Behavior of the GPS Timing Receivers in the Presence of Interference Faisal Ahmed Khan School of Electrical Engineering and Telecommunications, and School of Surveying and Spatial Information at University of New South Wales Biography: Faisal Ahmed Khan is currently a research student at University of New South Wales, pursuing the M. Phil. degree. His main area of research is interference effects and analysis in GPS environment. He holds a Bachelor of Engineering Degree in Electronics from NED University of Engineering & Technology, Pakistan. He has also gained hands-on experience in the field of Satellite Communications at Institute of Space Technology (IST), Pakistan and Pakistan Space and Upper Atmosphere Research Commission. Abstract: GPS timing receivers have been used for synchronizing telecommunications equipment since the early 9 s, currently providing an accuracy of up to 1ns. Such a high requirement of accuracy demands excellent operation from GPS timing receivers, avoiding the degradation of the synchronization process. Interference is an important threat to GPS performance. Such harmful interfering signals can originate from any electromagnetic radiation source producing its main signals or their harmonics in GPS bands. These sources may include signals from UHF/VHF TV transmissions, FM transmissions and VOR/ILS services. Any degradation in performance, due to introduction of interference can cause these receivers to provide a low quality timing solution, or to lose lock with incoming GPS signals altogether. This consideration motivates the study of the performance of GPS timing receivers in the presence of harmful interference. This paper proposes a hypothesis about interference effects on GPS timing receivers, and then confirms the hypothesis with experimental results. It also provides reasoning to explain this behavior of GPS receivers and identifies areas for further investigation. I. Introduction: Over recent years, telecommunications networks coverage, complexity, and data rates have increased manifold. This has called for a highly precise time reference for keeping these networks synchronized. According to IS-95 standard, CDMA networks expect an accuracy of 7µs per day from synchronization sources. According to a press release by Qualcomm, 1xEV-DO, which relies on GPS for their synchronization operations expect the provision of a data rates more than 4.9 mbps [1]. Although individual atomic standards like Rubidium or Cesium can be used for the job, GPS can uniquely synchronize geographically isolated networks, acting as the worldwide time disseminator. GPS is itself backed up by a large number of atomic clocks. GPS satellite clocks are continuously steered to remain synchronized with UTC (Coordinated Universal Time) [2]. GPS Based Clocks (GBC) therefore synchronize communications networks by steering the networks clocks to remain locked to world s most accurate time reference, direct connection to an atomic standard. GBC are capable of providing timing with an accuracy of up to 1 15ns accuracy [3]. Such high accuracy solutions require an electromagnetic environment that is free of interference, to avoid performance degradation. All GPS receivers are connected to GPS satellites via the air interface. The wireless nature of these links makes them vulnerable to interference. An interfering signal can affect the code correlation process reducing C/N o [4]. A reduced C/N o can cause tracking loop measurement errors, degrading the GPS position, velocity and time (PVT) solution [7] or can even cause tracking loops to lose lock altogether. Once this happens, network synchronization would be at the mercy of the local oscillator in the GPS timing receiver, which will drift relatively rapidly away from GPS/UTC time. Thus the study of interference effects on the performance of GPS timing receivers is crucial. This paper emphasizes the need for this study, investigates the effects of interference on GPS timing receivers, provides experiment results, and discusses these results to appreciate the problem. Section II defines the problem by shedding light on the importance of this study and providing some background information. Experimental methodology with some explanation for its adoption are provided in section III. This is followed by experimental results and discussion of these results in section IV. Finally, section V concludes this paper.
II. Background and Problem Definition: The evolution of communications technology, with the introduction of more complex and faster services, makes the provision of accurate timing more critical each day. An example of a high profile network failure has occurred: a national mobile network lost an entire region for a quarter of a day, thereby losing communication link with the entire world [5]. Other such examples are recorded in networks performance logs, but these are of course not discussed publicly because of the fear of losing business. Although these network casualties are relatively rare, the likelihood of them happening increases considerably if care is not taken to protect the network from adverse conditions and causes. The weak nature of GPS signal power levels makes them vulnerable to interfering signals. The International Telecommunications Union (ITU) has approved the 1559 161 MHz band for Radio-navigation Satellite (space-to-earth/space-to-space) services [6] so there should be no other signals in this band to provide interference. However, it is possible that GPS services can be disrupted by in-band or out-of-band interference. These may include interfering signals from intentional or unintentional sources like jammers, RF signal harmonics, near/far-band un-modulated carriers etc. GPS C/A code signals are spread using a 1ms periodic code. This gives them a line spectrum, with spectral lines 1 KHz apart [7]. Considering the case of a CW interfering signal, if its frequency coincides with any of these lines, it can leak through the baseband correlator circuit, disrupting the code search, acquisition and tracking procedures [7]. As the code spectral lines center frequencies change due to the Doppler Effect [8], this CW interference can affect a number of spectral lines over a period of time. If the interference is strong enough, it may cause the GPS receiver to lose lock with the particular satellite to which these lines belong, which in turn may degrade the PVT solution, or prevent positioning at all if there are insufficient satellites not affected by the interference. An example of a GPS timing receiver, the Motorola M12+ can work in Position-fix mode solving only for time, after determining position. It then produces a time solution for each satellite. If any of these individual solutions goes bad, it uses TRAIM (Time Receiver Autonomous Integrity Monitoring) to remove it from the cumulative solution. This removal of a satellite from the solution also causes the solution to degrade in accuracy. We proposed a hypothesis on theoretical grounds that in this Position-fix mode, effects of interference on the GPS timing solution could be in three phases (see Figure 1): Phase 1: Here interference encountered remains within manageable limits. The receiver will keep on solving only Figure 1 - Proposed hypothesis for behavior of GPS timing receiver in the presence of interference. The clock error is the difference between the receiver s time and UTC or GPS time for time, remaining in position-fix mode. Solution degradation should be negligible (within 1 2 ns). Phase-2: Here the effects of interference become noticeable. The timing solution will start degrading, as individual satellite measurements are degraded and some are lost. Phase-3: Here interference is increased to such an extent that the GPS receiver loses all the satellites. The time solution will be provided on the basis of local oscillator characteristics, i.e. a random walk with respect to GPS time or UTC. This hypothesis was then tested by experiments. This paper presents the results and justifies the results logically. It also identifies their agreement with hypothesis. Effects were studied using different types of interference signals including narrowband (CW), modulated (FM) signals and white noise signals. III. Experiment Methodology: Figure 2 shows the experimental setup adopted for testing the vulnerability of GPS timing receiver performance. Motorola s M12+ timing receiver was used as the test receiver and was compared against two identical SigNav receivers. The M12 was used as the test receiver because of its timing solution quality. It is capable of providing 1 pulse per second (PPS), with less than 2 ns 1-sigma average accuracy, if the PPS is corrected using clock granularity message. This has been fully calibrated to UTC at USNO [9]. Initially SigNav, NovAtel and Leica Mc5 were the candidates to serve as reference. From these, SigNav s MG series receiver was selected as the mean clock pulse difference of the two receivers output PPS was less than 15ns, in absence of interference. Two SigNav receivers were used in parallel, to make their solution more stable, by averaging out their results. First the M12 s output PPS was
Figure 2 - Experiment Setup compared against the SigNavs output PPS, without introducing interference, and the mean and standard deviation of the PPS phase offset was recorded between the two. These observations served as the reference for the experiments to follow. Later, interference signals were introduced in the GPS signal path to the M12+ only and the effect of this were recorded. Simulated signals, generated by SPIRENT GSS656 GPS Simulator, were used for the experiment to ensure the signals under test were consistent. The Hewlett-Packard HP8648B signal generator was used to generate interference signals. A number of experiments were performed for each case to confirm the results. IV. Experiment Results: Experiments were carried out to validate the hypothesis. As mentioned above, CW, FM and noise signals were used for testing the GPS timing receiver vulnerability. The following paragraphs present and discuss the results obtained. 1. Narrow-band Interference: CW interference centered at 1575.42 MHz (the GPS L1 frequency), with an initial power of -136 dbm was introduced in the path of GPS signal to the M12+ receiver. This was gradually increased, with irregular steps, until the M12+ receiver lost lock on all the satellites. Figure 3 shows the standard deviation of the phase offset in ns, between the PPS of the M12 and the average PPS of the SigNav receivers. For each point in the graph, the interference level was kept constant for 18 PPS readings, and mean and standard deviation were calculated for plotting. It can be noticed from the graph that standard deviation remains below 2 ns until the interfering signal power level reaches 87dBm. This point is common for all three runs, and can be considered as the power level below which interference is manageable by this GPS receiver. During this period (-136 to -87 dbm) the receiver is able to acquire and track satellites and the solution does not deviate considerably. This situation can be correlated to the Phase 1 region of the hypothesis in Figure 1. Beyond this point, it can be observed that the standard deviation starts increasing, and finally the timing receiver lose the lock. In this period (-87-82 dbm), timing solution can be seen to be degrading, consistent with Phase 2 of the hypothesis in Figure 1. We calculated the Allan Deviation of the collected data. Allan deviation shows the stability of a device with increasing averaging time. This averaging removes the noise, and shows the actual deviation of the device output. It can be computed as: σ ( τ ) = y 1 2( M 1) M 1 i= 1 ( y 2 i+ 1 yi ) where y i is the set of offset measurements and M is the number of data points in y, which is equally spaced by τ seconds. Figure 4 shows the Allan Deviation plotted for run 3. Each line indicates a particular value of interference power level. Solid lines indicate power levels from -136-87 dbm. It can be observed that these lines are in agreement with each other, showing that the timing solution is stable, although the interference power levels are increasing. This confirms the observations made from Figure 3, and relate to Phase 1 of the hypothesis. Dashed lines show lines for higher power levels. It can be noticed that these lines have a rising trend, which shows degradation of timing solution with time. 45 Run-1 Run-2 Run-3-136 dbm -116 dbm -96 dbm -87 dbm -86 dbm -84 dbm -82 dbm Std. Dev. of Clock Offset (ns) 4 35 3 25 2 15 σ (τ) 1 2 1 5 2 base -136-126 -116-16 -96-91 -89-87 -86-85 -84-83 -82 CW Interference @ 1575.42 MHz (dbm) Figure 3 - Standard deviation of phase offset between M12's PPS and SigNavs' PPS, due to CW interference. τ (s) Figure 4 - Allan deviation of the phase offset of M12 due to CW interference.
Also, it can be noticed that as the interference is increased, the solution is becoming increasingly unstable. It can be inferred that CW interference is leaking through correlator, over-riding the spectral lines. This causes the solution for individual satellites to degrade. TRAIM detects these satellites and removes them from the cumulative solution, causing a decrease in its accuracy. It can be seen that the solution accuracy degrades from less than 2 ns to more than 4ns. Also, these values may differ, on a case-to-case basis, depending on the observation time and type of receiver.. Because of this, the experiment was performed multiple times, and it can be seen that all the data runs are in relative agreement with each other and appear to be consistent. Once all satellites are lost, the GPS timing receiver outputs PPS on the basis of local oscillator characteristics, which follows a random walk. 2. FM Interference: A sine wave carrier at 1575.42 MHz, modulated by signals given in Table 1, was used as interfering signal. The same experiment was repeated as for the case of CW interference. Experiments were performed 5 times for each modulating signal, and their average values were plotted. This causes more satellite signals to be degraded simultaneously, causing them to be excluded from the cumulative solution by TRAIM. It can also be observed from the graph, that the third signal, which has the largest bandwidth (i.e. 1 KHz), has the graph with steepest slope. This confirms the points stated above, as this signals simultaneously affects more satellites than the other two, which causes the receiver s solution to degrade at a higher rate. Another observation which can be made in this case is that the receiver loses lock before much degradation occurs in the timing accuracy, and functions (i.e. Phase 2 in Figure 1 is narrower). Std. Dev (ns) 12 1 Center Frequency Bandwidth 2 KHz 4 Hz 4 KHz 4 Hz 2 KHz 1 KHz TABLE 1 - Modulating signal specifications 8 6 4 FM 2KHz 4Hz FM 4KHz 4Hz FM 2KHz 1Hz σ (τ) 1 2 Base -136 dbm -116 dbm -96 dbm -94 dbm -93 dbm -92 dbm 1-1 τ (s) Figure 6 - Allan Deviation of timing solution of M12 in presence of FM interference. Figure 6 shows Allan deviation of GPS timing in the presence for the third FM interference signal. It can be seen that for interference levels below -96 dbm, the Allan deviation has a decreasing trend (shown by solid lines). As identified in Figure 5, this represents Phase 1 of the hypothesis (-136 to -96 dbm). Above this interference power level (-95 to -92 dbm), Allan deviation lines (showed by dashed lines) have an increasing trend, showing degradation in the timing solution, which can readily be identified as Phase 2 of the hypothesis. It can be noticed that for any particular line, although the interference power level remains constant, the timing stability keeps on decreasing. 3. White Noise Interference: When a receiver s front end encounters noise, the AGC reduces the front end gain to maintain the original RMS level at the input of ADC [7]. We reduced the GPS signal power level by introducing 3dB attenuator between antenna cable and GPS receiver. This simulates the addition of white noise as an interferer. The AGC output in this case is the same as if white noise was added to the front end.. No external interfering signal was introduced for this experiment. It was observed in the case of FM interfering signals that when signals from multiple satellites were affected simultaneously, 6 ns 4 ns 2 ns -2 ns Mean Standard Deviation 2 base -136-126 -116-16 -96-95 -94-93 -92 FM Interference: Carrier @ 1575.42 MHz (dbm) Figure 5 Standard Deviation of Phase Offset between M12's PPS and SigNavs' PPS, due to FM Interference. -4 ns -6 ns Base 3 db Figure 7 Mean and Standard Deviation of Phase Offset in presence of 3 db attenuation
the GPS receiver loses lock more quickly, rather than experience a gradual degradation in the solution. An enhanced version of such an effect can be observed here, as here white noise affects all the spectral lines in the GPS signal spectrum, causing the solutions from all the satellites to degrade simultaneously. Figure 7 shows the variation in mean and standard deviation in phase offset of M12 s PPS. It can be noticed that standard deviation varies more than the mean. Also, in Figure 8, Allan deviation shows an increasing trend when 3dB attenuation is introduced. This introduction of 3dB attenuation emulates the effect in the presence of noise which causes the AGC to suppress the incoming signal by 3dB. When further 3 db attenuation was introduced, M12 was unable to acquire any satellites. Acknowledgements: The author would like to thank A/Prof Andrew G. Dempster, University of New South Wales, for his supervision, guidance and technical discussions that took place during this work. References: 1. Qualcomm, QUALCOMM Expects Commercialization of EV-DO Rev. B in 27, Press Release, (http://www.qualcomm.com/press/releases/26/6 47_expects_commercialization_ev.html) 2. USNO GPS Time Transfer, http://tycho.usno.navy.mil/gpstt.html ( ) (ns) 3dB Attenuation Base (No Attenuation) 1 2 Figure 8 - Allan Deviation of timing solution of M12 in presence of signal attenuation. Discussion and Conclusion: In this paper, we proposed a hypothesis about the behavior of GPS timing receivers in the presence of interference. This hypothesis was confirmed practically, and the results of the conducted experiments were presented. Narrowband, modulated and white noise signals were used for testing. It was established that highly accurate timing solution requirements may not be met by GPS timing receivers in the presence of interference. It can be observed from the results that although GPS receiver remains locked to the incoming GPS satellite signals, it still may not provide solution with the required accuracy. As proposed in the hypothesis, there will be levels of interference that do not affect timing, intermediate levels that, although the receiver may not lose lock, performance is degraded, and levels that prevent timing being locked to the UTC timing standard. This paper motivates needs for further examination of such effects and investigation of commercially available receivers for their interference rejection characteristics. In future, we plan to investigate the effects of degraded synchronization on communication networks. This will help us in recognizing the traffic handling ability of a poorly synchronized network. 3. Paul Skoog, Doug Arnold, Nanosecond-Level Precision Timing Comes to Military Applications COTS Journal, (http://www.cotsjournalonline.com/home/printthis.p hp?id=141 4. J. J. Spilker, Global Positioning System: Theory and Applications, chapter 3, Volume 1, American Institute of Aeronautics and Astronautics, 1996. 5. Curry, Charles, The Network Pacemakers, IRT Communications Engineer Journal, December/January 26/7. 6. Article 5, Radio Regulations, World Regional Conference, International Telecommunications Union (ITU), 23 7. Kaplan, Elliot D., Hegarty, Christopher J., GPS: Principles and Applications, 2nd Edition, Artech House Publication, 26. 8. Asghar T.Balaei, Andrew G. Dempster, Joel Barnes, A novel approach in detection and characterization of CW interference of GPS signal using receiver estimation of C/No presented in PLANS (ION IEEE) - April 26 9. Richard M. Hambly, Dr. Thomas A. Clark, Critical Evaluation of The Motorola M12+ GPS Timing Receiver Vs. The Master Clock at the United States Naval Observatory, Washington, Dc 34th Annual Precise Time and Time Interval (PTTI) Meeting, 22.