J.E. Evans, J. Capon, andD.A. Shnidman
Multipath Modeling for Simulatingthe Performance of the Microwave
Landing System
The Microwave Landing System (MLS) will be deployed throughout the world in the1990s to provide precision gUidance to aircraft for approach and landing at airports.At Lincoln Laboratory, we have developed a computer-based simulation that modelsthe performance ofMLS and takes into account the rnultipath effects of buildings, thesun-ounding ten-ain, and other aircraft in the vicinity. The simulation has provideduseful information about the effects of multipath on MLS performance.
The Microwave Landing System (MLS) is anew precision approach and landing navigationsystem that aircraft will soon be using at majorand small airports. MLS is a navigation systemthat operates conventionally: an aircraft determines iJs location by an on-board analysis of
~-I
measurements that the vehicle makes ofsignals-'
that are emitted from the ground. In particular,the aircraft calculates its elevation, azimuth,and distance with respect to separate elevation,azimuth, and distance-measuring-equipment(DME) [1] transmitters on the ground. MLS thusprovides navigation information that pilots canuse to land their aircraft, even in adverseweather conditions.
Endorsed by the International Civil AviationOrganization (ICAO), MLS has the following improvements over the current Instrument Landing System (ILS).
Precisionguidance and range information overa wide geographic area With such information,pilots can land their aircraft under instrumentflight rules (IFR), evenin adverseweatherconditions. Also, air traffic controllers can instructpilots to use curved and segmented approachpatterns in order to increase runway efficienciesand minimize noise levels around airports.
Electronic guidance using scanning microwave beams. Scanning microwave beams aremuchless susceptible to reflections from irregular terrain than are ILS beams, which require asmooth ground surface. Hence, for small airports and heliports situated in hilly regions,
The Lincoln Laboratory Journal, Vo[ume 2, Number 3 (1989)
MLS offers affordable approach and landingnavigation.
In adverse weather conditions, MLS is oftenthe only source of accurate navigational information that is available to pilots who are aboutto land their aircraft. Thus it is imperative thatMLS provide virtually error-free performance.
The basic accuracy of the system has beenvalidated by many tests. In addition, real-timemonitoring of performance by receivers in therunway area has further verified MLS's accuracy. Thus, because other system errors havebeen dealt with successfully, multipath is themajor potential causeofunacceptable angle andrange errors. The major multipath sources oferrors for MLS are signal reflection and diffraction. Figure 1 illustrates these two types ofscattering phenomena. Note that specular reflections can occur off the terrain, physicalstructures suchas buildings. and other aircraft.Shadowing, which causes beam diffraction, canresult from runway humps. other aircraft, andbuildings.
The Lincoln Laboratory program that ana1yzed the effects of multipath on MLS performance commenced with the Laboratory's participation in a NATO study [2] that looked at candidate MLS concepts for military use. During thestudy, itbecame clear thatmultipathwould playan important role in distingUishing betweenvarious candidates and in the use ofthe selectedsystem. The study also revealed that asubstantive test program was needed to achieve a real-
459
Evans et aI. - Multipath Modelingjor Simulating the Performanceojthe Microwave Landing System
Shadowing byOverflying
Aircraft 7
Reflections fromBuildings and
Parked Aircraft
ElevationTransmitter
Fig. 1-MLS multipath phenomena.
istic multipath-performance evaluation. Subsequently, the Federal Aviation Administration(FAA) commissioned Lincoln Laboratory to develop a multipath model for comparing variousMLS techniques that were under study by theFAA and/or proposed to the ICAO. The model,which was used extensively as an assessmenttool in ICAO's evaluation [3, 4], is now beingused to perform a variety of analyses for supporting MLS deployment. In particular, the following issues are being investigated:(1) Where should MLS be sited?(2) How should taxiing aircraft and otherve
hicles in the vicinity of MLS be constrained?
(3) Which ofthe various types ofMLS groundequipment should be used at a givensite?
(4) What impact would proposed airportchanges (e.g., the addition of a building)have on MLSdata quality?
Note that to investigate such issues by directmeasurement at major airports is logistically
460
very difficult and more costly than by computerbased simulation.
This article describes the methodology usedto develop a simulation that models both MLSperformance as well as the various multipathphenomena that affect MLS performance. Theexperimental validation of the simulation isdiscussed as well as the simulation's application for investigating locations where wide-bodyaircraft near the runway may cause unacceptable levels of reflection multipath.
Simulations for specific situations have fourprincipal elements:(1) an airport model that contains the loca
tions and characteristics ofreflecting andshadowing obstacles, terrain features,and the MLS antenna locations;
(2) a flight-profile model that describes theroutes flown by aircraft;
(3) a multipath model that takes into account the various reflection and diffraction paths and determines the radio signals that are received by the receiver for
The Lincoln Laboratory Journal. Volume 2. Number 3 (I 989)
Evans et aI. - Multipath ModelingJor Simulating the PerformanceoJthe Microwave Landing System
each evaluation point along a flight path;and
(4) a system model that determines themultipath system error for the specifiedMLS ground equipment and receiverprocessing algorithms used.
This article will focus on elements (3) and (4).The first two elements will be discussed briefly inthe context ofspecific issues that concern multipath sources and receiver modeling.
The development of our simulation wasgUided by a need to address the error sourcesrelevant to MLS. (This approach contrasts withgeneric simulations that can be used to simulatethe performance of all surveillance and navigation systems.) Therefore, this article is organizedin the following way. The section "MLS Features"describes the key attributes of the Time Reference Scanning Beam (TRSB) system, which theICAO adopted as its standard MLS. The multipath-mitigation features ofTRSB are discussed.Next, the section "Multipath Model Features"presents the key model features and gives examples of the experimental validation of thesystem for the various multipath sources. Thesection "MLS Model Features" then discussesthe validation ofthe receiver portion ofthe TRSBsystem model. We validated the model by comparing its output with the measured antennapatterns and the results ofbench tests on actualreceivers. The section "Simulation Applications"gives an example ofthe application of the simulation in addressing a multipath issue ofcurrentconcern. Lastly, the section "Conclusions"summarizes our results.
For two reasons we go into greater detail inthis paper than is customary in a review article.First, multipath is a problem for a number ofFAA surveillance and navigation systems thatoperate in the microwave bands. Our multipathmodel and the modeling insights gained in ourresearch are applicable to a number of thesesystems. Second, the military has been increasingly interested in bistatic surveillance systems,in which the transmitter and receiver are locatedseparately. MLS multipath effects can be regarded as a special case of bistatic scattering.Hence the model described in this article mightbe a useful starting point for a systems analysis
The Lincoln Laboratory Journal. Volume 2. Number 3 (1989)
of bistatic scattering.We should note that in our research it was
important to minimize the simulation computation time to a level that was practical andfeasible in the context of the computers thatwere available in the mid-1970s. This requirement led us to adopt a ray-theory model forhandling all of the reflection and diffractionphenomena of concern.
MLS Features
To provide the framework for a later discussion of multipath modeling, this section describes MLS and explains characteristics of thesystem's multipath-related errors.
The typical MLS region of coverage is a distance defined up to a range of 20 mni by anazimuth sector of±40° around the runway centerline and an elevation sector of +10 to +200
•
Outside these sectors, separate transmitterslocated to the side and back ofthe MLS transmitters warn pilots that they are flying to the left of,to the right of, or in back of the region ofcoverage.
The DME currently in use is a high-precisionversion of the conventional L-band DME, whichwas used for en route distance measurementsfor many years. The high-precision DME modifies the leading edge of the DME pulses toimprove the system's multipath immunity andbasic accuracy [5].
MLS obtains the angular locations of aircraftby electronically scanning a ground antenna'sfan beam to and fro so that the time separationof the received beam at the aircraft is proportional to 8
0, which is defined as the angle be
tween the runway centerline and an aircraft'sposition (Fig. 2). As a first approximation, thereceived beam envelope is eqUivalent to theground-antenna pattern as a function of 8. (If theground antenna rotated physically, the receivedenvelope would be identical to the antennapattern. In electronically scanned arrays, however, the received envelope is not exactly identical to the antenna pattern because the arraystypically have sidelobes that vary with scandirection resulting from phase-shifter quantization effects.)
461
Evans et aI. - MultipathModelingjor Simulating the Performanceojthe Microwave Landing System
AzimuthAntenna
Received ~Signals
(dB)
Measurement - - Threshold
(-3dB)
ScanBeam
Runway Centerline
---:,-- __ r-
.Scan Rate = 8
Time (liS).....
"Fro"ScanBeam
------~)... -
"To"Scan
Begins
,T1 T2
"--------...... ,------_/VTime Difference t::.T (Measurement Is
Directly Related to Azimuth Angle (0)
"Fro"ScanEnds
Fig. 2-TRSB bidirectional scan format. Note that t::.T is directly related to the azimuth angle eo'
The ground antennas typically have abeamwidth of 10 to 3 0 between the mainlobe'shalf-power points, and sidelobes that are approximately -25 dB with respect to the mainlobe's peak. Because the angle gUidance signalsare radiated at a high frequency (e-band), wecan readily design the antenna so that its radiation is confined to the desired coverage region.For example, to minimize ground reflections, theelevation pattern ofthe azimuth antenna can bedesigned to roll off rapidly at the horizon. Theairborne receiver determines the beam centroidto a fraction (typically 5%) of the beamwidth by
462
locating the -6-dB points on either side of thebeam or by using a split-gate tracker [6].
Let us now consider the effects of multipathon a received signal. When a multipath signal isat a scanned angular coordinate different fromthat of the direct signal, the received waveformwill consist of the coherent superposition of thetwo beam envelopes such that the centroids ofthe received beam shapes may no longer be atthe appropriate locations. For both the centroidand split-gate types of receivers, the error thatresults from multipath depends critically on themultipath source's angular location with re-
The Lincoln Laboratory Journal. Volume 2. Number 3 (1989)
Evans et a1. - Multipath Modelingjor Simulating the PerJonnanceoj the Microwave Landing System
spect to the angular location of the receivingaircraft in the scanned angle coordinate:(1) In beam When the angular separation (0)
between the direct and multipath signalsis less than 1.5 transmitter beamwidths(8), the multipath error can be as large asapproximately 0.5R8, in which R is themultipath amplitude divided by the direct-signal amplitude. For small 0, theworst-case error is proportional to oR.
(2) Out oJ beam. When 101 >1.58, the directsignal of the received envelope is perturbed by sidelobes that result frombeam scattering by the multipathsource. In this situation, the worst-caseerror is approximately R1]8, in which 1] isthe sidelobe amplitude ratio for thetransmitter antenna. The MLS trackingand acquisition logic may attempt tostart a track of a multipath signal if 1] isgreater than unity and if the multipathcondition exists for a long period of time,e.g., 20 s when MLS has been trackingthe direct signal for at least that amountof time.
Similarly, the DME measurement is accomplished by delay-and-compare processing [5]on the leading edge of the DME pulse suchthat multipath delays ('tj greater than approximately 300 ns will not cause errors. For shortermultipath delays relative to the direct signal, theworst-case error is approximately given by Rr.
The above multipath error characteristics ofMLS have been important in guiding our multipath modeling effort. In particular, we note thatout-of-beam multipath is oflittle concern for theangle gUidance subsystems unless the multipath level exceeds that of the direct signal for along period of time, e.g., for more than 5 s.Because multipath sources in airport terminalcomplexes are typically located in out-of-beamareas, low-level reflections from the many smallobjects (e.g., luggage carts) in those areas arenot of operational concern. Hence these smallobjects need not be considered in the modelingeffort. Similarly, scatterers (e.g., the flat terrainin front of an azimuth array) that give rise to inbeam multipath at a very small separation angletypically cause only small gUidance errors. Thusthose types of scatterers need not be modeled
Table 1. Principal MLS Multipath Sources of Concern
Azimuth• Building reflections and diffraction when aircraft are not on runway centerline• Aircraft reflections (especially when the scattering source is near the approach end of the runway)• Shadowing by taxiing and overflying aircraft• Shadowing by small objects in front of the antenna• Scattering from irregular terrain in front of the antenna
Elevation• Reflections from aircraft and buildings in coverage region• Reflections from sharply rising terrain• Building shadowing (when aircraft are not on runway centerline)
DME• Reflections from scatterers with multipath delays between 20 ns and 300 ns
The Lincoln Laboratory Journal, Volume 2, Number 3 (1989) 463
Evans et aI. - Multipath ModetingJor Simulating the PeiformanceoJthe Microwave Landing System
very accurately. Given the above error characteristics, Table 1 summarizes the scatterers ofmajor concern.
Multipath Model Features
This section describes some of the salientfeatures of the models we use to compute reflected and diffracted signals. Because the detailed mathematics of the models are availablein a series ofLincoln Laboratory reports [3, 4, 7],our objective in this article is to present some ofthe principal innovative ideas in the modelingand to show examples of the model validation.
First, however, a fewwords about the outputsof the scatterer models are in order. For a givengeometry that involves a transmitter, a receiver,and signal-scattering objects, each receivedsignal component is characterized [3, 4, 8] by its(1) amplitude (p),
(2) RF phase change due to scattering (1/>),(3) time delay relative to the direct signal (r),(4) elevation and azimuth relative to the
ground antenna (at' f3t ),
(5) fractional Doppler shift due to the motionof the receiver (OJ
sd), and
(6) arrival angles at the aircraft relative tothe vehicle's velocity vector (a, f3 ).
r r(Note that for a given transmitter-receiver-scat-terer geometry, there may be a number of received signal components due to the effects ofsecondary paths that involve ground reflectionsand/or the decomposition of a given scatterer'sreturn into several scattered or diffracted rays.)We assume that the receiver and the scatterersare far enough from the ground antenna so thatthe system model can represent the actual antenna patterns at each instant oftime during theground antenna scan by using the angles at' f3t'a , and f3 ' In addition to accounting for Doppler
r rshift of the received signal structure due toreceiver motion (a small effect with the ICAOstandard MLS), the term OJ
sdhandles changes in
the received signal phase between successiveantenna scans.
Another key element of our propagationmodel is that it takes into account the multiplebounce reflection paths that can result frombuildings, aircraft, and the surrounding terrain.
464
This feature is important because ground reflections can substantially reduce the effective azimuth and DME direct-signal level at low altitudes where the MLS accuracy requirements aremost demanding. The model handles multiplebounce effects by computing three additionalsignal components that correspond to the threeadditional paths that involve ground (G) reflections between the transmitter (T), scatteringobject (0), and receiver (R). That is, in additionto the standard T-O-R path, the paths T-G-O-R,T-O-G-R, and T-G-O-G-R are considered. Tokeep the entire computation manageable, weassume that the terrain of concern for thesesecondary ground reflections is flat so that theconventional method of images [7] can be used.
Specular Ground Reflection
Earlier, we noted that irregular terrain presents an important and direct multipath challenge to the performance ofMLS. (As discussedabove, reflections from homogeneous flat terrain is a secondary challenge. When the terrainis approximately flat and homogeneous, a standard simplified model for terrain reflections canbe invoked [9].) In irregular terrain conditions,the ground is considered to be a compositerough surface (as discussed by Beckmann [10])that has a small-scale roughness superimposedon a large-scale roughness. The large-scaleroughness, which describes a region's topographical features, is modeled by dividing theground surface of concern into a number oftriangular or rectangular plane surface elements, each with a homogeneous dielectricconstant. The small-scale roughness of thesurface elements is assumed to have a Gaussianheight distribution with rms roughness (Jh' Wefurther assume that (Jh is smooth enough so thatwe can apply the Beckman-Spizzichino [l0]approximation, in which the effect of the smallscale roughness is to reduce the reflected signalfor that plate.
Because the elevation patterns of the MLSantenna roll offrapidlynear the horizon, groundreflections are ofconcern principally at very lowelevation angles. At such angles, the Fresnelzone ellipses are often highly elongated so that
The Lincoln Laboratory Journal. Volume 2. Number 3 (1989)
Evans et aI. - Multipath Modelingjor Simulating the Performanceojthe Microwave Landing System
Fig. 3-Terrain height profile of Fort Devens, Mass., golfcourse.
I-Transmitter Antenna
IDitch .-".....
........... ~ -I
86
Building Reflections
-40 L.-....J.-_L.-.--.L._....l..-_..L..----'_-""_........._.L.....J
-30
-10
tion to the direct signal.
To model buildings and hangars, we use oneor more rectangular plates, each ofwhich has aspecified small-scale roughness height, a dielectric constant, and. in order to account for slopedroofs, a factor that represents the plate's tiltfrom the vertical. Each plate causes four scattered signals resulting from the secondary terrain-reflection effects discussed earlier.
We now consider the computation for a singleplate with a given transmitter-receiver position.By invoking Babinet's principle (7), we can showthat the scattered signal for a given plate can beapproximated by a product of Fresnel integralsthat corresponds to the horizontal and verticalintegration over the plate. This separable-
-8 -6 -4 -2 0 2 4
Elevation Angle (deg)
Fig. 4-Comparison ofC-bandpowerspectra from (a) fieldmeasurements and (b) computer model for the terrain ofFig. 3.
~ 120-; 110.g 100III~ 90iIi 80o 200 400 600 800
Distance away from Transmitter Antenna (tt)
Fresnel diffraction is no longer valid. Thus, inthe general case, a received signal component iscomputed for each surface element by using aslightly simplified full Fresnel-Kirchoff diffraction formula. (The main simplification is that theintegration is performed over a rectangularregion that corresponds to 2.8 Fresnel zones.This choice ofregion has been shown to comparewell with closed-form results for an infiniteconducting plane in a variety of transmitterreceiver geometries [7].l
We conducted extensive field tests to characterize how well our model represented the irregular terrain. The tests were experimentallychallenging because there were sometimes anumber of scattered signals whose angularseparation in elevation (i.e, a[l was less than theelevation beamwidth of the measurement system. Consequently, to analyze the experimentaldata we used several techniques [11) for estimating the high-resolution power spectra for themodel outputs.
Figure 3 shows one of the experimental sites,which is characterized by both downhill anduphill terrain with cross-slopes at severallocations. We approximated the site with 17 rectangular plates that corresponded to the site'slarge-scale topographical features. For the terrain of Fig. 3, Fig. 4 compares the model's datawith C-band measurements taken with a 60-Awave-front sampling aperture (12). In Fig. 4, BSis the classical beam-fOrming estimate, ME isthe maximum-entropy least-squares estimate,and ML is the maximum-likelihood estimate[11). Note that both the measured and simulatedestimates suggest the presence of approximately two scattered specular returns in addi-
TIle Lincoln Laboratory Journal. Volume 2. Number 3 (1989) 465
Evans et aI. - MultipathModelingJor Simulating the PerformanceoJthe Microwave Landing System
Height of Receiver Antenna (tt)
Aircraft Reflections
We should note that extensions to the modeldescribed above would be useful. A survey todetermine the nature of the building-reflectionphenomena at eight major U.S. airports foundthat many of the large buildings at the airportshad walls or doors made ofvertically corrugatedmetal in which the surface period of the corrugations was greater than a signal wavelength atC-band [8, 16]. Such surfaces act as a diffractiongrating in which reflections occur at the specular angle and at angles that correspond to solutions to the classical grating equation. Unfortunately, the shape of the corrugations is suchthat either laborious numerical calculations orexperimental measurements must be made todetermine the signal's power in each scatteredmode as a function of the angle of incidence.Furthermore, the plate-computation model willhave to be augmented to predict accurately theextent of the reflection region for each reflectionmode that does not correspond to the classicalspecular reflection.
Reflections from aircraft on the ground are ofpotential concern because these scatterers areoften situated on taxiways close enough to therunway so that the reflected signal causes inbeam multipath. Furthermore, aircraft surfacesare made ofcurved metal, which scatters beamsover a wide range of angles and thus creates alarger region of specular multipath than wouldbe produced by flat plates of the same sizes.Based on the results of our field tests andpublished data for aircraft cross sections versusaspect angles, we conclude that only the tail finand fuselage ofan aircraft need to be consideredfor our purposes.
Thus, following a suggestion by H.A. Wheeler[17], we modeled both the tail fin and fuselagewith portions of cylinders. The reflection levelswere estimated by taking the product of aFresnel integral that corresponded to integration along the cylinder, and a term to account forray divergence. The computation of ray divergence by integrating around a curved surface isa formidable task. Wheeler, however, cleverlynoted that he could accomplish the integration
••~ Measured
Data Points
Fresnel-diffraction assumption greatly increases our computational efficiency because itallows the use ofstandard subroutine packagesthat contain efficient numerical routines for theevaluation of the integrals.
Our model has been validated in both largescale qualitative and detailed quantitative tests.The basic model predicts the occurrence ofsizable building reflections only with buildinggeometries that yield specular reflections for thegiven transmitter-receiver locations. This prediction, which was initially verified in tests witha receiver on a moving van at Logan International Airport in Boston [13], was confirmed insubsequent testing with moving vans and aircraft at National Airport in Washington, D.C.;Philadelphia International Airport; Wright Patterson AFB in Ohio; Tulsa International Airportin Oklahoma~ and Kennedy Airport (JFK) inNew York City [3, 14, 15].
Using an instrumented van parked at a surveyed point, we took multipath amplitude measurements as a function of receiver height andcompared the measurements with our model'spredictions. Figure 5 shows the comparison forreflections from a DeltaAirlines hangar at LoganAirport. In this case, the transmitter and receiver were respectively 1,025 and 675 ft fromthe hangar, and the transmitter-to-hangarangle of incidence was 45° [13]. Similar agreement between real and simulated data wasalso obtained in tests involving the hangars atJFK airport [14, 15].
Fig. 5-Comparison of multipath model with C-band measurements for the Delta Airlines hangar at Boston's LoganInternational Airport (8 December 1974).
~ ~ 1.2 r----,--i--'---=:;::t:il::::;:-eriell ::JQi:= 1.0a:: a.
~ ~ 0.8::J_
:= ~ [email protected] .<l: CI( 0 4.c tl .~ ~.g- i5 0.2"S 0~ - 0.0 .........-----"---....---'----'-------''--'
50 54 58 62 66 70
466 The Lincoln Laboratory Journal. Volume 2. Number 3 (1989)
Evans et aI. - Multipath ModelingJor Simulating the PerformanceoJthe Microwave Landing System
•
Boeing
747
Transmitter
o 100 200 300 400 500
Scale (tt)
Fig. 6-Geometry for 747 aircraft at Boston's Logan InternationalAirport (12 December 1974). Note that because ofthe curvature of the aircraft's tail fin, the angle of incidenceis not equal to the angle of reflection.
for metallic cylinders by using the closed-formgeneralized divergence formula of Riblet andBarker [18].
We validated our model by a series of experiments (often in the middle of the night) thatinvolved parked wide-body aircraft at Boston'sLogan Airport. For those experiments, Fig. 6shows the geometry of the transmitter and receiver that were used to take measurements ofreflections off the tail section of a 747 aircraft.Note that the angle of the incoming ray asreferenced to the centerline ofthe aircraft is 20° ,and the angle of the outgoing ray is 35°. Thedifference results from the curvature of theaircraft's tail. For the geometry of Fig. 6, Fig. 7compares the multipath measured levelswith the simulated levels. We obtained similarmeasurements and results for DC-I0 and 727aircraft [7, 13).
Shadowing by Aircraft or Structuresnear the Line ofSight
Shadowing by obstacles near the transmitter-to-receiver line of sight (LOS) causes multipath errors through two mechanisms. Becauseof shadowing,(1)· the direct signal might be attenuated,
thus causing an increase in the relativeamplitude of(and error due to) multipath
The Lincoln Laboratory Journal. Volume 2. Number 3 (1989)
signals, and(2) the transmitted wave front might be dis
torted so that angular errors are directlyproduced even though little or no attenuation of the direct signal occurs.
Much of the existing literature on shadowinginvolves radio communication links in whichonly mechanism (1) was of concern. Thus manyradar and navigation system engineers havebeen surprised to find that sizable angular errors can occur in situations in which the transmitter-to-receiver LOS is not blocked.
We model all shadowing profiles as a collection of flat rectangular plates. For buildings,these plates are analogous to the plate modelsused for scattering computations. For aircraft,an appropriate plate collection for each viewingangle is chosen; i.e., the front-to-back and topto-bottom profiles consider the fuselage andwings while the side profile considers the tail finand fuselage. The user specifies the type of theshadowing aircraft and its movement characteristics; the model then determines an appropriate plate collection for the given shadowinggeometry. Thus the diffraction computationreduces to a calculation of separate signals foreach of the rectangular plates.
The key point in obtaining a ray-theory representation ofthe diffracted signal is the following:in determining the number and location of diffracted rays, the principal factor is the variationof the diffracted signal phase as a function of thepositions ofeach ofthe radiating elements in theaperture of the ground array antenna. Thus thebasic idea is to represent the diffracted signal as
(])iD>'0
'<ij -; 0 r----,----.-.....-r-----,----,--..-------r---,Qi""§l0:_(]) '3. -10
""§l E'3..:: -20E ~
~ ~ -30Cti~o..U'';:; ~ -40 '--"''--I_--l..._...L-__I.....---l.._....L._L-----l.-=-l.LI
'S (5 20 30 40 50 60 70 80 90~Q~ Receiver Height (tt)
Fig. 7-Comparison of multipath model with C-band measurements for the geometry of Fig. 6.
467
Evans et aI. - Multipath Modelingjor Simulating the PeTjonnanceojthe Microwave Landing System
Region_ Simulated _
Cia;
H~i:r§:s 0.00 -----------I<rIrl-t-"T'-----t+----I"C""".~ ~ -0.05> E~ 0w~
1092345678
Distance from Azimuth Site (nmi)
0.150.100.050.00
-0.05-0.10-0.15-0.20 L..-.......J'-----L_.....L.._...J...._....l..-_..l...-_.L..-.......J_---L_---J
o
Fig. a-Comparison ofmultipath modelwith field measurements foran aircraftapproach at JFK Airport in New York City. The approach was off the runwaycenterline and shadowing was caused bya large hangar near the MLS elevation antenna.
a Fresnel integral that is a function of theradiating elements' positions in the scanningdimension of the antenna's aperture. (For azimuth arrays, the position is horizontal. Forelevation arrays, the position is vertical.) Usingstandard expansions ofthe Fresnel integral, wethen approximate the integral representationwith a sum of plane waves. It can be shown (19)that this procedure yields a ray representationthat is a function of the shadowing plate's rectangular size and the LOS. Depending on thesize of the obstacle in the coordinate beingscanned, one, two, or even three diffraction raysmay be created by a given plate.
Figure 8 is a comparison of our model'sresults with elevation-error measurements of aflight atJFKAirport. The flight, which was off thecenterline of the runway, was shadowed by alarge hangar that was near the MLS elevationantenna [3). Figure 9 presents an example ofshadowing caused by an airborne aircraft at theFAA Technical Center [3] in Atlantic City, N.J.Given the likelihood of shadowing at majorairports, it is important to note how well the
output of our model agrees with the experimental data in both ofthese very different situations.
Humped-Runway Shadowing
When an aircraft is about to touch down,humps in the runway can shadow the vehicle'sreceiver from the azimuth transmitter or DME.This condition causes a significant loss in thedirect signal that the landing aircraft receives.
Initial solutions modeled humps as knifeedge creases in the ground. The representationwas simplistic in not taking into account the factthat forward reflections can occur offboth sidesofa hump, not just the side facing the transmitter. Consequently, we adopted the work ofWaitand Conda [20), who modeled humps as infinitedielectric cylinders. Wait and Conda showedthat they could represent the diffracted signal asthe sum of knife-edge diffraction (Le., a Fresnelintegral) and a correction term that takes intoaccount the radius ofcurvature ofthe hump andthe dielectric constant of the hump's material.Because the radii of runway humps are con-
468 The Lincoln Laboratory Journal. Volume 2. Number 3 (1989)
Evans et aI. - Multipath Modelingjor Simulating the Performanceojthe Microwave Landing System
MLS Model Features
results along a considerable length of the shadowed region. Similar agreement between ourmodel and experimental data was obtained fromfield tests at France's Coulommiers airport [7].In both the Bedford and French data, the modelgave the best results when it used the largestvalue for equivalent cylinder radius that couldbe justified by the runway profile.
The above model, which represents a diffracted signal with a single effective-direct-signal ray, may warrant refinement in the future.Normally, the major operational concern wouldbe whether a given geometric configurationyields an adequate SNRwhen aircraft are at verylow altitudes. In such conditions, however,MLS has an ample power budget margin if noshadowing is present. Thus we are principallyconcerned with model accuracy in shadowedregions; in those areas, the single-direct-rayrepresentation described above is adequate.When the receiver is above the geometricshadow region, however, the representationdoes not adequately handle the effects of aground-antenna-pattern gain that varies rapidly near the horizon. Therefore, it is desirable todevelop a multiray model that accurately represents the elevation-angle (Le., f3J distribution ofthe net received signal inside and outside thegeometric shadow region.
The MLS model, which is a very straightforward implementation of Fig. 2, computes thebeam envelope received by an aircraft as a beamscans by the vehicle. The functional form of thebeam wave is determined by the measured ortheoretical patterns of the ground antenna ofconcern. The superimposition ofbeam patternsthat correspond to the various signal pathsresults in the net received envelope. The remainder of the MLS model parallels the actual microprocessor-based receiver processing that MLSuses. A tracking gate centers on the largestconsistent envelope peak and the beam arrivalangle is calculated by finding the times at whichthe leading and trailing edges of the receivedenvelope cross a certain threshold. Variouschecks and tracking algorithms screen each
-
.-
-
-
-I
2.00 3.00
I I
Simulated-.. Portion ~
of Flight
0.10Eg 0.05~QiCl-oQ) 0 0.00:s~'-.r;0_ -0.05'- cow.Q-.r;=::
-0.10_ :::J
:::J~
E'N -0.15«
-0.20
OJ 0.10Q)
:s 0.05 -ew 0.00.r;'5E -0.05 -'N«
-0.10-0 -~:::J
-0.15en -coQ)
~ -0.20-1.00 0.00 1.00
Distance from End of Runway (nmi)
Fig. 9-Comparison ofmultipath model with field measurements taken at the FAA Technical Center in Atlantic City,N.J. Shadowing was caused by an airborne CV-580aircraft.
trolled to minimize their adverse effects on aircraft, we found that empirical tables could beused to determine the correction term [7] regardless of the transmitted signal's polarization.
It is important to note that the intended useof the Wait and Conda model is within or nearthe shadowing region. When an aircraft is highenough so that it is not shadowed by any humps,our multipath model reverts to the standard flatterrain representation, which considers theground reflection that occurs near the transmitter. In either the hump or flat-terrain case, theeffects of shadowing are treated as a change inthe complex amplitude of the direct signal.
We validated our model mainly by comparingits output with field data. Figure 10 comparesthe results ofthe model with field tests that weretaken on the main runway at the Royal AirEstablishment in Bedford, England. Althoughthe overall runway profile does not resemble adielectric cylinder, the model gives excellent
The Lincoln Laboratory Journal. Volume 2. Number 3 (1989) 469
Evans et aI. - MultipathModelingJor Simulating the PerformanceoJthe Microwave Landing System
10 11
Measured Signal Loss
2 3 4 5 6 7 8 9
Distance from Transmitter (10,000 tt)
Transmitter~Location
~ .""" ~~~~~~~~c~~~~~-,
Runway Profile ' - , - --
80
CIl 100CIl0
....Jco 120"0
140
CD>~ 300o.t=
.D ~«Qjc >o CD 280:.;:-1CIl CIl> CD~Cf) 260w
0
Fig. 1O-Comparison of multipath model with C-band measurements taken at the main runwayof the RoyalAirEstablishment in Bedford, England. Shadowing was causedbya runwayhump. The transmitter and receiver heights above the local ground level were 4.8 ft and 9.0ft, respectively.
measurement to ensure that only valid anglesand DME measurements are outputted (21).
The perfonnance ofthe receiver portion of theMLS model was validated primarily at a testfacility that the Calspan Corporation developedfor the FAA (6). The Calspan system could injectinto an actual MLS receiver a wavefonn thatcorresponded to the reception of a direct signaland a single multipath signal. The system hadfairly tight control over the characteristics of thedirect and multipath signals. e.g.. the amplitudes. RF phases. and angular separation between the two signals. Figure 11 compares ourMLS model's output with the Calspan Corp.system. Differences between the two sets ofdataare within the ±0.5-dB tolerance of the multipath-to-direct-signal-Ievel setting of theCalspan Corp. system.
We further validated the MLS model withtests in operational environments similar tothat described above. The tests were particularly useful in addressing sidelobe modeling.We found that at a given angular separationfrom the main-beam location. the sidelobesof an antenna vary with time due to varia-
tions in phase-shifter error. Additionally. thephase-shifter scan-control program alsocauses the errors at a given point in a scan tovary from scan to scan. Thus. it was not clearwhether we could accurately represent highlevel sidelobe multipath errors with a simplesidelobe model that consisted of an arrayexcitation pattern for the first few sidelobes anda sinusoid with a I-beamwidth spatial periodfor the remaining sidelobes.
The simple sidelobe model described aboveagreed reasonably well with results from fieldtests at the FAA Technical Center. (The sidelobemodel, however. overestimated the magnitudeof the multipath error by about 5 dB.) For theFAA tests. large reflecting screens were placedon a runway. The screens caused out-of-beammultipath that was greater than the direct signalover an extensive portion of the runway.
Predicting the effective sidelobe levels for theantennas ofvarious manufacturers presents anongoing challenge because the static antennapatterns tend to underestimate significantlythose sidelobes which are far removed from themainlobe. The dynamic beam envelopes. on the
470 The Lincoln Laboratory JournaL. Volume 2. Number 3 (J 989)
Evans et a1. - Multipath Modeiingfor Simulating the Performanceofthe Microwave Landing System
Simulation Applications
Azimuth Separation Angle (deg) BetweenMultipath and Direct Signals
other hand, tend to overestimate those samesidelobes.
method, however, requires a prohibitively largevolume ofcomputer runs because ofthe numerous multipath scatterer parameters (e.g., air
craft type, truck .size), types of ground antennasystems, and different approach parameters(e.g., ground velocity, airborne-antenna pattern, approach angle) that need to be considered. Consequently, we adopted the followingtwo-stage approach:(1) The worst-case error is determined as a
function of scatterer location for fixedground- and airborne-system parameters. Simple analytical models determinethe effects of receiver motion.
(2) Using the worst-case scatterer locationsfrom step (1), we run full-scale simulations. The simulations determine theway in which the different receiver-approach parameters affect the multipathparameters. In addition, the simulationsdetermine the operational nature of theresulting error.
This two-stage approach, which takes advantage of the modular nature of the overall simulation, permits the consideration ofa wide rangeof parameters for all of the principal variables.
We will now illustrate the above approach forthe specific case in which reflections offa taxiingaircraft cause azimuth errors for an airborneaircraft just before touchdown. In this example,the largest combined multipath level for fuselage and tail-fin reflections occurs when thetaxiing aircraft is turned so that the reflectionpoint is in the middle of the vehicle's fuselage.Using this knowledge, we can orient the taxiingaircraft to yield the maximum multipath level ateach point on the airport surface at which theworst-case error is to be calculated. The worstcase errors are individually computed as thesum of the absolute values of the errors thatresult from fuselage and tail-fin reflections. (Theerrors are calculated as the ratio between themultipath level and the direct-signal level. Boththe multipath and direct-signal levels take intoaccount ground-reflection effects.)
For the above example, Fig. 12 shows theO.03°-error contours as a function of aircraftposition for a Boeing 747 and 727 aircraft. Fromthe figure, we see that only the 747 aircraft is of
R =-3 dB
g> 0~
ew 0.1cell~ Calspan Data
0.2
0;Q)
~
e 0.1'ow
To date, one of the major applications of ourMLS multipath simulation has been in determining critical areas where restrictions must beplaced on the movement of aircraft and othervehicles to avoid excessive MLS gUidance errors.In this section, our discussion will deal onlywithreflection effects because shadowing effects canreadily be addressed by the same method.
One possible way to determine critical reflection areas is to carry out full-scale simulationsof aircraft approaches with a scatterer at eachpossible airport location of concern. This
Fig. 11-Comparison ofcomputermodel with Calspan datataken at a scalloping frequency of 0.6 Hz. The mean errorand standard deviation in error are calculated with respectto the RF-phase difference. R is the multipath level dividedby the direct-signal level.
The Lincoln Laboratory JournaL Volume 2. Number 3 (l989) 471
Evans et aI. - Multipath Modelingjor Simulating the Performanceojthe Microwave Landing System
concern. It is also important to note that themost severe multipath effects occur when thetaxiing aircraft are within a holding area adjacent to the end of the runway.
Because current ICAO gUidelines requiretaxiways that are parallel to a runway to be atleast 300 ft from the runway's centerline, Fig. 12suggests that existing taxiways can be usedduring MLS instrument approaches. This observation was further confirmed in full-scale simulations in which five worst-case-oriented 747aircraft were located at points 300 ft from therunway centerline. The vehicles produced maximum multipath levels at 8, 50, 100, and 200 ftabove ground level, and the peak azimuth errorsencountered dUring the full simulation run wereless than 0.01°. These results confirm the conservative nature ofthe worst-case-error calculation (8). It should be noted that 747s located 150ft off the runway in the middle of the errorcontour shown in Fig. 12 were found to produceunacceptable errors (0.07°) when the landingaircraft was near touchdown.
Similar calculations for MLS elevation measurements show that the area immediately infront ofthe MLS elevation system should be keptfree of 747 aircraft. However, on the oppositeside of the runway. aircraft on the surface mayoperate freely without any adverse effects (8).
Conclusions
This article describes a very ambitious simulation that can handle the full range ofmultipathphenomena of concern to MLS. We accomplished this comprehensiveness only by carefully considering the MLS error mechanisms atthe outset. We then focused the multipathmodel development and validation to emphasizethose factors which were of greatest concern.Fortunately. Lincoln Laboratory personnel werevery actively involved in the U.S. and ICAO MLSevaluation programs, which were being conducted throughout the major period when thesimulation was developed. Our involvementprovided us with the opportunity to interactclosely and frequently with a very knowledgeable group that was always available to critiqueour work. As a consequence of this close scru-
472
500 .------,-----,.-------.-----,;----.-----,
Scattering Aircraft
400 - 747 0.03°-Error Contours···..····727 0.03°-Error Contours
§:300OJUc:C1len 200(5
100
o~..L_.....l-_ ___..I.__...L-_......L..__L.....L;ii::=..:l
-1000 1000 3000 5000 7000 9000
Distance (ft)
Fig. 12-Plan view of simulated scattering conditions at ahypothetical runway. The simulation was performedfor twodifferent types of aircraft: a Boeing 747 and 727. Thecontours enclose areas where the MLS azimuth error is atleast 0.03°. .That is, if a taxiing 747 or 727 is inside itsrespective contour, the azimuth error would be at least0.03° for an aircraft about to land on the runway. For thisparticular example, the airborne aircraft is assumed to beapproaching along the runway centerline, 11,000 ft fromthe transmitter antenna and 50 ft above the ground.
tiny, the model received extensive validationboth by dedicated propagation measurementsin a variety oflocations as well as in comparisonwith the results ofoperationally oriented testingat airports in the United States and abroad.
We would like to call special attention to twoparticular features of the multipath model thatprevious work has often ignored. The consideration of ground reflections in determining boththe effective direct-signal level and effectivemultipath-signallevels is especially importantat low altitudes. where MLS gUidance is particu1arly critical. Second, the use of a ray-theorymodel for representing shadowing phenomenapermits the consideration of errors that resultfrom wave-front distortion as well as errors thatinvolve losses in the direct-signal level.
Lincoln Laboratory has also used the multipath model with the Mode-S (22) radar system topredict monopulse errors caused by shadowing.Thus the model can readily be adapted for otherMode-S propagation studies. In addition. thescattering assumptions we made are such thatthe model can also be used to evaluate multipath effects (in particular. angular errors due to
The Lincoln Laboratory Joumal. Volume 2. Number 3 (1989)
Evans et aI. - Multipath Modelingjor Simulating the Performanceoj the Microwave Landing System
shadowing, and false targets due to reflectionmultipath) for the ASR-9 (Advanced Surveillance Radar-9) and TDWR (Terminal DopplerWeather Radar) systems.
In the preceding sections, we noted severaldesirable refinements to the multipath model.Furthermore, we see the need to add the following two capabilities: consideration of out-ofcoverage and fly-Ieftjfly-right sector radiation signals, and explicit flagging of low SNRconditions.
Acknowledgments
The models described in this article are theresult of several years of research conducteddUring the 1970s by a dedicated group at Lincoln Laboratory. Principal contributors included Richard Orr, David Sun, Robert Burchsted, Steve Sussman, Samuel Dolinar, JanetReid, and Carol Martin. Work by Dr. HaroldWheeler of Hazeltine Corporation was very significant in our approach to multipath modelingand in our use ofanalytical system-error modelsto complement the results of formal calculations. We also benefited greatly from intensetechnical interactions with Robert Kelly ofBendix Corp., Mel Zeltser of Hazeltine andMITRE Corp., Lon Sanders of lIT Gilfillin, PaulFombonne of Thompson CSF, Jack Beneke ofCalspan Corp., the United Kingdom MLS groupat the Royal Air Establishment in Bedford,England, and the University ofBraunschweig inWest Germany.
Special mention should also go to the MLSprogram office of the Federal Aviation Administration. In particular, we thank FrankFrisbie, Joseph DelBalzo, Douglas Vickers, andGene Jensen for their support and encouragement throughout the MLS development andevaluation period.
References
1. M. Kayton and W.R Fried, eds., Avionics NavigationSystems. John Wiley & Sons, Inc.. New York (1969).
2. T. Breien "Computer Analysis of MLS in MultipathEnvironment," lEE Conjerence Publication No. 147(Nov.1976).
3. J.E. Evans, S.J. Dolinar, D.F. Sun, and D.A. Shnidman,"MLS Multipath Studies, Phase3. Final Report, Volume
The Lincoln Laboratory Journal. Volume 2. Number 3 (} 989)
II: Development and Validation ofModel for MLS Techniques," PrQject Report ATC-88. Lincoln Laboratory (7Feb. 1980). FAA-RD-79-21.
4. J.E. Evans, S.J. Dolinar, and D.A Shnidman, "MLSMultipath Studies. Phase 3. Final Report Volume III:Application of Models to MLS Assessment Issues."Project Report ATC-88, Lincoln Laboratory (8 June1981). FAA-RD-79-21.
5. RJ. Kelly and E.F.C. LaBerge, "Guidance AccuracyConsiderations for the Microwave Landing SystemPrecision DME," NaVigation, Journal ojthe Institute ojNavigation 27, 1 (1980).
6. J. Beneke, D. Wightman, A Offt. and C. Vallone. "TRSBMultimode Digital Processor," Calspan Corp. FinalReport (Apr. 1978) FAA-RD-78-84.
7. J. Capon, "Multipath Parameter Computations for theMLS Simulation Computer Program," Project ReportATC-68. Lincoln Laboratory (8 Apr. 1976). FAA-RD- 7655.
8. J.E. Evans, RB. Burchsted, J. Capon, RS. Orr. D.A.Shnidman, and S.M. Sussman, "MLS Multipath Studies, Volume I: Mathematical Models and Validation,Volume II: Application of Multipath Model to Key MLSPerformance Issues," Project Report ATC-63, LincolnLaboratory (25 Feb. 1976). FAA-RD-76-3.
9. K.M. Mitzner, "Change in Polarization on Reflectionfrom a Tilted Plane," Radio Science 1, 27 (1966).
10. P. Beckmann. "Scattering by Composite Rough Surfaces." Proc. IEEE 53, 1012 (1965).
11. J.E. Evans,J.RJohnson, andD.F. Sun, "Application ofAdvanced Signal Processing Techniques to Angle ofArrival Estimation in ATC Navigation and SurveillanceSystems." Technical ReportTR-582. Lincoln Laboratory(23 June 1982), FAA-RD-82-42.
12. D.F. Sun, "Experimental Measurements of Low AngleGround Reflection Characteristics at L- and C-Bandsfor Irregular Terrain." Project ReportATC-l 07, LincolnLaboratory (1 Nov. 1982). DOT/FAA/RD-81/65.
13. D.A. Shnidman, 'The Logan MLS Multipath Experiment." Project Report ATC-55, Lincoln Laboratory (23Sept. 1975). FAA-RD-75-130.
14. "Validation of Computer Simulation by Comparisonwith Tests at OperationalAirports."paper presented bythe United States at the International Civil AviationOrganization All Weather Operations Division Meeting,Montreal. Apr. 1978. AWO/78-WP/135.
15. J.E. Evans and P.H. Swett. "Results of L Band Multipath Measurements at Operational United States (U .S.)Airports in Support of Microwave Landing System(MLS) Precision Distance MeasunngEquipment (DME/P)." ProjectReportATC-109, Lincoln Laboratory (23 July1981), DOT/FAA/RD-81/63.
16. D.A. Shnidman. "Airport Survey for MLS MultipathIssues," Project ReportATC-58, Lincoln Laboratory (15Dec. 1975), FAA-RD-75-195.
17. H.A. Wheeler. "Multipath Effects in Doppler MLS," ascontained in Hazeltine Corp. report MicrowaveLandingSystem (MLSJ Development Plan as Proposed by Hazeltine Corp. dUring the Technique Analysis and ContractDefinition Phase oj the National MLS Development Program (Sept. 1972). FAA-RD-73-185.
18. H.J. Riblet and C.B. Barker, "A General DivergenceFormula," J. Appl. Phys. 19,63 (1948).
19. J.E. Evans. D.F.Sun,S.J. Dolinar, andD.AShnidman,"MLS Multipath Studies. Phase 3. Final Report, Volume[: Overview and Propagation Model Validation/RefinementStudies," Lincoln Laboratory Project ReportATC88. Lincoln Laboratory (25 Apr. 1979). FAA-RD-79-21.
20. J.R Wait and AM. Conda. "Diffraction of Electromagnetic Waves by Smooth Obstacles for Grazing Angles,"
473
Evans et aI. - Multipath Modelingjor Simulating the Performanceoj the Microwave Landing System
J. Res. NBS 63D, 181 (1959).21. R.J. Kelly, "Guidance Accuracy Considerations for the
Microwave Landing System," Navigation. Joumalojthe
JAMES E. EVANS is Leaderofthe Air Traffic SurveillanceGroup at Lincoln Laboratory.The group develops primaryradar weather and aircraftdetection systems for theFederal Aviation Administra
lion. At the Laboratory. Jim has worked on seismic discrimination, ELF communication systems. the Microwave Landing System. high-resolution array processing. and weatherradar systems. A senior member ofthe IEEE. he received anS.B., an S.M., and a Ph.D. from MIT in 1963, 1964. and1971, respectively. While at MIT, he received the ComptonAward and the Carleton E. Tucker award for teaching.
DAVID A. SHNIDMAN specializes in antenna researchand electronic directionfinding research at LincolnLaboratory's Surveillanceand Control Division. Beforejoining the Laboratory 18
years ago. Dave worked at Bell Telephone Labs in NorthAndover, Mass. He received a B.S. and an M.S. in electricalengineering from MIT. and a Ph.D. in applied mathematicsfrom Harvard University. From 1978 to 1979, Dave was thechairman of the Boston Chapter of the IEEE InformationTheory Group. He is a member ofEta Kappa Nu. Tau Beta Pi,and Sigma Xi.
474
Institute ojNavigation 24, 189 (1977).22. V.A. Orlando. "The Mode S Beacon Radar System."
Lincoln Laboratory Joumal2, 345 (1989).
JACK CAPON is a staffmember of the AdvancedTechniques Group in Lincoln Laboratory's Surveil-
"'- lance and Control Division.A 27-year employee of theLaboratory. Jack has spe
cialized in adaptive array systems. He received the followingdegrees in electrical engineering: a B.E.E. from the Collegeof the City of New York. an M.S.E.E. from MIT, and a Ph.D.from Columbia University. Jack is a Fellow ofthe IEEE andthe MAS. He received the IEEE Centennial Medal in 1984.and was elected for two terms to the Board of Governors ofthe IEEE Professional Group on Information Theory.
The Lincoln Laboratory Journal. Volume 2. Number 3 (J 989)