Autonomous Monitoring Stations. Phase II: Trials. Final Report. Autonomous Monitoring Stations Phase II Trials Final Report

Transcription

1 Phase II: Trials Final Report Phase II Trials Final Report Ofcom Riverside House 2a Southwark Bridge Road London SE1 9HA This report was commissioned by Ofcom to provide an independent view on issues relevant to its duties as regulator for the UK communications industry, for example on 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. Copyright Agilent Technologies UK Limited Page 1

2 HISTORY LOG Date Version Reason for change Written by Validated by Approved by 05/06/ First Issue Richard Holmes, Bob Cutler Distribution list: Christos Politis, Ofcom Ron Stanley, Ofcom Bob Cutler, Agilent Technologies. Leon Werenka, Agilent Technologies. Richard Holmes, Agilent Technologies. Don Hamilton, Agilent Technologies. Page 2

3 FORWARD Ofcom have a requirement to extend their existing monitoring systems into a more comprehensive network capable of monitoring the radio spectrum and detecting interfering sources over a large part of the country. As part of this work, Ofcom were interested to test the concept of Time Difference of Arrival (TDOA) for emitter location in an urban environment. Ofcom let a contract with Agilent Technologies to trial a TDOA based system in Central London Locating a transmitter using TDOA techniques would appear to offer a viable alternative to traditional direction-finding (DF) technology. DF technology tends to be expensive, and particularly hard to deploy in urban environments due to site requirements for the DF receivers and antennas, and by the unwillingness of building owners and managers to accommodate DF antennas and equipment. As a result, DF solutions are rarely deployed in urban environments in sufficient quantities to provide adequate coverage, especially for the newer higher frequency and wider bandwidth signals. TDOA technology allows greater flexibility in the selection and placement of antennas. With small, unobtrusive antennas, noncritical antenna placement, and compact receivers the number of available monitoring sites is greatly increased. Whilst not well suited to locating very narrow, or unmodulated signals, TDOA technology is shown to work with the vast majority of modulated signals in use today. Urban environments are unkind to both TDOA and DF technology. With high signal densities, co-channel signals, greater path loss, and significant multipath effects, the location accuracies that can be achieved are far from those achievable in more rural settings. This report details the performance of a TDOA system deployed in Central London, and provides insight into achievable accuracies using practical receiver densities. Throughout this report the term RF sensor, or simply sensor, will be used interchangeably with receiver. The usage of the term sensor reflects the sensor-network aspects of an urban spectrum monitoring system employing large quantities of receivers operating cooperatively and in synchronisation. RF sensors are not limited to TDOA applications as they incorporate digital IF technology. Spectrum monitoring and demodulation are two examples of applications well suited to a capable RF sensor Page 3

4 1 EXECUTIVE SUMMARY A trial was performed in Central London to test the performance and establish receiver densities for monitoring receivers capable of geolocating a signal using time-difference-of-arrival (TDOA) techniques. The trial was conducted in two phases. In the first phase, propagation measurements were made between transmitters at two different elevations, and a mobile receiver at street level. The measurements were compared to industry standard propagation models and were also used to plan the receiver spacing used during the TDOA phase of the trial. In the second, TDOA-phase of the trial, five receivers were deployed in three different configurations. Agilent recommended the use of five receivers over the theoretical minimum of three receivers required to estimate location. Using five receivers, geolocation accuracy was shown to improve significantly for narrowband signals when more than three receivers contributed to the result. Accuracy was not shown to improve using different combinations of three receivers. The five receivers (sensors) were principally deployed in two different grid configurations. This was done to vary both the size and geometry of the sensor network. In the large grid configuration sensor spacing varied from a minimum of 2.1 km to a maximum of 4.7 km. In the small-grid configuration spacing ranged from 1.5 km to 3.6 km. Whilst most of the measurements were made using these two configurations, one sensor was removed from its fixed location and made mobile allowing a third configuration to be tested. The mobile sensor employed UMTS for network connectivity. The TDOA trial included a mix of controlled experiments and realworld measurements. The controlled experiments used test-signal transmitters moved to various locations inside and outside of the coverage area defined by the sensor network configurations. The test signals varied in frequency, bandwidth, modulation type and radiated power, simulating a broad range of real-world signals. The real-world measurements used live, off-air signals specified by Ofcom. Some off-air and test signals were transmitted from locations outside of the sensor grids. The goal for most of these signals was to demonstrate that a line-of-bearing could be obtained using TDOA technology. This capability was reliably demonstrated on multiple signals and signal types. For one broadcast FM station with an overlapping pirate station, the system indicated two signals and two directions, though with some loss in accuracy. Page 4

5 There was more to be learnt by deploying a TDOA system in London than TDOA performance alone. The practical aspects of site selection and acquisition, equipment installation and networking are also documented in this report. This trial demonstrates that TDOA technology is forgiving of site selection as usable results were obtained without the availability of seven perfect sites. Small antennas and the ability to use less-than-perfect sites greatly simplified the temporary acquisition of the six fixed sites used during the trial. The fixed-sensor turned mobile-sensor was briefly located on the roof of a car park creating in effect a seventh site. No one networking technology was available to connect the sensors to the central server. Sensor were networked together using a combination of UMTS, DSL, and Wireless LAN. There were occasional issues arising from network congestion resulting in some measurements being limited to four sensors instead of five. This would sometimes occur when, for example, the UMTS links would slow down at the end of the work day. This trial was not designed to showcase TDOA technology. Had that been the goal, signals and locations favourable to accurate geolocation would have been chosen over the broad range of signals and locations actually used. Instead the trial was designed to determine both the potential accuracy of TDOA systems in dense urban environments and the sensor densities and other requirements necessary to achieve those accuracies under a wide variety of realworld conditions. The second goal, determining sensor density requirements, necessitated experiments operating at and beyond the expected limits of performance. Not unlike the process of determining the strength of a metal bar, the system was tested to the breaking point. Signals that are narrower in the frequency domain, have broader temporal characteristics making more difficult the precise determination of the cross-correlation peak used to estimate the difference in arrival times of a signal at two receivers. As was expected, narrowband signal accuracy was less than for wider bandwidth signal due to increased sensitivity to noise and multipath distortion. The sensor spacings recommend in this report reflect not only the need for adequate signal-to-noise, but also the improved accuracy obtained using four or five receivers instead of the theoretical minimum of three receivers. Results confirm that extra receivers can help mitigate the effects of multipath, improving location accuracy, especially for narrowband signals. Another trial goal was the establishment of system sensitivities. In a receiver, sensitivity measures are well understood. In a TDOA system, sensitivity measures must reflect the differences in signal levels received at a plurality of receivers. TDOA geolocation does Page 5

6 not require positive SNR at the receivers. What is required is a positive SNR in the cross-correlations between pairs of signals captured at three or more receivers. To deal with this added complexity, this report introduces a measure of system sensitivity call the Mean Cross-Correlation Signal-to-Noise Ratio (MXSNR). The relationship between MXSNR and geolocation accuracy is reported. The cross-correlation computation normally used in computation of TDOA may also be used to indicate the presence of a signal even when signals are not detectible above the noise at any one sensor. This coherent-detection technique is evaluated against traditional non-coherent techniques (e.g. spectral power thresholds) on the basis of detection sensitivity. Networking bandwidth and computational loads may limit this technique to specialized applications where detection performance is more important than scanning speed. The trial was comprehensive. It involved more than 1500 propagation measurements in phase one and 2300 TDOA measurements of both test and off-air signals in phase two. Test signals were transmitted from 22 different locations. In addition, more than 32 off-air signals were analyzed, not counting UMTS and GSM signals. Off-air signals included Aircraft AM, PMR, TV, Broadcast FM (including one pirate station) and one suspected RADAR signal. Frequencies ranged from 80 MHz to 2.8 GHz. Offair signals were a mix of continuous and burst (e.g. PTT) transmissions. The TDOA system used in the trial has been deployed in other, more rural environments. The inherent location accuracy of the system based on timing accuracy, and the demonstrated system performance in other environments are both better than the demonstrated performance in Central London. The primary limitation in accuracy in Central London is multipath followed by signal level. Ultimately, sensor spacing will be determined by the most critical application. For example, sensor spacing of 6-8 km would seem appropriate for PMR base stations. For mobile stations, the sensor spacing should be 3 km unless system sensitivity is raised. (Which could be accomplished by using antennas with gain at application specific frequencies, for example.) TDOA accuracy is affected by many parameters. The number of sensors contributing to a measurement, the height and power of the transmitter, the modulation bandwidth, the length of transmission, sensor spacing, and the relative geometries of the transmitter and sensors, to name a few. For this reason, attempting to specify accuracy using a single number is not advised. With that disclaimer, trial results indicate potential geolocation accuracies of 500 meters Page 6

7 and better for a reasonable percentage of measurements in a welldesigned sensor system. This accuracy was demonstrated with the strong, but narrowband off-air PMR base station signals. This level of accuracy was also demonstrated with the one-off measurements of a pirate radio station located outside of the sensor grid. Many of the test signal results also demonstrated this level of accuracy when the signal strength was sufficient. In summary, the TDOA system deployed in Central London was shown to be effective in detecting and locating both test and live offair signals under a wide range of conditions. TDOA accuracy was ultimately limited by multipath. However, accuracies significantly improved when multipath effects were overcome by combining results from four or five sensors. TDOA was shown to be relatively insensitive to antenna placement providing a greater number of potential monitoring sites. 1.1 KEY PERSONNEL Key personnel involved with this project are as follows: Agilent Technologies Richard Holmes Project Manager and Technical Consultant Don Hamilton Technical Consultant Brian Gillespie Account Manager Bob Cutler Chief Technologist Agilent Labs Leon Werenka R&D Manager Agilent Labs Page 7

8 Ofcom Christos Politis R&D External Projects Manager Charlotte Cole Business Support Coordinator 1.2 WORK BEING PERFORMED Agilent was contracted in a program of work to evaluate TDOA in the dense UK urban environment of Central London. This work being performed is based upon use of existing Agilent equipment and experience from similar TDOA field trials. This document is the Final Report detailing the work performed to setup the trial, the measurement performed, results analysis and final conclusions and recommendations. 1.3 FURTHER ASSISTANCE Should you require further clarification on any aspect of this document please contact Mr. Richard Holmes. This document is submitted for and on behalf of Agilent Technologies UK Limited by: Richard Holmes Project Manager Page 8

9 TABLE OF CONTENTS FORWARD EXECUTIVE SUMMARY KEY PERSONNEL WORK BEING PERFORMED FURTHER ASSISTANCE AGILENT TECHNOLOGIES INTRODUCTION AGILENT ORGANISATION REQUIREMENT SPECIFICATION APPLICABLE DOCUMENTS OVERVIEW INTRODUCTION PROJECT OBJECTIVES PREVIOUS EXPERIENCE INTRODUCTION TDOA EXPERIENCE Customer Trials Multi-path and Doppler Effects REMOTE MONITORING SENSOR DEVELOPMENT ANALYSIS OF AMS PHASE TDOA RESULTS NUMBER OF SENSORS SIGNAL TO NOISE Doppler Compensation Multi-path Summary WORK PROGRAM INTRODUCTION SENSOR CONFIGURATIONS Sensor Sites Communications PROJECT TASKS OVERVIEW Task 1- Project Start-up Task 2 RF Propagation Measurements Task 3 - Finalise Trials Plan and Sensor Locations Task 4 Sensor Location Setup Task 5 - TDOA Measurements Phase Task 6 TDOA Measurements Phase Task 7 Results Analysis Task 8 Final Report...34 Page 9

10 8 WORK PERFORMED TASK 1- PROJECT START-UP TASK 2 RF PROPAGATION MEASUREMENTS Transmit Frequencies Measurement Setup Transmit Sites Transmit Site Transmit Site Measurement Locations Measurement Methodology Amplitude Corrections Measurement Results TASK 3 - FINALISE TRIALS PLAN AND SENSOR LOCATIONS Trials Plan Features to be Tested Test Methodologies Testing Approach Test Data Transmit Site Selection Transmit Vehicle Antennas Sensor Specification Signal and Frequency Planning Pre-Existing Signals Finalised Sensor Locations Sensor Sites Selection Process Site Agreements Sensor Sites Large Sensor Configuration Small Sensor Configuration Site Site Site Site Site Site Site TASK 4 - SENSOR LOCATION SETUP Sensor Networking Networking Testing Site Sensor Installation Network Communications Site Sensor Installation Network Communications Site Sensor Installation Network Communications Site Sensor Installation Network Communications Site Sensor Installation Network Communications Site Sensor Installation...91 Page 10

11 Network Communications Site Sensor Installation Network Communications Site Setup Summary Sensor Sites Networking VPN Test Transmitter Transmitter Buggy Transmitter Trolley TASK 5 AND 6 TDOA MEASUREMENTS Transmit Locations Examples of Transmit Sites Transmit Site Transmit Site Transmit Site Transmit Site TECHNOLOGY BACKGROUND Basic Concepts Uses in TDOA and Signal Detection TEST SIGNALS: DESCRIPTION Signals 1 and 2: AM at 82 MHz and 169 MHz Signal 3: WBMF at 83 MHz Signals 4,5 and 6: NBFM at 169 and 454 MHz Signals 7 and 8: GMSK at 919 and 1802 MHz Signal 10: QAM at 919 MHz Signals 9 and 11: CDMA at 919 and 1802 MHz TEST SIGNALS: DETECTION Overview Non-Coherent Signal Detection Coherent Signal Detection Comparing Coherent and Non-Coherent Results Signal Detection Models Sensor Densities using Coherent and Non-Coherent Detection Detection Summary TEST SIGNAL GEOLOCATION Overview System Sensitivity SNR, MSNR, XSNR and MXSNR System Noise Levels Effects of Sensor Spacing Effects of Sensor Quantity Effects of Cross-Correlation Quantity Effects of Test Signal Type and Bandwidth AM 82 MHz and 169 MHz WBFM 83 MHz NBFM 169 and 454 MHz GSM 919 MHz and 1802 MHz MHz QAM Effects of Signal Power Multipath Mitigation with Diversity Multiple Transmitters FUNDAMENTAL LIMITATIONS IN SYSTEM ACCURACY OFF-AIR SIGNALS Page 11

12 Off-Air Summary AM 60W London Fire Brigade MHz PMR 2 NFM Barbican MHz PMR 6 NBFM Addison Lee SW PMR 9 Port of London Authority Millbank Tower MHz PMR 11 NBFM Hyde Park MHz Non-Stationary and Multiple Access Signals (AM and PMR) Other PMR Broadcast TV MHz Crystal Palace TV TV TV FM Pirate FM Pirate Radio Station Detection MOB 1 GSM 940 MHz MOB2 GSM UMTS MHz Bonus Signal CONCLUSIONS AND RECOMMENDATIONS Signal Detection TDOA Based Geolocation Sensor Density Sensor Density Overview Sensor Densities for Broadcast and Pirate Radio Sensor Densities for PMR Base Station Sensor Densities for PMR Mobile Station Sensor Density Summary RF Performance Size/Mounting Antennas and Filters Network and VPN Performance Summary Recommendations Page 12

13 ABBREVIATIONS ADSL Asymmetric Digital Subscriber Line AGILENT Agilent Technologies UK Limited AMS Autonomous Monitoring System DSL Digital Subscriber Line GPS Global Positioning System HP Hewlett-Packard LAN Local Area Network MSNR Mean Signal to Noise Ratio MXSNR Mean Cross-Correlation Signal to Noise Ratio PTA Pulse Triggered Measurement RF Radio Frequency TDOA Time Difference of Arrival S/N Signal to Noise SNR Signal to Noise Ratio UMS Unattended Monitoring System V0PN Virtual Private Network XSNR Cross-Correlation Signal to Noise Ratio Page 13

14 2 AGILENT TECHNOLOGIES 2.1 Introduction Agilent Technologies (Agilent) roots are in the earliest days of Hewlett-Packard Company (HP), which started as a test and measurement company in 1939 with the production of the company's first audio oscillator. On March 2, 1999, HP's board of directors announced its decision to create two new companies from the existing HP organisation. Agilent Technologies launched its initial public offering on November. 18, On June 2, 2000, Agilent Technologies became a fully independent company. Agilent is the world's leading designer, developer, manufacturer and provider of communications components as well as electronic and optical test, measurement and monitoring instruments, systems and solutions. The company's customers span key industries, including electronics, communications, semiconductors and life sciences. A truly international company, Agilent provides sales and services in over 100 countries. Research and manufacturing facilities are located throughout the U.S., Europe, Asia Pacific, Latin America, and Canada. The company employs 15,000 persons worldwide. Agilent embodies HP's heritage of innovation and contribution, uncompromising integrity, teamwork and trust and respect for the individual, while placing a renewed emphasis on customer focus, accountability and agility. Values: Based on the HP Way Technology: Agilent Technologies Labs and division labs Products: High-quality products, #1 in most markets served Customer Focus: Responsiveness, integrity, insight Employees: An employer of choice Page 14

15 2.2 Agilent Organisation Agilent Technologies consists of the following businesses: Electronic Measurement Agilent's electronic measurement business provides standard and customized electronic measurement instruments and systems, monitoring, management and optimization tools for communications networks and services, software design tools and related services that are used in the design, development, manufacture, installation, deployment and operation of electronics equipment and communications networks and services. Our electronic measurement business employed approximately 11,500 people worldwide as of October 31, Markets: The markets for our electronic measurement business include communications test and general purpose test. Product areas: Communication test products include products for the following types of networks and systems: fiber optics networks, transport networks, broadband and data networks, wireless communications and microwave networks. We provide assistance with installation and maintenance and operations support systems. General purpose test products include general purpose instruments, modular instruments and test software, digital design products, parametric test products, high frequency electronic design tools, electronics manufacturing test equipment and thin-film transistor array test equipment. Bio-Analytical Measurement Agilent's bio-analytical measurement business provides application-focused solutions that include instruments, software, consumables and services that enable customers to identify, quantify and analyze the physical and biological properties of substances and products. Our bio-analytical measurement business employed approximately 4,000 people worldwide as of October 31, Markets: The markets for our bio-analytical measurement business include life sciences, including the pharmaceutical analysis, gene expression and proteomics markets, and chemical analysis, including the petrochemical, environmental, homeland security and forensics and bio-agriculture and food safety markets. Product areas: The seven key product categories for the bioanalytical measurements business include: microarrays, microfluidics, gas chromatography, liquid chromatography, mass spectrometry, software and informatics, and related consumables, Page 15

16 reagents and services. Agilent Laboratories Agilent Labs is our central research organization based in Palo Alto, California, with satellite offices in Beijing, China; Everett, Washington; Leuven, Belgium; and South Queensferry, Scotland. Agilent Labs conducts three kinds of research in electronic and bioanalytical measurement to meet the needs of Agilent s customers across a range of markets and industries: Research that will lead to evolutionary, revolutionary and disruptive technologies to grow Agilent s existing businesses in electronic and bio-analytical measurement systems, Research that leads to technologies that create new businesses outside Agilent s current markets but within Agilent s field of interest, and Basic research that contributes to the fundamental understanding of areas critical to Agilent s future. Electronic measurement research focuses on technologies that address trends such as the emergence of modular architectures for instruments and sensors, the growing network complexity and the growth of Internet-enabled data services. Areas of research include physical measurements, mixed-signal devices, signal processing and analysis, modular and distributed measurements, and complex software systems. Bio-analytical measurement research addresses solutions for chemical analysis and life sciences to enable new understanding of living systems, more precise analysis of organic and inorganic compounds, accelerated drug development, and medical research for the diagnosis and treatment of disease. This research benefits from the trends of miniaturization and integration that are leading to nanoscale science and engineering. Areas of bio-analytical research include chemical and environmental solutions; separations, such as gas and liquid chromatography, and capillary electrophoresis; microfluidics for faster, integrated and less-costly analyses; inorganic and organic spectroscopy; life-science tools such as proteomics and genomics reagents, systems biology and software tools, and molecular diagnostics for personalized medicine. Page 16

17 3 REQUIREMENT SPECIFICATION This proposal has been produced in response to a customer supplied requirement document and is based upon Agilent s understanding of these requirements. 3.1 Applicable Documents The following documents cover all the contractual requirements imposed on the project. This list is ordered from the highest to the lowest priority. Name Identification Version Date Autonomous Monitoring Station Phase II (This Document) /06/2007 Final Report Autonomous Monitoring Station Phase II Trials Plan OF04TP10.doc /11/2006 Autonomous Monitoring Station Phase II Trials Proposal OF04PP10.doc /08/2006 Invitation to Tender for Autonomous Monitoring Stations Phase II Trials Program (SES ) (SES ) N/A 31/07/2006 Page 17

18 4 OVERVIEW 4.1 Introduction This document is the Final Report detailing the work performed to setup the trial, the measurements, results analysis and final conclusions and recommendations. 4.2 Project Objectives Agilent Technologies proposed to perform an extensive range of trials and analysis to evaluate the use of TDOA techniques in Central London. The objectives of this work were: Profile actual RF propagation (at different frequencies) in the environment where the measurements are to be performed. Perform measurements to determine sensitivity and sensor density required for each test signal type in an urban environment, taking into account both signal detection and TDOA accuracy. Demonstrate the ability of the system to indicate a line-of bearing for emissions outside of the sensor network. Page 18

19 5 PREVIOUS EXPERIENCE 5.1 Introduction Agilent Technologies has considerable experience in Radio Monitoring equipment including the Blackbird system which is used extensively by Government Agencies around the World. Agilent also developed 20 Unattended Monitoring Systems (UMS) used by Ofcom in Agilent has delivered a large number of RF monitoring systems including several systems using TDOA techniques to physically locate emitters. 5.2 TDOA Experience Agilent Technologies has been actively investigating the use of TDOA for emitter location for a period of 4 years. Research was performed using a network of modified versions of the Agilent Performance Spectrum Analysers (PSA). This was used to develop and implement TDOA algorithms and perform a wide range of testing. The PSA is a Digital based Spectrum Analyser with the same general architecture to the units used in the Phase 1. The PSA provides a convenient platform to make a broad range of measurement capabilities and offers leading-edge performance Customer Trials In additional internal testing, Agilent has performed over 10 field trials for customers in a mixture of urban, suburban, rural, and very challenging indoor scenarios. Networks have ranged in size from 3 to 5 Sensors interconnected over various TCP/IP based communication including Wireless LAN, ADSL/DSL and cable modems. GPS is used to provide accurate time synchronization between the Sensors in the network, Figure 1 shows a typical network of Sensors. Page 19

20 Figure 1: TDOA Sensor Network Page 20

21 A Windows PC controls and performs the measurements including the computation of the TDOA algorithms. This PC can be located either locally or remotely Multi-path and Doppler Effects Multi-path and Doppler are examples of propagation effects which heavily influence successful operation of TDOA in an urban environment. This has been a particular focus of Agilent s research. The impact of multi-path on TDOA measurements and TDOA accuracy depends on both the strength of the multi-path signals, and the delay spread relative to the signal bandwidth. For example, a GSM signal has a bandwidth of 200 khz. The correlation pulse width is then on the order of 1/200 khz or 5 usec. Signals with large delay spreads (in this case >> 5 usec) can be more easily compensated as the correlation peaks from the direct and indirect paths are more easily distinguished. In Figure 2, the peak of the composite correlation for the short-delayspread signal has shifted as a result of the combination of direct and indirect signal paths of nearly equal amplitude. For the long-delayspread signal, the peaks are in the correct location for both the weaker and stronger signal paths. From a single cross-correlation pair it s impossible to tell which peak is associated with the direct path. Figure 2: Multi-path Example In high-multi-path, short-delay-spread conditions, which are common in urban and indoor environments, Agilent s experience has been that increasing the number of Sensors improves TDOA accuracy by taking advantage of the fact that multi-path is different at every sensor. Page 21

22 Figure 3 below shows the correlations between three TDOA Sensors where multi-path is being experienced due to the transmitter being located next to a large building. The multi-path is demonstrated by the multiple peaks present in the traces. In the first cross-correlation plot two distinct peaks of almost equal amplitude are clearly visible. Figure 3: Multi-path example Page 22

23 Figure 4 below shows the TDOA result from the correlation data. Figure 4: TDOA Result Despite the severe multi-path, and poor sensor geometry in relation to the transmitter (located outside the network of Sensors), the Agilent algorithms were able to accurately locate the emitter. Page 23

24 Figures 5 and 6 show the five corresponding spectrums and ten cross-correlation pairings for a five Sensor setup. While only two of the spectrums show positive signal to noise, nearly all the Sensors contributed to this measurement highlighting that Signal to Noise (S/N) is not the only factor in making successful TDOA measurements. Two of the correlation pairings, shown in grey, were automatically identified by the Agilent algorithms as not contributing to the measurement. Noise and multi-path distortion are also clearly visible in the remaining eight pairings displayed. As the delays spread is quite small, the distortion does not show up as discrete correlation peaks as in Figure 8, but rather as asymmetrical pulse shapes. Despite the weak signal, and significant multi-path in the building, the measurement was accurate to a few metres. Figure 5: Spectrum plot for all five Sensors Page 24

25 Figure 6: Ten Cross-Correlations 5.3 Remote Monitoring Sensor Development As detailed above, Agilent has been investigating TDOA for a period of 4 years. The trial was performed using Agilent s TDOA development platform based upon Agilent s Performance Spectrum Analysers (PSAs). The PSA contains all the main blocks required for an Autonomous Remote Station including the receiver, IF processing, data processing and various communication interfaces. Agilent considers that the development of the required functionality in a small low-cost AMS type unit is a realistic possibility. Page 25

26 Such a unit could be designed to support a wide range of potential deployment scenarios including antennas directly connected to the unit and separate antennas (e.g. on the roof of a building). Mounting an antenna on the roof of the building with good lineof-sight conditions typically provides good location performance against higher level emitters particularly at the lower frequencies, e.g. pirate radio stations, Private Mobile Radios etc, with a relatively low number of Sensors. An antenna located lower down e.g. on building walls or item of street furniture such as a lamp-post, typically gives more selective performance and sensitivity against lower level and/or high frequency emitters (due to free-space path loss.) In both cases the small size of the unit significantly reduces installation and maintenance complexity and therefore cost. Agilent believes the cost of installation and site rental of such a AMS will be significantly less that those predicted in the Phase 1 Study. Organising the sites for this trials would initially appear to confirm the viability of this approach. More details on the sites used for the trial are provided in Section below, Page 26

27 6 ANALYSIS OF AMS Phase 1 Agilent studied the Phase 1 report in considerable detail and made a number of observations. 6.1 TDOA results The TDOA results achieved in Phase 1 were broadly consistent with Agilent s experience with similar measurements across a predominantly open area; against reasonably high level signals and low RF propagation effects like multi-path. The measurements were not however performed in a typical urban environment. 6.2 Number of Sensors The geo-location measurements performed in Phase 1 all appeared to be based upon three Sensors, i.e. the minimum number of Sensors required to make a TDOA measurement and obtain a location for the emitter. Agilent experience in TDOA is that three Sensors is generally sufficient in an open area or where the antennas are on elevated positions, where reasonable line-of-sight conditions and against a reasonably high level signal. In an urban environment however, effects like multi-path makes a significant impact on measurements and the number of Sensors add significantly to the operation of the system. This is mostly since three Sensors will at best result in three correlations pairs and they all have to be good to locate the emitter. Agilent proposed performing trials in an urban environment with five Sensors. Only with more than three Sensors it is possible to understand the impact of propagation effects like multi-path and determine how the number of Sensors truly impacts the measurement. Performing measurements in an urban environment with three Sensors is likely to result in more questions than answers. Page 27

28 6.3 Signal to Noise The Ofcom Phase I trial report appears to over-emphasize sensitivity in the sensor network planning and TDOA accuracy. While certainly a factor, Agilent s own experience suggest that multipath tends to dominate the ability to make measurements before Signal to Noise Ratio (SNR) particularly in an urban environment Doppler Compensation The Phase 1 report does not make any mention of the algorithms performing Doppler compensation. In TDOA systems, signals must be compensated for Doppler offset as the transmitter signal may be frequency shifted. This can occur even when the transmitter is stationary. Doppler shifted signals may also result from moving signal reflectors. This is especially true in dense urban environments where street-level transmitters are often next to busy arterials, and there are no directpath signals to work with only reflected signals from buildings and vehicles Multi-path The Phase 1 report makes mention of Multi-path but the system developed does not appear to attempt to address this issue. Agilent s experience is that such a solution must address this when operating in an urban environment. More details on multi-path are provided in Section above Summary Agilent considered than any urban trial need to be performed with more than three Sensors and additional algorithms/techniques (beyond those described in the Phase 1 report) are required to obtain a representative study of the true performance a TDOA in an urban environment. Page 28

29 7 WORK PROGRAM 7.1 Introduction Agilent proposed the following program of work to evaluate TDOA in the dense UK urban environment of Central London. This was intended to give Ofcom the most comprehensive trial while minimising cost and risk. The main objectives of were to: Profile actual RF propagation (at different frequencies) in the environment where the measurements are to be performed. Perform measurements to determine sensitivity and sensor density required for each test signal type in an urban environment, taking into account both signal detection and TDOA accuracy Demonstrate the ability of the system to indicate a line-ofbearing for emissions outside of the sensor network. Agilent proposed to establish system sensitivity in terms of Carrierto-Noise (C/N) as request by Ofcom. Agilent did however highlight that while this is a useful metric in terms of a traditional Direction Finding (DF) system, in a TDOA system sensitivity is more complicated as the system performance is not just a function of antenna gain and receiver performance. 7.2 Sensor Configurations Based upon similar field trials, Agilent believed the optimal configuration for the trial was a square of four Sensors (one in each corner) with an additional sensor in the centre. This allows the results to be analysed as if there were three or four Sensors by recomputing the TDOA algorithms on the original data retained from each measurement. In addition, Agilent proposed to make tests with two sensor separations. In Figure 7, the red circles indicate the provisional sites to cover the full 6km square configuration. The sensor sites were selected to include Guys Hospital and The Barbican as known sources of off-air signals. Page 29

30 Figure 7: Provisional Sensor Locations Page 30

31 A second configuration was on a smaller spacing based upon reusing three of the original sites and moving two Sensors to new sites. An example of such an arrangement is shown by the blue crosses. The actual location of the smaller grid was to be determined during the proposed work. While the use of mobile based Sensors was not ruled-out, Agilent focussed on the use of a great number of fixed sites as it was considered this would provide more reliable communications. Agilent s experience from previous trials is that five Sensors and two different spacing provides more than sufficient information to analyse different sensor geometries Sensor Sites Communications Where existing network connectively was not available Agilent proposed to install an ADSL line where possible. ADSL was expected to provided a reasonably reliable network connection. In the event of using a mobile based sensor Agilent proposed to use Wireless LAN or 3G. Agilent has previous experience in the use of 3G for data communication in Central London and also experience in the use of Wireless LAN on previous trials. Page 31

32 7.3 Project Tasks Overview Agilent proposed a task based approach to the work as detailed below Task 1- Project Start-up A start-up meeting will be held between Ofcom and Agilent to review the work to be performed and discuss the signals to be tested in more detail. At this stage Ofcom were to provide a list of known emitters and also frequencies, bandwidths and power levels that may be used for test emitters Task 2 RF Propagation Measurements Prior to making TDOA measurements or finalising the Sensor sites, Agilent considered it important to have real propagation measurements at the frequencies where measurements will be performed. While various propagation models do exist, Agilent s experience has been that real-life conditions vary significantly depending upon the environment and is impacted by many factors including the size and spacing of buildings. These relatively simple propagation measurements were to provide critical information in finalising the TDOA sensor locations and the analysis of the TDOA measurements Task 3 - Finalise Trials Plan and Sensor Locations Using the results of the propagation measurements Agilent then proposed to finalise the trials plan and sensor locations with Ofcom. These plans would include simulations of sensor network sensitivity for each of the test signals (frequency, modulation, power level). An example of a coverage map for a given signal is shown in Figure 8 below. The probability of locating the signal is represented in colour with blue being >80%, green 60-80%, magenta 40-60% and Red <30% Page 32

33 Figure 8: Example TDOA Coverage Map Task 4 Sensor Location Setup This task was to install the infrastructure required for all the sensor sites in the large and small area configurations. This includes the RF and GPS antennas, provision of power and the network communication (e.g. ADSL) Task 5 - TDOA Measurements Phase 1 This task was to perform the TDOA measurements with the Sensors in the large configuration. A range of measurements were to be made against known emitters and also self generated test signals to cover the range of emitters (listed in below) distributed within and some outside the sensor area. It was expected the Phase 1 measurements would take approximately 4 days however a total of 7 days were allowed in the overall project schedule to cover unexpected issues that can occur with field based measurements. An additional 3 days were allowed for setup time. Page 33

34 7.3.6 Task 6 TDOA Measurements Phase 2 This task was to perform TDOA measurement with the Sensors redeployed in the smaller configuration. The test procedure was similar to Phase 1. As with Phase 1 the measurements were expected to take approximately 4 days with 7 days allowed in the overall project schedule to handle unforeseen events. Additional time was also allowed for moving two of the Sensors to the new sites. Agilent did not rule-out the use of mobile based sensor, however Agilent s experience from previous trials is that five Sensors and two spacing generally makes this unnecessary. Agilent planned to perform initial analysis of the Phase 1 and 2 measurements and decide if further measurements with a mobile based sensor were required. If necessary these were to be performed during the early stages of the Results Analysis Task using addition resources Task 7 Results Analysis While some results would be available immediately, comprehensive results would not be available until more detailed analysis has been performed. Results were to include the following: Typical TDOA accuracy including key performance numbers will also be extracted. For example: X % percent of the measurements were accurate to Y metres. Analysis of contributing factors to the TDOA results. This would include accuracy as a function of frequency, bandwidth, location, and signal level, for example. Analysis of live-signal measurements specified by Ofcom Recommendations for sensor density and other performance related system and sensor specifications Task 8 Final Report A final report was to be provided detailing the work completed. Page 34

35 8 WORK PERFORMED 8.1 Task 1- Project Start-up A start-up meeting with Ofcom to review the work to be performed and discuss the signals to be tested in more detail. Ofcom provided a list of known emitters and also a list of frequencies, bandwidths and power levels that may be used for test emitters. Agilent then applied for a Temporary License from Ofcom to transmit on these frequencies for the trial. 8.2 Task 2 RF Propagation Measurements Prior to making TDOA measurements or finalising the Sensor sites, Agilent considered it important to have real propagation measurements at the frequencies where the TDOA measurements would be performed. This work was performed using a signal source at a fixed site and a receiver mounted on a vehicle to measure actual path loss. A GPS receiver was used to record the location of the car for each measurement. This information was then entered into Agilent s Sensor Planning tool to build an accurate model of the environment Transmit Frequencies Agilent obtained permission from Ofcom to transmit on a list of frequencies as detailed below. Frequency Max Bandwidth MHz 12.5 khz MHz 12.5 khz MHz 12.5 khz MHz 2 MHz MHz 2 MHz MHz 250 khz The test signal was a modulated periodic signal allowing more accurate power measurements under poor Carrier to Noise (C/N) conditions. Page 35

36 8.2.2 Measurement Setup Figures 9 to 11 show the measurement setup for the propagation measurements. Figure 9 shows the vehicle used with the antenna on the roof. The antenna is shown in more detail in Figure 10. Note the antenna used is the same antenna proposed for the Sensor sites. Figure 11 shows the Spectrum Analyser used to measure the received signal. This was powered from the vehicle s 12Volt supply using a DC to AC inverter. A diesel vehicle with a large alternator was selected to ensure sufficient power was available. Figure 9: Transmitting Car (antenna on roof) Page 36

37 Figure 10: Close-up of Receive Antenna Figure 11: Spectrum Analyser in the car Page 37

38 8.2.3 Transmit Sites Agilent selected two locations for the signal source based upon the expected locations for the TDOA Sensors Transmit Site 1 The first site was the Ofcom building. This was selected since it was almost certainly going to be one of the Sensor sites. It was also in the centre of the City and also representative of a tall building. The antenna setup is shown in Figure 12 below. A panoramic view from the roof level is also shown in Section Figure 12: Ofcom building antenna Transmit Site 2 The second transmit site was based upon a car park in the Pall Mall area of London. This was selected since it was more representative of an antenna located at or around the typical roof level of buildings anticipated for the some of the sensor sites. The site and the antenna setup are shown in Figure 13 below. A 40ft antenna mast was used on top of the car park elevating the antenna to a level similar to the buildings to the North and East which were the areas of interest. Buildings to the south were higher than the antenna but this direction was not of particular interest. A panoramic view from the car park level is also shown in Section Page 38

39 Figure 13: Transmit Site 2 Page 39

40 8.2.4 Measurement Locations Figures 14 and 15 show the measurement locations for Transmit Site 1 (Ofcom) and Site 2 (Car Park) respectively. The red circle shows the location of the transmitter. Measurements for Site 1 (centrally located to the planned trial area) focused on a 180 degree span to the West, North and East. Measurements for site 2 (located to the West of the trials area) focused on a 90 degree span to the North and East Measurement Methodology The traditional way of measuring power is to record the peak power of an un-modulated carrier or the sum of the power with the bandwidth of a modulated signal. This is typically performed with the marker functions on a spectrum analyser. The main problem with this approach is that the measurement is the sum of the signal power and any noise. At lower Signal-to-Noise Ratios (SNR) the measurement can be heavily biased by the noise floor of both the measuring equipment and environmental noise. SNR can be improved by using an un-modulated carrier and with narrow Resolution Bandwidth (RBW) filters on the spectrum Analyser. For propagation measurements this approach is however highly susceptible to error due to frequency selective fading conditions found in high-multipath environments. Doppler effects from moving vehicles, even when the receiver is stationary, can also introduce error as the signal is frequency shifted away from the centre of the measurement filter. For these reasons a correlation approach was used for the propagation measurements. Most TDOA technologies use some form of cross correlation to establish the timing relationships between two signals. The cross correlation is performed between the signals from two receivers, or between a signal from a receiver and a reference signal. In the crosscorrelation process, signals that are uncorrelated are attenuated relative to common signals. For example, the noise observed by two monitoring receivers is not likely to come from a common source it is uncorrelated and is therefore attenuated. To determine path loss, a modulated signal was transmitted, and subsequently received and digitized by the receiver. The digitized signal was then compensated for Doppler and correlated against a reference waveform representing the ideal transmitted signal. This correlation approach allowed for accurate path-loss measurements under low signal-to-noise and high multipath Page 40

41 conditions. Using a spectrum analyzer approach, the data would have asymptotically approached the receiver noise floor with distance. In the correlated power results there s no obvious flattening of the curve at the lower detected power levels. To predict sensor network performance where multiple receivers with overlapping coverage are required to detect a signal, it becomes important to understand not only the mean path loss, but also higherorder statistics. The graph shows high variability in the measured results, not due to measurement error, but caused instead by the significant amount of multipath and shadowing in dense urban environments. Channel reciprocity is assumed. During this phase of the trial, logistics were simplified by moving the receiver, and locating the transmitter at potential sensor sites Amplitude Corrections The table below shows the corrections for each frequency for each site. Correction compensates for antenna gain, mismatches etc. Cable loss is assumed to be ideal (perfect match conditions) and antennas are assumed to be ideal isotropic radiators. System is calibrated to obtain the expected free-space path loss at 50 meters. DAY 1 CONFIGURATION (Short Coax, Short Mount) ESG Amp TX TX Ant RX Ant Cal Free Space Cable Expected Measured Correction FREQ Pwr Gain Power Gain Gain Distance Path Loss Loss Pwr Power Required 8.26E E E E DAY 2 CONFIGURATION (Long Coax, Tower) ESG Amp TX TX Ant RX Ant Cal Free Space Cable Expected Measured Correction FREQ Pwr Gain Power Gain Gain Distance Path Loss Loss Pwr Power Required 8.26E E E E Page 41

42 Drive Test, TX at Ofcom Sensor Sites Pre-Trial Drive Test Figure 14: Measurement Locations for Transmit Site 1 Page 42

43 Drive Test, TX at Car Park Sensor Sites Pre-Trial Drive Test Figure 15: Measurement Locations for Transmit Site 2 Page 43

44 8.2.7 Measurement Results The graph below shows a typical result of amplitude against distance. This example is for the 169.6MHz test frequency. There are four data sets depicted in the graph. For each data set, there is also a best-fit log curve. The diamonds indicate amplitudecorrected, correlated power measurements. The squares are the uncorrected measurements. The triangles and crosses depict two industry-standard path loss models, the Hata model and the WIM NLOS model MHz dbm Correlated Pw r WIM NLOS Hata Uncorrected Log. (Correlated Pw r) Log. (Uncorrected) Log. (WIM NLOS) dist (m) The graphs below show results for four frequencies for Site 1 and Site 2 respectively. In general, there is good agreement between industry standard models and the measured results for nominal path loss (without higher order statistics). The one exception was the data for site 1 (Ofcom building) at 919 MHz. In this result, the shortrange path loss was lower, but with a greater slope. Page 44

45 MHz MHz dbm Corrleated Pw r WIM NLOS Hata Uncorrected Log. (Corrleated Pw r) Log. (Uncorrected) Log. (WIM NLOS) dbm Correlated Pw r WIM NLOS Hata Uncorrected Log. (Correlated Pw r) Log. (Uncorrected) Log. (WIM NLOS) dist (m) dist (m) MHz 919 MHz dbm Correlated Pw r WIM NLOS Hata Uncorrected Log. (Correlated Pw r) dbm Correlated Pw r WIM NLOS Hata Uncorrected Log. (Correlated Pw r) Log. (WIM NLOS) dist (m) -140 dist (m) Site 1 : Measurement Results Page 45

46 MHz MHz Correlateed Pw r -40 Correlated Pw r -60 WIM NLOS Hata -60 WIM NLOS Hata dbm Uncorrected Log. (Correlateed Pw r) Log. (WIM NLOS) dbm Uncorrected Log. (Correlated Pw r) Log. (Uncorrected) -120 Log. (Uncorrected) -120 Log. (WIM NLOS) dist (m) dist (m) MHz 919 MHz dbm Correlated Pw r WIM NLOS Hata Uncorrected Log. (Correlated Pw r) Log. (Uncorrected) Log. (WIM NLOS) dbm Correlated Pw r WIM NLOS Hata Uncorrected Log. (Correlated Pw r) Log. (Uncorrected) Log. (WIM NLOS) dist (m ) dist (m ) Site 2 : Measurement Results Page 46

47 8.3 Task 3 - Finalise Trials Plan and Sensor Locations Task 3 was to finalise the trials plan and sensor locations Trials Plan Agilent developed a Trials Plan for the TDOA measurements to be performed. For completeness the main sections of that document are provided here. The trials plan defined the test signals and the general areas where test signals would be transmitted (relative to the Sensors). It also defined pre-existing signals to be measured and more detail on how the trial would be performed Features to be Tested The objective was to detect and geo-locate signals with the following characteristics: These signals are representative of a wide range of signals including FM/AM, DSSS, TDMA, Broadcast, etc. Bandwidths 10kHz to 2MHz Frequency 20MHz to 3GHz (Test frequencies at 82, 169,454,919 and 1802MHz) Modulation Formats Narrowband FM Aircraft AM FM Broadcast 270kHz GMST BT-0.3 DSSS (1.25MHz CDMA) Continuous 2MHz BW QPSK Pre-existing Signals Selected by Ofcom Signal Format Amplitude Location Other Continuous (CDMA), Single Event (PTT FM) 10W (0dBi Antenna) Some test will invoice dropping power to determine system limits Generally inside the perimeter defined by the Sensors. Identify direction of signals originating outside the perimeter of the Sensors. Including broadcast TV/FM.AM, Fixed LMR etc. Page 47

48 Measurements do not include basic monitoring of the spectrum Test Methodologies Testing Approach Testing requires a minimum of three people, one to operate the system, and one to move and control an RF emitter and a qualified driver. The System Operator (SO) will coordinate the testing, and, verify and annotate data as it's collected. The Mobile Operator (MO) will physically move an RF emitter to each pre-determined location and verify each location by recording the LAT/LON using a handheld GPS receiver. The MO will also document site characteristics. The mobile transmitter will be a signal generator with an amplifier and antenna. All transmissions will be made using omni-directional antennas. During a portion of the testing, two transmitters will be used. A second MO will be required to operate the second transmitter. The second transmitter will remain at a fixed location selected during the trial. This site is likely to be one of the other sensor sites not being used in the current (larger vs. smaller) measurement configuration or could be from another site Test Data All test data will be stored in the central server as it's collected. The database will be backed up after each day of active TDOA testing. Both SO and MO will keep paper logs as the data is collected. These logs will note: time, location, signal characteristics and measurement ID. If it does not slow data acquisition, the SO will add appropriate comments to the signals in the database as they are collected. The MO will document the environment. This includes distances to nearby buildings, and a description of the line-of-sight conditions to each sensor site. If appropriate, photographs of the TX site in the direction of each sensor will be taken. If not, a written description of the surroundings is required. One of the objectives of the trial is to gather requirements and use scenarios. As problems occur, shortcomings are discovered and ideas for improvements are generated, these should be documented by the SO, MO or Project Manager Page 48

49 Transmit Site Selection During the trial, testing was to be performed from approximately 17 locations distributed over 4 geographic regions as shown in Figure 27 below. Geographic Region # of TX Locations Figure 16: Transmit Site Regions Transmit Vehicle Agilent estimated the time to perform the measurements at each location (for all frequency and modulation formats etc.) could easily exceed 1 hour. From the experience of the propagation measurements, Agilent had concerns about being able to park a vehicle for this length of time in one place in Central London. Since this was not practical and/or likely to disrupt the trial, Agilent planned for the transmitter to be on a larger Stroller which can be pushed along the pavement. A suitable vehicle was then to be used to transport this at the beginning and end of each day. Page 49