GIDL: Generalized Interference Detection andLocalization System

Konstantin Gromov, Dennis Akos, Sam Pullen, Per Enge, and Bradford ParkinsonDepartment of Aeronautics and Astronautics, Stanford University, Stanford, CA

BIOGRAPHY

Konstantin Gromov obtained an M.S. degree in ElectricalEngineering in Leningrad Electrical Engineering Institutein 1992, B.S. degree in mathematics from Leningrad StateUniversity in 1991, and M.S. degree in Astronautics andAeronautics at Stanford University in 1994. Currently, he isworking on his Ph.D. degree in Aeronautics and Astronau-tics at Stanford University under Prof. Bradford Parkinson.As an undergraduate and graduate student, he has workedon variety of topics involving GPS, in particular LAAS.

Dr. Dennis M. Akos completed the Ph.D. degree in Electri-cal Engineering at Ohio University, conducting his gradu-ate research within the Avionics Engineering Center. Aftergraduating he has served as a faculty member with LuleaTechnical University, Sweden and is currently a researchassociate with the GPS Laboratory at Stanford University.His research interests include GPS/CDMA receiver archi-tectures, RF design, and software radios.

Dr. Sam Pullen completed his Ph.D. in Aeronautics andAstronautics at Stanford University, where he is now theTechnical Manager of Stanford’s Local Area AugmentationSystem project. His research includes performance predic-tion, optimal system architectures, and integrity algorithmsfor both the Wide Area and Local Area Augmentation Sys-tems.

Dr. Per Enge is an Associate Professor in the Departmentof Aeronautics and Astronautics at Stanford University. Hereceived his Ph.D from the University of Illinois. Prof.Enge’s research currently centers around the use of GPSas a navigation sensor for aviation.

Dr. Bradford W. Parkinson is the Edward C. WellsProfessor of Aeronautics and Astronautics at StanfordUniversity. He served as the first Program Directorof the Joint Program Office and was instrumental in

44

GPS system development.

ABSTRACT

The Local Area Augmentation System (LAAS) and theWide Area Augmentation System (WAAS) are being devel-oped by the U.S. Federal Aviation Administration (FAA)to provide satellite navigation performance compliant withthe stringent requirements for aircraft precision approachand landing. A primary design goal of both systems is toinsure that signal-in-space failures are detected by groundfacilities and affected measurements are excluded beforedifferential corrections are broadcast to users. One suchfailure is unintentional interference or intentional jammingin the GPS frequency band. To protect integrity, LAASand WAAS ground facilities must quickly detect the pres-ence of interference that fall within the restricted zone de-fined by LAAS and WAAS system requirements and thusmay be hazardous to users. To protect availability, groundpersonnel must also be able to locate and deactivate the in-terference source.

To serve this purpose, Stanford University developed andtested a 1-D Interference Direction Finding system and re-ported its results at ION GPS-99. This prototype demon-strated the ability to locate wideband interference sourcesto within a few centimeters along a 12-meter track betweenthe two IDF antennas. This system has been significantlyexpanded and enhanced to form the Generalized Interfer-ence Detection and Localization System (GIDL) prototype,which includes four antennas and RF sections slaved to acommon clock to allow three-dimensional interference lo-cation. Measurements of differential signal propagation de-lays across the multiple baselines between the GIDL anten-nas are combined to estimate the location of the undesiredsignal transmitter in a manner analogous to GPS positiondetermination. The GIDL can be implemented in parallelwith a three or four-receiver LAAS ground facility (sharing

7

some components with the LAAS reference receivers andprocessors) or as a separate installation to support nearbyLAAS sites and WAAS approaches.

This paper describes the GIDL receiver design and derivestheoretical predictions of the ability of the GIDL to accu-rately locate interference sources. For the “star” configu-ration of GIDL antennas, where the antennas are separatedby 100 meters (as would be typical of a LAAS ground fa-cility), the GIDL should be able to locate a relatively stronginterference source 3 km from the center of the GIDL an-tennas to within 100 meters longitudinally and 5 meters lat-erally. Better accuracy can be obtained from longer antennaseparations. In addition, the paper includes test results froma real-time demonstration of the GIDL system in a largefield on the Stanford campus. In this test, a calibrated wide-band noise source at -70 dBW/MHz is moved around thefield between a set of pre-calibrated locations, and the real-time GIDL display provides up-to-date estimates of the lo-cation of this noise source to within the limits predicted bythe accuracy analysis. Further testing with a wider varietyof interference sources is planned in the near future.

INTRODUCTION

The Local Area Augmentation System (LAAS) is aground-based differential GPS system being implementedby the Federal Aviation Administration (FAA) for aircraftprecision approach and landing. LAAS will adequatelyprovide Category I service for those airports that are notcovered by the FAA’s Wide Area Augmentation System(WAAS) and is intended to provide Category II and Cat-egory III performance in the future [1]. A primary de-sign goal for LAAS is to insure that failures occurring inthe ground or space segments of GPS be eliminated by theground system before differential corrections are broadcastto users. One such failure is unintentional interference orintentional jamming in the GPS frequency band. To protectintegrity, the ground and air must quickly detect the pres-ence of interference. To protect availability, ground person-nel must also be able to locate and disable the interferencesource. The GIDL system would be able to assist groundpersonnel in finding interference sources, by providing es-timated locations of any interfering signals that lie out-side the tolerable LAAS interference environment, whichis specified in Appendix H of the RTCA LAAS MASPS[2].

In order to serve this purpose, a specialized prototypeknown as GIDL has been developed. GIDL can improveLAAS availability by accurately estimating the location ofthe source, particularly its direction from the GIDL lo-cation. GIDL activities can be implemented in parallelwith reference receiver functions and could share compo-

44

Receiver Mnf. Lost No pos. All SVs1 SV, m fix, m Reacq., m

Survey A 10 0 17Survey B 12 5 20Survey B 9 3 11LAAS Ref. A 15 3 19LAAS Ref. C 4 4 7LAAS Ref. C 12 6 18LAAS Ref. C 20 8 10Military (C/A) B 4 2 7Consumer C 6 4 7Aviation B 27 14 19Aviation B 12 6 14Consumer B 9 6 11Automotive B N/A 4 N/A

Table 1: Receiver Jamming Test Results

nents with the reference receivers and processors in LAASground stations.

There are several ways to implement sources of interfer-ence localization: interferometry, time-of-arrival differen-tial system, spatial spectrum estimation, phase antenna ar-ray, etc. Multiple direction and location finding algorithmscan share common hardware. Significant work has beendone on estimating direction of the arrival of signals. Thissubject has been studied almost since invention of the radio.There are significant number of algorithms which could beused in the GIDL system.

At Stanford University, an ongoing effort is focused on re-search, development, implementation, and testing of LAASarchitectures and architecture subsystems. This paper con-tinues to consider the development of the GIDL receiverand algorithms as an adjunct to the Stanford LAAS proto-type. Currently, most of our efforts concentrated on the in-terferomerical, or time-of-arrival techniques. Analysis andexperimental results are described in this paper.

We have also tested a number of GPS receivers to the suc-ceptability of interference to have a bench mark on rangeof the jammer. The interference source which has beenused for the test is a white noise source with power of -70dBW/MHz. Results of this experiment are present in theTable 1. The interference source was moved closer to thereceiver, and the jammer location where the first satelliteand all satellites are lost has been recorded. Then, the jam-mer was moved away from the receiver, and we recordedthe range when all the satellites were reacquired. These re-sults show that at a range of about 10 meters, this jammerbecomes a threat. We can scale this results to a jammer ofa different power.

8

Antenna 2

Baseline R

Some Signal

Propagation Delay τ

Antenna 1

To the Source ofInterference

θ

Figure 1: GIDL Concept — Interferometry

A1A2

A3

2-Dimensional Case-at least 3 Antennas are Needed

BL1

BL2BL3

Figure 2: GIDL Idea “Inverted Loran”

GIDL THEORY

One idea for localization, known as an “interferometrical”approach, illustrates the concept and simple algorithms fordirection finding. Assume that the source of interferenceis located far from the reception antenna so that we canassume planar wavefronts. Interfering signal propagationwould hit Antenna 2 first and then with some propagationdelay it would enter Antenna 1, as illustrated in Figure 1.If we can somehow estimate this propagation delay � , fromsimple trigonometry we would be able estimate the direc-tion to the signal source. There is an ambiguity associatedwith the single baseline, regarding which side of the base-line the source is located, but this ambiguity can be simplyresolved using multiple baselines. For many signals, it ispossible to estimate propagation delay � by correlating sig-nals received by Antenna 1 and Antenna 2.

44

Jammers

ReceiveAntenna Array

Receiver andOptimal Processing

Algorithms

Number of Jammers

Jammer Locations

Figure 3: GIDL Realization Concept

Another idea, which we call “inverted Loran”, also is quiteeasy to grasp, and it does not need the assumption of pla-nar wavefronts, as shown in Figure 2. In this case, measure-ments include signal propagation delay times between eachpair of antennas. For each baseline, the measured propa-gation delay would correspond to a hyperbola of possiblejammer locations. Two baselines would generate two suchhyperbolas. By finding their intersection, we would be ableto estimate jammer location. There is an ambiguity in thistechnique, as in the general case two hyperbolas could in-tersect at two points, but it is possible to resolve this am-biguity by various techniques, for example by adding extrabaselines.

GIDL HARDWARE OVERVIEW

One of the ways to build a prototype of such a system isshown in Figure 3. It is possible to build a receiver withfour RF sections (i.e. with ability to connect to four anten-nas). It would be operated from one common clock, so itwould be a completely coherent system. There are somelimitations to this prototype. Namely, it would have onlyfour antennas, and the antenna locations would be limitedby cable length from the receiver to the antenna.

This section summarizes the GIDL prototype hardware thatwe have built and tested. For a more detailed descriptionplease see our previous paper [3]. The GIDL receiver isimplemented with a single analog down-conversion/mixing

9

RF=1575.42 MHzLO=1527.68 MHzIF=47.74 MHz

IF

Figure 4: GIDL Frequency Plan: Analog Mixing

Fs = 38.192 MHz

IF = 47.74 MHz

Fs/2 = 19.096 MHz

0 MHz Fs/4 = 9.548 MHz

3Fs/2 = 57.288 MHz

Fs/2 = 19.096 MHz

6.548 MHz 12.548 MHz

IF = 5Fs/4 Fs = LO/40 = 1527.68 MHz/40 = 38.192MHz

Figure 5: GIDL Frequency Plan: Sampling With Aliasing

stage. A second down-conversion is done by aliasing dur-ing A/D conversion. The remainder of the signal process-ing is completely digital. The system RF bandwidth is 24MHz, and the IF bandwidth is 6 MHz, with a possible ex-pansion to 17 MHz to protect wideband LAAS referencereceivers (see Figure 4 and Figure 5).

The GIDL hardware setup is depicted in Figures 6, 7, and 8.A picture of the completed system is shown in Figure 9. Analuminum chassis hosts the RF/IF downconversion stages,master clock and synthesizer, and power supply. The A/Dsare PCI cards which are installed into a data processingPC, which is set up below the RF/IF stages. In the pic-ture in Figure 9, only antenna 1 and antenna 2 inputs areconnected.

EXPECTED SYSTEM PERFORMANCE

There are two major characteristics to the GIDL system:(1) maximum range and coverage; (2) the accuracy of thejammer location. The first describes what maximum rangethe system should be able to detect a jammer of a givenpower, and that range better be greater than effective radiusof the jammer. Another characteristics describes how ac-curately we can locate the jammer if we can detect it. Inthis section we look at both characteristics, state expectedperformance, and list aspects which can improve it.

45

RF Section and AtoDRF Section and AtoD

RF Section and AtoDRF Section and A/D

To data collector and(post)processing

Ref. Osc. and Synthesizer

LO to RF Mixers Clock to A/Ds

Figure 6: Diagram of GIDL Hardware Setup

FromAntenna

+19dB +19dB +19dB +20dB+20dB A/D: 12 bitsMixer

1575.42 MHz, BW=24 MHz 47.74 MHz, BW=6 MHz

To Processing

LO1527.68 MHz

A/D Clock: 38.192 MHz

AGCBPF BPF

Figure 7: Diagram of RF Section and A/D

TCXO

PLL Synthesizer

LO Out1527.68 MHz

Div. 40

A/D Clock Out38.192 MHz

16.368 MHz

Figure 8: Common Clock Diagram

0

1

Figure 9: Picture of the Completed GIDL Receiver

Maximum range and coverage of the system is determinedby looking at the post-processing signal-to-noise ratio ofthe received signals. That ratio is defined by setting proba-bilities of detection and false alarm.

Assume that we have a single baseline system as shown inFigure 10. Assuming that the desired post-processing SNR(q2out cor ik) is set than using link equation we can derive

that for an antenna pair Ai, Ak, the maximum range andcoverage determined by

kR� Lik2kR� Lkk

2 � F 4

1 ik;

orqout iqout k

q2out cor ik

� 1:

Where

F 4

1 ik = i(NG)Gi�

2

4�kTe� i

� k(NG)Gk�

2

4�kTe� k

�1

q2out cor ik

=

(NG)2GiGk�4 i k

(4�)2k2Te� iTe� kq2

out cor ik

:

This equation describes Cassini’s ovals with the center atthe center of the baseline between two antennas. For longranges these ovals approximate circles.

45

NG

NG

O - origin

Gi

Gk

Ai

Ak

iL

kL

R

Jammer

Figure 10: Single Baseline System

Equation 1 is the general equation, let’s see what results wewould get if we plug in some numbers which are of interestto us. In Figure 12 the maximum range and coverage forspecific baselines and jammer powers are plotted. Figure11 shows antenna array configuration which has been usedto generate each plot. This is the so called “Star Antenna”configuration where master antenna is located in the centerand three other antennas are located on some distance fromthe central antenna and spaced 120o apart. This configu-ration produces three independent baselines, on the eachplot three Cassini’s ovals are shown, one per each baseline,intersection of these ovals is the coverage of the whole sys-tem. The top row of the plots corresponds to the baselinesof 12 meters long, with the bottom row to the baselinesof 100 m. Note that for the weakest jammer and 100 mbaseline there is no coverage. The jammer is simple tooweak to be detected by two antennas simultaneously withthe given system parameters. Other assumptions whichused in generation of this plots: Gi = Gk = 1 — An-tennas are omnidirectional with unity gain; postprocessingSNR q2

out corr= 50 and it is set by probability of detection

Pd � 0:75 and probability of false alarm Pfa = 10�6);background white noise is kTe� = 4 � 10�21 W/Hz; andprocessing losses are i = k = 0:3

Comparison of these plots with Table 1 shows that the cov-erage of the GIDL is larger than effective radius of the jam-mer, which is taken as a loss of the one satellite, and onaverage is about 10 m for -70 dBW/MHz jammer (it is pos-sible to scale that number to a jammer of any power).

Everything what has been said in the previous paragraphis true for a GPS receiver using the same antennas as theGIDL system, but by using different antennas it is possi-ble to get a differential advantage of GIDL versus a GPSreceiver.

By using different antennas for GIDL and the GPS receiver

Figure 11: Antenna Configuration for Analysis

100m; 0.1W/20MHz 100m; 1W/20MHz

12m; 0.1W/20MHz 12m; 1W/20MHz

-50 0 50-50

0

50

North, m

Wes

t, m

12m; -70dBW/MHz

-2 0 2-2

-1

0

1

2

North, km

Wes

t, km

12m; 1mW/20MHz

-20 0 20-20

-10

0

10

20

North, km

Wes

t, km

-50 0 50-50

0

50

North, km

Wes

t, km

-200 0 200-200

-100

0

100

200

North, m

Wes

t, m

100m; -70dBW/MHz

-2 0 2-2

-1

0

1

2

North, km

Wes

t, km

100m; 1mW/20MHz

-20 0 20

x 104

-20

-10

0

10

20

North, km

Wes

t, km

-50 0 50-50

0

50

North, km

Wes

t, km

No coverage

Figure 12: Maximum Range and Coverage of the GIDLSystem With Baselines 12 m and 100 m and Star Configu-ration for Jammers of Various Power

45

Figure 13: GPS Receiver Antenna

Figure 14: GIDL Antenna

it is possible to improve the range of the GIDL and lessenthe effective range of the ground jammer. A typical an-tenna gain pattern of the GPS receiver is shown in Figure13. It has low gain at low elevations, exactly where thejammers most likely would happen be. On the contrary ifwe use antenna patterns similar to that shown in Figure 14we would get extra antenna gain in the most probable jam-mer direction. Thus the small gain of the GPS antenna atlow elevations shrinks the jammer effective radius for GPSreceiver, and the high gain of GIDL antenna in horizontalplane improves the jammer detection capability.

The Stanford experimental GIDL system has only 4 an-tennas, which form 3 independent baselines. Jammer lo-cations would lie on the intersection of three hyperbolasdefined by these three baselines. In this special case it ispossible to derive an analytical solution for the jammer lo-cation. Four antennas would form three baselines, so threerange difference �Ri measurements would provide:

�Ri = R (1�q1� 2li

R(cos � cos �i cos(� � �i) + sin � sin �i) +

l2i

R2

�;

with unknowns: R range to the jammer, � elevation of thejammer, � elevation of the jammer; and system parameters(constants): li Radius-vector of the i-th antenna, � eleva-tion of the i-th antenna, � elevation of the i-th antenna.

We developed an analytical solution to the jammer position(R, �, �) based on the measured signal delays �Ri, andknown system parameters.

2

3

It is possible to estimate the error covariance matrix of pos-sible jammer location. Jammer location ECM is defined byBloc = (HtB�1

�RiH)�1, where B�Ri

is ECM of measure-ments �Ri, and depends on SNR in each antenna.

H =@�Ri

@R; @�; @�

For a constant power jammer, depending on the jammerpower and location, one would have to calculate SNR foreach antenna, then compare it with the detection threshold,and only if it is above such a threshold, use it to estimatemeasurement errors.

In order to estimate ECM of the measurements we firstwould use the link equation to estimate the SNR of the jam-mer as seen by each antenna

q2i =PjGi�

2 i

(2�)2kTe� ijR� Lij2

And then use these estimates to calculate ECM of the mea-surements:

B�Ri=

c2

�fT

3

2�2�f21 + q21 + : : :+ q24q21+ : : :+ q2

4

�

0B@0B@

1

q22

0 0

0 1

q23

0

0 0 1

q24

1CA+

1

q21

0@ 1 1 1

1 1 11 1 1

1A1CA

Now when we know B�Riwe can calculate ECM of the

jammer location estimate: Bloc = (HtB�1�Ri

H)�1, sigmasare on the diagonal.

Knowing ECM of the jammer position estimations we canplot the expected system accuracy, in terms of error el-lipses. Here we present the results of the analysis forthree types of the jammer (relatively strong, weak and veryweak). Results are shown in the Figures 15, 16, 17. It is as-sumed star antenna configuration with baselines of 100m.Jammers of the specific power have been placed in the var-ious locations around the antenna array, and a horizontalprojection of the error ellipses has been plotted. Each lineon the plot is actually an error ellipse, narrow in azimuthaldirection and wide in the range. Thus for the relativelystrong jammer one can see individual ellipses, and the res-olution in range is good. For the weak jammer, we are stillobtaining good azimuthal resolution, but the ellipses elon-gates in range direction (they start to overlap), and for thevery weak jammer we plotted “x” at the range which thespecified power of the jammer is undetectable.

GIDL CALIBRATION USING GPS SIGNALS

Currently the basic GIDL measurement is TDOA of thesignals to the antennas. Jammer direction and location is

45

8000 6000 4000 2000 0 2000 4000 6000 80008000

6000

4000

2000

0

2000

4000

6000

8000

North direction, m

Wes

t dire

ctio

n, m

One Sigma Error Ellipses for Star GIDL Configuration (l=100m) Pj = 5e-09 W/Hz

Figure 15: Error Ellipses for Constant Power Jammer:10mW/20MHz

1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.8 1

x 104

1

0.8

0.6

0.4

0.2

0

0.2

0.4

0.6

0.8

1x 10

4

North direction, m

Wes

t dire

ctio

n, m

One Sigma Error Ellipses for Star GIDL Configuration (l=100m) Pj = 5e-10 W/Hz

Figure 16: Error Ellipses for Constant Power Jammer:1mW/20MHz

2 1.5 1 0.5 0 0.5 1 1.5 2

x 104

2

1.5

1

0.5

0

0.5

1

1.5

2x 10

4

North direction, m

Wes

t dire

ctio

n, m

One Sigma Error Ellipses for Star GIDL Configuration (l=100m) Pj = 5e-11 W/Hz

Figure 17: Error Ellipses for Constant Power Jammer:0.1mW/20MHz

Antenna 2Antenna 1

GPS SVJammer

TCXO GIDL

Cable 1

Cable 2

Figure 18: GIDL Calibration Setup

45

calculated on the basis of these measurements. These mea-surements have to be done relative to the antennas, so un-known delays in the cable and in the system should be cal-ibrated out. Also because a system clock is been used tocalculate propagation delay, it also has to be calibrated.Thus the major system unknowns, which should be cali-brated are: relative delay in the antenna cables; offset inthe master clock (TCXO); variations with temperature andother factors.

The situation exists that the jamming signal is not presenta majority of the time, and GPS signal is coming from theknown sources at known locations — GPS satellites. Thuswe can utilize that signal to calibrate the system clock andsystem delays. Algorithms have been developed to utilizethat signal for the calibration of the GIDL system.

EXPERIMENTS AND RESULTS

Numerous experiments have been performed with theGIDL hardware and software. Most of the effort has beenon developing calibration software and performing calibra-tion of the system. Experimental results of the system de-lay calibration and the clock calibration are shown in theFigure 19. Fifty data sets have been collected a few min-utes apart. Each data set used to calculate system clockerror and system delay. Antenna locations and satellite ge-ometry are known, expected delay and the expected signalfrequency is know. The satellite frequency is measured us-ing our system and signal time difference of arrival calcu-lated. Then the expected results are subtracted from themeasured and the errors are plotted. The first plot showsaverage clock error for each given data run. Averaging isdone on all satellites in view. The next plot shows averagesystem delay and also the average of all satellites in view.The next plot shows number of satellites used, and then de-lay error on the basis of each individual satellite, and clockerror based on each satellite measurement.

Figure 20 shows the running average of the clock estima-tion and running average of the delay estimation. It alsoshows three different ways which we are trying to estimatethe top of the GPS correlation peak: maximum, discrimina-tor function, and median. All three of these measurementsare valid measurements, but we decided to go with the max-imum measurement, which is least susceptible to multipatherrors. Figure 21 shows a running average for the num-ber of runs of the system delay, based on the maximum ofthe correlation peak measurement, and it is converging toapproximately 5.75m. That error was primary defined bythe cables of the different length used in that experimentalsetup.

This results shows that we can calibrate the system clock

4

2.216 2.217 2.218 2.219 2.22 2.221 2.222 2.223 2.224

x 105

−47

−46.5

−46

−45.5

GPS Time, secMea

n cl

ock

erro

r fo

r th

e ru

n, H

z Calibration of GIDL by GPS signals (GIDLD

emo/22082000C

alibr1)

2.216 2.217 2.218 2.219 2.22 2.221 2.222 2.223 2.224

x 105

−5

0

5

10

GPS Time, sec

Mea

n de

lay

for

the

run,

m

2.216 2.217 2.218 2.219 2.22 2.221 2.222 2.223 2.224

x 105

3

3.5

4

GPS Time, sec

Num

ber

of s

atel

lites

use

d

2.216 2.217 2.218 2.219 2.22 2.221 2.222 2.223 2.224

x 105

−20

−10

0

10

20

GPS Time, sec

Del

ay fo

r th

e ea

ch S

V, m

2.216 2.217 2.218 2.219 2.22 2.221 2.222 2.223 2.224

x 105

−47

−46.5

−46

−45.5

−45

GPS Time, secClo

ck e

rror

for

the

each

SV

, Hz

Figure 19: GIDL Calibration: Results of Clock and DelayCalibration

1.0625 1.063 1.0635 1.064 1.0645 1.065 1.0655

x 105

44.04

44.02

44

43.98

43.96

43.94

GPS time, sec

Hz

Running average of DeltaFref

1.0625 1.063 1.0635 1.064 1.0645 1.065 1.0655

x 105

6

5.5

5

4.5

GPS time, sec

m

Running average of Error(Maximum)

1.0625 1.063 1.0635 1.064 1.0645 1.065 1.0655

x 105

6

5.5

5

GPS time, sec

m

Running average of Error(Centroid)

1.0625 1.063 1.0635 1.064 1.0645 1.065 1.0655

x 105

10

9.5

9

8.5

8

7.5

7

GPS time, sec

m

Running average of Error(Disc.Func)

Figure 20: GIDL Calibration: Results for Data Runs 1—15; Number of SVs per Each Run: 5 5 4 4 4 6 6 5 5 4 5 5 44 4

4

1.0625 1.063 1.0635 1.064 1.0645 1.065 1.0655 1.066

x 105

6

5.5

5

GPS time, sec.

Err

or (

Del

ay),

m

Running average of Error (Maximum), over 20 runs

Figure 21: GIDL Calibration: Results for the 20 Data Runs

−3000 −2000 −1000 0 1000 2000

−2000

−1500

−1000

−500

0

500

1000

1500

Nor

th d

irect

ion,

m

East direction, m

Jammer Location and Error Ellipse

Figure 22: Real-Time GIDL Display Software (Simula-tion)

and system delays by the means of the GPS signals.

Another piece of software which has been developed in thelab is GIDL jammer positioning display software. It dis-plays jammer position in the real time, and it has an optionof tracking the location of the moving jammer. There isalso an option to run it from simulated data or from realdata. In Figure 22 the display shows locations of the jam-mer as they are appear in time, so it is easy to see thatjammer performs spiral motion (data has been supplied bysimulation code for this demonstration). In addition to jam-mer locations this display also indicates calculated error el-lipses (which looks like a little strikes), which would assistin searching for the jammer.

Figure 23 shows results of one of the real experiments. An-tennas have been set up at the bed of the dry lake Lagunita

55

5

−100 −50 0 50 100−100

−80

−60

−40

−20

0

20

40

60

80

100

Nor

th d

irect

ion,

m

East direction, m

Jammer Location and Error Ellipse

Figure 23: Real-Time GIDL Display Software (Experi-mental Results)

on the Stanford Campus, and their locations have been sur-veyed. They are marked by circles on the display. Alsothe locations for the jammer have been surveyed, they aremarked by the “x”. The system has been set up, and cali-brated. Then jammers are placed in each survey location,and display shows estimated jammer location and errors.Errors and jammer location are shown in this figure. Theyare close to the theoretically predicted results. For the com-plete and certain results, as well as meaningful statistics,additional data collections are needed.

FUTURE DEVELOPMENT: CONCEPT OF DIS-TRIBUTED GIDL

In this paper we also would like to introduce concept ofthe distributed GIDL system, shown in Figure 24. This ispossible extension of GIDL. In the current setup all hard-ware and software processing in the GIDL is colocated. Soantenna configuration and coverage pattern are limited bythe length of RF cables, which cannot exceed few hundredmeters.

In the proposed concept of the Distributed GIDL systemwe would put data transmission and time synchronizationon a network, which would have number of nodes — Sin-gle channel GIDL receivers, or Bit grabbers, which wouldcollect data simultaneously, and then send it to the mainprocessing station, or even would do preliminary process-ing on each node.

This system has number of the advantages compare to theoriginal multichannel GIDL system: Each receiver is inex-pensive; covers large area; can cover specific areas; easyto extend; improved accuracy due to the better geome-try. But there is a major problem: how to synchronize all

4

Jammers

CentralProcessing

Station

Number of Jammers

Jammer Locations

Receiver/Bit Grabber

Receiver/Bit Grabber

Receiver/Bit Grabber

Receiver/Bit Grabber

Data transfer/Time synchronization network

Figure 24: Distributed GIDL System

these receivers, and how to transmit large amounts of datawhich they are generate. Another issue is generating com-putationally efficient algorithms to process all this wasteamount of data.

CONCLUSIONS

We have developed a 4-channel, common-clock digitalreceiver which operates in the L1 GPS band. The pri-mary intended use for this receiver is Generalized Interfer-ence Detection and Localization System and developmentof new localization algorithms. Also jammer localizationand GIDL interface display software has been developed.Experimental results correspond with predicted theoreticalperformance. GIDL can improve overall LAAS availabilityby finding direction or location of any interference sources.

FUTURE PLANS

There are a number of aspects which are planned to do withdeveloped hardware and to improve system performance.These include: Develop short baseline direction finder andantenna array; Develop and test distributed GIDL system;Verify system performance on different types of interfer-ence sources, and different configurations of antenna ar-rays; Develop detection algorithms to estimate number ofinterferers; Develop operational procedures for GIDL oper-ation; Integrate GIDL subsystem with Integrity MonitoringTestbed and evaluate overall system performance.

6

ACKNOWLEDGMENTS

We would like to thank our colleagues for the help theyprovide to us while we are working on the project. We alsowould like to thank Stanford University and academic com-munity for the nice environment to work and communicatewith people. And our special thanks to the Federal AviationAdministration for the support and funding of this project.

REFERENCES

1. Swider, R., Recommended LAAS Architecture, Pre-sented to RTCA SC-159, WG-4, Anaheim, CA, November12, 1996.

2. Minimum Aviation System Performance Standards forthe Local Area Augmentation System. Washington, D.C.:RTCA DO-245, September 28, 1998.

3. Gromov K., et al. “Interference Direction Finding forAviation Applications of GPS,” Proceedings of ION GPS-99, Nashville, TN, Sept. 14-17, 1999.

4. “Coherence and Time Delay Estimation”, edited by C.Carter, IEEE Press, 1993

457