ANTIJAMMING TECHNOLOGYINNOVATION

38 GPS World FEBRUARY 2004 www.gpsworld.com

amming suppression techniquescan be separated into enhance-ments to the receiver itself, and

enhancements to the antenna providing in-puts to the receiver. Part I of this article dis-cussed receiver enhancements. We now turnto antenna enhancements.

Antenna enhancements are attractive inthat they can often be applied to an exist-ing GPS installation by upgrading only theantenna. By serving as an appliqué or sim-ple “add on,” users are saved the cost of in-tegrating a different receiver into their sys-tem.

Antenna enhancements can be furtherbroken down into analog and digital im-plementations. The analog implementa-tions involve mixing the radio frequency(RF) signals from multiple antennas into asingle weighted signal, which is then digi-tized and used by the receiver digital sec-tion. As the term implies, digital imple-mentations first digitize the RF signals usingan analog-to-digital (A/D) converter, thenperform the signal processing in the digitaldomain.

Basic Antenna EnhancementsFigure 1 shows block diagrams of severalpopular techniques for antenna enhance-ments, presented in such a way to highlightthe differences as the complexity increases.

Cancellers. One of the simplest and eas-iest to understand antenna enhancementsis the canceller. As shown in Figure 1A, thecanceller uses two antenna inputs. Thepremise of the canceller is that one antennahas visibility to the jammer, but poor (or no)visibility to the composite GPS signal; thesecond antenna has visibility to both.

An example of such an installation couldbe an aircraft with an antenna on the topsurface of the fuselage with its boresightnominally pointing in the zenith directionand another on the lower surface, point-ing to the ground. The jammer has suffi-cient power to reach both antennas, but theGPS signal is only visible to the upper an-tenna.

For this system, if the gain (amplitude)and phase of the output of the jammer-onlyantenna is adjusted with a complex (gain andphase) weight, W, to be equal and oppositeto the output of the jammer-plus-signal an-tenna, when added together, the jammerwill cancel out. This is possible only becausethe GPS signal is well below the noise floorof the antenna, and would not affect the de-termination of W. The calculation of thevalue of W can be performed in analog cir-cuitry that varies W to minimize the powerat the summer output.

Cancellers also can prove very effectivewhen the jamming comes from a local trans-mitter. Consider an application where a sig-nal (e.g., a communications signal) is beingtransmitted from the same location as theGPS receiver, such as an integratedGPS/communications link. In this case,called co-site cancellation, a second antennais not required since an accurate sample ofthe jammer-only signal is available from thetransmitter itself. Co-site cancellers alsocan eliminate the effects of spurious signalsgenerated within the GPS receiver RF sec-tion itself.

Cancellers are most effective against asingle jammer source (although they alsohave proven effective in providing bands ofprotection such as against an array of

ground-based jammers viewed from the air).Multiple jammers would provide differentsignatures to the two antennas, measuredas gain variations and time-of-arrival vari-ations. However, the complex weights ap-propriate for one jammer are not generallyappropriate for the second jammer, sincethe gain and time-of-arrival variations aregenerally different when the jamming sig-nals emanate from two different locations.To circumvent this limitation, some ap-proaches utilize multiple channel cancellersthat are able to mitigate the effects of morethan one jammer. However, the complex-ity of these designs increases rapidly withthe number of channels.

Polarimeters. Another two-channel im-plementation of antenna enhancement isthe polarimeter (see Figure 1B). Here, asingle antenna output is fed through twooutput feed elements to generate two out-put channels with different polarizations,typically right-hand circularly polarized andleft-hand circularly polarized. These twoelements will observe the jammer differ-ently, providing a signal difference that canbe exploited to eliminate the jammer effects.As with the canceller, the trick is to calcu-late the two complex weights, W1 and W2,which minimize the jammer power. Again,it is convenient to define W1=1 as was doneimplicitly for the canceller, simplifying theproblem to the determination of a singleweight.

As with the canceller, polarimeters aremost effective against a single jammer, al-though for specific geometries they can beeffective against multiple jammers.

Spectral Filters. Figure 1C shows ablock diagram of a spectral filter, with mspectral taps. Such a device is also called atemporal filter, a notch filter, or in some im-plementations an adaptive transversal filter.Although shown in the time domain forcomparison with subsequent implementa-tions, this filter is often most easily visual-ized in the frequency domain.

Part II: Antenna EnhancementsJamming Protection of GPS ReceiversSteve Rounds, Interstate Electronics Corporation

THIS MONTH’S COLUMN features the second installment of a two-part article on jamming protection for GPS receivers. In Part I, we looked at enhancements to receiver design – primarily by integratingthe GPS sensor with an inertial measuring unit. In this month’s column, we examine a variety of antennaenhancements. — R.B.L.

J

We consider a CW jammer (see sidebaron jammer types). If the antenna output isconverted to the frequency domain by pass-ing it through a fast Fourier transform(FFT), the output will be a white-noise-likefloor with a single large spike at the jam-mer’s frequency. This is the common pic-ture that would be seen if looking at thejammer signal on a spectrum analyzer. Ina digital implementation, it is trivial to re-duce the gain on the frequency bin (or bins)that correspond to this spike in order to“whiten” the output in the frequency do-main. After notching out this spike, an in-verse transform is applied to return the out-put signal to the time domain.

Unlike previous implementations withmultiple taps, the spectral filter can removemultiple jammers. In the frequency domain,this corresponds to reducing the gain onseveral frequency bins. However, each timethe gain on a frequency bin is reduced, thegain on the GPS signal content in that binis correspondingly reduced. Therefore,while the spectral filter can mitigate the ef-fects of more than one jammer, too manyjammers can overwhelm this filter. Thenumber of jammers that can be effectivelyeliminated is determined by the number andspacing of the individual temporal taps/fre-quency bins.

Whereas the spectral filter can reduce theeffects of multiple jammers, it is completelyineffective against broadband jammers.Thinking again in the frequency domain, abroadband jammer does not appear as a sin-gle spike, but is spread over the entire GPSspectrum. With no spikes to notch out, thespectral filter is of no use.

Spatial Nulling. Spatial nulling is oneof the most effective forms to mitigate theeffects of GPS jamming, but it comes withthe attendant cost and complexity. Figure1D shows an implementation of a spatialnuller using a multi-element antenna.

To more fully understand this importanttechnique, consider Figure 2 which showsa diagram of a simple 2-channel spatialnuller with a single GPS signal and a singlejammer. The outputs of antenna elements1 and 2, A1 and A2, can be written as inEquation 1, where G1 and G2 represent thedifferent gains of the two channels, �1 and�2 are the independent noise values on those

same two channels, and the effects of an-tenna gain variations have not been con-sidered. S(t) and J(t) represent the GPS andjammer signals, D is the distance betweenthe antennas, and �1 and �2 are the eleva-tion angles of the GPS satellite and jammer.Parameters c and � are the speed of light

and the GPS signal angular frequency. If wechoose W1 =1 and optimize the value of W2,the spatial nuller output, W1 A1 � W2 A2can be written as Equation 2.

Examining this equation, we see that thechoice of W2 has caused the jammer to van-ish, while both the signal and noise have

W

J+S

J

Σ

� FIGURE 1 Various antenna enhancementmechanizations: (A) canceller, (B) polarimeter,(C) time domain and frequency domain spec-tral filters, and (D) spatial nuller. Each mecha-nization uses complex signal weights, W, orfrequency-bin weights, K.

W1

Σ

W2

RHCP

LHCP

Z-1

W1

Σ

Z-1

W2

Σ

W3

Z-1

Σ

Wm

ΣFFT

ω1

ω2

ωm

K11

K12

K1m

IFFT

Time Domain Implementation

Frequency Domain Implementation

www.gpsworld.com FEBRUARY 2004 GPS World 39

(A)(B)

(C)

W1

Σ

Wn

(D)

40 GPS World FEBRUARY 2004 www.gpsworld.com

been modified — possibly increased, possi-bly decreased. An important observation isthat if �2 � �1, the GPS signal also vanishes.In physical terms, this means that if thejamming signal comes from the same di-rection as the satellite, the spatial nuller willinadvertently but unavoidably null the satel-lite as well. As a corollary, the same weightsare equally effective (in an average sense)for all satellites.

If there were a second jammer from a dif-ferent direction, no value of W1 and W2would null both jammers (except the trivialcase of W1 � W2 � 0, which also nulls allsatellites). However, were a third antennaelement added, a solution could be found.This points out a fundamental property ofspatial nullers: an n-channel spatial nuller

can null n�1 jammers.With the exception of the

very accurate approximationof the first equation (that is,the phase delay of the jam-mer signals across the an-tenna elements is frequency-independent), spatial nullers

are equallyadept atn u l l i n gbroadbandand CWjammers.

In theabove two-element ex-ample, it

was easy to identify the value of W2 to nullthe jammer. However, in practice the coef-ficients of the individual channel equationsare not known, and the complexity increasesfor increasing numbers of elements. Thereare multiple approaches for solving theseequations, but with the increasing capabil-ity of modern processors, current applica-tions trend from analog solutions towarddigital solutions. Even here, multiple ap-proaches can solve the more general equa-tion set, including least mean square, re-cursive least squares, and sample matrixinversion techniques. Each of these com-putational algorithms has its advantages anddisadvantages, particularly with regard tothroughput requirements and the resultantlatency of the weights, and scalability to

larger and larger antenna arrays.The equations above imply that the jam-

mer can be completely eliminated, but ofcourse there are limitations in the real-worldimplementation. These limitations includethe antenna size effects (related to the ap-proximation in the first equation), loss ofaccuracy in the A/D conversion, and latencyeffects. This last category arises from typi-cal implementations that sample the jam-ming environment in one cycle, calculatethe weights, and apply the weights in thesubsequent cycle. Changes in the jammingenvironment, such as caused by pulsed jam-mers, can reduce the algorithm effective-ness. Nonetheless, spatial nulling typicallyprovides 25–40 dB of jamming suppressionagainst all jammer types.

The effectiveness of the spatial nuller isoften presented in an antenna gain chart,such as shown in Figure 3. This standardazimuth/elevation-angle chart, with the an-tenna zenith at the center of the circle andthe horizon around the perimeter, showsresults for a seven-element antenna systemwith five broadband jammers. The locationsof the jammers and satellites are identified,and the colors represent the effect of theidealized weights in nulling the jamming foreach direction in space. Notice the deep andnarrow nulls that surround the jammers(represented in brown), with very little lossin effective antenna gain in the direction ofthe satellites.

STAPWhat if you could combine the power ofthe spatial nuller against all jammer typeswith the relatively inexpensive ability of thetemporal filters to eliminate large numbersof CW jammers? The resulting technique,called spatial temporal adaptive processing(STAP), is generally recognized as the mostpowerful technique for jamming suppres-sion. Figure 4 shows a block diagram of sucha technique. By comparing this figure withthe Figures 1C and 1D, we see that STAPcan be thought of conceptually as the op-timal combination of spatial and temporalprocessing with n antennas and m spectraltaps. Calculating all n � m weights in par-allel is critical in the implementation. Thisallows use of the temporal taps to attack theCW jammers, reserving the spatial capa-

INNOVATION ANTIJAMMING TECHNOLOGY

� FIGURE 2 Simplified geometry for 2-element antenna used for spatially nulling a jammer

� EQUATION 1

A t G S t J t

A t G S tD

cJ t

D

c

G S t e J t ei D

c

i D

c

1 1 1

2 21 2

2

2 2

1 2

( ) ( ) ( )

( ) (cos( )

) (cos( )

)

( ) ( )cos( ) cos( )

= + +[ ]= − + − +

≈ + +

− −

ν

θ θν

νω θ ω θ

� EQUATION 2

W

W W

2

1 2

= −

+ = −

+ −

− −( )

G

Ge

A A G S t e e

i D

c

i D

c

i D

c

1

2

1 2 1 1 2

2

1 2 2

1

ω θ

ω θ θ ω θ

ν ν

cos( )

cos( ) cos( ) cos( )

( )

D

GPS signal, S

Jammer signal, J

Antennaelement 2

Antennaelement 1

θ1

θ2

INNOVATION ANTIJAMMING TECHNOLOGY

bilities to attack the broadband (and other)jammers.

Of course, if all n � m weights must becalculated in parallel, this implies matricesof the order n � m by n � m, and these ma-trices must be multiplied and inverted. Forn � 7 and m � 15, for example, this wouldimply inverting a 105 � 105 matrix. As withthe deep integration processing discussedin last month’s installment, the trick in im-plementing STAP is identifying simplifica-tions in the algorithms that bring the pro-cessing requirements within reason, whileretaining performance as close to optimalas possible.

Figure 5 provides an example of the ad-vantages of STAP processing compared tospatial nulling alone. Both systems use a7-element antenna and the STAP systemalso includes 7 temporal taps. The jammingscenarios for each case are identical — thesame 5 broadband jammers used in Figure3, with 5 additional CW jammers. As canbe seen from the figure, the spatial nuller isoverloaded — it has more jammers than itsn�1 degrees of freedom. This is evidentfrom the larger areas of deep nulls. TheSTAP system, however, is not overloaded,as the 5 CW jammers are handled by thetemporal taps, leaving 7 spatial elements toattack the 5 broadband jammers. The STAPsystem shows sharp nulls around the jam-mers, while the nulls from the spatial nullerare broad and not as deep.

SFAPJust as the temporal nuller could be visu-alized and implemented in either the timeor the frequency domain, so theSTAP process can be alternatively im-plemented in the frequency domain– a technique known as space fre-quency adaptive processing (SFAP),shown in Figure 6.

Whereas frequency-domain im-plementations are often more effi-cient than the corresponding time-domain implementations, even thissimplification often does not reducethe processing requirements suffi-ciently. However, the SFAP imple-mentation does offer an approxima-tion that can substantially reducethese requirements. Considering a

single frequency bin in Figure 6,the weights can be calculated forthe n different antenna elementswithout considering the otherj�1 frequency bins, and theprocess repeated for each bin.Since the matrix operations tendto go as the cube of the matrixdimensions, this reduces theorder of the computations from(j�n)3 to j�n3. This approxima-tion does come at the expense ofsome performance degradation,though often of small magni-tude. This can be offset by in-creasing the number of tempo-ral taps (j) for SFAP versus thenumber (m) for STAP. With thisadjustment, SFAP canbe more effectiveagainst large numbersof CW jammers, whileminimizing the perfor-mance degradationselsewhere.

Figure 7 continuesthe comparison of theSTAP system (n � 7, m= 7) of Figure 5, thistime with SFAP (n � 7,j � 128) against thesame 5 broadband and5 CW jammers. Asseen from this figure,the STAP and SFAPperformances are verysimilar.

� FIGURE 4 Spatial temporal adaptive processing. This technique is acombination of spatial nulling with the time domain implementation ofspectral filtering.

Z-1

W11

Σ

Z-1

W12

Σ

W13

Z-1

Σ

W1m

Z-1

Wn1

Σ

Z-1

Wn2

Σ

Wn3

Z-1

Σ

Wnm

Σ

� FIGURE 5 Antenna gain charts comparing (A) spatial nulling with (B) spatial temporal adaptive processing

� FIGURE 3 Antenna gain chart example of spatial nulling(7-element antenna with 5 broadband jammers)

(A) (B)

42 GPS World FEBRUARY 2004 www.gpsworld.com

STAP, SFAP Added BenefitsWhile STAP and SFAP are generally rec-ognized to be the most powerful methodsto minimize the effects of jammers, thesetechniques have additional benefits.

Beamforming. Beamforming can bethought of as the inverse of spatial nulling:instead of minimizing the gain in the di-rection of the jammer, beamformers maxi-mize the gain in the direction of a satellite.This technique is most effective when theantenna array is not fully loaded, that is, thenumber of jammers is less than the maxi-mum of n�1.

Unlike nulling, however, beamformingcannot observe the satellite signal and adap-tively maximize the antenna gain in that di-rection since the satellite signal is below the

noise floor and not ob-servable. (A small num-ber of implementationscalculate the weightsafter correlation withthe satellite code, andtherefore the satellitesignal has been restoredto be above the noisefloor. As with mosttradeoffs, this abilitycomes at some expense,in this case substantiallyincreased hardware re-quirements.) There-fore, the beamformingalgorithms must beprovided with theknown azimuth and el-

evation angle of the satellite, trans-formed to the antenna coordinate sys-tem. When examined through thedetailed mathematics of calculatingthe complex weights, beamformingrepresents a straightforward addi-tional operation of multiplying by asteering vector to the satellite.

Since the direction to each satel-lite is different, the process must berepeated for each satellite. This meansthat the weights applied to the an-tenna outputs, Wij in Figure 4, are dif-ferent for each satellite. Correspond-ingly, there are multiple outputs ofthe beamforming/STAP system, eachoptimized for a distinct satellite. If a

receiver has only four input channels, abeamforming/STAP system might developthree outputs optimized for three distinctsatellites, with the fourth channel optimizedto maximize the overhead gain.

In typical applications, beamforming canincrease the strength of the received satel-lite signal by about 3–6 dB. Note also thatsince this technique makes use of the spa-tial properties of the STAP and SFAPprocess, it is equally effective with a spatialnuller.

Self-Equalization. One of the real-world issues that limit the effectiveness ofspatial nulling is the need to match the RFchannels for each antenna element. Gainmismatches, phase delay mismatches, andbandwidth mismatches all limit the ultimate

system performance. One of the strengthsof STAP and SFAP is that the temporal tapscan substantially reduce the sensitivity tothese errors. While it is difficult to quantifythe sensitivity of the various algorithms tothese effects, degradations in threshold per-formance of 7–15 dB are not uncommonfor poorly matched spatial nulling systems,while these same effects are greatly reducedfor STAP and SFAP systems.

Spatial nullers circumvent this limitationby utilizing high precision RF components,or by performing calibration and compen-sation of the RF characteristics. While thishas proven feasible, utilizing the temporaltaps of the STAP/SFAP algorithms to min-imize the effects provides a much more cost-effective solution.

Other Benefits. Just as multipath cancause problems with precision tracking ofthe satellite signals, multipath can cause asingle-jammer signal to arrive from multi-ple directions with different delays. In a spa-tial nuller, this appears to the nulling algo-rithms as two distinct jammers, andtherefore consumes two of the “n�1” de-grees of freedom from the antenna array.For a STAP/SFAP implementation, becausethe temporal weights are optimized in com-bination with the spatial weights, the mul-tipath jammer signal can be recognized as atime delay of the direct jammer signal.

Another benefit of STAP/SFAP and spa-tial nulling, when implemented in a digitaldesign, is the ability to locate the jammingsources. As shown in Figure 2, the weightsare a function of the angle of arrival of thejammer. Once the weights have been cal-culated, it is possible to invert the equationsto determine the angle of arrival of the jam-mers that are being nulled. The derived an-gles represent the azimuth and elevationangle of the jammer in antenna coordinates,which typically must then be rotated toEarth-fixed coordinates to identify the jam-mer location. One such implementation isthe Multiple Signal Classification (MUSIC)algorithm, developed at Stanford Univer-sity. The effectiveness of algorithms such asthese depends strongly on the antennageometry.

Receiver Impacts Antenna enhancements, even when imple-

� FIGURE 7 Antenna gain chart illustrating spatialfrequency adaptive processing (compare with Figure 5)

� FIGURE 6 Spatial frequency adaptive processing. This technique is acombination of spatial nulling with the frequency domain implementationof spectral filtering.

Σ

FFT

ω1

ω2

ωj

K11

K12

K1j

FFT

ω1

ω2

ωj

Kn1

Kn2

Knj

Σ

Σ

IFFT

www.gpsworld.com FEBRUARY 2004 GPS World 43

44 GPS World FEBRUARY 2004 www.gpsworld.com

mented as an appliqué, are not completelytransparent to the receiver. Two importantissues are phase perturbations and RF dy-namic range.

Phase Perturbations. Most of thetechniques discussed include multiplyingthe output of the antenna elements by com-plex weights in order to null the jammer(s).Furthermore, these weights change con-tinuously as the antenna rotates and the jam-mer waveform is otherwise modified. Sincethe weights are complex, this implies thattime-varying phase changes will be appliedto the signal provided to the GPS receiverfor tracking.

Inside the GPS receiver, the Costas looptracks the phase of the GPS signal, usingthe measurements for high accuracy posi-tioning (in real-time kinematic applications)or for delta pseudorange observations to im-prove velocity performance. Phase modifi-cations introduced by the nulling algorithmswill be erroneously interpreted as phasechanges of the GPS signal. Some imple-mentations go so far as to avoid nulling until

the jammer level reaches a threshold, andthen discontinuing phase tracking for higherjamming levels.

In a digital implementation of the nulling,the weights and therefore the applied phaseperturbations are known. In theory, thesecan be compensated, although issues of la-tency become important. Additionally, thephase compensations are unique for eachsatellite being tracked.

RF Dynamic Range. Because of thevery high performance capabilities of someantenna enhancement techniques, often theRF front end ultimately determines the limitof system performance. For the nulling al-gorithms to work, the RF channels must re-main linear at the peak jamming levels. Infact, for the receiver to continue to work ef-fectively against most waveforms, the RFlinearity must extend to 12 dB above theaverage jamming levels. With the capabili-ties afforded by deep integration andSTAP/SFAP, this implies linearity of the RFfront end to exceedingly high input powerlevels. This in turn becomes a major driver

in system power, size, and cost.

ConclusionsJamming of GPS receivers is real. The sus-ceptibility to GPS jammers is very high, dueboth to the very weak GPS signal and thesensitivity of the short C/A-code to CW in-terference. Unintentional jamming has beenobserved in civilian applications. Militaryforces that depend on GPS are rapidly try-ing to develop protection for their receiversfrom known, intentional jamming threats.Industry is responding with a wide varietyof schemes for jamming protection, as out-lined in this article, with each technique hav-ing its particular strengths and weaknesses.

AcknowledgmentsThe author would like to thank his manycolleagues in government and industry fortheir contributions to the development andunderstanding of the many topics in this article, particularly his current colleagues atIEC. �

INNOVATION ANTIJAMMING TECHNOLOGY

JAMMERSAny transmitter whose fundamental fre-quency or radiated harmonics are withinthe L1 or L2 frequency bands has thepotential to interfere with or jam the GPSsignal. Military forces use purpose-builtjamming transmitters, or jammers forshort, as weapons in electronic warfare todeny an adversary the use of the radiospectrum whether it be used for commu-nications, radar, or navigation.

A jammer is typically characterized byits signal spectrum – its carrier and howit is modulated. The basic jammer typesare CW, white-noise, pulsed, and FM.

CW. A continuous wave (CW) jammertransmits a continuous unmodulated car-rier. If the CW jamming signal has thesame frequency as the “target” signal butis not in phase with it, the jamming isnon-coherent. If the jamming signal is inphase with the target, the jamming iscoherent. It is also possible to sweep theCW jamming signal over a range of fre-quencies centered on the target signal’sfrequency (SCW). For example, every mil-

lisecond the jammer frequency could beswept linearly from 50 kHz below the tar-get frequency to 50 kHz above it and backagain.

White-noise. For jamming purposes,the phase of a carrier can be randomlyvaried to create a white-noise jammer.(White noise has a Gaussian voltage dis-tribution with a zero mean and a uniformphase distribution between 0 and 2π andessentially contains the contributions of awide band of frequencies akin to howwhite light is composed of a wide band ofcolors.) The rapidity with which the phasechanges are made will determine thebandwidth of the jammer. If the band-width were 20 MHz, for example, the jam-ming signal would essentially overlay thewhole GPS signal. This type of jammer isalso known as a wideband or broadbandjammer. In a variation on the white-noisejammer, the carrier is modulated by apseudorandom noise code. The resultingwideband signal has a structure similar tothat of a GPS signal and, in fact, the codesequence and chipping rate can be

selected to match one of the GPS signals.Such a jammer can be quite effective.

Pulsed. Interrupting a continuouswave by periodically keying the transmit-ter on and off produces a sequence ofpulses. The pulsing might be regular witha duty cycle of say 50 percent (as generat-ed by a square-wave sequence) or ran-dom with varying pulse width. Radars arepulsed transmitters with very short regu-lar duty cycles. A pulsed signal is a basicform of amplitude modulation (AM).

FM. In a frequency-modulated (FM)signal, the instantaneous carrier frequen-cy changes with the frequency and ampli-tude of a modulating waveform. Thetransmitted bandwidth is governed by theratio between the frequency deviation(how much the instantaneous frequencydeparts from the carrier frequency) andthe modulating frequency. An FM jam-ming signal might use a pure sinusoidalmodulating waveform with a frequency of1 kHz and with a frequency deviation of±50 kHz. – R.B.L.

www.gpsworld.com FEBRUARY 2004 GPS World 45

Steve Rounds is a senior director in theAdvanced Technology Department at InterstateElectronics Corporation (IEC), a division of L-3Communications, in Anaheim, California. In addition tohis support of antijamming development, he is respon-sible for the identification and integration of new tech-nologies at IEC. He holds a B.S. degree in physics fromStevens Institute of Technology in Hoboken, New Jersey,and an M.S. degree from Yale University in New Haven,Connecticut.

Further Reading For general introductions to GPS and jamming as well as references for receiver antijammingenhancements, seeFurther Reading in Part I of thisarticle,“Jamming Protection of GPSReceivers – Part I: Receiver Enhance-ments” by S. Rounds in GPS World, Vol.15, No. 1, January 2004, pp. 54-59.

For an introduction to antennas and how they work, see“A Primer on GPS Antennas” by R.B.Langley in GPS World, Vol. 9, No. 7, July1998, pp. 50-54.

For further information on adaptive antenna arrays, see Adaptive Antennas: Concepts andPerformance by R.T. Compton, Jr., published by Prentice Hall, EnglewoodCliffs, New Jersey, 1988.

Introduction to Adaptive Arrays by R.A.Monzingo and T.W. Miller, published byJohn Wiley & Sons, New York, NewYork,1980.

“Adaptive Multisensor Detection andEstimation” by A. Steinhardt in AdaptiveRadar Detection and Estimation, S.S.Haykin and A. Steinhardt, eds., publishedby John Wiley & Sons, New York, NewYork, 1992, pp. 91-160.

For details of the MUSIC algorithm, see“Multiple Emitter Location and SignalParameter Estimation” by R.O. Schmidt in IEEE Transactions on Antennas andPropagation,Vol. 34, No. 3, 1986, pp. 276-280.

“Innovation”is a regular columnfeaturing discussions aboutrecent advances in GPS technolo-gy and its applications as well asthe fundamentals of GPS position-ing.The column is coordinated byRICHARD LANGLEY of the

Department of Geodesy and Geomatics Engineering at theUniversity of New Brunswick, who appreciates receivingyour comments and topic suggestions.To contact him,see the “Columnists”section on page 2 of this issue.

CIRCLE 17 CIRCLE 18