A STUDY OF THE EFFECT OF TERRAINONHELICOPTER NOISE PROPAGATIONBY

ACOUSTICAL MODELING

FINAL TECHNICAL REPORT

HARRY STERNFELD, JR.

MARCH 23, 1981

U,S. ARMY RESEARCH OFFICE

CONTRACT DAA629-78-C-0002

BOEING VERTOL COMPANY

.> -

APPROVED FOR PUBLIC RELEASE.DISTRIBUTION UNLIMITED,

, ;.. ,i

I L. ..... .. ... .. ''' ° P".. -T = ,:. ... ...: ;'" ; ' - ..o.. .. ... - ...* ... ,-8 1 4-, 1 3 ., ,.(.' ; :,.;,:-,", ' 3 .

.1!

THE VIEW, OPINION, AND/OR FINDINGS CONTAINED IN THIS REPORT ARE*1 .THOSE OF THE AUTHOR(S) AND SHOULD NOT BE CONSTRUED AS AN OFFICIALI, DEPARTMENT OF THE ARMY POSITION, POLICY, OR DECISION, UNLESS SO

DESIGNATED BY OTHER DOCUMENTATION1

UNCL SSIFIED89CUNIYY Ct.ASSIFICAI@N OF TMIS PACE Xonw Datae. I.0wd)

REPORT DOCUMENTATION PAGE 39FAD 9 COUT04 ORM

4- TITLE find *ube1aI* 3. TYPE OF REPORT A PERIOD O VERED

A Study of the Effect of Terrain onHelicopter Noise Propagation by R.RPOTNME

~- Acoustical Modeling popai m. oayNMC

-7. AUTHORfi) .ONAN P NMAU6RS

Harry St@4ern.eldt Jr/ 9-8CA0 A/(~~ g ') /i jj 0 AAG29-8C0

1.PROMtING 3111111 ATISN NAMR AN* AW 418 TAS

ii. Boeing Vertol CompanyP .0. Box 16858

It. CONTO6NO OPIK NAM6 AND ADORBSS .I

U. S. AnW Research Office A A

Post 0tt1ne Box 12211. 13- M.UMMER O A@ 5

Research Trian le Park NC 27702_____________-T4. XOITRI XM Mi4 Ag;US(p gua11f.ult rMm asnee.II1n 011=6) I.SEURI CLASS. (of this ,p

Approed fr pubic rleas; disriUuionaulsmifded

17. DISTRIBUTION STATEENT IP he .baai 011111" #8 11tek it, It ditffomt hm teRpt)

NA

Is. SUP10.W MNY NOlES iThe iw opinions, and/or findings contained In this report are those of theauthor s) and should not be construed as an official Department of the ArmyPosition, policy, or decision, unless so designated by other documentation.

Ke WOR feenCie n tevewas old* 0I woos wO se identity NY 6108 numso?)

Acoustical Modeling

Noise Propagation

~i ( DetectionZ.' AftNRAOT ?= - mem wo 11 nevis-o WidmIP b look nmmit)

IIA An experimental program was conducted to evaluate the applicability6i of acoustical modeling techniques to study the effects of terrain

on helicopter noise propagation.

Comparison of model results with flyover data of a full scale tIH-1helicopter showed very good correlation with 500 ft altitude dataand moderately good correlation with 50 ft altitude data,----

(over)

DD I 1073 EonwlOFO I Nov ItIsOOLEI UNCLASSIFIED '*O

-&'lh V .- .. .-- - . . --....

20. ABSTRACT (continued)

* Model studies of the effects of the blocking and channelingof sound by barriers, such as hills, shows good correlationwith expected results for simple cases with the model pro-

. viding useful reuults for more complex cases.

IFurther full scale verification is recommended to furtherexchange confidence in the modeling technique for applicationto helicopter detection studies.

L. ,

44

J

I

I. .,,--

TABLE OF CONTENTS

PAGE

1.0 INTRODUCTION 1

f., 2.0 MODEL DEVELOPMENT 2

7 2.1 General Approach 22.2 Sound Sources 32.3 Instrumentation 42.4 Model Construction 5

3.0 MODEL VERIFICATION

3.1 Model Teat 73.2 Correction for Atmospheric Attenuation 73.3 Correction for Directivity 73.4 Validation Test Results 8

4.0 BARRIERS 10

5.0 CHANNELING OF SOUND 14 A

5.1 Line of Sight 145.2 Bends 17

6.0 CONCLUSIONS AND RECOMMENDATIONS 18

REFERENCES 19PERSONNEL 21ILLUSTRATIONS 22

ZNTIS GRA&tDTIC TAB3U

I Juntie,()%ou ...... ....

Avriilabi.liltY d:nAvvlA

ii~L

1.0 INTRODUCTION

The increasing use of helicopters in tactical situations hasled to the development of two types of low altitude operationsreferred to as 'Nap of the Earth' Flying arnd 'Contour Flying'.In contour flying the helicopter operates at altitudes whichare above tree-top level but generally lower than that of themain terrain features such an hills. Nap of the Earth opera-tions are conducted at even lower altitudes where the helicop-ter is often flying below three-top level using the trees andsmall terrain features to hide behind. The purpose of suchtechniques are to minimize radar, visual, and hopefully auraldetection.

In considering aural detection, there are three major elements;noise generation at the source, propagation to the observer,and detection and recognition of the signal. The subject ofhelicopter noise generation is very complex and has been thesubject of considerable study for several years. A comprehen-sive review of the state of this particular art can be found Iin several publications such as Reference 1. Detection of hel-icopters by aural preception, although not studied in as greatdetail as noise generation, had also been the subject of re-search programs such as those reported in Reference 2, 3, 4,and 5.

Up to the present time, the studies of propagation of aircraftnoise through the atmosphere and over terrain have concentratedon airplane noise where the aircraft is operating at substan-tial heights above the ground and Reference 6 describes a gen-erally accepted method for calculating atmospheric attenuationfor this stiuation. The unique problem associated with heli-copter operations is that the noise source (the helicopter) isnot remofe from the terrain but actually operating in it suchthat the ground surfaces can act as acoustical barriers and/orreflectors thereby attenuating or amplifying the noise receivedat the observer compared with that which would be expected ifthere werm only a direct (line or sight) path from source toreceiver. Figure 1 illustrates some of the important effectswhich must be considered in accounting for terrain effectswhen predicting the aural detection of helicopters.

The development of prediction techniques for the effects ofterrain on helicopter sound propagation would be extremelydifficult and costly if one were to attempt the traditionalapproach of correlation of analytical prediction with fullscale test data. Systematic variations in terrain features,such as changing surface impedance without changing the ter-rain geometry, or changing the width of a 'hill' without al-tering its height are virtually impossible. Even the detaileddescription of terrain, to any degree finer than a good con-tour map, would require an inordinate amount of effort and re-sources. It is for these reasons that acoustical modeling

L1

techniques, such as have been employed for architecturalacoustical studies, highway noise studies, and ai.-port noisestudies, presents a unique capability for conducting thedesired investigation.

The objectives of this program were to adapt existing acous-tical modeling techniques so that they would be optimum for

L. application to helicopter noise propagation, verify the appli-cability, and then use the modeling approach to evaluate thesensitivity of the noise propagation to systematic changes !.n'terrain'.

2.0 MODEL flSVELOPMENT

2.1 Oenergl ApDroftch

An acoustical model in a physical model in which the scalingfactor is wavelength. For example, if one constructs a modelwhich is i/X times the size of full scale, then one must usenoise sources whose wavelengths are I/X times that of thenoise source to be studied. Since wavelength is related tofrequency by: "

Where I rFetuenaya m 2.ed of foundi .iavelonth

it is simpler to think of model scaling as a I Xth scale modelrequires noise sources whose frequencies are X times that ofthe full scale noise source. In a similar manner, acousticalproperties of the materials used to construct a model mustalso be frequency scaled. For example, if a full scale mater-ial has an absorption coefficient of .7 at 1000 H2 a 50thscale model material would be required to have an absorptioncoefficient of .7 at 50,000 H2 .

The physical scale of the model is diztated by two parameters:the size of the full scale situation and the room availablefor the model. In the case of this program, studies of heli-copter detection have indicated that the critical range foraural detection ranges from about 1/2 to 2 kilometers, al-though with very quiet ambient conditions certain helicopterscan be detected at considerably greater distances. The iJodelitself may be of any size as limited by practical economLc con-siderations and available room size. As a practical matter,it is desirable to keep the model size to one where all partscould be reached without walking on the model surface and to asize which would not require extraordinary sized rooms, so thatthese technique. could be employed at reasonable cost. It wasdecided, for this study, to limit the model long dimension to16 ft. since the ration of 2 km to 16 ft. is 410 to 1 an acous-tical scale of 400 to 1 was selected for these experiments asbeing typical of that which might be employed for helicopterdetection studies. It should be pointed out that most acous-tical modeling which has been performed to date has employed

2

El

scales of 100 to I or less and in this aspect alone the subjectstudy is pushing the state-of-the-art.

Two types of measurement techniques are generally employed inacoustical modeling. The first technique uses steady statenoise such as a loudspeaker or an air jet as a source and theresultant sound pressure levels are measured at various loca-tions of interest on the model. While this method does pro-vide information on amplification or attenuation of sound, itdoes not give any insight as to the relative contributions frommultiple propagation paths, if they exist. If a very shortduration impulsive source, such as an electric spark is used,

the separate arrivals of the directly radiated and reflectedi time paths can be sensed and separated to provide such infor-mation. Both techniques, as they were applied to this program,will be described more fully in the remainder of this section.22 Sound Sources

A sound source can be described by three primary character-4 istics: its spectrum in the frequency domain, its patterning

in the time domain (steady, random or periodic) and itsspatial directivity. An ideal model noise source would bescaled by the inverse size scale of the model, in the frequen-cy and time domains, and should replicate the directivity ofthe original source. In addition, it may be desirable thatthe physical size of'the model sound source also be somewhatin scale. A helicopter rotor acoustical siqnature is extreme-ly complex with respect to all of the required descriptors.Figure 2 (from Reference 7) depicts several sources of helicop-ter rotor noise classified by their time domain characteristicof periodic or broadband. Figure 3 illustrates the spectraand waveforms associated with several of the periodic sources.Directivity of rotor noise is also a function of the generatingmechanism. Impulsive noise, for example, tends to maximizein, or near, the plane of the rotor while rotational and broad-band sources are greatest below the disk. The cases which areof interest in studying helicopter detection are those whichare generally impulsive in waveform and whose spectra aredominated by harmonic response. Some initial effort was madeto develop a model sound source which would replicate the fullscale signature using a frequently shifted magnetic tape re-cording of an actual helicopter as a source. In the first case,the source was used to drive a condenser microphone as aspeaker. Although fidelity was reasonable, the signal was tooweak for practical use at distances greater than about one foot.Two other attempts were made using pizeaelectric crystals asthe transducer, In one case the vibrating crystal was used todrive a small piston inside a casing and in the other, the ex-posed faces of a 3 inch diameter and a 6 inch diameter crystalwere used as pressure sources, Neither the vibrator nor the 3

.inch crystal provided sufficient acoustical power, and allthree sytems exhibited such marked resonance characteristics

.. .-..... ,

that it became apparent that excessive manipulation of the testresults would be required, and in fact, one would be betteroff with a powerful noise source with a rather smooth spectrum.The data could then be read out at any desired frequency. Sucha procedure is valid if it can be assumed that in detectingsounds people sense on sound pressure level and frequency butnot on the interrelationship between levels at different fre-quencies. Another way of stating the above is that detectionis amplitude and frequency dependent but is not dependent onphasing between frequency components. Fortunately, this turnsout to be the case as was determined by Fidel and Pearsons(Reference 8) and more recently verified by A. Ahumada of theStanford Research Institute who specifically studied the do- itection of the HU-1 helicopter and found harmonic phasing notsignificant. In fact, the procedures for predicting detection(References 9 and 10), which have been vertified by field tests(References 11 and 12), do not consider a spectrum definitionwhich is any finer than one-third octave, or at very low fre-quencies the critical bandwidth of the ear.

It was threfore decided to use a broadband noise source of suf-ficient intensity to do the Job and to read the data at thefrequencies which corresponded to the approrpriate helicopterrotor harmonics. An air Jet, similar to one which had beendesigned at MIT consisting of four impinging nozzles, was con-structed (Figure 4a). When operating with 80 psi of air prom-sure, the level with a useful spectrum leveled out to 80 KX_(shown in Figure 4b). The impulsive source was an electriczspark generated by a Grozier Technical Systems Inc. OTS51 SparkGenerator (Figure 4c) which discharges a high voltage capacitorwhen the electrode gap is ionized by a 30 KV low energy spark.The restullant acoustical signature is an extremely short dur-ation 'N' wave.

2.3 Instrumentation

Data was measured using a Bruel and Kjaer 4135 1/4 inch dia-meter condenser microphone mounted on a Type 2619 cathode fol-lower powered by a Type 2807 Power Supply. The 1/4 inch micro-phone was selected as an optimum considering the trade-offbetween frequency response and sensitivity. Frequency spectrafrom the air-jet source were obtained using a Federal Scien-tific UA-500A "Ubiquitious' Spectrum Analyzer which producesspectra such as that illustrated in Figure 4b. When required,a Federil Scientific Model 30A Range Translator was used toisolate and expand specific portions of the spectrum. Thiswas done not only to obtain finer frequency resolution but topermit greater amplification of the high frequency end of thespectrum without saturating the higher amplitude lower fre-quency data.

While the frequency analysis system used with the air-jet wasrelatively conventional, the analysis equipment used to extractinformation from the spark signal is more specific to this

4

particular measurement. The elements of the system perform .

the following functions: the signal from the microphone powersupply is amplified using a low noise, high gain scaling ampli-fier (Grozier Model OTS20) and the signal amplified an required

to provide an input for a Grozier Model GTS33 Waveburst Processorwhich has the capacity to separate the sequential impulsesthrough variable short term averaging and, if desired, to mea-sure the energy of each impulse. The Waveburst Processor alsoincorporate, a trigger and delay system which is used to acti-vatse the system when the spark is generated but blanks out thereception of the electromagnetic energy associated with thespark generation while admitting the more slowly travelingacoustical energy. The output of the Waveburst Processorwas captured and measured using a Nicolet Model 2090-1I1 'Ex-plorer II' Oscilloscope. This device digitizes the input sig-nal, captures it in a memory, and displays the information ona screen. Digital readouts of the amplitude (voltage) and timecorresponding to any point on the signal can be obtained bysuitable positioning of the cursor(s). The device can also beused to expand the data along either axis to enhance readingaccuracy. Disk storage is also available for preserving therecords. With this device, reading accuracies to .001 milli-second and .1 millivolt are easily obtainable. Typical outputdisplays are shown in Figure 17.

2.4Model Construction

• In order to conduct a successful acoustical modeling experiment,it is necessary to ensure that there are no paths, extraneousto the model, by which acoustical energy can travel from thesound source to the microphone. such paths might include re-flections of room walls and ceilings, lighting fixtures, equip-ment in the room, or the model support structure. This pro-gram was carried out in a test chamber of the Boeing VertolAcoustical Laboratory which is a large (35' x 17' x 20') roomwith all four walls and ceiling lined with an acousticallyabsorptive treatment and a carpeted floor. The chamber alsohas its own air contitioning system incorporating mufflers tominimize ambient noise. The criteria used to declare the roomsatisfactory was that using the spark test technique any re-flection from the room must be at least 20 dB below that ofthe direct path. This test chamber was adequate by itself butwhen the plywood necessary for building the model base was in-troduced, these surfaces were unacceptably reflective evenwhen covered with several layers of very fine filament (PFI05)fiberglass. A satisfactory solution was found by constructinga second inner surface of "foamcore', a paper faced rigid foamspaced by an 'eggbox' separator from the plywood and then liningthis inner surface with 3" layers of 1/2"1 thick fiberglass.Figure 5a illustrates this final configuration the result of aspark test,within the enclosure showed it to be essentiallyanechoic and therefore satiafactory for further testing.Selection of model materials to simulate terrain has not been

-L5

reduced to a rigorous scientific basis and in fact, the bestproof of proper selection probably lies in correlation betweenmodel and full scale data. As discussed in Reference 13, theabsorption of acoustical energy by the ground is probably ofless importance than the impedance which determines the phasingbetween the directly propagated path and the ground reflection.This in illustrated in Figure 6. In order to show the effectsof surface impedance, two materials were selected: the firstwas very high impedance steel plate and the second was a lowimpedance material. In as much a. possible, it was desired tomake this lower impedance material a reasonable scaled approxi-mation for relatively firm ground such as was on the surfaceof the airfield when measurements were made. The absorbtion

A •coeffecient and/or acoustic impedance of ground is very diffi-'r . cult to measure and generally somewhat meaningless since most

natural surfaces are far from homogeneous. Work such as Ref-erences 14 and 15, however, indicate that maximum attentuationsdue to ground effects of the range of 10 dB to 15 dB might bei i expected. An experiment was met up in which the spark was used

as a source and the difference between direct and reflectedpath sound pressure levels measured. Several candidate mater-ials were tested including cork, 1/4" foam, poster board, ply-wood and two types of fiberglass insulating board. Of these+i materials, the one which most closely seatched the desiredcharacteristic was Owens Corning Linear Glass Board. Thismaterial consists of fiberglass fibers with a binding materialwhich forms a moderately rigid board and faced on one sidewith a porous protective facing.

3.0 MODEL VERIFICATION

The model, and modeling techniques, were verified, for applica-tion to helicopter noise studies, by comparison of model andfull scale data of a HU-l helicopter flying over open terrainat altitudes of 50 ft. and 500 ft. In making such comparisons,the following adjustments are made to the model data:

SPL a SPL +

Where I-D

SPL m Corrected Model Sound Pressure LevelcSPLM = Measured Model Sound Pressure Level

a Correction For Atmospheric AttenuationD Correction For Directivity

?.6

3.1 Model Test

Since the terrain being modeled in this case was flat groundwith line of sight propagation path, the model which is illus-trated in Figure 7a was simply a flat surface 4' wide by 16'long thus permitting a full scale approach distance of about6000'. The microphone was mounted at one end of the surfacewith its diaphragm vertical and in the plane determined by the'flight path' and the normal to the gound plane. In this man-ner microphone directivity sensitivity did not vary as duringthe 'flyover'. The microphone center was .15" above the sur-face which corresponded to the 5' height of the full scalemicrophone. It must be recognized, however, that in the model-ing scale being used, the microphone cartridge diameter repre-sented 8.3'. The air-jet was placed at the appropriate heightabove the ground surface and spectra of the type illustratedin Figure 4 were recorded for each of twenty 'aircraft posi-tions' until the source was directly over the microphone.

3.2 Correction for Atmospheric Attenuation

It is well known that as sound propagates throuqh the atmos-phere its sound pressure level in the far field decreases by 6d3 per doubling of distance due to simple spreading of theenergy over the surface of an expanding sphere plus additionallosses due to heat conduction and viscosity (classical absorp-tion) and to molecular abuorbtion which results from a vibra-tory relaxation process of oxygen and nitrogen molecules. Adiscussion of atmospheric losses can be found in References 3and 6 along with tables for estimating these losses as'a func-,tion of temperature and relating humidity. Atmospheric atten-uation is very sensitive to frequency and the available tables,, do not cover the ultrasonic range which was employed in themodeling. In order to provide this information, tests wereconducted by measuring the noise in incremental distances fromone to sixteen feet from the source. This data, shown in Fig-ure 8, indicates that between 1 and 3 ft. from the source theattenuation rate is greater than 6 dB per doubling of distan-.ewhich is typical in the 'near field' where the noise generationI cannot be considered as coming from a poit source. Beyond 3ft. the slope of the lower frequencies closely approaches the6 dB per doubling of distance with increasing deviation asfrequency increases. This data was used to correct the modeldata while the charts of Reference 6 were used for adjustingthe full scale data.

.3.3 Correction for Directivity

The directivity of the full scale helicopter harmonic noisewas taken from the work of Schmitz and Boxwell (Reference 14)who measured free field noise cround a HU-lH helicopter, inforward flight, by means of microphones mounted on a quiet air-plane. Directivity of the model source (air-jet) was tound to

.. ... ..

be quite sensitive to the angle and alingment of the four smallimpinging air-jet tubes. It was not practical, however, toproduce a model source as directional as the helicopter and sothe tubes were positoned to provide as omnidirectional sourceas possible. Figure 8a, which compares the directivity of thehelicopter harmonic noise with that of the model source, wasused to develop the required corrections. An assumption im-plicit in this procedure is that on the helicopter all harmonicamplitudes have the same directivity. Since the generation ofhigh amplitude harmonic noise is caused by discrete azimuthalfunctions, such as local blade-vortex intersections and/or theazimuth locations of highest resulting advancing top speed,and do not occur randomly around that azimuth, such an ausump-tion is not illogical.3.4 Validation Test Results

Model testing was conducted to simulate flybys of a UH-lH heli-

copter at altitudes of 50' and 500' and a sideline distance of200' in order to correspond with full scale data which was al-ready available. The full icale data had been recorded at aquiet open airfield with closely mowed vegitation. Two terrainsurfaces wore used on the model in order to demonstrate impe-dence effects, one was a 'hard' metal surface and the otherwas the 'soft' linear fiberglass board discussed previouslywhich it was thought would approximate the terrain.

There air-jet was locted at increments along the 'flight path'corresponding to full scale increments of 90 meters as illus-trated in Figure 7a. Figure 7b displays a typical spectrameasured at the microphone and indicates how the spectrumcould be read to predict the effects on rotor harmonic levelsof desired helicopter.

The full scale data was reduced by playing the magnetic taperecording through a 1/3 octave band analyzer onto a stripchart recorder. One third octave bandwidth was selected be-cause this is the basis for analytical detection predictions(References 9 and 10) This data was read at intervals corres-ponding to every 100 meters of aircraft location.

The data was reduced and corrected for atmospheric and directiv-ity effects. Figure 9 shows a typical set of corrections andindicates the pronounced role of directivity, particularly atclose range. The full scale data must be corrected in the timedomain because although the aircraft position is initially cor-related with the sound at the microphone, this is not its loca-tion at the time the sound was generated. The full scale data,therefore, must be retarded by the time which it took the moundto travel from the aircraft to the microphone.

L8

Figures 10 and 11 present the results of the model validationtest. Each figure shows the model test results as measuredand corrected. Also shown is the time retarded full scaledata. There is no direct relationship between the absoluteamplitudes of the model and full scale data and they are to becompared only for trend. In order to aid such comparison, thefull scale data has been normalized to the model data. Thenormalization points were selected as 500 meters in order tobe far enough away to past the range qf very rapid directivitychanges and close enough to avoid excessive atmospheric orterrain effects. The data has been shown for the center fre-quencies of preferred octave bands but comparisons could have

[- been made for any specific frequency such as that correspond-ing to a particular rotor harmonic. (Figure 7).

The general trend of sound pressure level variation with dia-tance results from the trade-off between atmospheric attenua-tion and directivity. At low altitude the distance parameterdominates as the aircraft passes close to the microphone andthe drop off due to directivity does not occur until the heli-copter is almost overhead. At altitude, however, the direc-tivity effect, which tends to maximize in the rotor plane, isrelatively stronger and the maximum value occur. further away.

The agreement between model and full scale data for the 500'altitude (Figure 11) is quite good.

The model hard surface data is higher than that with the softsurface. With a coincidence dip of the type discussed earlier,and illustrated in Figure 6, can be seen when the model had ahard surface. These points which are indicated (*) on Figure11 are not evident with the 'soft ground' model nor in the datathereby verifying the selection of the glass board as a reason-able for simulation of ground impedance.

The match of full scale data with model data for low altitudeflight (Figure 10) shows greater anomalies than does the higheraltitude data. At low altitude, a dip in the full scale dataoccurs beyond about 1000 meters. This is believed to be dueto the fact that the airfield at which the measurements were

*made bordered by a dense growth of high trees which blockedi the line of sight to the helicopter at approximately that din-tance. Therefore, the low altitude comparisons will be con-sidered valid within 1000 meters only. The discrepancies be-tween model and test data appear to occur when the grazingincidenge angle of the acoustic propagation becomes loes thanabout I. At thise low angles, extra terrain losses, due tovegitation, and scattering due to small objects becomes quiteimportant and it appears that a more textured (but not lowerimpedance) terrain simulation might have improved the correla-tion. This might be achicived by using a grid or netting overthe glass board to get more scattering at greater distance.

9

The coincidence dips and large variations in'souid pressurelevel as the aircraft approaches che overhoad location are ofcunsiderable importance to civil concerns, such as effects ofground surface on aircraft noise certifioat-1n, because theyaffect th measured noise levels at or near their maximumlevels. These conditiors are of little *iktereit t9 ,,he ques-tion of detection, however, because theaircraft is not onlyplainly yimible but the warning times are dot. to a ftry fewseconds.

The effect of thin barrier, walls on the attenuation of noisehas studied quits extensively. since barriers which are en-countered in natural terrain-formations such as hills andmountains are, if generally are extremely thick and withoutwell defirid edges. It tae dasired to employ acousticalmodoling techniques'to-examine what happens as barriersdeviace •fL*no ths classical 'knife edge' and how one mightapproximate a wide barrier for analytical purposes.

These tests were conducted using the lined enclosure shown inFigure S. Barriers were inserted to separate the source andmicrophone as illustrated in Figure 12. In the 400:1 modelingscale used'previously the barrier simulated was 800' high, thesource 200' from the ground and both the source and receiver(microphone) 2600' from their respective barrier faces.

Rectangular barriers simulating 13', 100', 200', 3001, 400',and 500' were used along with a 5001 wide barrier whose topwas a 250' radius circle. The corresponding model dimensions - 1will be used throughout the remainder of this discussion,.

The test technique employed and illustrated in Figure 12 wasas follows:

1. The spark generator was iocated at the source loca-tion S1 and the time for sound to travel over thebarrier to the microphone was measured.

2. The spark was replaced with the air Jet and a fre-quency spectrum at the receiver location was re-corded.

3. The barrier was removed and the spark fired todetermine the time for the shortest path from sourceto receiver.

4. The spark generator was then moved back until the'time of the direct sound path equalled that of thepath over the barrier determined in step I (pointS2). This could be done to an accuracy of .001millisecond or a distance of about .001 ft.

10

5. The spark was replaced with the air jet and a secondspectrum obtained.

6. The difference in spectra obtained in steps 4 and 5was the insertion loss of the barrier since allatmospheric propagation losses had been cancelledout by this procedure.

Figure 13a shows the spectrum received with no barrier inplace while Figure 13B compares this with the spectra obtainedwith each of the walls. The results of the tests are tabu-lated in Table I which shows barrier attenuation as a functionof frequency, the mnasured delay tiej the speed of soundduring the test, and the computed difference in path length.

In order to evaluate these results they were first comparedwith that which would be expected by calculation using thethin barrier theory of Maekawa (Ref. 16) who developed a re-lationship between barrier attenuation in the "Shadow Zone"and the Fresnel Number N where N * (AL)

where xa Wavelengtha La Difference between path

over barrier and shortestpath.

The attenuation due to a barrier using Maekawas formulation ispresented graphically in Figure 14a. In order to apply this

I to thick barriers a screen of 'equivalent height' was used asillustrated on Figure 14b. Using this approach attenuationswould be predicted for each test as shown in Table II. Aquick comparison of the test result. of Table I and this sim-plified analytical approach yields poor correlation, withrespect to both absolute value and sensitivity with frequency,and a better analytical simulation of the barrier must besought.

The shortcomings of applying thin screen approaches to widebarriers have been recognized and discussed in papers such asRef. 17. Following the thinking of Pierce but still using theattenuation data developed by Maekawa leads to the approachillustrated below:

}C'

II;; 11L. -I

M~ ( N LA OI I N N N N N I

I I L . L n in I

A 0 LA nM~I I N N N N I I 0 c

IIc

ta LM

II N N N N I I N 0% I0 M !, N 0

N0 N0 N II

0 P14

4.0 .A 0 MN

-4 ,P P, N.1 N I N F

Lo ' M A

P-4 r-n H4 N- LAN

... en

0 "

1.4 N LA m 4 4 LM 0c N- c1J~ 4 LA 0 '. LA A 44fig"

_ _ _ _ _ _ _ __12

TABLt It&- CALCULATION Of V'Tm M,~OTHI EQUIVALLNT REZOKE' 5A5ihZ

.4 75.24 24.45 60.39 74.16 19.4.2 2.03

3 79.72 34.50 61.92 79.27 159,11 1.08

1 1.44 35.1 *3.47 60.56 162.1. 2.13

9 63.2.5 25.61 55.42 52.A5 165.11 2.17

t12 64.62 21.96 56.94 63.35 164,12. 2.32.

is 6.,57 36-42 60.95 84.43 1.71,.11

TAIL Ilb -SAMIER ATTENUATION - 'QUIVALENT KBIOHTt SARRISR

EQUIVALENT MIOHT

411 21 tit 1:~ 111~ -18 - CaliT1 o t 0. au12.27 1 a 1 28 a1229.1

frequencyIOMe N Ab N Ab N Ab N Ah N Ab N Ab

10 3.0 20 3,1 20 3.1 20 3.2 20 3.3 21 3.4 21

20 6.0 29 6,2 22 6.3 22 6.4 22 6.6 23 d.6 23

30 9.0 23 9.3 33 9.3 23 9.8 24 9,9 24 10.2 24

40 12.0 24 12.4 24 12.4 24 12.6 24 13.2 21 1.3.6 25

s0 11.0 25 15.5 25 15.5 25 1.6.0 26 156 46 17.0 26

60 15.0 26 1.5.6 36 1.6 24 19.2 27 19.5 27 90.4 27

70 21.0 27 21.7 27 21.7 27 22.4A 26 23.1. 38 23,6 39

190 24.0 25 24.6 29 24.9 JO 25.6 20 26.4 28 27.2 20)

U.., .. ~ ~ .. ~fla ,.J 4tI~,,f~13

Instead of treating the entire barrier as a refracting edgeconsider each corner separately with a virtual source locatedat the center such that the level at V is first calculated,using Maekawa's method based on propagation from the originalsource and then using these levels as the source for pro aga-tion to the final receiver. Calculations are shown in Tagle ZII.

A comparison of the measured attenuation with calculatedvalues using this 'Virtual Point' concept are shown in Figure15 and shows good correlation at the lower frequencies butoverprediction at the higher end of the spectrum. Comparisonof data for the square edged and rounded top barrier indicatedno significant difference in attenuation.

5.0 CHANN LING OF SOUND

The term channeling of sound, as used in this report, is meantto describe situations, much as helicopters flying throughvalleys or canyons, where the acoustical energy is reflectedback towards the line of flight instead of spreading unimpededin the original direction of propagation.

Two conditions were evaluated, a straight channel in which thesource and receiver maintained 'line of sight' and a channelwith a ninety degree bend.

5.1 Line of Sighi

The model used was, the same one used for the barrier testswith the barriers removed. The variables were the materialsused to face the murfaces. The soft fiberglass (PF-105) was

4used at the ends of the model, being aneclioic they simulatedan infinitely long channel length. The materials used to facethe floor and walls were similar to those used in the flyoververification test except that aluminum was used instead ofsteel for the 'hard' surface. Once again it was felt that thefiberglass board and metal surfaces would bracket the rangewhich might be encountered with natural terrain. Tests wereconducted with the walls vertical and sloped outwards so thatthe effect of this variable could be identified. Spark test-ing was conducted to determine the relative strength of thedirectly radiated path and the'reflected paths. The air jetnoise source was then 'flown down the canyon' by moving it inincrements and measuring spectra at the receiver location, aswas done with the verification tests. The test setup was asillustrated in Figure 16. The noise source was kept at an'altitude' of 6" (200' full scale) and the flight path wasdown the centerline of the model. Unsymmetrical microphonelocations were- used so that different path lengths and delaytimes would result from re.lections from the two walls andthus could be separated and identified. Figure 17 shows theresults of spark measurements with three configurations ofincreasing reflectivity. In Figure A7a the direct path radia-

14

TAILS I Z- BAI|IR ATTMnATION 'VITU AL POINT' MOD

3Y RV

.4 10,000 .0037 4 0141I 5 920,000 .0074 4 :0262 5 9

30,000 .011~. 402 s$40,000 .0144 5 ,0364 1 1050,000 .01 0 $ 50505 s 1060,000 .0222 1 0644 6 1.70,000 .0259 1 .0067 7 11

- 00,000 .2.31. S 1 - 6

3 10.000 .059 7 .1047 7 1420,000 ,15i 7 .2133 7 140,,000 .177 7 .3300 a is

40,000 .236 7 .4266 a 1530,000 .25. 5 .332 9 1740,000 .354 6 .6400 t0 is70,000 .413 9 .746 12 20

0 10,000 .LI4 7 .227 7120,000 .22 7 .453 10 V?30,000 .342 a .076 11 1940,000 .45 1 I .104 12 2250,000 .iL 10 1.13 13 23

2 0,000 .244 7 .354 10 4

70,000 14 9 1.01 13 2230,000 .549 4 9 a4 35

"0,000 .37 1 ,016 16 150,000 ,106 .077 1W5 23

40,000 1.320 14 049 . 10 i70000 1.5 14 3.556 1 260,000 1.072 15 4.064 19 34

15 01,000 1.170 73 253 17 30

20,000 .341 1 2.349 i4 2330,000 .12 7 2.0 10 2730,000 .443 L9 1.09 17 29

50,000 1.33 14 3,372 l4 3260,000 1.62 14 4.040 19 370,000 l."74 15 4.022 19 3400,000 2.4%2 16 1.400 it 3

403

tion is clearly evident at 4.73msec delay time, with one re-flective surface (17b) the additional path appears, when twofacing walls become reflective (17c) multiple paths arise fromreflections back and forth between the two facing walls.

Figure I8 shows a typical sent of spectra which were measuredwith a configuration employing one reflective sidewall. It isinteresting to note that when the source was located 12 inchesfrom the sidewall a very strong pattern typical of constructiveand destructive interference appears with a separation of about2200 Hz. This corresponds to a calculated interference inter-val of 1903 Hz as illustrated in Figure 19, and consideringprecision of measurement is probably reasonably close. At agreater distance from the wall (19") the frequency, as ex-pected, decreases but it is not clear why the depth of modu-, lation is so much less or why it in not evident with a dis-

21 tance of 26". At the very close distance of 2" the inter-ference pattern does not appear because the frequency hasbecome extremely high and this interval between dips exceedsthe resolution limits of the analyzer.

The air jet noise source was 'flown' down the model canyonwith the several configurations of ground and wall materialswhich were indicated in Figure 16. The test results for threefrequencies of noise are presented in Figure 20 which showsthat at close range the directly radiated path dominates(within about 500 feet full scale and that as distance in-creases the hardness of the terrain becomes more importantwith the level increasing significantly with increased terrainhardness. None of these trends is unexpected and but themodel does demonstrate the great effect which terrain canexert on the sound pressure level at the observer. In generalthe level measured at the maximum distance with the 'hard'terrain would not be reached until the aircraft range was onlyone third that distance with the 'soft' terrain.

Since the configuration with high absorption throughout inessentially anechoic the effect of successively increasing thereflectivity of the setup can be obtained by subtracting thelevels measured with the former configuration from the latterones resulting in the amplifications shown in Figure 21. Thisis a good method to use with model testing since all directiv-ity and distance attenuation effects are automatically can-celled out. An idea of the strengths of the various propa-gation paths was obtained by also 'flying' the spark andmeasuring the amplitudes of the successively time delayedimpulses such as those illustrated in Figure 17, Figure 22shows the results for two reflective cases.Although the spark results encompass all frequencies theycontain some interesting information. It appears that thefirst reflection is almost as strong a. the direct radiationand that all others, with the possible exception ot thesecond, could be neglected but even if there were no rein-forcement by any reflected paths the magnitude of difference

16

between most and least reflective configurations of 15 dB asmeasured with the air Jet could not be accounted for by sum-ming of acoustic ray propagation paths. Figure 22b demon-strates clearly that all higher order reflections involved thesecond wall.

Another way of looking at the situation of the canyon model asan enclosed room with absorption coefficients of 1.0 at thetop and ends and absorption coefficients corresponding to theappropriate material and frequencies for the sidewalls andground surfaces. Using the method for calculating totalsteady state sound energy density of Rof. 18 Section 10.14 topredict the buildup within a room yields the results shown inFigure 23. Considering that the room volume of the model isat least an order of magnitude smaller than the applicationsfor which the method was developed and considering the sim-plistic assumptions which were made one ought not to expecttoo good an agreement. But the fact remains that the testresults indicate a stronger affect of terrain absorption thanwould be expected.

The effect of sloping the walls outward at approximately 450is shown in Figure 24. This is, of course a more naturalsituation because truly vertical walls are rarely encounteredin nature.' indications are that additional attenuation willbe experienced at moderately close ranges but at large dis-tances a vertical wall model may suffice. The condition shownis for the most reflective configuration and the effect ofslope would be expected to be even less with more absorptivesurfaces.

The line of sight model was converted to the configurationillustrated in Figure 25 which results in a considerably morecomplex experiment. The results shown in Figure 26 indicatethat the attenuation due to the bend, which is very evidentwith soft terrain, can be virtually cancelled but if the wallsare reverberant. It also shows that when flying at very lowaltitude the effect of the ground is less than that due to thebend but the ground provides considerable attenuation of lineof sight propagation. It is also quite evident that theseeffects increase with frequency.

Curving the outer wall of the bend had little effect on thedata simulating N.O.E. operation as shown in Figure 27 butFigure 28 reveals an interesting effect on the data whichsimulated 200 ft. flying. The effect of the curviture is toreflect the pressure waves, as illustrated, so that they tendto bend around the corner instead of being reflected back.That this is more effective on higher frequencies, of shorterwavelength is indicated by the data which shows that the levelwith the curved wall increases from a level below that of thesquare bend at 20 KHZ(50 Khz full scale) to considerably higherthan the square bend at 60 Khz.

17

One of the propagation paths illustrated in Figure 2. is thatdiffracted over the edges of the 'canyon' rather than propa-gating around the bend. In order to determine the effective-ness of this path a lid was put on the approach leg o themodel so that this path was blocked. The results (Figure 29)clearly show that with soft terrain the path over the edges ofthe 'canyon' was a very substantial one and also increasedwith frequency and that failure to consider such propagationpaths could lead to major errors in predicting detection dis-tances. Note that with hard terrain the paths through themodel still predominate

8.0 CONCL.USIONS-AND REOMM!ENDATIONS

The work described in this report has demonstrated that acous-tical modeling techniques can be applied to studies of theeffects of terrain on the propagation of helicopter noise.Experiments to separate complex effects, which are virtually

2 impossible to perform in full snals are very easy to do withmodels. Although the techniques themselves are general theycan be tailored to prediction of any particular helicopterprovided its acoustical spectrum and directivity are known,either from measured data or analytical prediction. The testswhich were conducted pretty much bounded the range of acousticalimpedance and absorption which would be encountered in natureand demonstrated very large effects on noise propagation.

*1 iThe model verification was conducted only for the classically,

simple case of flat, soft, terrain and for this case showedgood results when 'flying' 500 ft. above the ground but onlymoderate correlation with nap of the earth data. The modelapproach shows strong promise but needs further verificationby application to more complex full scale situations for whichboth data and description of the terrain are known. One ofthe major questions to be answered is the detail to which ter-rain features must be modeled).

More research is also needed into modeling materials forproper simulation of various types of terrain.

ai

.1J

I

1 B . .

REFERENCES

1. Anon., Helicopter Acoustics, NASA Conference Publication2052, May 1978.

2. Loewy, R. 0., Aural Detection of HolicoRters in TacticalSituations, Journal of the American Helicopter Society,

October 1963, pp. 36-53.

3. Ollerhead, J. B., Helicopter Aural DetectabilityUSAAMRDL Technical Report 71-33, Eustis Directorate,U.S. Army Air Mobility Research and DevelopmentLaboratory, July 1971 AD 730788.

4. Hartmran, L., and Sternfeld, H., An Experitment in AuralDo D Cton gof Heicopter, USAAMRDL Technical Report73-50, Eustis Directorate U.S. Army Air MobilityResearch and Development Laboratory, December 1973,AD 917355L.

S. Scruggs, B. W., and Hampton, Kenneth D., An AnalyticalRotor Tip Speed to Reduce Helicopter Acoustic DetectiowUSARTL Technical Note TN-371 U.S. Army Applied TechnologyLaboratory, August 1979 AD A076961.

6. Anon., Standard Val4es of Atmosoheric AbsorDtion as aFuncionR T mperature and Humidity, SAE. ARP 866A,Society of Automotive Engineers, Inc., March 1975.

7. Pegg, R. J., A Summary and Ivaluation of Semi-EmDigicalMethods for the Predirtion of Helicopter Rotor Noise,,NASA TM 80,200 December 1979.

8. Fidel, S., and Pearuons, K.; Study of the Audibilityof Impulsive Sources NASA CR1598, May 1970.

9. Ollerhead, J. B., Helicopter Aural Detectability,USAAMADL Technical Report 71-33, Eustis Directorate,U.S. Army Air Mobility Research and DevelopmentLaboratory, July 1971.

10. Ungar, E., et al, A Golds fgr Prqgi~tinS the AuraQD~egeJo of AircrAft, AFFDL-TR-71-22 Air ForceFlight Dynamics Lo0atory, Air Force Systems Command,March 1972.

11. Hartman, L., and Sternfeld, H., An Experiment In Aura!Detection 04 He.icopterg, USAAMRDL Technical Report73-50, Eustis Directorate U, S. Army Air MobilityResearch and Development Laboratory, December 1973.

12. Wylie Laboratories Inc., Corre ation of Actual andAnal ticsa Helicopt r Aural Deection C iterla, USAAMRDL-TR-74-102A EustisD rectorate, U. S. Army Afr MobilityResearch and Development Laboratory, January 1975.

19

..~ .........

13. Lyon, R. H., Acoustical Modeling IHa1dckoo11, MassachUsettsinstitute of Technology.

14. Schmitz, F. A., and Boxwell, D. A., In Flight rar-FieldMeasurement of HeliSogter Impuluive Noil.; Journalof the American Helicopter Society; Volume 21 No. 4October 1976.

15. Parkin, P. H. and Scholes, W. E., Zgggtonq ~nStHafelJournal of Sound and Virto o.2

Pq .353,1965.

16. Anon, Ac~ti mgct Podue by a Belii FlgSociety of Automotive Engineers AIR 1327, January 1975.

17. Maekawa, Z., Noise Reduction by Screens, AppliedAcoustics, 1968.

lB.PieceA. .,Diffraction of Sund Arougd Corners

a nd over Wide Bgrrien , Journ~al of the Acousticalhocety of America Vol. 55 No. 5, May 1974.

19. Beranek, L. L., Acoustics, McGraw Hill, 1954.

-. ----- ----

PERSONNEL

The following scientific personnel 2articipated in the conduct

of this studyi

Principal investigator - Mr. Harry Sternfeld, Jr.

Experimental testing was conducted byt

Mr. Bernard BorekMi Lima Richardson (Intern student M.I.T.)Mr. Edward Schaefer

Illustrations were prepared by:

Mr. Mark Miller (Co-op student Widener College)

21

21

FW

Fiue lac fTeri o e icotrNief rpgto224r NO

ii-

44

23B..I .................

:ROTATIONAL NON-IMPULSIVEI odI

n W, ,ob . .

. HIGH ADVANCING TIP SPEED

I OdI

Illtl Y[00 lit lo 1

TANDEM ROTOR BLADE-VORTX INTERSECTION,-

I

FIGURE 3. ACOUSTICAL SIGNATURES 01' HARMONIC ROTOR NOISE

24

ri

F'igur 4-o NieSore

25

lawI

inner

* Spark Instrumentation IsligVStorage oscilloscopeTtuur

Scaling AmplifIVYerte

I Wave BurutoSup r

ElProcessor Structure

R-1 SoftFiberglass4Aneoh oi.cLi ninrg

Figure 5 Model Construction

26

THSPATH LONGER BY%'-*

WAV E GROUND nIMPEDANCE IMPEDANCE fwb

PHASE REVERSAL WHEN GROUND IMPEDANCE<41AVE IMPEDANCE

-10

-15

-20 ~LOGF

- --- -GROUND IMPEDANCE> WAVE IMPEDANCE

GROUND IMPEDANCE<4WAVE IMPEDANCE6 as so a EXPECTED LEVEL RELATIVE TO FREE SPACE

Figure 6 SOURCE-RECEIVER ARRANGEMENT LEADING TO GROUND EFFECTAND THEORETICAL VALUES OF EXCESS ATTENUATION(REDRAWN FROM REF,.13)

27

'Flight Path'

A (2 0' full scale)Ar Microphone

F *r

Ground Surface

SCALE DISTANCE 6000f

10 d0dB Ref

0cHz (Model Scale)ncy KHZ250Hz (Full Scale)

~ai~S io Kt-l Helicopter RotorHarmonic Prequencies(mode. scale)

b-Typical Spectrum measured at microphones

Figure 7 Validation Test

28

0I0

~L~'4-kA ~ .

ap -TOA07 va*feUSd PunOS VAT41TOU

4J 4'

4j1

N~in

Figure ~ ~ ~ ~ ~ ~ ~ ~ ~ I 8 DrciivadAmopei orcin

29 '

200 Hz

4S0' Alt

Soft Terrain

Corrected ForAtmospheric Effect

30-

'Raw' Test Data

4). 20-

10 -

Corrected For Atmospheric

o .. and Directivity Effects-16.405 -13.124 -9.843 -6,562 -3,281 0

Model Distance- Feet

/CORRECTION FOR ATMOSPHERIC EFFECT

CORRECTION FOR DIRECTIVITY

Figure 9-Model Data Corrections

30

SO.~~ so...~u~0 8 a 4 fee ob

_. s, 4 .)#

we. in: tg~o 111 WI t I

Ile0.s

no' olIIII I0

Msi N0

10'-

~ano amIn N m a

A 'HrAaon

310

.. .....

we I s

go _____Ion__in._low__ISO__10_in

ao iou Of"Mlyu us

10M 1.1 UnoauiUPs~s

It~45

as

Aa 4

to.

0 a. Ato..J ~ S A

to iu o

IN A onidfc Di 'adarud

o0 aSf rud

Figure pul SclDt(Normalized to Test Date)

32

MA.FES

-u -M-

8278

788" 81

Figure 12 Setup F~or Barrier Tests

33

&-Spectrum with no wall$

AROI PROJECT 0-0OK

80-

600 60

0 50 1.00

FREQUENCY, KHZ

N Wall

12" Wall

15", Wall

F'igure 13 b-Effect of Barrier Thickcneuu

34

... - .. . . . . . . . . . .... -.. .

30-- - -

S25

'4 20-

0~ 110 0

NaFresna1 Number

Equivalent Height Screen~0

\&wavelength'

Figure 14-Prediction of Barrier Attenuation by Maekawalu Method(Re f 16)

-------- ---- -

Wall Thickniss-inches005 10 15

2 0 kZ~z

-10

-20

-30

040 kHz

-10

-30 oi

4J4

060 kHz

~ ~ -4-- Measured From Test -

-4-Calculated 1:y'Virtual Imagel Method

Figure 15 Compa~rison of Measured and Predicted Barrier Attenuation

.....6

0)iLnL

,0 1'

Figure 16-Teat Setup-Line of Sight Channeling

37

V

18-

(Fig 16,Confiq I)

4-0-

Im -0IU, (b)One Wall 'Soft'

C)Uone Wall 'neHard

3 9)

Delay)On Wall MiSoftond

Figure 17 Typical Spark Test Results

38

._..__..__.__..__.._..__.._

DSTAN'CE TO REFLECTIVE WALL

2" MODEL67' FULL SCALE

UE~~~hhE12" MODEL400' FUtLL SCALE

a., 19" MODEL

101 633' FULL SCALE

26" MODEL867' FU7LL SCALE

* .MODEL SCALE 100,000

0 FULL SCALE 2SU

FREQUENCY - Ha

Figure 18 -Effect of Wall Distance on Acoustic Signal

39

I'I

F01

Ir

The destructive interference between oaths 0Orand or will occur at freayuencieu which differ muchthat their wavelengtv w lrve- differences of . 593'

c duingthis exoeriment wasn 1124 ft/soc

interference frequency intervolmjI29a 1903 Hz593

Figure 19 calculation of interference Frequency

40

20 kHz-(50 Hz Pull scale)Distance Ft-(Full Scale)

-zto ICc - , 'At

ZIch* a (ModelI)

40 kH.-(l00 Hz Flu3l scale)

0

----- ------

~e d! 45 4 us s A a gi10a.

Hard (l50~lo 1z Hall Scale)ot ol~ ~ (Fig416 03)

Hard, -al otWll Fg1 2

-of Har S(F oo -6 Had01)~U (rj1

Figure 2O-Effect of Acoustical Absorption on~ Line of Sight Propagationl

41

*SS~~ * t2kHz

0

b 2, h Q ~'70kHz

bb tb 40 40 40 %Dsace-nce

Fiue l-ApifctinMadWal adFlo

42

'~ * SI, * **** .. 40kHz

I I

bo Go 6Asb

! 70kHz

443

Direct Path

lt Reflection

00 2nd Reflection

A164I

4th Reflection

- 5th Reflection Single Reflections

47 ;M 4S ' 16 $3 3o .1 J1i~Distance-inches

Figure 22.-Propagation Path Amplitudes Two 'Hard'Wallu

44

Direct Path

~ / lmt Reflection

2nd efletionA.,3rd Reflection

Single Reflection$ .

4.. S~~~th Reflection6t elcif

Distance-inchas

F~igure 22b-Propagation Path Amplitudes one 'Hari'Wall

45V

.~~~ .......

All Surf coo Soft

Floor,l Wall HardI Wal l~soft

Floor.Hard IWalls Soft

t. All Surface

HA~H

V lMe fa Ro m old,

~.Auort~o Ceffiien

Fiur 23 mlfiaini aS 1~ Ca yoate

0j 21 easure

.,~2' Ca-=-~ -- - -u

.04

0Cm 40a

.4)4

4 04

Figure 24 Zffect of Wall Slope- 200 kHz

47

,.ah zone Line Of Sight

- ~Cu:vod

Doubl.e Barrier at

(Over Edges of 'Moder--

%microphone

Figure 25- Ben~t Chanel Configuration and Propagation Paths

48

- 4-.

011

cI~ .11

EI TV4 9= 1 92 pU

Figure 26 feto al'ades49

.........

0

A U N

00

I

I

CD

o 0

0

.qt

0

~0

"1 IN 0

C,

0 0

K0

U,

~~% '~ "' V

p 0I , /0

, IN A

1.0 0

N I I..0 0

0 o~I ~ N

No N

0 ) 0I \ \~ 0 0

0 ~ 0 0

~P -tSAS1 S~3US.~d pUr~CS

Figure 28- Effect of Bend curviture (200). Altitude- 'Hard Walls

51

- -- - ~*~3

N ~

NN I

CC

r*4 0)

I.-* - TAG 9= 69d0

Figir 29 rpgto ve deo al

520