AIRCRAFT WAKE DETECTION

DEVELOPMENT

T

Aircraft Wake Detection Using Bistatic Radar:Analysis of Experimental Results

Russell J. Iannuzzelli, Charles E. Schemm, Frank J. Marcotte, Laurence P. Manzi, Harold E.Gilreath, James M. Hanson, Jr., David A. Frostbutter, Alexander S. Hughes, Allen D. Bric,Dennis L. Kershner, and Leo E. McKenzie

he Applied Physics Laboratory has built and fielded an experimental bistatic X-band continuous-wave radar system for studies in aircraft wake vortex detection. Theradar, along with an acoustic pumping system, was installed at the Baltimore-Washington International Airport in the fall of 1996. Controlled experiments usinga NASA C-130 aircraft were conducted on 23 September, 29 October, and 10December. Results from two specific runs on 29 October are presented and discussedalong with numerical simulations of those runs.(Keywords: Bistatic radar, Radioacoustic sounding, Wake vortex.)

INTRODUCTIONWake vortices produced by the lifting surfaces of

large commercial transport aircraft can have cata-strophic effects on smaller planes following closelybehind. Many incidents have been blamed on wing-tip vortices in the past 15 years. The Federal AviationAdministration (FAA) and NASA are working withindustry to develop an intelligent system called theAircraft Vortex Spacing System. This system woulddetermine the optimum aircraft spacing within 3.2 kmof each runway based on a combination of vortexsensors located near the glide slope and a real-timepredictive model of the vortex environment in theatmosphere. The sensors would be used initially tovalidate this model. A vortex detection system suchas this would not only enhance airport productivity byallowing adaptive spacing, but would increase thesafety of all aircraft operating around the airport byalerting controllers that hazardous conditions mayexist near the runways.

In 1996, APL initiated an internal research projectto investigate the usefulness of a bistatic X-bandcontinuous-wave radar system for detecting wingtipvortices around the end of a runway. The bistatic

JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 19, NUMBER 3 (

radar, built by APL, was installed at the Baltimore-Washington International Airport (BWI) on eitherside of the south end of runway 33L. Data were takenon a C-130 emitting smoke from the wingtips on 23September, 29 October, and 10 December 1996. Inaddition to the radar, an acoustic illumination systemwas built and installed to evaluate acoustic pumpingfor possible signal detection enhancement.

A previous article on the vortex detection radarappeared in the Technical Digest in 1997.1 This secondarticle gives an expanded view of the acoustic illumi-nation and radioacoustic detection as well as the re-ceiver processing. In addition, we present and analyzeresults from both the BWI experiments and severalsimulation runs from the APL-developed Wake Evo-lution Code (AWEC).

INSTRUMENTATION RADARDESIGN ANALYSIS

The weather radar range equation for estimatingreceived power PR for a continuous-wave bistatic radar,

1998) 299

R. J. IANNUZZELLI ET AL.

in which the transmitter and receiver are spatiallyseparated, can be shown to be

PP G G

R RV VR

T T R

T Rv C V

( )( ) ,= ∩

l

ph

2

3 2 24(1)

where

PT = transmitter power,GT and GR = transmitter and receiver antenna gains,

respectively,l = wavelength,RT and RR = ranges from the transmitter and receiver

to the intersection of the two antenna beams,respectively,

hv =reflectivity of volume clutter (expressed in m2/m3),

VC = the common volume, which is contained withinthe intersection of the transmitter’s andreceiver’s conical beams, and

VV = vortex volume.

For a monostatic radar, the ranges are identical.From Ref. 2, clear air turbulence reflectivity hv can

be expressed as

h pv S[ sin( / )] ,= k k2 2 24F u (2)

where k = 2p/l is the radar wavenumber, and uS is thescattering angle, i.e., the angle at which the radarenergy is bent at the common volume from the trans-mitter toward the receiver (180° for the monostaticcase). The spectral representation of the turbulence isgiven by

F( ) . ,n/k k= −0 033 2 11 3C (3)

where k is the spatial wavenumber, and Cn2 is a struc-

ture parameter that measures the intensity of turbulentfluctuations. For the special case of homogeneous iso-tropic turbulence, Cn

2 can be expressed in terms of themore general structure Dn(r) using

C r D rn/

n( )2 2 3= − , (4)

where

D r n x r n xn( ) ( ) ( )≡ + −[ ]2 (5)

is the structure function, and r is a spatial increment. Thevariable n in Eq. 5 is the refractive index of air, whichis related to meteorological variables (temperature,

300 JOHN

pressure, humidity). The overbar indicates an ensem-ble average. Values of Cn

2 can range from 10216 m22/3

for weak turbulence to 10212 m22/3 for very strongturbulence.3,4

Vortex detection by radar poses several challengesto the radar designer. Most of the problems arise be-cause the vortex provides little radar return. Sincemost of the refraction will be away from the radartransmitter (Ref. 1), a bistatic arrangement is favored.The biggest problem with the bistatic continuous-wave radar is the performance degradation due tophase noise from the spillover signal. Unlike pulseradars, a continuous-wave radar is both transmittingand receiving all the time. Time gating, which isolatesa pulse radar receiver from the transmitter, does notapply with a continuous-wave radar, so other means ofisolating the high spillover signal from the transmitterinto the receiver must be used. We applied severaltechniques to reduce spillover at BWI, including con-struction of fences around the transmitter and receivershelter and separation and pointing of the receiver/transmitter so that sidelobe patterns could help nullout as much of the direct spillover as possible. Inaddition, we experimented with an active nuller1 andadaptive processing, which will be described later inthis article. Figure 1 depicts a hypothetical receivedsignal spectrum, with a finite-passband noiselike vor-tex signal component, a receiver noise component,and a spillover component with phase noise.

RADIOACOUSTIC DETECTIONOF AIRCRAFT WAKES:ACOUSTIC PUMPING

In 1961, the Midwest Research Institute developeda radioacoustic detection system, later known as RASS(radioacoustic sounding system), as a means formeasuring the temperature profile of the lower

–60

–80

–100

–120

–140

–160

–180

–200–150 –100 –50 0 50 100 150

Spillover

Vortex signal

Phase noise

Receiver noise

Doppler frequency (Hz)

Pow

er r

ecei

ved

(dB

m/H

z)

Figure 1. Illustration of received vortex signal spectrum showing themajor components. Turbulent fluctuations, Cn

2 , were 10212, 10214,and 10216 m22/3 for strong, moderate, and weak turbulences,respectively. Radar specifications were transmitter power = 400 W,beamwidth = 1.6°, transmitter/receiver separation = 1620 m, andvortex height = 61 m.

S HOPKINS APL TECHNICAL DIGEST, VOLUME 19, NUMBER 3 (1998)

atmosphere.5 The method was later refined and ex-tended elsewhere, notably at Stanford University,6,7 inthe former Soviet Union,8–10 and most recently inNorway.3 In such a system, acoustic waves are trans-mitted into the atmosphere to produce traveling per-turbations in the index of refraction that can be sensedby a bistatic radar. The design depends on establishingthe Bragg scattering condition, in which the RF energythat is scattered from successive acoustic wavefrontsadds coherently at the receiver. For a simple collinearconfiguration, with the transmitter and receiver equi-distant from the acoustic source (cf. Fig. 2), this con-dition occurs when

k ka Ssin( / ) ,= 2 2u (6)

where ka is the acoustic wavenumber.By virtue of the movement of the scattering wave-

fronts, the frequency of the received RF signal is shift-ed by an amount equal to the Doppler frequency fd,which, for a quiescent atmosphere, equals the frequen-cy of the acoustic source fa:

f = f k vd a a s ,= − (7)

where vs is the sound speed. A change in the speed ofthe wavefronts, which can be due to the advection ofthe waves by air currents or by a variation in soundspeed, changes the Doppler frequency, and trackingthis change is the basis for measuring wind speed ortemperature.

An idealized version of a radioacoustic system hasbeen analyzed2 for the case of Fraunhofer scatteringfrom advecting perturbations in the refractive index.Applying this analysis to the configuration shown inFig. 2, and arranging the results in the form of thebistatic radar equation, we obtain

z

Acoustic sourceTransmitter Receiver

Commonvolume

x

h

b u

s/2

s

u

Figure 2. Radioacoustic detection of aircraft vortices (us = scat-tering angle, b = antenna beamwidth, and h = height.)

JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 19, NUMBER 3 (1

AIRCRAFT WAKE DETECTION

P = P g Rr T / ,2 2 3 44l p s( ) ( )

(8)

where Pr = echo power, g = antenna gain, and R =range.

The variable s is an effective cross section (radarcross section of the entire volume), which is shownto be

s p= ( ) ( )

C P g k ha a / ,4 2 24 (9)

where Pa is the power of the acoustic source, ga is theacoustic antenna gain, h is the distance from the acous-tic source to the center of the common volume, theconstant C is given by

C (K ) /( ) . m /W ,s= v2 3 69 102 3 15 2r r ≈ × − (10)

and K is the Gladstone-Dale constant (see the boxedinsert).

For the radioacoustic configuration used in the BWIexperiments, the acoustic line in the Doppler spectrumwas readily detected under ambient conditions at theproper frequency (Fig. 3). However, the echo powerpredicted using Eq. 10 for the effective cross sectionwas found to greatly exceed the level actually mea-sured. The reason for the large difference is not clear,but it might stem from the coarseness of the initialtheoretical approximations. The assumption of Fraun-hofer conditions implies that the phase fronts of therefracted RF waves are planar throughout the commonvolume, and requires that the receiver be in the farfield of this volume. Neither constraint is satisfiedin the BWI configuration, but the role these factorsplay in the discrepancy remains unresolved. The lim-itations of time and scope have prevented furtherinvestigation.

As noted earlier, changes in the speed of the wave-fronts, and thereby changes in the Doppler frequency,can be caused either by alterations in the local speedof sound or by the advection of the waves by air cur-rents within the common volume. Both of these effectsoccur when a vortex wake descends into the field ofview of the radar. In general, the change in Dopplerfrequency equals

D Df = k ( v + wd a s ) ,− (11)

where the perturbation in sound speed Dvs depends onchanges in temperature and composition, and w is the

998) 301

R. J. IANNUZZELLI ET AL.

THE CONNECTION BETWEEN SOUND PRESSURE LEVEL AND RADAR CROSS SECTIONApplying the assumptions and analysis in Ref. 2, the

variable s can be related to the change in index of refrac-tion induced by the sound waves by

s p s= ( / ) ( ) cos .C h1 4 4 2 2 2k V nD u

In this equation, VC ~ (xb)3/us is the common volumedefined by the RF transmitter and receiver beam patternsin Fig. 2, and k is the RF wavenumber. The term Dnh is theamplitude of the index of refraction variations within thisvolume, which are represented as a train of plane waves,

Dn = Dnhcos [ka(z 2 vs t)] ,

where vs is the sound speed and ka is the acoustic wavenum-ber. The change in index of refraction can be related to theacoustic pressure by

Dn = K(Dr) = K(Dp)/vs ,

where K (by definition) is the Gladstone-Dale constant andr is the air density. The acoustic pressure (Dp), in turn, isrelated to the intensity of the sound waves in the commonvolume by

Dp = 2(Dvs)I ,

where the intensity is given by

I = [Paga/(4ph2)] exp(22ah).

Here, Pa is the power of the acoustic source, ga is theacoustic antenna gain, and h is the distance from the acous-tic source to the center of the common volume. The variablea is the absorption coefficient, which depends on the acous-tic frequency, the transport properties of air, and the relativehumidity. For the range of altitudes and frequencies of in-terest in the present work, a is negligibly small, even at100% relative humidity. The bracketed expression given inEq. 9 in the text results from using the above relationships,with a set equal to zero.

Acousticline “Zero Doppler”

spillover

T = 63.69 s

Figure 3. Spectrogram of run 9 on 29 October 1996 (10-kHz bandwidth).

302 JOHNS HOPKINS APL TECH

component of the vortex velocityfield parallel to the acoustic axis.

Because these variables are dis-tributed nonuniformly, we mightexpect to see the Doppler spec-trum broaden from a line to acontinuous spectrum of width(Dfd)max as the vortex wake movesinto the common volume. Thiswas not the case, however; obser-vations showed a large signaldropout correlating with the arriv-al of the descending wake (see theResults section for a discussion).

PROCESSING

Receiver ProcessingFigure 4 is a block diagram of

the receiver processing. Bothchannels—the main vortex chan-nel that is collecting informationfrom the common volume (as wellas spillover from sidelobes) and thereference channel—are mixed toapproximately 25 kHz. From there,both channels are sent through abandpass anti-alias filter. The data

NICAL DIGEST, VOLUME 19, NUMBER 3 (1998)

JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 19, NUMBER 3 (1998)

spectrogram fo1024, FFTdsizebandwidth of along with a Hthe spectral reproximately 15mately 0.13 s,integration thaFig. 3, time protime scale is a fin the display mode set, thissimultaneouslypretation of thsection.

Adaptive ProThe spillov

is several ordespherically basmon volume. Tzero-Doppler fsome of the spiadaptively cantor phase noiseshould cancel, contained in t

A/D32K

samples/sDigital

receiver

Referencechannel

Metrumdata

recorderClock

Digitaldata

recorder

Spectraland

adaptiveprocessing

Display

Anti-aliasbandpass

filterDigital

receiver

Vortexchannel

Oscillator Tunable

A/D32K

samples/s

Anti-aliasbandpass

filterX

X

Center @ 25 kHz

Center @ 25 kHz

Figure 4. Receiver processing block diagram.

are then recorded on a Metrum RSR-512 rotary stor-age data recorder at an 80-kHz sampling rate and aresent for real-time data processing on a VME system ata sampling rate of 32 kHz. This latter rate, along witha downconversion/filtering scheme, will preserve up to8 kHz of complex digital bandwidth, depending onwhich digital filter is used. All the digital filters areflat for approximately 75 to 85% of the passband, without-of-band rejection greater than 90 dB. The fourdifferent digital filters, at ±4096, 2048, 1024, 512 Hz,use between 128 and 512 taps. From there, a time/frequency spectrogram is performed on the data, witha variety of time/frequency resolutions possible. Thespectrogram, which is displayed in real time on a PC,is described next.

Spectrogram Processing and DisplayBriefly, a spectrogram is a specific type of time/

frequency distribution based on the periodogram, withextensions built in for time resolution. Many types oftime/frequency distributions have received much at-tention; however, for real-time display, none can com-pare to the fast Fourier transform (FFT)-based distri-bution in terms of computational speed.

Figure 5 is a block diagram of overlapping FFTprocessing known as the spectrogram. After the digitalreceivers, the complex data are buffered. This allowsthe FFT to be overlapped by anydesired amount. Three parameterscontrol the spectral processing. (1)FFTsize is the size of the FFTs used(typically 256, 512, or 1024points). (2) FFTdsize is the datasize portion of the FFTsize. Typical-ly, this is at most the FFTsize or 1to 2 powers of 2 smaller. This pa-rameter, along with the samplingrate and the specific digital filterused in the digital receiver, willestablish the spectral resolution. If

I/Qdata

WData buffering

several secondsof data

FFT

FFTndsize Possiblezero padding

FFTsizedata

points

FFTdsizedata

points

Figure 5. Block diagram of spectrogresponse).

303

AIRCRAFT WAKE DETECTION

a windowing function is used, itmay also affect the resolution. (3)FFTndsize is the new data sizeportion of the FFTdsize for FFTprocessing, which will typically beat most the FFTdsize or perhaps 1to 4 powers of 2 smaller. This pa-rameter will establish the time res-olution of the spectrogram. In ad-dition to these parameters, thenoise floor parameter can be set tobetween 210 and 2150 dB, andvariable infinite impulse response-type averaging can be performedon the postmagnitude operation.

Figure 3 is an example of ther run 9 on 29 October, with FFTsize = = 512, and FFTndsize = 128 in a digital±5000 Hz. With these parameters set,anning windowing scheme that reducessolution, the spectral resolution is ap- Hz. The time resolution is approxi-

since this is the amount of spectralt occurs between successive FFTs. Ingresses downward, where extent of the

unction of the number of graphics pixelswindow. With a 1024 3 768 display

window yields approximately 78 s of viewable spectrogram data. The inter-e data will be discussed in the Results

cessinger from the transmitter to the receiverrs of magnitude larger than any atmo-ed signal reflected down from the com-

his makes any detections close to therequency difficult to obtain. To cancelllover, the reference channel is used tocel anything that is common. Illumina- is common to both channels and thuswhereas the vortex signal, which is nothe reference channel, should become

20 log ( )IIRfilter

indowing

Averagetime constant

Lowerlimit

Real dataComplex data

Output FFT line(FFTndsize * sampleperiod in seconds)

ram FFT processing (IIR = infinite impulse

R. J. IANNUZZELLI ET AL.

more apparent after adaptive pro-cessing. The formulation for thisadaptive process assumes that themain vortex channel has two com-ponents: the common volumecomponent and the direct spill-over component. The referencechannel is first phase-shifted andscaled to match the zero-Dopplermagnitude and phase of the vortexchannel peak, and then subtract-ed. The subtraction is performed inthe frequency domain before themagnitude is taken. This has typ-ically reduced the power in thezero-Doppler component in themain vortex channel by as much as30 dB. Figure 6 is a block diagramof the adaptive processing, whichcan be contrasted to the spectrogram processing de-picted in Fig. 5. Many runs were looked at with thisprocess, and for those of importance on 29 October,adaptive processing was not needed since the returnswere quite large.

THE BWI EXPERIMENTBoth the Maryland Aviation Administration and

the FAA where extremely helpful and receptive to theidea of allowing APL to install and operate the bistaticradar on the south side of runway 33L (Fig. 7). Asmentioned previously, because of the proximity of thetest site to a highway, wire screen fences (Fig. 8) wereinstalled around both locations. This cut down reflec-tions from cars as well as the direct path signal fromthe transmitter to the receiver.

The traffic interference can be recognized in water-fall spectra as a double S-shaped curve with an inflec-tion point at zero Doppler (Fig. 3). This region of zeroDoppler occurs when the vehicle is midway betweenthe transmitter and receiver and the rate of change ofthe sum of the ray paths from the transmitter and thereceiver to the vehicle is zero. At times prior to thisthere is a decreasing positive rate of change (positiveDoppler); after the midway point there is an increasingnegative Doppler. The maximum positive and nega-tive velocity is limited to that of a monostatic radarwith a completely radial component. Traffic interfer-ence can be seen in the actual waterfall data of Fig.3 as depicted in Fig. 9.

Initial checkout of the system consisted of antennaalignment testing followed by measurement of receiversignal levels in clear air, both with and without acous-tic pumping. Also, with the runway out of service, theacoustic transmitter was moved to several locationsalong its centerline, and receiver signal levels were

VortexI/Q

data

Data bufferingseveral seconds

of data

FFTndsizez

FF

p

Data bufferingseveral seconds

of data

ReferenceI/Q

data

Figure 6. Block diag

304

20 log ( )IIRfilter

WindowingFFT

Possibleero padding

Averagetime constant

Lowerlimit

Real dataComplex data

FFTsizedata

points

Tdsizedataoints

Output FFT line(FFTndsize * sampleperiod in seconds)

FFT

Reference channelphase-shifted and

subtracted from mainvortex channel

ram of adaptive spectrogram FFT processing.

monitored. During this phase of testing, the acoustictransmitter consisted of a Peavy 44T compression driv-er loaded with a 60° horn, which was typically drivenwith less than 20 W. A digital synthesizer was used togenerate a sinusoidal input to the amplifier, whichcould be varied in frequency.

Tests confirmed that the optimal orientation of theacoustic source was vertical and directly in line be-tween the transmitter and receiver along the center-line of the runway; i.e., at about 79 m from the southend of the runway. Unfortunately, commercial aircraftare at a very low altitude as they pass over this point,perhaps ≈50 m or less. At this altitude, a vortex canbe unstable and dissipate quickly. To allow better con-trol of the test environment, arrangements were madefor a NASA C-130 aircraft from Wallops Island, Vir-ginia, to overfly the runway on several days. Theseflights were funded by NASA as a contribution to theAPL research efforts.

Test AircraftThe C-130 aircraft was equipped with smoke

generators (smokers) installed on both wingtips. Thesmokers burned corvus or canopus oil at the wingtip,which caused the smoke to became entrapped in thevortex core (see Fig. 10). The entrapped smoke al-lowed visual tracking of the otherwise invisible vortex.The right-tip smoker was inoperative for the first twotest dates owing to a defective switch in the cockpit;however, both tip smokers were operational during thethird and final test date.

Pretest instructions required the C-130 crew to flythe glide slope until reaching the requested altitude,which was to be held until the aircraft passed 305 mbeyond the runway threshold. To conserve the corvusoil, the smoke was to be turned on 305 m prior to the

JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 19, NUMBER 3 (1998)

runway threshold and turned off 305 m beyond thethreshold. A NASA video of the C-130 taken duringa previous test at Wallops Island revealed that the tipvortices would break up if the landing flaps were used.Therefore, the C-130 was instructed to fly by the APLtest sight in a “clean” configuration (gear and flaps up)at a speed of 1.3 Vso (120–130 kt). (Vso = stall speedor minimum steady flight speed at which the aircraftis controllable in the landing configuration.) At thisspeed and at maximum fuel weight, the strength of thevortices generated by the C-130 equaled that gener-ated by a Boeing 727.

Runway 33L

MD Rte. 176 Receiver Transmitter

(a)

(b)

TransmitterReceiver

Radar fence

Radar fence

15–120-mcommonvolumeheight

Acousticpumpingdevice

600-mtransmitter/receiver

separation

Touchdown area

Runway33L

Figure 7. BWI vortex detection test site on runway 33L: (a) aerial photograph and (b)schematic view.

Figure 8. Wake vorte

JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 19, NUMBER 3

AIRCRAFT WAKE DETECTION

During the test, the C-130 re-ceived instruction via a UHF radiothat was based at the radar receiverlocation. APL would transmit al-titude and lateral displacement in-structions to the crew before eachflyby. The co-pilot received in-structions from the APL test site,while the pilot communicatedwith the control tower at BWI.

Data Collection ProceduresA typical run of the C-130 in-

volved at least six people: four atthe receiver container, one at thetransmit container, and one at theacoustic site. One person (localcontroller) was in voice communi-cation with the C-130 at all times.The local controller and at leastthree other people were at the re-ceiver container and would specifya height above the runway anddistance from side to side offsetfrom the runway to the crew. A“mark” would be given to the twoor three people manning the re-ceiver equipment that the C-130was about 305 m out. Data collec-tion usually started at this point.Visual contact was also made bythe person voice-annotating theMetrum data. The receiver con-tainer, the transmit container,and the person manning theacoustic illuminator were all inradio communication. The personat the acoustic setup, which wasin-line with the runway, wouldgive a mark when the C-130 wasover Dorsey Road and then when

x receiver at BWI.

(1998) 305

R. J. IANNUZZELLI ET AL.

S

306 JOHN

over the common volume. At this point, most of thevoice annotation centered on the vortex smoke trailsas seen by the receiver video camera. When the trans-mit container saw smoke in the transmit monitor, radiocontact was made and notes taken at the receivercontainer specifying the degree of visual collaborationof smoke in the cameras at the receiver and transmitsites. This procedure worked well in establishing whichof the many data runs were worth further investigation.

Observed Vortex BehaviorThe behavior of the wingtip vortices was video-

taped during the three test periods. Near-surface tem-perature and wind data were recorded at the radarreceiver site. Recording devices were not allowed nearor above the runway surface. Operations during theevening hours provided the most stable atmosphericconditions for the first test period. Because of the timechange to Eastern Standard Time after the first testsand the heavy airline activity, the second and thirdtests were conducted during the morning and after-noon hours. A general description of the wind condi-tions and the vortex behavior during each test periodis given in the following paragraphs.

The first test began at 5:20 p.m. on 23 September.Winds were 5 to 10 kt along the runway centerline.The C-130 made 20 passes by the APL test site ataltitudes ranging from 15 to about 107 m above groundlevel (AGL). As noted earlier, the left wingtip smokerworked, but the right tip had malfunctioned. The left-tip vortex lasted more than 2 min after the C-130 flyby.The vortex remained near the runway centerline atapproximately 12 m AGL. In some cases, the vortexpassed through the boresight camera’s view and wouldreappear shortly thereafter, giving the false impressionthat the vortex was rising. Instead, it remained nearabout 12 m AGL but underwent a wavelike oscillation.The peak of the wave motion reappeared in the bore-sight camera’s view.

The first test date provided the best atmosphericconditions for vortex longevity, but posttest analysis

Tim

e (s

)

40

1500

20

0

–20

–4010005000–500–1000–1500

Doppler frequency (Hz)

Figure 9. Interference shown for a vehicle traveling 40 mph onRte. 176.

HOPKINS APL TECHNICAL DIGEST, VOLUME 19, NUMBER 3 (1998)

determined that the radar’s common volume was toolow and the local traffic on Dorsey Road caused toomuch interference in the radar data.

The second test started at 9:50 a.m. on 29 October.Again, the right-tip smoker had malfunctioned. Ini-tially, winds were approximately 15 to 20 kt across therunway. In some cases, the C-130 was instructed to flyat two wingspans about 80 m to the right of the runwaycenterline in hopes that the vortices would passthrough the radar’s common volume. After six passes,the winds decreased to light and variable and theC-130 was instructed to fly over the runway centerline.Fifteen passes were made by the C-130 until the BWItower terminated testing owing to increasing commer-cial air traffic. The second half of this test provided thebest radar data.

The third and final test date was 10 December. Themalfunction in the right-tip smoker had been correct-ed, thereby allowing the observation of the interaction

(a)

(b)

Figure 10. The NASA C-130 research aircraft with generator onleft wing. (a) Trailing vortex seen on approach to BWI. (b) Smoke-marked vortex as seen several seconds after the C-130 passedby (a).

between the left and right vortices. The C-130 wascommitted to a full day of testing, which included alanding at BWI with the smokers on. Wind conditionswere less than ideal, with estimated crosswinds atvelocities of 20 to 25 kt. Lateral positioning of the C-130 was difficult owing to the variation in the windvelocity and direction. Even under these conditionsthe tip vortices would remain intact for over 1 min.The vortices appeared in the boresight camera’s viewfor only a few seconds because of the lateral displace-ment caused by the crosswinds.

Testing on 10 December began at 10:19 a.m. TheC-130 made 13 passes before the BWI controllers ter-minated the test because of an increase in commercialair traffic. At this time, the C-130 landed for refuelingwith the smokers on. The tip vortices were destroyedwithin two C-130 body lengths, with the landing flapsfully deflected.

Afternoon testing began at 1:35 p.m. The windshad not subsided, and lateral positioning remained aproblem. The radar’s common volume was moved inan attempt to track the vortices as they were displacedby the crosswinds. At 3:00 p.m., the C-130 made its40th and final pass of the afternoon and returned toWallops Island.

RESULTSNo significant Doppler signatures could be correlat-

ed with the times and runs when smoke was seen inthe video cameras on 23 September or 10 December.On the earlier date, the common volume was at 15 m,which we later realized was a poor choice because ofground effects. On 10 December, the winds made itdifficult to impossible to position the C-130 so thatsmoke would drop through the common volume.

Two data runs taken on 29 October are detailed inthe following paragraphs. Both proved to have defin-itive Doppler signatures when smoke was observed tobe dropping through the common volume. (A moredetailed presentation of the data from all the runs maybe found in Ref. 11.)

The radioacoustic return showed a large signaldropout correlated with the arrival time of the de-scending wake (cf. Fig. 11a). The major qualitativedifference between these results and the simplifiedpicture presented earlier is likely due to a combinationof effects associated with three phenomena:

1. Wake turbulence. Fluctuations in velocity and tem-perature distort the acoustic wavefronts, scatteringthe acoustic power and causing a loss of coherence inthe wavetrain.

2. Cross-currents. Velocity components perpendicularto the acoustic axis tilt the Bragg grid, changing itsgeometry with respect to the RF beams, and possiblyde-tuning the system.

JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 19, NUMBER 3 (

AIRCRAFT WAKE DETECTION

3. Vertical shear. The presence of shear means thatwaves are advected by differing amounts in differentparts of the wavetrain. When the acoustic axis isvertical, the fractional change in wavenumber alonga central ray is directly proportional to ∂ w/∂ s, i.e.,

∂ ∂ − ∂ ∂( k)/ s = w s v + wln / /( ) ,s (12)

where s is the distance along the ray. The resultingrefraction is strongest in the core region of the vorticesand would tend to de-tune the system when the coresoccupy the common volume.

Recall Fig. 3, a spectrogram of the data from themain vortex channel from run 9 (time running fromoldest to newest, top to bottom). Figure 11b is a re-duced bandwidth view of the same run. Here the fre-quency resolution is approximately 1 Hz as opposed to15 Hz in Fig. 3. Marks have been added to the time axiswhere the smoke was observed in the cameras.A Doppler event within these marks is evident and isbelieved to represent the drop rate of the vortex core.The positive side of the Doppler spectrum correspondsto the “drop side” Doppler, i.e., the path length is beingshortened. Calculations of the drop rate from the re-ceiver video after the fact seem to generally agree withthe drop rate indicated by the spectrogram and with theresults of the numerical model calculations (cf. Fig. 15).Knowing the video camera’s field of view and theapproximate distance to the smoke, the video drop ratewas estimated to range from 1.3 to 3.5 m/s. This cor-responds to a Doppler frequency at the specific X-bandfrequency used of approximately 30 Hz (2 m/s).

Figure 11c is a cut through the spectrogram at thetime when the vortex signature was the strongest.From this figure, the estimated Doppler offset is ap-proximately 40 Hz. Two other traces are depicted here,attempting to place bounds on Cn

2 using differentmeasurement methods. The top trace (upper bound)assumes that the full common volume (about 1000 m3)is contributing to Cn

2 , and the energy is concentratedin the narrow frequency band of the vortex signalshown in the figure. The bottom trace (lower bound)assumes that the received energy is spread over thefrequency band corresponding to the estimated turbu-lent velocities contained in the vortex. The figureshows that Cn

2 is several orders of magnitude greaterthan the upper bound, suggesting that vortices mayhave Cn

2 values much greater than expected forstrong, clear air turbulence. Another explanation isthat the smoke contained in the vortex is contributingto overall vortex reflectivity. Compare Fig. 11c withFig. 11a, a peak amplitude versus time plot for theacoustic line. Here, the dropout from the passage of theplane is apparent; however, little else could be countedon because of the wide swings in received power.

1998) 307

R. J. IANNUZZELLI ET AL.

Iannuzzelli-11a

45

–30

–40

–50

–60

–70

–80Aco

ustic

tone

spe

ctra

l pow

er (

dB)

4035302520151050–5Elapsed time from C-130 overflight of common volume (s)

(a)

“Zero Doppler”spillover

Vortexsignature

T = 63.69 s

(b)

Spe

ctra

l pow

er (

dB)

–100

–80

–60

–40

–20

150100500–50–100

SpilloverVortex signal

2.2 m/s

Phase noise

Doppler frequency (Hz)

C2 = 10–12 fullvolume; power across

measured band

n

C2 = 10–12 vortex volume;power across

vortex circulation

n

(c)

Figure 11. Run 9 on 29 October 1996: (a) acoustic line power, (b) spectrogram, and (c) single line through spectrogram.

Figure 12 shows the acoustic amplitude, spectro-gram, and single line spectrum for run 15. Again, thereader will find details on additional runs in Ref. 11.

MODEL PREDICTIONS OF AIRCRAFTVORTEX WAKES

Model calculations have been carried out for runs9 and 15 from 29 October. Our objectives were (1) todetermine the extent to which the model could predictthe early evolution (prior to breakup) of aircraft vor-tices and (2) to help interpret the radar measurementsdescribed in previous sections.

308 JOHN

Obviously, some success in model validation is re-quired if the model is to be used to provide estimatesof the location, strength, and internal structure of theatmospheric flow signatures believed to be responsiblefor the measured radar returns. Accordingly, our ap-proach was to first examine the predicted wake vortextrajectories and then to compare the predicted time ofarrival of the vortex cores at the radar common volumewith the time at which radar signatures were observed.Since no ground-truth measurements of atmosphericmotions were available during the BWI experiment,these were the only comparisons that could be used tohelp establish the model’s credibility. The next step in

S HOPKINS APL TECHNICAL DIGEST, VOLUME 19, NUMBER 3 (1998)

AIRCRAFT WAKE DETECTION

–30

–40

–50

–60

–70

–80Aco

ustic

tone

spe

ctra

l pow

er (

dB)

“Zero Doppler”spillover

Vortexsignature

T = 31.85 s

Spe

ctra

l pow

er (

dB)

–100

–80

–60

–40

–20

150100500–50–100

Spillover

Vortex signal~4.8 m/s

Phase noise

Doppler frequency (Hz)

200

C2 = 10–12 vortexvolume; poweracross vortex

circulation

n

C2 = 10–12 fullvolume; power

acrossmeasured band

n

403020100

Elapsed time from C-130 overflight of common volume (s)

(a) (b)

(c)

Figure 12. Run 15 on 29 October: (a) acoustic line power, (b) spectrogram, and (c) single line through spectrograph.

the analysis process was to examine features of thepredicted vortex wakes (e.g., overall size, internalstructure) to better understand what parts of the sig-nature might be responsible for the measured radarreturns.

Predictions of vortex wake evaluation were ob-tained using the recently developed APL Wake Eval-uation Code (AWEC). AWEC solves steady-state non-linear equations for the velocity

rV , density r,

turbulence kinetic energy K, and turbulence dissipa-tion rate e, all in a coordinate system that is fixed withrespect to a moving source (in this case, an aircraft)

JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 19, NUMBER 3 (1

and moves linearly with that source. A second-orderturbulence closure scheme, based largely on the Mellorand Yamada Level 2.5 model,12 is used to evaluate theremaining turbulence correlations.

The principal approximation in AWEC, and theone that renders the problem tractable, is the so-calledparabolic approximation. The physical assumptionbehind this approximation, which is valid for almostall wake flows, is that changes in the downtrack or xdirection are small compared with changes in thecrosstrack and vertical (y and z) directions. Derivativesof x other than those in advective terms involving the

998) 309

R. J. IANNUZZELLI ET AL.

mean speed of advance U are neglected, thereby allow-ing the equations to be reformulated and solved as amarching problem in x. The need to store modelvariables at locations along the x axis is eliminated,thus reducing the problem from three to effectivelytwo spatial dimensions.

To initialize AWEC model calculations, estimates ofdata are needed at a vertical (y–z) plane close to theorigin. For submarine wake predictions, a separateprogram, FWI,13 is used to generate estimates of com-ponent wakes for the many sources that contribute tothe composite wake signature. FWI stores relevantgeometrical variables for different submarines and usesthese values along with operating conditions providedby the user to compute estimates for variables such asthe size and strength (circulation) of vortices shed bythe hull and control surface. For the present aircraftvortex calculations, it was a fairly straightforwardexercise to add a file containing the wingspan andother relevant geometrical variables for a C-130 trans-port and to use existing software to calculate the re-quired initial plane data. Since the inviscid theoryapplied to derive the vortex circulation equations inFWI was originally developed for the aircraft vortexproblem, there was little risk in applying those sameequations to the problem at hand. And since, unlikesubmarine wakes, aircraft wakes are dominated by asingle component, all sources other than wing vorticescould safely be ignored.

Run 9During run 9, the aircraft was assumed to be in level

flight at an altitude of 83.8 m, with an air speed U of64.3 m/s. The plane’s weight was taken to be 143,000lb (64,865 kg), and the lift force L was assumed equaland opposite to the weight. The wingspan B was about40.4 m.

The initial vortex circulation G0 was estimatedfrom the inviscid theory. Assuming elliptical loading,

G00

2312=L

Ub= /

rm s , (13)

where the density of air r = 1 kg/m3, and the vortexcore separation b0 is given by

b = B =0 431 7

p. m . (14)

The diameter of the vortex core DV is given theoret-ically by

D = bV . ,0 197 0 (15)

310 JOHN

which yields an estimate of DV = 6.25 m for theC-130. This value was considerably larger than theobserved core diameter of approximately 1 m. Thelatter value was determined by the visible boundary ofthe entrained smoke in the photographs. Despite thisdiscrepancy, the larger theoretical value was used inthe model calculations. Use of the smaller value wouldhave required a minimum grid separation distancemuch smaller than the 0.5 m used in the presentcalculations, which would have resulted in a nearlyprohibitive increase in the computational burden.

The net downward vertical velocity w0 induced byone vortex acting on another is given by

w =b00

021 57

−= −

Γp

. m/s . (16)

The atmosphere was assumed to be weakly stratifiedwith a potential temperature gradient of 6.7 3 1023

°C/m. Values of potential temperature were convertedto density for use by AWEC. Ambient levels of turbu-lence kinetic energy and turbulence kinetic energydissipation rate were chosen to be 0.1 m2/s2 and 2 31023 W/kg, respectively.

Figure 13 depicts vector velocities at times late t of1.4 (the initial plane), 9.0, 21.0, and 31.0 s. Since thewing was positioned near the top of the fuselage (ef-fective diameter of approximately 5.2 m) on theC-130, the vortex cores initially were situated 86.4 mAGL. Figure 14 shows contours of the turbulencekinetic energy for two of these times late, 1.4 and 31.0.The position of the radar common volume, which iscentered at 45.7 m AGL, is also shown.

Figure 15a plots the position of the vortex cores(solid curve) as a function of t. The heavy dashed lineindicates where the vortex positions would be if thewake descended at a constant velocity given by theinduced velocity computed from Eq. 16. The actualvortex descent begins near this speed but then slows.The fixed upper and lower boundaries of the commonvolume are indicated by a pair of solid horizontal lines.The outlying lighter dotted lines indicate the firstcontour level above ambient of the turbulence kineticenergy distribution.

From Fig. 15a it is estimated that the leading edgeof the vortex wake first reached the common volumeapproximately 12 s after aircraft passage and that thevortex cores were within the common volume betweent ≈ 29 and 33 s after aircraft passage. Radar results dis-cussed previously show a distinct signature at t = 31 s,which appears consistent with the model prediction.This result, however, could be fortuitous as it dependson the aircraft altitude at the time of passage, and thereis considerable uncertainty in the estimates of altitudeobtained from the C-130 pilot.

S HOPKINS APL TECHNICAL DIGEST, VOLUME 19, NUMBER 3 (1998)

AIRCRAFT WAKE DETECTION

–60 –40 –20 0 20 40

120

100

80

60

40

20

0

Hei

ght (

m)

Crosstrack (m)–60 –40 –20 0 20 40

120

100

80

60

40

20

0

Hei

ght (

m)

Crosstrack (m)

–60 –40 –20 0 20 40

120

100

80

60

40

20

0

Hei

ght (

m)

Crosstrack (m)–60 –40 –20 0 20 40

120

100

80

60

40

20

0

Hei

ght (

m)

Crosstrack (m)

(a) (b)

(c) (d)

Figure 13. Run 9 vector velocities at times late of (a) 1.4 s, (b) 9.0 s, (c) 21.0 s, and (d) 31.0 s.

Run 15Except for the altitude, all model inputs for run 15

were the same as those for run 9. Although the re-ported altitude was 64 m, a video tape of the aircraftflying over the common volume shows that the vortexcores as measured by smoke are in the common volumeat the time of passage. Accordingly, an initial vortex-core height of 53.3 m was adopted for the run 15 modelcalculation, which resulted in better agreement withexperimental data.

Figure 15b plots the position of the vortex cores asa function of time. Again, the actual vortex motiondescent rate is slower than predicted by inviscid theory.Vortex cores are predicted to be within the common

JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 19, NUMBER 3 (

volume between 6 and 12 s after aircraft passage. Thisagrees with the radar observations—an expected resulthere since the radar data along with the video wereused to estimate the initial altitude.

DISCUSSIONRadio waves are scattered by turbulent irregularities

in the refractive index of air. In the initial stages of aturbulent mixing process, air parcels are rapidly dis-placed from each other and their identities are pre-served. It takes a while for diffusion to act and for themixed region to become nearly homogeneous again.The stronger the initial gradients and the turbulence,

1998) 311

R. J. IANNUZZELLI ET AL.

–60 –40 –20 0 20 40

120

100

80

60

40

20

0

Hei

ght (

m)

Crosstrack (m)60

(a)

–60 –40 –20 0 20 40

120

100

80

60

40

20

0

Hei

ght (

m)

Crosstrack (m)60

(b)

Figure 14. Contours of turbulence kinetic energy for run 9 attimes late of (a) 1.4 s, and (b) 31.0 s.

the larger are these temporary local variations in therefractive index.

The physical variable most identified with radarsignatures is the structure parameter Cn

2 discussed pre-viously. To effectively use the model predictions tointerpret the radar observations, a means must befound to relate Cn

2 to the mean-flow and turbulencevariables predicted by AWEC.

An algebraic expression for Cn2 in terms of selected

AWEC model variables has been derived for a specialcase of zero moisture content in the atmosphere,11

C vy

wz1/3n

a ( . R),2

2 2

4

77 6

10= − ′ ′ ∂

∂− ′ ′ ∂

∂

«

rr

rr

(17)

312 JOHN

Figure 15. Vortex trajectories for (a) run 9 and (b) run 15. Theposition of the vortex cores (solid curve) is plotted as a function oftime late. The heavy dashed line indicates where the vortexpositions would be if the wake descended at a constant velocitygiven by the induced velocity computed from Eq. 16. The actualvortex descent begins near this speed but then slows. The fixedupper and lower boundaries of the common volume are indicatedby a pair of solid horizontal lines. The outlying lighter dotted linesindicate the first contour level above ambient of the turbulencekinetic energy distribution.

where « is the turbulence dissipation rate, ′ ′r v and′ ′r w are turbulence density-velocity correlations, and

∂ ∂r / y and ∂ ∂r / z are gradients of the mean potentialdensity14 in the horizontal (cross-track or y) and ver-tical directions, respectively. Constants appearing inEq. 17 are the universal gas constant R and a2, whichranges between 0.4 and 4. A value of a2 = 1.0 waschosen for estimating Cn

2 in this study.Figure 16a shows that the wake boundary as deter-

mined by the structure parameter roughly coincideswith the outer contour of turbulence kinetic energy forrun 15, which is shown for comparison in Fig. 16b.Unlike the turbulence kinetic energy, however, theplot of Cn

2 shows evidence of structure within thevortex wake. Values of Cn

2 range from 230 to 270 dBover the extent of the recirculation cell, with the

6040200Time (s)

120

100

80

60

40

20

0

Hei

ght (

m)

6040200Time (s)

120

100

80

60

40

20

0

Hei

ght (

m)

(a)

(b)

S HOPKINS APL TECHNICAL DIGEST, VOLUME 19, NUMBER 3 (1998)

AIRCRAFT WAKE DETECTION

largest values occurring near the vortex cores. The plotof Cn

2 also shows a well-defined leading edge of thewake.

It is unclear whether mean velocities (as measuredby the vector plots) or turbulent perturbations to therefractive index (as measured by the structure param-eter) are primarily responsible for the observed radarsignatures. The coincidence of regions of high velocitywith regions of high Cn

2 in the model predictionsmakes it difficult to use model predictions to resolvethis issue.

The interpretation of the radar measurements isalso made difficult by the absence of ambientcrosstrack wind data. The model predictions, whichassumed no ambient wind, show most of the vortexwake (including both cores) passing through the radarcommon volume; however, even a modest crosstrackwind could cause an offset that could result in only oneof the cores being detected. In the former case, thelarge rotational velocities would likely be averaged

120

100

80

60

40

20

0

Hei

ght (

m)

–60 –40 –20 0 20 40 60

Crosstrack (m)

–30

–40

–50

–60

–70

Cn (dB)2(a)

–60 –40 –20 0 20 40

120

100

80

60

40

20

0

Hei

ght (

m)

Crosstrack (m)

60

(b)

Figure 16. Run 15 at 9.0-s time late: (a) gray-scale plot of Cn2 and

(b) contours of turbulence kinetic energy.

JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 19, NUMBER 3 (

out, leaving only the much smaller (≈1 m/s) net down-ward-migration velocity to be measured. In the lattercase, the radar could be detecting the composite of thevortical velocity and the migration velocity, which,according to the model predictions, could be as highas 7.4 m/s in regions where the two components add.Clearly, future experiments would benefit by directmeasurements of the atmospheric variables of interest.

SUMMARYAPL has designed and constructed a bistatic con-

tinuous-wave radar system for experimentation in air-craft wake vortex detection. Many hurdles, both tech-nical and logistical, were overcome in developing thefacility. Acoustic pumping was evaluated as a possibletechnique to enhance overall detection performance.The radar and acoustic systems were installed at BWIin the fall of 1996, and controlled experiments usinga NASA C-130 aircraft were conducted.

From these experiments, Doppler signatures fromthe main vortex signal were found in several data runs.The acoustic signal showed amplitude dropouts thatcoincided with the passage of the aircraft through thecommon volume and could be used for a coarse, butunreliable, detection of vortex passage. We speculatethat with the wider-band acoustic illumination, Dop-pler phase modulations will also be visible.

APL has carried out a number of wake vortex sim-ulations in support of these experiments. Model resultshave been used to help interpret the data observations.Predicted vortex trajectories were found to agree withthe data when uncertainties in the aircraft altitudewere taken into account. The model predictions havealso provided insight into which part of the overallwake signature might be producing the observed radarreturns.

The main result from these experiments is thatvortices are detectable with the APL X-band bistaticradar system. In addition, given the tremendous vol-ume that must be scanned in an operational environ-ment, the current system parameters (power, resolu-tion) appear to be more than suitable for a fieldablevortex detection radar.

One possible future direction would be to add theability to track vortices in and around the end of therunway. This, along with a search mode, would berequired to make the current radar a valuable sensorin an operational environment. Another possibilitywould be to add a millimeter-wave radar, since mod-eling has shown that it may be possible to observeintercore velocities with shorter wavelengths. Finally,we would like to extend our experimentation withvarious acoustic waveforms and techniques to furtherour understanding of the radioacoustic effects.

1998) 313

R. J. IANNUZZELLI ET AL.

CONTRIBUTORS

REFERENCES1Hanson, J. M., and Marcotte, F. J., “Aircraft Wake Vortex Detection Using

Continuous-Wave Radar,” Johns Hopkins APL Tech. Dig. 18(3), 348–357(1997).

2 Doviak, R., and Zrnic, D., Doppler Radar and Weather Observations, 2nd Ed.,Academic Press, San Diego, CA (1993).

3Measurement of Atmospheric Parameters with a Bistatic Radar Supplemented byan Acoustic Source, Environmental Surveillance Technology Programme,Oslo, Norway (Oct 1995).

4Stewart, E. C., A Study of the Interaction Between a Wake Vortex and anEncountering Airplane, AIAA Paper 93-3642, Atmospheric Flight MechanicsConference, Monterey, CA (Aug 1993).

5 Smith, P. L., Jr., Remote Measurement of Wind Velocity by the ElectromagneticAcoustic Probe, Midwest Research Institute, Kansas City, MO (1961).

6 Marshall, J. M., Peterson, A. M., and Barnes, A. A., Jr., “Combined Radar-Acoustic Sounding System,” Appl. Opt. 11(1), 108–112 (Jan 1972).

7 Frankel, M. S., and Peterson, A. M., “Remote Temperature Profiling in theLower Troposphere,” Radio Sci. 11(3), 157–166 (Mar 1976).

8 Kon, A. I., “A Bi-Static Radar-Acoustic Atmospheric Sounding System,”Izvestiya (Atmospheric and Ocean Physics) 17(6), 481–484 (1981).

9Kon, A. I., “Signal Power in Radioacoustic Sounding of a TurbulentAtmosphere,” Izvestiya (Atmospheric and Ocean Physics) 20(2), 131–135(1984).

10Kon, A. I., “Combined Effect of Turbulence and Wind on the SignalIntensity in Radioacoustic Sounding of the Atmosphere,” Izvestiya (Atmo-spheric and Ocean Physics) 21(12), 942–947 (1985).

11Iannuzzelli, R. J., Marcotte, F. J., Hanson, Jr., J. M., Manzi, L. P., Schemm,C. E., et al., Final Report: Results of the APL Bistatic X-Band Vortex DetectionRadar from C-130 Flyovers at BWI Airport, JHU/APL, Laurel, MD (1997).

12Mellor, G. P., and Yamada, T., “Development of a Turbulence ClosureModel for Geophysical Fluid Problems,” Rev. Geophys. Space Phys. 20, 851–875 (1982).

13Rosenberg, A. P., FASTWAKE Initialization and Flow Evolution ProgramUser’s Guide, STD-R-1577, JHU/APL, Laurel, MD (Nov 1987).

14Gill, A. E., Atmosphere-Ocean Dynamics, Academic Press, New York (1982).

ACKNOWLEDGMENTS: The authors gratefully acknowledge thecontributions of Robert Neece and Ann McKenzie, NASA, Langley,C-130 funding; Doug Young and the C-130 crew at NASA, WallopsIsland; Ben Martinez, MAA, BWI; and Jerry Dudley, FAA, BWI.The BWI experiment would not have been possible without thediligent efforts of Bernie Kluga and Bob Geller and the air trafficcontrollers of the FAA at the BWI Tower.

(Back row, left to right) David A. Frostbutter, William R. Geller, Laurence P. Manzi.(Front row, left to right) Charles E. Schemm, Harold E. Gilreath, Russell J. Iannuzzelli, JamesM. Hanson, Jr., Frank J. Marcotte, Dennis L. Kershner, Alexander S. Hughes, Allen J. Bric,Bernard E. Kluga. Leo E. McKenzie is not pictured.

314 JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 19, NUMBER 3 (1998)