developed over the last 7 years to
Interface (HMI) control of local and networked sensors manage and control multiple wideband
networked sensors. including data scanning sensors in an intuitive manner.
Displays for spectrum management, and
monitoring, location automatic modes for signal
and network (sensor) identification and location
control. without operator interaction.
Scan 20MHz to 3GHz 20 - 3000MHz in 20 - 3000MHz. By Default The proposed hardware is capable of
frequency range. Scan 40MHz blocks - capture will be a 40MHz supporting a scan rate of 40GHz a second.
control. approx 10s per scan. block, but with an option to However, there must be a compromise
reduce this in high signal between detection sensitivity and speed. The
level environments. software approach allows the flexibility to
Variable scan, down to 1s. optimise for different scenarios, and to
upgrade as more sensitive detection
algorithms are developed.
Networked Sensors to be remotely Remote monitoring of Full control of remote The final system will be capable of providing
monitoring configurable. spectrum / emissions. sensors. Network hierarchy full control of remote sensors from either
operation Spectrum Occupancy. for data management regional or national control centres. Some
Tasking. features are bandwidth intensive (e.g. real time
spectrum / emissions displays), so might only
be available on higher bandwidth links.
TDOA Location capability. System consists of 3 Control of all surrounding Only three demonstrator units are being built,
location sensors, each one can sensors. Optimal so the location technique is limited to
AoA was discounted as task the other two combination of intersection of two TDOA lines. Improved
an option after sensors to co- measurements to calculate methods have been investigated for optimal
discussion at a progress operatively perform a location from multiple combination of measurements from an array of
meeting because of location. sensors. sensors.
antenna size, so only
TDOA is being
considered as an option.
Modulation Modulation recognition. QinetiQ signal QinetiQ signal classification The QinetiQ modulation classifier proposed for
Recognition classification software. software. the AMS unit would have the capabilities to
process the following modulation types:
Analogue: CW, AM, FM/PM, FM, SSB, DSB,

USB, LSB.
Digital: ASK, MSK, FSK, PSK, QAM, ASK/PSK

Hybrids
Additional Signal Type Descriptors: Voice,

Tone, Morse, Radio Teletype (RTTY), CTCSS
and Chirp detection is also applied to suitable
signals.
Spectrum Short term emissions As for demonstrator, but

occupancy history, showing all with the ability to work over
emissions against a network connection.
time. Long term view Occupancy data files are
records average large so search a facility
occupancy over 15min will allow only selected data
intervals. Available to be downloaded.
bandwidth locator
allows identification of
unoccupied bandwidth
against user defined
criteria.
Database Archiving and retrieving Results of spectrum Data will be stored in a

data. occupancy and signal format compatible with the
recognition stored in a Ofcom database. The
predefined text format database allows for
allowing search or automatic tasking of signal
import into database classifier or location on
packages. unlicensed signals.
Table 5.2, Features of the AMS
5.3 The proposed solution
The remainder of this chapter gives an overview of a proposed system that

addresses the issues listed above, and explains some of the design
considerations leading to this solution. A high level system diagram is shown in
Figure 5.1.
The modelling study concluded that TDOA was the preferred option (see Section
4.6). Hence this section concentrates on the design of a TDOA system.
LOW BAND HIGH BAND

ANTENNA ANTENNA
ANTENNA
SWITCH
GPS RECEIVER
UNIT
ADC DIGITISER
CLOCK
PROCESSSOR INTERNET
AND CONNECTION
SOFTWARE
Figure 5.1, Systems architecture of the proposed solution
5.4 Solution Architecture
The original QinetiQ proposal suggested the system would consist of antennas
and two separate units, one located at the mast top next to the antennas and a
second indoor mounted unit connected to the mast top unit using a single coaxial
cable. During the project, a single-unit solution has been discussed for the
operational equipment.
The response to Ofcom’s ITT proposed a candidate system design. The design
shows antennas to capture the signals, receivers for down-conversion to IF, ADC
for digitising the signals and high specification computers for processing.
Additionally, a GPS is included for timing and an oscillator to provide and
accurate clock to ADC.
The units are arranged into two separate units. The outdoor unit contains the
receivers to keep them as close as possible to the antennas to reduce RF cable
losses. Signals are transferred to the indoor unit at IF where cable losses are
less of a problem. By keeping the majority of the equipment indoors the
environmental requirements are eased.
However, with an IF of 76.8MHz (as used by the TRL receivers) interference from
in-band transmissions on long cable runs is likely to be considerable, and with
the sensitivity of a correlation based location system this is not considered to be
a good option. Hence, now a single-unit solution is being proposed for the
operational equipment
5.4.1 ‘Two-Unit’ Architecture
In this approach, the RF functionality is placed in a unit to be mounted near the

antennas – ‘outdoors’. A second unit containing the remainder of the
functionality will typically be located ’indoors’ (Figure 5.2).
Figure 5.2, Two-Unit Architecture
The outdoor unit would contain antenna inputs, antenna switching matrix,
receivers, GPS, and a diplexer for sending information to the second ground
mounted unit. The lower unit would contain the unit’s power supply, digitiser, and
computer for signal processing, command and control and logging, and diplexer
for interfacing to the other unit.
Figure 5.3 expands on the content of each unit.
LOW BAND HIGH BAND

ANTENNA ANTENNA
NETWORK
ANTENNA PC
SWITCH GPS UNIT
1pps
RF
10MHz
RECEIVER ADC DIGITISER

CLOCK
IF IF
10 MHz RS232 10MHz RS 232
ANTENNA MOUNTED UNIT BOTTOM UNIT
10 MHz, IF, RS232
Figure 5.3, Details of the two unit architecture
A variant increases the functionality in the outdoor unit by positioning the digitiser
there but this is offset by the disadvantage of adding a high speed digital link
between the units. The indoor unit becomes a computer processing unit and
power supply for the outdoor unit.
Design Advantages
• Separate analogue and digital units simplifies the internal RFI design,
• Outdoor unit is physically smaller,
• Solar heat gain and internal power dissipation in outdoor unit will be lower.
Design Drawbacks
• Inter-unit communications link requires a custom design.
• Analogue inter-unit communications link - a connection from the receiver IF
(78MHz centre frequency) to the ADC is required. Losses in this cable run
could be large for the potentially long distance between units. Furthermore,
mobile radio transmissions and the low frequency end of band II FM
broadcasting are within this IF bandwidth and are likely to result in direct
signal pick up on the cable appearing at the ADC input at unacceptable
levels. The use of double or triple screened coaxial cable for the IF link will
increase the installation cost.
• A design variation would involve moving the digitiser to the outdoor unit. This
requires a digital inter-unit communications link - the raw data rate would be
102.4Msamples/sec at 14bits and allowing a margin for synchronisation and
management, a data rate of the order of 1.5Gbps results. This would require
the addition of an optical fibre link to replace the existing coaxial cable link.
5.4.2 ‘One-Unit’ Architecture
An alternative architecture (Figure 5.4) with all the components located in a

single unit will avoid the problems of IF interference. Although nominated a ‘one-
unit’ solution, this architecture will require two units:
• Outdoor unit with all the functionality - to be mounted on the mast,

• Inside unit consisting of a power supply unit to provide low-voltage power to
the outdoor unit.

LOW BAND HIGH BAND
ANTENNA ANTENNA
ANTENNA
SWITCH
10MHz RECEIVER
GPS UNIT
1pps
10MHz
ADC CLOCK
DIGITISER
102MHz
Network
PC COMMS
Figure 5.4, One unit architecture
Design Advantages
Apart from reducing the susceptibility to interference at IF, this design approach
is simpler in that it consists of a single unit, which
• Minimises manufacturing costs,
• Simplifies installation,
• Minimal indoor accommodation for power supply,
• Reduces losses and potential signal degradation between outdoor and indoor
units,
• Reduces design effort and risk by eliminating the intercommunications link
between units.

Design Drawbacks
This approach puts the all functionality in the outdoor unit with the following
drawbacks:
• Thermal management - greater functionality in the outdoor unit produces
additional heat,
• Analogue and digital processing in one unit will require careful internal RFI
design,
• Greater power to be routed to mast.
On balance the one unit solution is the preferred choice, both for its improved
performance and ease of installation. The architecture will still require two units –
one with the functionality to be mounted outside near the antenna, and a second,
a power supply unit located inside a weatherproof structure.
5.4.3 Outdoor Unit
It has been estimated that approximately 100w would be dissipated by the

modules within the outdoor unit. There will be some hot spots, such as the
processors. It is proposed that a fan be used to circulate the air within the unit,
and heatsinks be provided on the inner and outer surfaces to maximise heat
transfer into the environment.
An additional solar protection shield should be provided on the appropriate faces
to minimise the effects of solar warming.
The outdoor unit should be designed to minimise the EMC risks and be
environmentally sealed to avoid the need for protecting the individual modules.
This unit would be of a generic design suitable for mast and wall mounting and
large enough for maintenance purposes in potentially difficult access situations.
Space must be available for adequate air circulation for cooling purposes.
A complementary suite of mounting brackets for the antennas will be needed.
5.4.4 Thermal Analysis: Power Dissipation
The estimated power dissipation budget for each unit is shown below in Table
5.3. Each element is described in subsequent sections.

Item Watts
MRX Receiver 8
Clocking, RF Switching 5
Digitiser 5
GPS & reference clock 5
PC & Co-processor 50
Disk drive 5
DC to DC converter PSU 25
Total 103
Table 5.3, Estimated power dissipation budget
Forced air cooling has traditionally been considered an unattractive option due to
the maintenance issues with fans. However, fans are available with a quoted life
expectancy before failure of 100,000 hours at 60oC - over 11 years continuous
operation. Turning it off when not required will extend this.
Ideally such a fan would operate on a cold-wall external heatsink. IP67 rated
fans are available and a cowling could be arranged such that the direct impact of
bad weather and ingress of foreign matter which might cause a blockage, are
avoided.
Some benefit would be attained by the use of an internal fan in a sealed unit by
redistributing heat from hot spots.
There are a number of alternative approaches: Finned heatsinks could be
provided on external surfaces. This requires the heat generating components to
be mounted on external walls to maximise thermal transfer to the external
heatsink. Natural convection could provide heat dispersion. Any solar shield
may impair natural convection.
Alternatively, the unit could contain louvers and internal heatsinks, which would
make them more thermally efficient. However, an opening unit dictates that
sealed or conformally coated component parts be used. EMC problems may be
exacerbated. See further discussion on materials in Section 5.4.10.
The size of the unit is also a factor – a densely packed unit has less ability to
distribute heat from hot spots by convection. A larger unit than mechanically
necessary to contain the operational components may be appropriate to provide
a greater degree of internal heat dispersion. However, an enlarged unit is likely
to incur greater solar heating effects, as described below.

5.4.5 Thermal Analysis: Solar Generated Heat
An outdoor mounted unit is likely to be subjected to solar gain. This effect is

estimated to be:
Outside Unit
Solar gain w/m2 1000
Unit mm 500
mm 400
mm 150
Sun on largest area w 200

side only
Allowance for other w 100
sides
Total w 300
Table 5.4, Effect of solar gain on outside unit
Judicious positioning of the unit can be used to minimise the solar gain effect –
positioning on north facing walls or in the shadow of other fixings.
The solar gain effects may also be reduced by the introduction of a second skin
to the unit, positioned some 20 -30mm from the unit surface. This becomes the
primary absorber of solar warming.
Dissipation of these levels of power will require careful thermal management
design.
5.4.6 Form Factor
The proposed unit is dimensioned to contain the proposed functionality. Thermal

management considerations are primary and may dictate that a larger volume be
employed to allow internal air flows to assist cooling.
Maintenance considerations may also dictate a less compact configuration within
the unit to ease the replacement of internal modules in the difficult working
conditions that could be imposed by the units’ locations.

5.4.7 Mounting Options
The outdoor unit will be of generic design for mast and wall mounting. It is
proposed that a selection of separate fixing plates be designed for diverse
mounting scenarios, e.g. wall, girder, mast etc.
A complementary suite of mounting brackets for the antennas will be needed.
5.4.8 Connectors
It is proposed to mount the connectors on the base of the unit to minimise

moisture ingress. A shielding sheet will be provided give a degree of protection
from direct driving rain.
Connectors will be required for:
• Low frequency antenna,
• High frequency antenna,
• GPS antenna,
• Link to LAN/WAN network,
• Link to GMS/3G antenna if no LAN/WAN,
• Power.
5.4.9 Indicators
The utility of indicators on the outdoor unit will be determined at the detailed
design stage. It is not expected to be located where any such indicator will be
readily visible. Any indicators could be provided inside the unit, visible to an
engineer with the unit open, thereby reducing the number of piercings and
potential leaks.
Basic indications, such as power and antenna selection could be provided.
5.4.10 Materials & Construction
Both polycarbonate and stainless steel units have been successfully deployed in
similar applications. Both provided adequate environmental protection, with
stainless steel additionally providing EMC shielding.
It is proposed stainless steel be used to exploit the EMC shielding effects.
5.4.11 Environmental Specification
The following environmental specifications have been derived by comparison

with the specifications of similar in service equipment.

Operating/Deployed
Item Limits Commentary

o
Temperature - high +60 C
Temperature - low -40oC
Humidity 100% Condensing
Altitude 4,500m
Vibration 0.91 g (5 to 500 Hz)
Shock 10 g 10msec duration
Driving rain 75mm/Hr Max
Table 5.5, Environmental Specification for Operating/Deployed Equipment
Storage (Un-powered)
Item Limits Commentary

Temperature - high +80oC
Temperature – low -50oC
Humidity 95% Non-condensing
Altitude 10,000m
Vibration 2.4 g (5 to 500 Hz)
Shock 40 g 10msec duration
Table 5.6, Environmental Specification for Stored Equipment
5.4.12 EMC Specification
The equipment is categorised as “IT and general equipment” and as such, the
appropriate harmonised standards for EMC performance are:
EN55022 Emissions
EN55024 Immunity
Specifications under the RTTE directives may also be appropriate for the outside
unit.

5.4.13 Power Supply Unit
This power supply unit provides a safe voltage supply to the outside unit.
The indoor unit need not be environmentally sealed or protected to such a high
degree as the outdoor unit. Fan cooling should not pose a problem. Solar
thermal gain is not expected to be a problem.
The outside unit is predicted to dissipate about 100W. Some margin for future
expansion may be allowed at minimal cost and size impact. Thus a minimum
250W at 24V supply is recommended, which may be implemented using a COTS
power block enclosed in a customer case that contains appropriate terminations
of cabling to the outside unit.
To cope with diverse installation requirements, wall mounted and 19 inch rack
mounted options may be appropriate.
5.4.14 Summary
The original QinetiQ proposal suggested the system would consist of antennas
and two small boxes, one located at the mast top next to the antennas and a
second unit ground mounted connected to the mast top unit using a single
coaxial cable.
During the design of the demonstration system units however it was concluded
that the two box solution was likely to suffer from RF pickup on the long IF
(intermediate frequency) cables, and that this would adversely effect the
performance of the system. Instead it would be preferable to accommodate all of
the functionality in a single unit that could be mounted close to the antennas.
This avoids having long cables at IF, and complex communication between units,
but does increase the size of the outdoor unit, and introduces extra complexity
with weatherproofing and cooling.
Although nominated a ‘one-unit’ solution, this architecture will require two units –
one with the functionality to be mounted outside, and a second, a power supply
unit located inside.
5.5 Antenna
Choice of antennas is important to the performance of the system, both for

detection sensitivity and location accuracy. The requirement is challenging to
meet in that the system is to cover a large frequency range (20MHz to 3GHz)
with ideally a flat gain of around 0dBi. Furthermore, the size of antennas is
important as this could affect where the systems can be located, both for
structural reasons, and to meet local planning permission rules. The smaller the
antenna is the less likely it is to cause concern.

Antennas for the final system must be robust to survive outdoor conditions
without degradation, and have lightning protection circuitry to protect the rest of
the AMS unit.
Testing of some candidate antennas suggests that to cover the this frequency
range (20MHz – 3GHz) two separate antennas will be required, and that disc-
cone antennas provide the best performance over a wide frequency range
without being too large.
5.5.1 Testing
Tests were carried out using the configuration shown in Figure 5.5.
Figure 5.5, Antenna test configuration
A Rohde and Schwartz (R&S) signal generator and Minicircuits RF power

amplifier were used as the transmit (TX) source. A telescopic whip was used as
the receive antenna and connected to the Hewlett Packard spectrum analyser.
Both antennas were mounted at a height of 2m and at approximately 50m apart
outside on unobstructed ground.
The isotropic gain (dBi) of the antennas under test (AUT) was measured using
the method of gain substitution. The following calibration antennas were used to
accomplish this:
• AH Systems bicone (20MHz to 325MHz);
• R&S 4035.8755.02 log periodic (400MHz to 3.5GHz).
The received signal level present on the spectrum analyser was recorded for
each AUT over the measurement frequency range. The procedure was then
carried out for the calibration antennas. The isotropic gains of the antennas under
test were then calculated from the results.

The following three plots summarise the results.
Ofcom Antenna Com parison
10.0
0.0
0 10 20 30 40 50 60 70 80 90 100
-10.0
Gain (dBi)
-20.0
-30.0
-40.0 AOR DA753 Discone

European Ant. XP03V-500-1300/034
Thunderpole Eurostick
-50.0 Mag Mount
AOR DA5000 Discone
QinetiQ Discone
-60.0
Frequency (MHz)
Figure 5.6, Summary of results. Gain versus frequency (5MHz to 100MHz).
10.0
0.0
100 200 300 400 500 600 700 800 900 1000
-10.0
Gain (dBi)
-20.0
-30.0

-50.0 Mag Mount
AOR DA5000 Discone
QinetiQ Discone
-60.0
Frequency (MHz)
Figure 5.7, Summary of results. Gain versus frequency (100MHz to 1GHz)

10.0
0.0
1000 1500 2000 2500 3000 3500
-10.0
Gain (dBi)
-20.0
-30.0

-50.0 Mag Mount
AOR DA5000 Discone
QinetiQ Discone
-60.0
Frequency (MHz)
Figure 5.8, Summary of results. Gain versus frequency (1GHz to 3.5GHz).
5.5.2 Antenna choices
At the higher frequencies, of the antennas tested, the AOR 5000 discone (Figure
5.9) has the best performance. This is a robust, well made, compact antenna,
and is recommended as a good choice for the final system.
Figure 5.9, AOR DA5000 discone antenna

At the lower frequencies the “QinetiQ Discone” shows the best performance. This
is a QinetiQ manufactured antenna very similar in style to the Scanking Royal
discone antenna (Figure 5.10) used in the demonstrator system. These antenna
offer good performance, but are fairly large (1.5m high by 0.6m across), so there
may be some locations where it would be more appropriate to accept a
compromise on the performance in favour of a lower profile antenna.
Figure 5.10 - Scanking Royal discone antenna
5.5.3 Switching
Latest technology RF switches are required to minimise degradation to sensitivity

through switching losses, and maintain linearity. Lightning protection circuitry
should be utilised on the antenna inputs to protect the rest of the system.
5.6 Functional modules
5.6.1 Receiver
The requirement implicitly calls for the AMS receiver to have a wideband, fast
tuning receiver, both to accommodate the widest bandwidth signals and to

achieve the spectrum revisit time. This will also ensure a high probability of
intercept against fleeting or intermittent interference.
With the proposed large number of AMS units to be deployed, the underlying
requirement; mapping directly to RF performance, is its affordability, both in
terms of component cost and DC power consumption.
There are many technical challenges to the design of the AMS receiver, not least
the performance demanded from a low cost receiver.
An obvious limitation of the ultra wide instantaneous bandwidth solution offered is

the performance in strong signal and cluttered EM environments.
To overcome the effect of multiple strong signals in the selected observation

bandwidth, the AMS must engage front end gain controlling attenuators to
position the available dynamic range (circa 80dB) such that the strong signals do
not overdrive the maximum input voltage of the ADC.
Whilst this function is essential to prevent unwanted clipping and distortion in the
ADC circuitry, and consequent severe corruption of the FFT spectra, it is
undesirable from a sensitivity perspective; adding front end attenuation directly
impacts the receiver noise figure and therefore the ultimate system sensitivity.
In essence, there is a finite dynamic range available in the receiving and

digitising circuits and this must be positioned to accommodate the power of all
the signals in the selected bandwidth. This has the effect of degrading the
sensitivity of the system over the entire observation bandwidth.
This problem becomes more apparent as the selected bandwidth increases.

Firstly, by nature of increased bandwidth there are there are by definition more
signals to consider, each of which may in their own right be only of moderate
power level but combine to overdrive the ADC.
Secondly, the subsequent degradation of system sensitivity by engaging the gain

control circuits to compensate for the strong signal environment is spread across
a wider bandwidth. For example, in the extreme, a single high power signal in the
receive spectra will cause the degradation of an entire 40MHz of bandwidth
A more subtle and possibly worse effect can occur when strong signals are
outside of the final tuned IF bandwidth but within the front end pre-selected
bandwidth; under these conditions, the aggregate power at the ADC input may
be within the dynamic range of the ADC. Thus the system will not select any front
end attenuation. This is problematic as the troublesome high power signals have
to be linearly processed in the first stages of the receiving circuits without any
protective attenuation. Failure to process these, out of band signals in a linear
fashion will result in the generation of numerous harmonic inter-modulation
spurious terms, some of which will undoubtedly appear in the subsequent IF
bandwidths following the receiver front end.

For any receiver to operate under these conditions would require a front end, out
of band dynamic range at least equal to the in-band dynamic range plus the
available front end attenuation. In this case requiring a receiver front end
equivalent input intercept point 40dB stronger than the equivalent in-band input
intercept point; around +40dBm in this example. Such performance can be
approached but comes with diminishing returns against a steeply escalating cost
and power.
TRL MRX receiver suitability

The demonstration unit uses the B-Model MRX3500A receiver which will have
formally entered production by the time that the Automated Monitoring System is
rolled out. The formal factor, EMC and environmental specifications are not
expected to change.
The MRX receiver provides an input of 40MHz bandwidth to the ADC with an
instantaneous two tone in-band dynamic range of 75dB, capable of operation
with a total signal power of up to -40dBm in the selected bandwidth at the
antenna input. Once input signal power exceeds -40dBm, attenuators are
automatically switched in to prevent overloading the ADC input. This reduces the
sensitivity of the receiver by moving the available 75dB instantaneous dynamic
range up to the level required to process the increased signal power.
In its MRX3500B format, it is housed in a sealed connectorised chassis, which
will meet the environmental requirements.
Any relatively low cost style receiver represents a compromise between size,
weight and power consumption (SWaP), cost and ultimately RF performance.
The most noteworthy compromise is between cost DC power consumption and
RF performance. DC power consumption, increases in line with the RF power
handling capability.
Results from the central London trials (see Section 7.2.1) indicate that the signal
handling performance of the RF tray may be too weak to pass on 80dB of
dynamic range available in the ADC circuits to the RF environment when there
are strong signals.
The situation can be partially and relatively easily improved by hardware

modification to the receiver with no increase in cost or power consumption. A
reduction in nominal gain of the receiver by 14dB whilst maintaining the noise
figure would yield a comparable increase in dynamic range of the system under
in band strong signal conditions.
Significant improvements to the strong out of band signal handling capability can
only be achieved with a substantially increased out of band input intercept. This
type of design change comes with steeply escalating cost both in terms of DC
power consumption and money.
Realistically there are two options for achieving an increased high signal
performance:

• Instantaneous capture bandwidth reduction,
• Higher performance (and cost) receivers for urban environments.
Reducing the instantaneous capture bandwidth reduces not only the number of
simultaneous signals driving the AGC but also the bandwidth over which
sensitivity is lost once the AGC is engaged. The solution does however have the
disadvantage of increased scan time.
A universal increase in the performance of the receiver is undesirable from a cost

aspect but may still be more cost effective than maintaining different performance
grade receivers depending upon the strength of the EM environment for the
chosen installation.
5.6.2 Digitiser
The digitiser will capture IF signals from the receivers and convert them to a
binary format for processing.
The digitiser will be required to work in two modes. It needs to have a wideband
mode for rapid scanning when doing spectrum monitoring and detection, and a
narrowband mode for modulation identification and location processing.
The narrow band signal can be derived from a wideband signal by digital signal
processing (DSP) on the computer, but this is not considered a good option as it
increases the work load on the PC, and increases the data throughput
requirement across the bus onto the PC. For example, the digitiser might be
capturing at 100MHz with a required narrow band capture rate (chosen to match
the captured signal bandwidth) of 200kHz. This would mean that 500 times more
data than is required would have to be transferred to and processed by the PC.
This would not be sustainable across a PCI bus for anything other than a very
short sample period.
Two options are available to provide the narrow band mode. Digitisers are
available commercially with on-board digital tuners, generally based on the
popular Graychip 4016 as used in the demonstrator system. This will tune and
filter on a selected frequency and give a real-time IQ data stream for a wide
range of bandwidths. The second, more flexible option is to use a digitiser with an
on-board FPGA (field programmable gate array) or DSP capability. This would
allow a similar capability to that of the Graychip to be programmed, but would
also give the option of doing some of the wideband signal monitoring and signal
detection processing on the digitiser. This would reduce the requirements on the
PC processor considerably which would mean an overall reduction in unit power
consumption and heat dissipation.
The FPGA/DSP option is recommended for the production system because of
the overall reduction in power consumption and size, but the advantages must be
carefully balanced against the increased development costs and more complex
upgrade path. Upgrading PC based software is very straightforward; as faster
processors are developed they are generally backwards compatible, so old

software can be transferred straight onto them. Conversely, upgrading to a new
DSP chip could well mean a complete re-write.
16-bit ADC devices at 100Msps are now available and would satisfy the
requirements for the wideband capture mode. The number of bits is less
important in the narrow band mode than in the wide band mode (TDOA
processing has been shown to work with as few as 4 bits with little loss of
accuracy). However, in the wideband mode the number of bits affects the
dynamic range of the system, and must be matched to the dynamic range of the
receiver. This is particularly important if the sensitivity to low level signals is not to
be compromised when captured within the same band as a high power
transmitter.
5.6.3 Processor
A PC is required to process the data acquired by the digitiser, to act as an

interface and control point to the rest of the network and to provide local storage
for captured data. The main requirements of the PC are sufficient processing
power, and small size and power.
A modest performance PC as the host, with a DSP co-processor to carry out the
signal processing functions, will be the most power efficient implementation.
The final design could use a generic processor as these are readily available at
reasonable cost. The risk associated with a custom design is high and is difficult
to justify. A PC104 format is proposed, being compact in format and rugged in
design with ample options for connecting to external devices through various
interface types.
It is recommended that the DSP processor be implemented as a co-processor to
the main processor. This can be via a standard interface such as PMC or via the
PCI bus. A number of standard cards are available, but a custom design may be
desirable to integrate the A/D converter and signal conditioning. An FPGA could
also be incorporated to soak up any other utility functions.
Porting the signal processing algorithms to a DSP co-processor involves some
risk and incurs system testing and verification but this should be a relatively low
risk task, as the code was originally written to operate with a co-processor of this
type. The FFT transforms and averaging are well established and available ‘off-
the-peg’ functionality.
The demonstration system software is written in C++ to execute under PC
Windows XP. This provides a flexible upgrade path as PC processor speeds
increase.
The following processing tasks are identified:

• RF tray management,
• Command & control interface,
• Network interface,

• Signal processing,
• Spectral processing,
• Modulation classification.
These former three tasks are generic and can be efficiently carried out by the PC
processor.
The signal processing task includes Fast Fourier Transform of the spectral
samples into frequency spectrum followed by averaging. While this task can be
carried out on the generic PC, this is not the most efficient solution. Digital Signal
Processors are specifically optimised for such tasks. In terms of power
consumption, a similar function can be achieved with approximately 10% of the
power dissipation.
5.6.4 GPS
The key to successful TDOA processing is synchronising capture of data

between the sensors. This means we need accurate timing, and an accurate
frequency reference. GPS should be used to provide this as it gives superior long
term stability to crystal based oscillators.
The GPS will:
• Supply a 10 MHz reference to the receiver such that all system units measure
the same frequency,
• Supply the 10 MHz reference to the digitiser ADC clock generator such that
all units sample at the same rate,
• Supply a 1 pulse per second (1pps) to allow all system units to perform data
acquisition at the same time,
• Supply unit location for location purposes.
A timing accuracy of 20ns is achievable from a good commercial GPS unit.

These units use the fact that it is stationary to allow averaging over a period of
half an hour to improve the timing accuracy. This equates to a location error of
approximately 10m. This is a reasonable level to aim for.
For the production quantity of systems, a custom design would be cost effective,
but it must use the averaging techniques of the commercial systems if the
required accuracy is to be achieved.
5.6.5 ADC Clock
A sampling clock provides the input clock frequency to the digitiser / digital
receiver. This needs to have an extremely low jitter (below 0.5ps) if the full
dynamic range of a 14 bit digitiser is to be achieved at a 100MHz sampling rate.
The output of the clock must be impedance matched to the input impedance of
the digitiser.

5.7 TDOA architectures
5.7.1 Introduction
Although it is possible to perform TDOA geolocation using only three sensors, the
position fix produced can sometimes be ambiguous. It can be made
unambiguous by using a fourth sensor. Using four or more sensors can also
reduce the error in the position fix due to multipath or random measurement
error.
Three sensors will give two TDOA hyperbolas which will cross at 0, 1 or 2 points.
Although these points will not correspond exactly to the emitter location due to
measurement errors, they will each be well-defined. However with more sensors,
there will be three or more hyperbolas. If there are any errors in the TDOA
measurements, these will not all intersect at a single point so, to obtain a position
fix, the lines must be combined by some optimisation algorithm.
The most important performance metrics for a multi-sensor geolocation algorithm
are the expected accuracy of the position fix and how well the accuracy of this fix
may be estimated. Several algorithms are proposed in this report and their
performance against these metrics is discussed qualitatively for a multipath co-
channel environment.
5.7.2 Sensor Selection
If a very large array of sensors is available, the first decision which needs to be
made is which ones should be used to perform the position fix and how the
correlations should be performed?
Figure 5.11 –The master sensor should task all of its nearest neighbours
The optimum geometry for three sensors is always obtained by sensors which
surround the target emitter (i.e. the emitter lies within the triangle formed by the

sensors). However when locating a weak signal, such as a source of
interference, it may only be possible to detect its presence at one of the sensors.
The sensor which has the strongest SNR will be referred to as the master sensor
and is marked in red in Figure 5.11. The master sensor is likely to be the one
which is closest to the emitter. This master sensor and all of its ‘nearest
neighbours’ should then be tasked to collect data simultaneously. This should
ensure that at least one set of 3 tasked sensors surrounds the target emitter.
With a large array of sensors, not every sensor will be required to locate a
transmitter, and many may be too far from the transmitter to be of any use. A
decision must also be made on which sensor pairs to use in the correlation.
The signal strength is likely to be greatest at the master sensor so captures from
each of the other sensors in turn should be correlated with the capture from the
master sensor. Thresholding of the correlation SNR may be used to eliminate
any sensors where the signal strength is too low to produce a reliable correlation.
A position fix using three sensors could be ambiguous, so ideally four or more
sensors should be used.
5.7.3 Timing errors
Each TDOA measurement from a pair of sensors will be subject to a random

error, ∂t. If all the errors are independent and normally distributed, then
increasing the number of measurements used should improve the accuracy of
the position fix. However, if a very large ∂t is included in the calculation then the
error in the position found may be much worse than if it was not included. If
different sensor pairs are expected to give different errors, perhaps because the
sensors are of different types or because some of them are in poor multipath
areas, the algorithm should give greatest weight to those measured TDOAs
which are expected to have the smallest errors.
5.7.4 Multipath
A multipath environment can add echoes or delays onto a signal captured at a

sensor, effecting the measured TDOAs and hence the calculated locations.
Multipath is likely to be the greatest source of geolocation error.
If a sensor does not have clear line-of-sight to an emitter than the signal from the
emitter must reflect off other objects on its way to the sensor. This increases the
path length so it delays the time at which the signal arrives. If a correlation is
being performed between sensors A and B, where A has line–of-sight to the
emitter and B does not, the effect of the multipath will be to make it look as if the
emitter is further from sensor B than it really is so the resulting TDOA line will not
pass through the true emitter location. It is likely that several components of the
signal will be received at sensor B, each with different amplitudes and different
delays. Each component will produce its own peak in the correlation but the time
differences between the peaks are likely to be small compared with the peak
width so they combine to give a single peak whose precise TDOA value is an
average of the TDOAs resulting from each multipath component.

If the sensors are all at roughly equal distances from the emitter then it may be
reasonable to expect that the signals received at each of the sensors will have
been subjected to similar amounts of delay so the multipath effects will cancel out
to some extent. The effects of multipath will be worst if, within each sensor pair,
one sensor experiences multipath and the other does not. This could be the case
in Figure 5.9 since the master sensor is significantly closer to the emitter than
any of the other sensors. If the received signal levels suggest that the emitter is
very close to the master sensor, it might be advantageous if the master sensor
tasked its neighbours to capture and correlate signals but did not itself take part
in the geolocation process.
5.8 Emitter Location techniques
The location algorithm must take information from a set of correlation surfaces
and combine it to produce the positions of any emitters. It should also provide an
estimate of the confidence of each position which it finds.
The AMS demonstrator system uses an algorithm which assumes that there are
always three sensors available. The algorithms discussed below consider the use
of four or more sensors to solve the potential ambiguities which arise when only
three sensors are used. The algorithms also have the potential to improve on the
accuracy by the use of more sensors.
The Exhaustive Search Algorithm and the Optimal Solution Algorithm, are
designed to include information from all available sensors to reduce the random
errors in the locations. Careful modelling is required to determine if there really is
much advantage to be gained from this approach. It is the nature of this approach
that many of the sensor pairs will have poor geometries and SNRs for locating
the emitter, so are unlikely to give much improvement on the PF.
Provided the errors are truly random, a similar improvement in accuracy could be
obtained from a longer capture or by repeat locations using only 3 sensors.
However, if there are errors due to multipath or co-channel signals, then all the
repeated measurements will be subject to the same effect so the errors will
remain.
5.8.1 Calculation of TDOA position loci
Correlating data between a pair of sensors will give a single TDOA value which
can be mapped onto a hyperbola on the ground. Using the TDOAs from the
various sensor pairs, multiple lines of possible emitter positions can be
calculated. At least two lines are required to define a position, as in the three-
sensor demonstrator but, if there are more sensors, more lines can be used to
refine the position. As errors are inevitable, there will not be a single point of
intersection of all the lines, but a user can estimate the location of the emitter,
and a measure of the confidence from how close the lines come to all
intersecting. This technique is the most straightforward to implement, and is
computationally efficient.

Figure 5.12 - Each pair of sensors gives a "Line of Position"
5.8.2 The extended three sensor algorithm
A problem with using only three sensors is location ambiguities. If only one
emitter is present, the two TDOA hyperbolas may intersect in two places, only
one of which will be the correct location. If there are two or more signals present,
then it will not be known which of the hyperbolas from one sensor pair
correspond to which of the hyperbolas from the other pair, which produces more
ambiguities.
Ambiguities can be removed by using four sensors, shown as green circles in
Figure 5.13, and running the 3 sensor algorithm twice. First the algorithm is run
on sensor set 1, giving the positions shown by the red crosses. The algorithm is
then run again on sensor set 2. Another set of possible emitter locations, shown
by the blue crosses, will be produced. Those position fixes which correspond to
real emitters will appear in the same place in both sets, to within the size of the
error ellipse, whereas the ambiguities will not.

Set 2 Set 1
Figure 5.13 - Using a fourth sensor to remove ambiguities
5.8.3 Maximum likelihood search
Rather than trying to associate peaks from different sensor pairs and calculate
the point where multiple hyperbolas come closest to intersecting, it may be
simpler, although more computationally expensive, to do a maximum likelihood
search.
First, a search area where all the emitters are expected to lie is identified on the
ground. A grid of points is then overlaid on this area. For the first point on the
grid, the TDOA which would be expected for each sensor pair for an emitter at
that position is calculated. The error between the calculated TDOA and the
closest measured TDOA value is then found for each sensor pair and the RMS of
these errors is used as a measure of the likelihood of an emitter being present at
that position; the lower the RMS error, the more likely it is. The same calculation
is then performed for each of the other points on the grid. The result will be an
error surface which has a minimum wherever there is an emitter.
Alternatively, to eliminate the need for a set of TDOA values to be found from
each correlation, for each point on the grid the expected TDOA can be calculated
for each sensor pair as before, and the value of the correlation function at this
TDOA value can then be looked up. The resulting correlation strengths for all of
the sensor pairs can be combined to produce a likelihood value for that point on
the grid.
This second method has the advantage that information about the SNR and
width of each correlation peak is automatically pulled through into the likelihood
surface but it is computationally intensive, and therefore slow. To ensure that the
peak is not missed, the search must be performed at point spacing less than half
the expected width of the correlation peak. The higher the bandwidth, the
narrower the correlation peaks will be. For instance, a bandwidth of 1MHz would

give a correlation peak width of about 1000ns so the peak on the likelihood
surface would have a width of about 300m. This would restrict the grid point
spacing to a maximum of 150m so a grid of 400 by 400 points would be required
in order to cover a 60km square.
To speed up the search, the captured signals could initially be filtered to a
reduced bandwidth so that a course grid could be used to find the approximate
position. Then a finer grid over a smaller area can be searched using the full
bandwidth to refine the location.
An advantage of doing a grid search is that the calculation of the expected
TDOAs for each point can make use of terrain mapping data, whereas the
algorithms which involve calculating the equation of each TDOA line are
restricted to solutions in one plane.
The speed of a maximum likelihood algorithm may be improved by using a
simplex search as an alternative to the exhaustive grid search to find the
maximum of the likelihood surface.
5.8.4 Simplex Search
Rather than evaluating all points in a search area, a starting point, or seed value
is used. The error gradient at this point is used to suggest the best direction to
move to find the next test point. An analogy to this method could be standing on
a hill side and walking in the direction of the steepest gradient as a way of finding
the top. The simplex search will only give one position so it cannot be used if
multiple emitters are expected to be present.
Care must be taken to ensure that the global maximum is found, rather than
some local maximum which is dependent on the initial position. A simple way to
avoid this problem is to run simplex searches from a few different starting points.
Increasing the number of sensors tends to make the surface smoother, reducing
the risk of there being local maxima.
5.9 Communication between sensors
The internet will form the backbone of the sensor network. It is anticipated that
the final system will use whatever network medium is available to connect to the
internet. This should be broadband wherever possible.
The final system will be capable of providing full control of remote sensors from
either regional or national control centres. A network hierarchy for data
management is essential to prevent the maximum available bandwidth from
being exceeded for the control centres.
Some features are bandwidth intensive (e.g. real time spectrum / emissions
displays), so might only be available on higher bandwidth links. Similarly location
of weak signals where long sample lengths are required could have latency
across a narrow band link.

The AMS system will use the TCP/IP protocol across the Internet. This allows the
use of nearly any modern communication bearer or interface and significantly
lowers the communications hardware requirements as the backbone of the
network is already in place. Furthermore it allows multiple systems to be
controlled simultaneously from a single control centre as opposed to only a single
one with a point to point link.
5.9.1 Network control
With a large network of sensors, as is proposed for the AMS system, it is

essential to keep tight control over the configuration. This will be administrated by
a central communications service to which all sensors and control stations have
access. The sensors register their position, status and most importantly external
IP address with this service. This information is then available for control stations
to access information on the sensors, and for other sensors to identify their
neighbours for TDOA tasking.
Figure 5.14 suggests a block diagram for the network design. Multiple sensors
connect via the Internet to the primary and secondary controllers supplying their
BIT (built in test) status, latitude and longitude location, TCP/IP address and port.
AMS users then connect to the primary or secondary controller to determine
available sensors in the correct locality for signal observation or tasking for
location processing.
SENSOR
SENSOR
SENSOR
INTERNET
SENSOR SENSOR
USERS
USERS
SECONDRY
PRIMARY
CONTROLLER
CONTROLLER
Figure 5.14, Internet connectivity

Primary and secondary controllers will be required to provide a backup in the
event of a failure. The access to the controllers can be controlled, via IP
addresses or MAC addresses or usernames and passwords, and this effectively
results in access to the sensor system being secured. To maximise system
security all sensors should run anti-virus software and operate firewalls.
The sensors will connect to the Internet using individual modems and a user
tasking the system will be able to communicate simultaneously with multiple
sensors using a broadband internet connection. This Internet infrastructure usage
eliminates the need for multiple modems at the users’ locations to handle the
multiple simultaneous information flows.
5.9.2 Data storage
Database repositories will be required in different parts of the sensor network to

record the data being collected by the sensors. The database repositories
required are:
• Sensor repository,
• Regional repository,
• Regional backup repository,
• Central repository,
• Central backup repository.
Figure 5.15 shows an example part of the hierarchy of the AMS collection
system. In region 1 a number of sensors are distributed in a local area. A number
of similar regions will collectively cover the whole country. Emission data
collected from the sensors in the region are sent to the regional servers via the
internet. The secondary regional server continually mirrors the primary server so
no emission data and sensor information is lost in the event of a failure. The
central server collates the emission and position data from the regional servers.
The sensors in all regions can be monitored to ensure they are functioning
effectively. Again a secondary server will mirror the central server so no emission
data and sensor information is lost if the primary server fails. All sensors can be
monitored to ensure they are functioning effectively.

Region 1
7x 8x 9x 10x 11x 12x 7x 8x 9x 10x 11x 12x
Router
Ethernet
C
7 8 9 101112
A 12 34 56 1x 2x 3x 4x 5x 6x 1x 2x 3x 4x 5x 6x
A B
TALK / DATA
TALK RS CS TR RD TD CD
ADSL Modem TALK / DATA
TALK RS CS TR RD TD CD
ADSL Modem
WWW Server
AMS Sensor 1 AMS Sensor n
Secondary
Central
Central
server
Server
Local datase Local datase

Internet
Hardware Firewall
Secondary
Central
Central
7x 8x 9x 10x 11x 12x 7x 8x 9x 10x 11x 12x
Router/Firewall
Ethernet
C
7 8 9 101112
Database
A 1 234 56 1x 2x 3x 4x 5x 6x 1x 2x 3x 4x 5x 6x
A B
Database
Primary Secondary
Regional Regional
Server Server
Regional
Regional
Backup
Database
Database
Figure 5.15, Example network layout
Sensor repository
The sensor repository will continually record emissions details and occupancy
detected by the sensor. These emission details are:
• Emission ID,
• Frequency,
• Bandwidth,
• Amplitude,
• Start time,
• Duration,
• Modulation type.
Additionally position fix results will be stored in the database. The parameters
that are recorded are:
• Task ID,
• Task time,
• Sensors tasked to obtain the position fix,
• Capture parameters returned from the sensors tasked to achieve the position
fix (e.g. position, actual capture time),

• The calculated position,
• Estimate of confidence in the position fix.
Regional repository
The regional repository will collate data recorded by individual sensors (as
described in section 0). Many sensors will detect the same emissions so it is the
responsibility of the regional repository to remove duplicate entries, and to form a
picture of the regions spectrum usage and occupancy, and to provide a means of
archiving collected data. Data transfers can take place during quiet periods to
ease network congestion.
Database reporting will be available to provide statistics on spectrum usage and
identifying illegal or interfering emissions.
Regional backup repository
The regional backup repository will provide a live hot swappable backup of the
regional repository if it fails.
Central repository
The central repository will collate summary occupancy data from the regional
repositories providing a picture of the UK spectrum usage.
Database reporting will be available to provide statistics on spectrum usage and
identify illegal or interfering emissions.
The database will be accessible in a read only mode via a web server to allow
access to government, educational and other institutions that have an interest in
the data collected.
Central repository backup
The central backup repository will provide a live hot swappable backup of the
central regional repository if it fails.
License database
Database entries are to be cross referenced against to determine if transmissions
are in accordance with agreed terms in the Ofcom licensed user database.
5.9.3 Data Transfer Messages

Messages will be sent from the sensor to control centre, and control centre to
sensor to describe emitters and their emissions for subsequent display as well as
sensor tasking commands. A key requirement for the AMS design is that the unit
be operable remotely. The summation of the data messages per unit time must
not exceed the available communications bit rate.
The data required to be sent over the link are:

• Service location. To allow the discovery of the unit upon the Internet.
• Data types required. This message specifies what data (spectrum plots,
emissions, modulation classification results) to send to the control centre, for

example all automatic data items can be stopped in the event of a low
bandwidth ink.
• Emission detections. These messages identify a unique emitter that has gone
on, off, or has timed out, its amplitude and timing parameters.
• Spectrum display. This shows in real time the radio frequency environment
that the sensor is seeing.
• Modulation classification results. These are strings that describe the
modulation of an emission at a specific time and frequency.
• Location request messages. These specify the time, frequency, and
bandwidth of an IQ data capture for the purposes of emitter location via time
difference of arrival techniques.
• Location data replies. Specifying the location request and the data to be
returned.
• Data selection message. This specifies which of the available data items
above are to be sent from sensor to HMI.
5.10 Software
5.10.1 Current QinetiQ capability
Over the last seven years QinetiQ have been developing high specification
spectrum monitoring and modulation classifier applications capable of working
analogue and digital signal types. The software will manage and control multiple
wideband scanning sensors and is designed to operate in an intuitive manner
with minimal operator training. It is recommended that the AMS spectrum
monitoring and modulation classification functionality is based upon QinetiQ’s
existing capability.
This software has the ability to:
• Display real time and historical results from multiple sensors separated
geographically but connected by a network,
• Determine the presence of networks of emitters by matching of signal
parameters such as frequency, modulation type and locations,
• Log sensor results to hard disk at regular intervals,
• Manage the transfer of the information between sensors to optimally achieve
location of all emitters,
• Calculate position fixes based on TDOA techniques based on information
from multiple sensors,
• Show position fixes on a digital map,
• Report results dependent upon pre-set filters.

The software at each sensor logs emissions from emitters as they occur and
saves these to disk after a configurable interval. Requests on emissions that
have occurred are made by a designated master to the other AMS applications
running on remote sensors, and these then mine their history databases to find
time and frequency correlating emissions.
Narrow band captures can be tasked across multiple sensors and data returned
to a single point to allow correlation and position fixing.
The QinetiQ modulation classifier proposed for the AMS unit has the capability to
process the following modulation types:
Analogue signal types:
CW, AM, FM/PM, FM, SSB, DSB, USB, LSB.
Digital signal types:
ASK, MSK, FSK, PSK, QAM, ASK/PSK Hybrids.
Additional Signal Type Descriptors: Voice, Tone, Morse, Radio Teletype (RTTY),
CTCSS and Chirp detection is also applied to suitable signals
5.10.2 Key functionality
Table 5.7 highlights the key components that should be included within the
software for the final AMS system. Refer also to section 6.6.3 for sample display
plots.
Component Description
Real time display The AMS component needs to support real time
displays to show spectrum plots and detected
emissions over a selected frequency range from
any sensor on the network. Additionally historical
displays should be viewable in the form of waterfall
plots.
Mapping The mapping component allows the application to

view position fixing results as well as sensor
locations and tracks.
Various standard mapping formats are available

commercially and should be supported.
Location Fixing The location component allows the sensors to

operate co-operatively to calculate the location of an
emitter. This links in with the mapping component to
display location results, and will plot also show
estimates of errors in the form of confidence

Component Description
ellipses.
Spectrum The spectrum occupancy component allows usage

Occupancy as a function of time to be plotted in the form of
waterfall plots. This view also allows percentage
occupancy to be monitored.
Clear channel data should also be shown either

graphically or in tabular form.
Filtering The filtering component sits between any data store

within the application and the user displays with
functionality that prevents data being displayed that
does not pass the filtering criteria.
This is applicable to spectrum views to reduce the

numbers of emissions being displayed and map
views to only display locations from selected
emitters.
Results Display The result display component displays to the user

information he/she has requested. This is used to
display, for example, detailed emitter perameters or
location results.
Targets A target component allows identification of signals

of interest for monitoring, signal classification or
location.
Handoff Receiver The handoff receiver component is responsible for

controlling handoff receivers, setting, frequency,
mode and scanning.
Audio Streaming The audio-streamed component allows

demodulated audio streams to be transferred
across the network from a remote sensor to the
control centre.
Table 5.7, Key software components

Qinetiq Report On Geolocation

Uploaded by

Copyright:

Available Formats

Qinetiq Report On Geolocation

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

Qinetiq Report On Geolocation

Uploaded by

Copyright:

Available Formats

q

QinetiQ; PR Edmonds , B Ashforth, DI Elliner & S Madsen

The radio spectrum is an increasingly important national asset, and so efficient

It is expected that an AMS network would have the following benefits:

• Enhanced scope of spectrum monitoring, both geographical and transmission

• Greater precision in identifying and locating sources of interference,

• Improved responsiveness to interference problems,

• More detailed information on spectrum usage,

• More detailed information on interference patterns.

Review of current state-of-the-art automatic interference monitoring

Work Package 1 required a literature review of automatic interference monitoring

• exhibit particularly poor performance in multipath environments which limits their

• will not support fast scan rates;

Identification and Analysis of System Benefits

AMS System Design

The purpose of Work Package 4 was to design an unattended monitoring system

Hardware and Software development of the demonstrator system.

Sensor performance and trials of the AMS

• Modifying the rules governing unwanted emissions or modifying the parameters of

• Monitoring if the licence-exempt spectrum is becoming congested or whether the

1.2 An Automatic Monitoring System

To understand spectrum usage, Ofcom commissioned a study into the design of an

• Greater precision in identifying and locating sources of interference,

• Improved responsiveness to interference problems,

• More detailed information on spectrum usage,

• More detailed information on interference patterns.

The consortium, which has already established a track record of providing

1.5 Report Overview

The structure of the document is as follows:

Work Package 1 required a literature review of automatic interference monitoring

The design of an unattended monitoring system whose primary purpose is to

• QinetiQ ltd using its own sources of information.

2.2 Commercially available systems

2.2.1 Literature search summary

A summary of the search of current commercial and research literature of automatic

2.2.2 Equipment summary

Mode of A narrowband Wideband and Narrowband Wideband 10 Narrowband up to Wideband

Antenna Up to 6 One Multiple Options Various options Multiple options for

Noise Figure 9dB (HF) 15 dB 12 dB Typical 14dB typical (HF),

No. Narrowband 6 to 8 Unspecified, cost

Modulation FM, AM, Phase / ITU-R SM 1268-1

Modes AM, SSB/CW, AM/FM/SSB AM/FM/LSB/USB/ AM:A2A, A2B,

Manual Tasking Yes Yes Yes Yes Yes Yes

No No No Licence Violation Licence Violation Can integrate with No

Client Server Yes Yes Yes Yes

Table 2.1, Commercially available spectrum monitoring systems

System Internet URL

Sat Corp http://www.sat.com/products/terrestrial/sigmon_options/sigmon_1000/

Rohde & http://www.spectrummonitoring.rohde-schwarz.com

2.3 Press Releases

In addition to a review of the commercial and research literature for automatic

2.3.1 Conatel Signs $10Mil TCI Spectrum Monitoring Contract 07/02/99

CARACAS, VENEZUELA 1999 JUL 2 (NB) -- By Staff, GNCS Business News.

2.3.2 Dielectric receives $11 million contract

2.3.3 TCI International, Inc. Announces Final System Acceptance of Spectrum

2.3.4 CTS announce systems for India

• exhibit particularly poor performance in multipath environments which limits

• cannot cope with co-channel signals and have limited classification

• will not support fast scan (sample) rates;