Propagation of noise along a finite impedance boundary

C. I. Chessell

AustralianDefenceScientificService,Departmentof Defence,WeaponsResearchEstablishment,
Salisbury,South Australia
(Received 20 January 197.,7;revised 28 June 1977)

The purposeof this paper is to show that the essentialproperties•of the propagationof noiseover real soil
surfacescan be modeled using the propertiesof fibrous absorbentmaterials and the appropriatetheory of
soundpropagationalong a finite impedanceboundary.This work is an expansionof the original proposals
of Delany and Bazley [J. SoundVib. 16, 315-322 (1971)]. The model predictionsare shownto be in
good agreementwith a variety of experimental ground absorptionmeasurements.The application of the
model to the measurement of aircraft reverse thrust noise is described.

PACS numbers:43.28.Fp, 43.50.Vt

INTRODUCTION terials. The model is specified by a single parameter,

the specific flow resistance per unit thickness crand the
The development of techniques for the control and re-
parameter value for a particular situation can be readily
duction of noise in the community is dependent upon an
determined by simple acoustic propagation measure-
understanding of the way in which sound propagates in
ments over the surface in question. The use of such a
the atmosphere, particularly near the earth's surface.
modelwas originally proposedby Delany and Bazley9
The principal factors determining the sound field around
following their earlier measurement of the acoustic
a point source, aside from spherical spreading, are
properties of fibrous absorbentmaterials. •0 However,
generally held to be
in comparing their model results with experimental
(a) ground absorption losses, measurements, Delany and Bazley considered plane-
wave reflections at the boundary and the model predic-
(b) atmospheric refraction in wind and temperature
tions were only in fair agreement with experiment. The
gradients,
current work is essentially an extension of their origi-
(c) atmospheric turbulence which leads to scattering nal proposal. It is st{ownthat in manycasesof practi-
losses and random fluctuations in the sound parameters, cal interest it is necessary to consider the reflection of
and a spherical wave at the boundary. Using this more ac-
curate theory it is shown that the model predicts ground
(d) atmospheric absorption losses.
absorption losses which are in good agreement with ex-
These factors depend in a complicated way on the perimental measurements. Finally the applicationof
source-receiver geometry, on the atmospheric param- the model to a particular situation, that of measuring
eters (wind, temperature, and humidity) and their ran- the spectral characteristics of aircraft reverse thrust
dom variations, and on the nature of the surface and the noise, is described.
terrain geometry. The factors are also known to be
strongly frequency dependent. For example, the fre-
I. SUMMARY OF THEORY
quency dependence of atmospheric absorption due to
classical and molecular absorption has been extensively Consider a point source located at height hs above an
investigated and values to be used in practice are rec- interface between air and a semi-infinite layer of mate-
ommended in SAE ARP-866. • rial with finite characteristic impedanceZa (see Fig.
1). The pressure reflection coefficient for a plane sound
The propagation of sound along a boundary having a
wave traveling in the air (characteristic impedanceZ•)
finite impedancewas originally studiedby Ingard•' and
and incident at an angle •b• at the material is
Rudnick3'4assumingsomewhatdifferent boundarycondi-
tions. The modification of the sound field by the bound-
ary gives rise to the so-called ground absorption losses
R•=Z •.sin•;- Z1sinqb•.
Z•. sinqb• + Z•. sin qb•.
, (1)
referred to in (a) above. A number of subsequenttheo-
where •b•.is the angle to the horizontal subtended by the
retical and experimental studies (e.g., Refs. 5-8) have refracted wave in the second medium. The incident and
delineated many of the essential features of this propa-
refracted angles may be related by
gation. However one of the major problems in any ap-
plication of these results outdoors is the requirement k•.cos•. = k• cos• , (2)
for a knowledge of the acoustic impederice and propaga-
where k• and k•. are the propagation coefficients in the
tion coefficient of the boundary. These quantities are
two media. The plane wave reflection coefficient may
difficult to measure, particularly over the wide fre- then be written as
quency ranges involved (see, for example, the discus-
sion in Ref. 8). _ sin•b
1- Z•./Z•.(1- •' •'cos
•' •'/•'
The main purpose of this paper is to show that the es-
R• sm•+Z•/Z•(1
-
- k•/k•
Cos
•) • • g 1/2 ß
(3)
sential properties of real soil surfaces can be modeled Rudnick
• •d Ingard• useddifferent assumptionsabout
using the acoustic properties of fibrous absorbent ma- the bo•dary conditions in developing their solutions,

825 J. Acoust.Soc.Am., Vol. 62, No. 4, October1977 Copyright

¸ 1977 by the Acoustical
Societyof America 825

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 99.244.138.70 On: Sun, 18 May 2014 03:09:18
 826 C.I. Chessell'
Noisepropagation
alongan impedance
boundary 826

SL r•

R •

FIG. 1. Source-receiver geometry.

Rudnick considering the case of two semi-infinite media

in which the reflection coefficient is written as in Eq. for reasonably small values of w. For larger values of
(3) above. Ingard considered reflection from a normal w (Iwl > 10) the asymptotic series
impedance boundary in which propagation in the second
medium is ignored (the case of local reaction).
pression for the reflection coefficient then reduces to
The ex- F(w)
=-• 2"n!
(2w)"
sinqb
1- Zi/Z•. =- •-•+(-•'• + (2• +'" (11)
/{•=sin4>t
+Zt/Z•.' (4) is used.
Model attenuation predictions have been evaluated nu-
merically under both assumptions and the results are For grazing incidence and with Z•. >>Z• or klr • small
discussed in Sec. IIA below. (or both), w will be small and by Eq. (9), F(w) • 1, so
that the image source strength Q-• + 1, independent of
Considering the source-receiver geometry in Fig. 1,
the value of R•, thus resolving the anomaly observed
the velocity potential 4, at the receiver is often expressed with the use of Eq. (5). In the other limiting case, that
as
for higher frequencies, longer ranges and smaller-sur-
face impedances, w >>1 and by Eq. (11), F- 0 and Eq.
4,= exp(iktrO
• • exp(iklr•.)
(6) reduces to the plane wave result of Eq. (5).
Ingard found an identical result for the image source
_•exp(ik•r•)
fl
[1+R•exp(ik
lar)], (5) strength to that given in Eq. (7) when considering the
locally reacting case, however the numerical distance
where Ar is the path length difference, the differential w reduces to
inverse distance loss along the two paths has been ne-
glected, and a time dependenceof the form exp(-i•o t) , (sine+Z•/Z•) •
has been omitted. For acoustically hard surfaces Z•. w=• ik:rz(1+sine
' Z•/z•)' (12)
may exceed Z• by several orders of magnitude so that As stated above, numerical results obtained under the
Rp• 1 for all but small values of qb. Ne.ar grazing inci- two boundary assumptions will be compared in See. III A
dencehowever,sinqb
• 0, so thatRp• - 1 [byEq. (3)], below.
regardless of the surface material. From Eq. (5), the
acoustic signal would be expected to vanish since A r--• 0
II. THE GROUND ABSORPTION MODEL
for grazing incidence. This result is contrary to exper-
ience. In order to determine the sound level above a boundary
using Eq. (6), the characteristic impedance and propa-
The apparent anomaly is resolved when the reflection
gation coefficients of the boundary material must be
of a spherical wave at the boundary is considered, rather
known. Delany and Bazley•ømeasured the acoustic
than a plane wave as in Eq. (5). Using the development
properties of a wide range of fibrous absorbent materi-
of Rudnick,s the velocity potentialin the case of a spher-
als. The measured values of characteristic impedance
ical incident wave can be shown to be approximated by
and propagation coefficient were shown to normalize as
a function of frequency divided by specific flow resis-
4,-"•
exp(iktrt)
r1
[1+Qexp(iktAr)
], (6) tance per unit thickness (rr) and to be able to be repre-
where
sented by simple power-law functions.
Expressing the characteristic impedance Z of the ma-
Q =R• + F(w) (1 - R•) . . (7) te rial as
Here Q is knownas the image source strength, F(w) as
Z = l{ + ix (i3)
the boundary loss factor and w is the so called "numeri-
cal distance." This terminology largely follows the early and the propagation coefficient k as
radio propagation work, particularly that of Sommer-
k=a+i• , (i4)
field. The numerical distance is defined by (using cor-
rections to Rudnick given in Ref. 11) the experimentally determined power-law relations are
as follows:

2iklr•'
w= (1- R•) •Z--l•'
\ Z•./ (1• k•
• cos% cøs•'½
z )'
kz (8) ///poCo= 1 + 9.08(f/e) 'ø'• , (i5)
X/poCo - il 0(//it) 'ø'•a (i6)
The following asymptotic series are convenient for the
numerical evaluation of F(w) a/(w/c o): 1+ 10.8(/fir) 'ø"ø , (i7)

J. Acoust. Soc. Am., Vol. 62, No. 4, October 1977

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 99.244.138.70 On: Sun, 18 May 2014 03:09:18
 827 C.I. Chessell'
Noisepropagation
alongan impedance
boundary 827

60
tended reaction case of two semi-infinite media due to
(a)
Rudnick[Eqs. (3) and (8)] and the local reaction case
due to Ingard [Eqs. (4) and (12)]. Calculationswere
made for the following range of parameters: .flow re-
O z
sistance 50-500 cgs units, source-receiver separation
20-1000 m, source and receiver heights 0.05-5 m. In
all cases the results were found to be in close agree-
0
ment, the maximum difference found being approximate-
UJ
ly 0.3 dB. For the comparisons with experimental re-
suits to be described later in this paper this difference
<:I:•
0.8
n,-' is not significant. Therefore, although for the remain-
z
der of the paper the predicted results for one case only,
nO/,
the two semi-infinite media assumption, are used, it
should be noted that the local reaction assumption pro-
i i I I I I i vides essentially identical results. Further the good
0 5• lOO 200 500 1K 2K 5K 10K
FREQUENCY- Hz
agreement found between the model predictions and mea-
sured propagation values of excess attenuation would in-
dicate that real soil surfaces can be adequately repre-
6O i i i ! i i sented by the simpler local reaction boundary condition
(b)
assumption.

B. Frequency dependence
o•
o The excess attenuation due to the presence of a bound-
ary over the case of free space attenuation is given by
[from Eq. (6)]

Ae=20log•01
1+(r•/r•.)Qexp(ik•A
r) l . (19)
<I:

o i
06
ExpressingQ as IQleie Ae canbe written as
200' lO'••o

0
50
I
100
I
200 500 2K K 10K
where•' =•/•[. In later sectionscomparisonsof the
FREQUENCY- Hz model predictions with experimentally measured absorp-
1
tion losses will be made which are averaged over •-
FIG. 2. (a) Model characteristic impedance and (b) propagation octave band intervals. The effect of this finite b•d-
coefficient values as a function of frequency. Flow resistance
width is dependent upon the spectrum shape but results
(cgs units) is the parameter.
reported in Ref. 12 indicate that the assumptionof white
noise introduces only small errors for a wide r•ge of
•/((.0/½0)
= 10,3(f/(7)-0'$9, (18) spectrum shapes. Equation (20) is therefore modified
to the following form for averaging over •-octave bands'
where co is the velocity of soundin air, •0= 2•T/ and PoCo
is the characteristic impedance of air. The measure-
mentswere madeovertherange5-1000 g'Zcm• ofthe ratio
f/(•. (Note that cgs units have been used here to simplify
referrals to Ref. 10), In developing the ground attenua-
tion model in this paper, extrapolation of these formulae
to lower values (f/(•--0.5 g-Zcm•) is required, the ex- where

cellent agreement obtained with experimental observa-

= 27/27 , (22a)
tions in this lower range justifying the extrapolation.
Some examples of the variation of the model character-
istic impedance and propagation coefficients with fre-
•d the Q•, •i, and Xi Are the values corresponding to
quency are shown in Fig. 2 with the flow resistance (• as
the center frequency of the •-octAve bands. This ap-
a parameter. These model values may now be used tb
proximation is used throughout where •-octAve band lev-
predict the magnitude and frequency dependence of the els are considered.
ground absorption losses.
The general characteristics of the ground Absorption
III. MODEL GROUND ABSORPTION PRODICTIONS losses predicted using Eq, (21) And the model values of
the surface parameters c• best be studied by a particu-
A. Comparison of boundary assumptions
lar example. Consider the case of hs = 1 m, h• = 1.52 m
Model excess attenuation predictions due to the pres- (see Fig. 1) and the flow resisteance • = 200 cgs units,
ence of a boundary were calculated for the two boundary The excess attenuation due to the presence of the ground
condition assumptions described above, namely, the ex- plane is shown in Fig, 3 for the case of two semi-infinite

J. Acoust. Soc. Am., Vol. 62, No. 4, October 1977

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 99.244.138.70 On: Sun, 18 May 2014 03:09:18
 828 C.I. Chessell:
Noisepropagation
alongan impedance
boundary 828

three different ranges. The ground is again character-

ized by a flow resistance of e = 200 cgs units and the re-
ceiver height is maintained at 1.52 m. As the source
height increases the minimum in the excess attenuation
curve shifts to lower frequencies, this shift being more
pronounced at shorter ranges. At longer ranges there
is also a rapid decrease in the peak excess attenuation
with increasing source height. Thus the importance of
ground attenuation outdoors is heavily dependent upon
the elevation angle of the source at the receiver. Simi-
lar effects are obtained if the source height is held con-
stant and the receiver height varied.

D. Effect of varyinggroundparameter
The effect of varying the model ground characteristics
is illustrated in Fig. 5 where the excess attenuation as
a function of frequency is shown for model flow resis-
tances in the range 50-400 cgs units and for three
ranges. The source and receiver heights are maintained
K 1OK at 1 m and 1.52 m, respectively. Changing the model
parameter primarily affects the excess attenuation at
FIG. 3. (a) Modelresultsof excessattenuation,
(b)contribu- lower frequencies, the most important for long distance
tion to phase shift from path difference Ar/k and image source propagation. An increase in the flow resistance causes
0, and, (c) IF(w)I, for •-octave bandlevels. Sourceheight1 a reduction in the excess attenuation and a shift in the
m; receiver height 1.52 m; model parameter (• = 200 cgs units.
peak attenuation value to higher frequencies. The strong
Parameter is source--receiver separation R(m).
dependence on • at low frequencies is used in later sec-
tionsto selecttheappropriatemodelparameter•o best
media for two distances (R = 20 m and 1000 m for each fit experimental measurements made at a particular
1 site.
r-octave band between 50 Hz and 10 Khz. Also shown
in Fig. 3 are the phase difference contributions from
the phasepathdifferencea r/X and the image source• E. Effects of atmospheric turbulence
and also the values of IF(w) l for the two distances as a In calculating the ground absorption effects in previous
function of frequency. sections, coherent addition of the direct and reflected
The main properties of these curves can be summa- waves has been assumed. In the real atmosphere ran-
rized as follows' dom fluctuations of wind and temperature will give rise
to fluctuations in the amplitude and phase of each signal
(a) The excess attenuation curves for both distances at the receiver. The effect of the fluctuations would be
show negative excess attenuation values at low frequen-
cies, i.e., in-phase addition of the direct and reflected
wave s.

(b) Both curves exhibit a minimum (a maximum ex-

cess
attenuation)
atapproximately
500
Hz.Atlarger 0
distances this minimum is primarily causedby the phase
delay • of the reflected wave [see Fig. 3(b)]. The depth [o]
i I i I I I i
of this minimum increases with source-receiver sepa- , -lC
ration. z

(c) With further increase in frequency, both curves

recover to near zero excess attenuation, this recovery
being due to the increasing importance of the small path
difference between the direct and reflected waves. The
larger path difference at the shorter range then leads to
oscillations in the excess attenuation curves as (6 + 2•r/X)
= - n7r, n = 2, 3, 4, ....
(d) The values of IF(w)[ shownin Fig. 3(c) show that
even for a range of 1000 m, this factor cannot be ne- i i i ! i &

glected, particularly at low frequencies. 50 100 200 500 1K 2K 5K 101<

FREQUENCY- Hz

C. Effect of varying source height FIG. 4. Model excess attenuation for three ranges showing
effect of varying source height (a) R =20 m, (b) R =200 m, and
Model results obtained by varying .the height of the (c) R = 1000 m. Receiver height is 1.52 m, model parameter
source (h8) abovethe boundaryare shownin Fig. 4 for (y=200 cgs units. Parameter is source height hs(m).

J. Acoust. Soc. Am., Vol. 62, No. 4, October 1977

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 99.244.138.70 On: Sun, 18 May 2014 03:09:18
 829 C.I. Chessell:
Noisepropagation
alongan impedance
boundary 829

differencefluctuations
(YA•will be writtenas
=cf• •/•' (25)
where C is a constantfor particular atmospheric condi-
tions and the weak frequency dependenceof the last term
in brackets in Eq. (23) has been ignored. The time av-
eraged excessattenuation•A•) can be written as [using
Eq. (20)]

/ ( ,,,i
where • is now a random variable with mean value zero
•d st•dard deviation•a•. If • is assumed
to be nor-
mally distributed, then Eq. (26) can be written as (see
Ref. 14)

<A•>
=10log,o 1IQ]•'
1+r--;•
FIG. 5. Model excess attenuationsshowingthe effect of varyin•
the model parameter <•. (a) R =20 m, (b) R =200 m, and (c) R
2 ) 3. (27)

=1000 m. Source height is i m, receiver height 1.52 m. <Ae>has been evaluated for a number of values of the
Parameter is <• (cgs units).
"turbulencestrength" parameter C in Eq. (25) and the
results are shown plotted as a function of frequency in
Fig. 6. The model parameter is again <y=200 cgs units
to reduce the ground attenuation from the values calcu-
and hs=l m, hR=1.52 m. At longer ranges the effect of
lated for a quiet atmosphere. The extent of this reduc-
turbulence is to reduce the maximum excess attenuation
tion will depend upon the strength of atmospheric tur-
and to cause a shift in the peak excess attenuation to
bulence and the extent to which the fluctuations along
each path are correlated. Here we consider the effect
lower frequencies. As would be expectedfrom Eq. (25),
the effect is very much smaller for shorter ranges and
of phase path fluctuations only as a first approximation
lower frequencies. The inclusion of amplitude fluctua-
of the effects of a turbulent atmosphere.
tions as well as phase path fluctuationswould be expected
Chernov
xsderivedthe followingrelationfor the stan- to reduce the peak attenuation values still further.
dard deviation of the phase path fluctuations, assuming
a Gaussian spatial correlation coefficient for the acous- IV. COMPARISON OF MODEL WITH FIELD
tic refractive index fluctuations MEASUREMENTS

• •/•.w(%/cg)(RL)•/•.(l+arctanD/D)
ty• =•r •/•' , (23) In this section excess attenuation values obtained us-
ing the above theory and the fibrous material model are
where L is the atmospheric scale size, R the path
compared with field measurements of excess attenuation
length, •½ the standard deviation of the acoustic velocity
fluctuationsaboutthe meanvalue Co, andD is the wave
parameter defined by

D=R•/2L •' , (24)

-• 0
where X is the acoustic wavelength. A number of as-
sumptions made in deriving this relation may not be
<[
valid for propagation close to the stirface, e.g., iso- z

tropic turbulence, outer scale size greater than a wave-

length; however, the range and frequency dependence ', ; I I I I I

of Eq. (23) will be usedto obtainan approximateesti-

mate of the effects of turbulence.
\ ,6.,o•/' //•/ ./•'
The fluctuations in phase path difference between the ß

• o/-/8o/.• o/.• .•

direct and reflected wave will also be dependentupon <3:.20

the degree of correlation between the fluctuations along
each path. In many situations of propagation close to ••.•/ [•]
the surface the two paths will be close together and this ;o • ' '
2• '
5• 2 ;K
1 '
5K 10K
correlation would be expectedto be high. Chernovts FREQUENCY- Hz
has derived formulae which describe this correlation in
FIG. 6. Time-averted model excess attenuationsshowing
terms of the path separation, however in view of the
effect of varying turbulence stren•h parameter C. (a) R =200
doubts concerning the validity of this theory near the m, •)• =1000 m. Sourceheightis i m, receiver height1.52
ground, direct application of these relations would not m, model parameter •=200 cgs units. Parameter is turbulence
seem justified. Here the standard deviation of the path strength C (r• see m-•/2).

J. Acoust. Soc. Am., Vol. 62, No. 4, October 1977

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 99.244.138.70 On: Sun, 18 May 2014 03:09:18
 830 C.I. Chessell'
Noisepropagation
along
animpedance
bounda• 830

ceiver heights for the source on the ground are shown

in Fig. 8 (takenfrom Fig. 5 of Ref. 8). The model re-
sults with (• = 300 cgs units for [hese two cases (hs = 0.03
m and 1.22 m, R= 15.24 m) are also shownin Fig. 8.
In calculating these attenuations the reference level has

• -10 V "Model
Z
been assumed to be the sound levels observed at a dis-
0 -
tance of 0.30 m (1 ft) with both source and receiver on
-20
200
I, ,!
400
,I I
800 1K
I
2K
•
4K
the ground(hs=h•=0.03 m). Goodagreementis again
obtained between the model predictions and the field
FREQUENCY- Hz
measurements within the experimental error involved.
FIG. 7. Comparison between model results and field mea- The gradual reduction in the excess attenuation for the
surements taken from Ref. 8. (•Ref. 8, ¸--¸ m0del.) model values for the 1.22-m receiver height at frequen-
Source height is 1.22 m, receiver height 1.22 m, range is cies above i kHz is strongly dependent upon the source
15.24 m and model parameter is •= 300 cgs units. Reference
and receiver heights. The source height of 0.03 m is
level is that due to inverse square law alone.
particularly difficult to quantify exactly over grass and
a small reduction in the stated height would provide a
for sound propagation near the ground. In each case the better fit of the model to the measured values in this
procedure is to determine whether a flow resistance pa- region.
rameter value can be found which results in model ex-
cess attenuation values which agree with observation. The field measurementspublishedby Embleton et al. s
If good agreement can be obtained then the particular are somewhat unique in that they include the results of
extensive direct measurements of the characteristic
parameter defines frequency dependentground charac-
teristic impedance and propagation coefficient functions impedance of grass covered surfaces (presumably the
which characterize the particular surface and which can same surfaces as those used for the propagation mea-
be usedto estimate the effeci of groundabsorptionat surements already considered). These measurements
other ranges, source-receiver geometries, etc. were madeusingimpedance
tubeandinterferenceat
oblique incidence techniques and are shown plotted as a
A. Canadian measurements function of frequency in Fig. 9 (measurements taken
from Fig. 8 of Ref. 8). The lines of best fit to the ob-
Embleton el al. 8 have described a series of measure- servational data for the real and imaginary components
ments of ground absorption made at N. R.C. in Canada of the impedanceare als0 shown,as givenby Embleton
for relatively short source-receiver separation and et al. The model parameter providing the best agree-
propagation very close to the ground. The measured ment with the observed propagation data has been shown
sound level as a function of frequency for a source and above to be (• = 300 cgs units. The model values of the
receiver each 1.22 m (4 ft) above a grass covered sur- normalized characteristic impedance components for
face and separatedby a distance of 15.24 m (50 ft) is this flow resistance are plotted in Fig. 9. The model
shown as the full line curve in Fig. 7 (values taken from values are in close agreement with the measured values
Fig. 6 of Ref. 8). The reference sound level was that throughout the frequency range. This excellent agree-
due to inverse square law alone and the source was flat ment confirms the ability of the fibrous absorbent model
to within+ 1.5 dB over the frequency range 0. i to 4.5 kHz.
Excess attenuations using the fibrous material model
were calculated for a range of flow resistances using
I I I I I I
Eq. (20) and this source-receiver geometry. The best
fit to the observations was obtained for a value of (• = 300 -

cgs units. The model values are shown superimposed

hR=O
03
on the experimental values in Fig. 7. They are in good
agreement with the observed sound levels, particularly
at low frequencies and up to the first interference mini- t -10

mum which is primarily due to the phase shift of the re-

_

flected wave (see Fig. 3). The positions of the higher

order minima are critically dependent upon the path •= 1.22
z•_20 - _

length difference and the small discrepancy between the o

NRC
position of the observed and model minima may be at-
tributed to uncertainty in the exact geometry of the ex-
perimental set up, e.g., effective height of source and
receiver above a grass covered surface.
-30
-

Having established the model parameter which best I I I I I I

fits the grass covered surface used for the measure- 100 200 400 800 1K 2K

ments in Ref. 8 ((• = 300 cgs units), further comparisons FREQUENCY- Hz

with observations can be made. The effect of lowering FIG. 8. Comparison between model results and field measure-
first the source and then the receiver to near the ground meQts taken from Re[. 8. Source height
while maintaining the separation at 15.24 m was mea- [5.2• m, a•d model pa•amete• [s •=300 c•s u•[ts. ReEe•eQce
sured by Embleton et al. and the results for two re- level [s that due to [•ve•se s•ua•e law •o•e.

J. Acoust. Soc. Am., Vol. 62, No. 4, October 1977

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 99.244.138.70 On: Sun, 18 May 2014 03:09:18
 831 C.I. Chessell:
Noisepropagation
alongan impedance
boundary 831

Summer, vector wind +5 m/sec; (2)Summer, vector

30
.....
i • I• I 1
/
)L
2(],,,,•,
•.•.•-•,
,,
•
NRC
roo Inclined track
•,©=Impedance
•-•
tube
Model
wind zero; (3) Winter, vector wind + 5 m/sec; and (4)
Winter, vector wind zero.
observed soundlevels at the closest microphone (R = 19.5
m) as the reference for calculating the excess attenua-
Parkin and Scholes used the

tion in each case. The measured values have been cor-

U - . O n rected for the effects of classical and molecular atmo-
Z x - -.13,._..•_
•
• O-
o c/u • a"o
-
spheric absorption.

The excess attenuations calculated using the model

[Eq. (21)]andthis source-receivergeometryare shown
plottedfor each«-octavebandin Fig. 11. A modelpa-
rameter of • = 150 cgs units has been used for these cal-
culations. The calculations for the shortest range
0 -20•
Z
- (R = 19.5 m) showthat significantpositive and negative
excess attenuations would be expected to occur even at
--3 -
this range so that the reference level 'usedby Parkin and
i i i i i i
100 200 400 800 1K 2K 4K Scholes does not accurately represent the source noise
FREQUENCY- Hz spectra. The excess attenuation curves given by these
authors would therefore not be expected to reflect the
FIG. 9. Comparison between model and field measurements
total excess attenuation at each range. Using these
of the characteristic impedance of grass covered soil taken
from Ref. 8. (• line of best fit to experimental data.) curves the excess attenuations for each range have been
Model parameter is cr= 300 cgs units. calculated using the values at R = 19.5 m as a reference.
The results are superimposedon the experimentally
measuredattenuations
in Fig. 10. Themodelparame-
to represent acoustic propagation conditions close to a ter of • = 150 cgs units was chosen to give the best over-
ground surface of finite impedance.

B. Measurements of jet noise by Parkin and Scholes R= 61 6m

•,--- ,•, •..

I•arkin and Scholes•-•7 carried out an extensive series
of measurements of the propagation of jet engine noise ß.],,•, ,/, '• .-o
close to the ground at Radlett and Hatfield in the United -10

Kingdom. The measurements were made over essential-

ly flat grass covered ground at ranges of 19.5, 34.7, -20

61.6, 110, 195, 347, 615, and 1097 m from a jet engine
noise source using microphones mounted 1.52 m (5 ft)
-10
j : I I' I I I i I I i I
above the ground. Various source heights were used,
here we compare the model predictions with the Radlett [. R=110m '"" •
,,
R=615m

0 ø ' -s,
data (reported in Ref. 15) for which the source height
was approximately 1.83 m (6 ft). The Radlett measure-
ments were made at weekly intervals over the period of .10
one year at times near midday and the results are con-
fined to temperature lapse or near neutral conditions. .20

Strong correlation was found between the observed at-

tenuations and the vector wind (the componentof the wind - : • I I I I

velocity along the line of microphone, positive if from ,,',. R=195m • ,•,
source to microphone). '• .'[:•N /' ',,, ;'',
'2,
Parkin and Scholes present excess attenuation curves
.10
for Summer and Winter conditions which are broken
downinto the three cases of vector winds of + 5 m/sec,
0, and - 5 m/sec. The attenuationsfor the - 5 m/sec RADLETT

case for both Summer and Winter are very much larger Summer 5 m/s
0
than the other cases, particularly at higher frequencies W•nter 5 m/s • --•

and longer ranges. This may be attributed to the for- 0 •--•

MODEL • :

mation of acoustic shadow zones due to upward refrac-

+•0
50 I
100 •
200 •
500 •
1K 2SK 51K
50 100 I I
200 500 I 11K21K 5KI
tion of the sound waves and these cases will not be con-
FREQUENCY - Hz
sidered further. The measured results for the first
FIG. 10. Comparison between model and field measurements
two cases, namely vector winds of zero and + 5 m/sec
of the excess attenuation in «-octave bandstaken from Ref. 15
are plotted in Fig. 10 (taken from Figs. 3 and 4 of Ref.
for various ranges. Experimental results are for Summer,
15) for the six largest ranges. For each range the ob- vector wind: +5 m/sec and 0 andWinter, vector wind: +5 m/
served excess attenuations in each •-octave
• band in the
sec and 0. Model results are for the parameter cr= 150 cgs
range 50 Hz to 4 kHz are shownfor the four cases; (1) units. Source height is 1.83 m, receiver height 1.52 m.

J. Acoust. Soc. Am., Vol. 62, No. 4, October 1977

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 99.244.138.70 On: Sun, 18 May 2014 03:09:18
 832 C.I. Chessell'
Noisepropagation
alongan impedance
boundary 832

the angle of incidence to the horizontal will be decreased

•-10
/I I ; • • • and the ground absorption correspondingly increased.
This effect may contribute to the large excess attenua-
tions observed for the vector wind-5 m/sec cases in
'o,,
'., ß / ....- .o
Refo 15.

Atmospheric turbulence would also be expected to con-

tribute to a reduction in the observed excess attenuation
values,-as discussed in Sec. IIID above. The presence
of the deep nulls in the observed sound levels indicates
',".!/ ."'
.,.,.,/,o
,9
•
• 616

that phase coherence has been maintained out to range

-"ø e•e

ß-----•
19S

34'7

of 1000 m, at least at low frequencies. An estimate of

•.- -..o 109'7
the effects of turbulence on the model excess attenua-

i
tions can be foundby using Eq. (27)ø The calculated re-
ß40 50 100200 500 1K 21K 5K 1QK sults for the longest range used by Parkin and Scholes
FREQUENCY- Hz (1097 m) are shown in Fig. 12 for several values of the
FIG. 11. Mode! excess attenuations _for the source-receiver turbulence strength parameter C. The experimental
geometryusedin Re_f.15 _foreach«-octaveband. Mode! measurements for this range are also shown for com-
parameter is or= 150 cgs u•its. Parameter is range. parisono The inclusion of turbulence in the model can
result in excess attenuations which are in better agree-
ment with the observation, however the general form of
all fit to the measured data. Note that the observation
the curves does not agree well with observation and it
of negative excess attenuations by I•arkin and Scholes is concluded that either refraction is more important
(as in Fig. 10) was suggestedby Wenze118
to be possible than turbulence in determining the excess attenuation at
evidence for the existence of surface waves. However
long distances or that the simplified method of allowing
the model results of Figs. 10 and 11 show that negative for turbulencedevelopedhere [Eq. (27)] is not adequate
excess attenuations are predicted by the consideration to describe the real atmosphere.
of the reflection of a spherical •vave alone.
The excess attenuations predicted by the model are in V. APPLICATION OF THE MODEL TO REVERSE
good general agreement with the observed values. The THRUST NOISE MEASUREMENTS
observed values are averages taken over six monthly In this section the application of the ground absorption
periods when presumably the properties of the surface model to a particular case of outdoor noise measure-
may have changed due to differing water content, nature ment is described to provide an illustration of the way
of grass covering, etc. The model results give best in which the model may be used. The effect of ground
agreement at low frequencies and shorter ranges where absorption is important in many cases of outdoor noise
the effects of atmospheric tu.rbulence and refraction measurements, e.g., measurement of vehicle noise,
would be expected to be reduced. At longer ranges the aircraft noise, etc. Here we consider the case of the
best agreement is with both the summer and winter vec- measurement of aircraft reverse thrust noise. It was
tor wind zero cases. The probable explanation for the desired to measure the "free field" spectral content of
lower excess attenuations for the + 5 m/sec vector wind the noise made by an aircraft applying reverse thrust
cases at longer ranges is the effect of refraction in the
downwind direction. If the simple two path geometry de-
picted in Fig. i is maintained, downwind refraction R=1097
m '
would be expected to increase the angle of incidence to
the horizontal of the reflected wave, thus reducing the
excess attenuation (see Fig. 4). Refraction may also
-10
! ' , ,, , , -.
I

contribute to an increase in the energy at the receiver

by focussing of the direct wave. If the appropriate tem-
zo ;"•
'"•'
•,v'•x
•0-4
perature and wind profiles were available it might be
possible to calculate the importance of these effects by
ray tracing. However, the application of ray tracing to
situations where significant changes in the acoustic re-
fractive index occur in distances of the order of a wave-
length (hs=1.83 m, hR= 1.52 m) is dubiousand will not
be pursued here. Refraction can also introduce the pos-
sibility of a number of other paths between the source
•

s0
HODEL
t

•00 200 sm 2K
0
0

and receiver (see, for example, Ref. 8) which will fur- FREQUENCY- Hz
ther complicate the excess attenuation. However the
FIG. 12. Comparison between model excess attenuation and
good general agreement of the observations with the
the observations of Ref. 15 for a range of 1097 m. Experi-
simple two path model suggests that these effects may mental results are for Summer, vector wind: 0 and + 5 m/sec;
only become important at much longer ranges. It should and Winter, vector wind: 0 and +5 m/sec. Model parameter
also be noted that in the upwind direction the effect of •r= 150 cgs units. Parameter is turbulence strength C (tad
refraction will be reversed. For the reflected wave sec m-•12).

J. Acoust. Soc. Am., Vol. 62, No. 4, October 1977

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 99.244.138.70 On: Sun, 18 May 2014 03:09:18
 833 C.I. Chessell:
Noisepropagation
alonganimpedance
boundary 833

10•

iii 90

70

5o I
100 J
200 ,
500 ,
1K I
2K 5•K 1OK
,
FREQUENCY- Hz

FIG. 13. Observations of reverse thrust noise of DC-9 air- FIG. 15. Corrected "free field" «-octave band reverse thrust
craft at Sydney Airport at two different microphone heights. noise levels as observed at microphone heights of 6 and I m.
Range from runway 165 m.

after touchdown. An observation position was set up a Calculations of the differential ground absorption be-
distance of 165 m from the East-West runway at Sydney tween the two microphone heights were carried out us-
Airport at a point opposite the position where aircraft ing the ground absorption model and a range of flow re-
normally apply maximum reverse thrust. Two micro- sistance values. In calculating these values some al-
lowance for the extended nature of the noise source was
phones with associated sound level meters and tape
recorders were used, one mounted at a height of 6 m made using the technique outline in Ref. 12. A source
extended in the vertical direction a distance D can be
above the ground and the second at a height of 1 m. The
terrain between the runway and the recording position considered as a vertical distribution of n sources sepa-
was approximately level and covered with short grass rated by a distance d suchthat d/h,,,= 0.1, where h,, is
of height 15-25 mm. the mean height of the extended source. These n ele-
1
mentary sources extend out to a distance • D above and
A series of reverse thrust events were recorded and below h,,. The effect of the extended source is to smooth
subsequently
replayedthrougha «-octavebandanalyzer. out the interference maxima and minima in the observed
The noise levels for each aircraft corresponding to the frequency spectrum. The n sources are combined on an
time when the aircraft was located directly opposite the incoherent basis using Eq. (21). In the case of reverse
recording position were identified. The sets of noise thrust noise, the nature and extent of the noise source
levels for each aircraft type were then normalized to is not well understood. As a first approximation in cal-
the same total sound pressure level and arithmetically culating the model results, the source was considered
averaged. Typic•al results for a DC-9 aircraft landing to extend a distance of 1 m above and below the engine
in an easterly direction are shown in Fig. 13 for the two height.
microphone heights. These curves are the result of av-
eraging five events recorded over a period of about 2 h. The closest agreement with the observations was ob-
The mean surface temperature during this period was tained with a model parameter of • = 150 cgs units. The
9 oc, the mean humidity80%andthe mean vector wind calculated differential attenuation between the two micro-
in the direction of propagation0.5 m/sec. The different phone heights is plotted in Fig. 14 for this model param-
effects of ground attenuation at the two heights is clearly eter. The model values are in good agreement with the
evident. The difference between the two curves was cal- observations, the main difference being at high frequen-
l
culated and is plotted in Fig. 14 for each r-octave band cies where the maxima and minima in the model curve
level. are not observed in practice. This suggests that the
noise source may probably be considered to be further
extended in the vertical than has been assumed. The
overall good agreement between the model results and

ß
10•.... ' • ' observation means that the model
ues may be used in this and other cases to correct the
observed sound levels to the corresponding "free space"
1
excess attenuation val-

levels. The resultant r-octave band levels for the two

microphones are shown in Fig. 15. If the mean of the
two sets of observations is taken as the "free field"

• ; MODEL
noise level in each band, these mean levels are within
-2C
+ 2 dB of the observed "free field" levels at the two mi-
crophone heights after correcting for ground effects.
• 200
50 100 ' • I'K 2•K $1K10
500 'K Thus the ground absorption model can be used to obtain
FR EOU E NCY - Hz a quite accurate estimate of the reverse thrust noise
F%G. •4. Comparison between model and observed dffœerenoes level frequency spectrum. A more accurate estimate
between noise •evels at microphone heiEhts of 6 a•d • m œor could be obtained by including recordings on a third mi-
each«-octave band. Model parameter is (r=Z50 oEs units, crophone at an intermediate height.

J. Acoust. Soc. Am., Vol. 62, No. 4, October 1977

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 99.244.138.70 On: Sun, 18 May 2014 03:09:18
 834 C.I. Chessell:
Noisepropagation
alongan impedance
boundary 834

Vl. DISCUSSION Aerospace Recommended Practice 866 (31 August 1964).

2U. Ingard, "On the Reflection of a Spherical SoundWave from
In this paper it is proposed that the propagation of an Infinite Plan," J. Acoust. Soc. Am. 23, 329-335 (1951).
sound over the ground be modeled by the use of acousti- 3I. Rudnick,"ThePropagation
of anAcousticWavealonga
cal properties of fibrous absorbent materials together Boundary," J. Acoust. Soc. Am. 19, 348-356 (1947).
with the appropriate theory of sound propagation along 4R. 13. Lawhead
andI. Rudnick,"Acoustic
WavePropagation
a boundary. The model predictions of excess attenua- Along a Constant Normal Impedance Boundary," J. Acoust.
tion have been compared to experimentally measured at- Soc. Am. 23, 546-549 (1951).
,

tenuations for propagation outdoors and shown to be in 5M. E. DelanyandE. N. Bazley, "MonopoleRadiationin the
Presence of an Absorbing Plane," J. Sound Vib. 13, 269-
good general agreement, both for short ranges and for 279 (1970).
ranges out to about 1000 m. Clearly the model could 6p. B. Oncley,"Propagation
of Jet EngineNoiseNear a Porous
be applied to the problem of predicting the ground atten- Surface," J. Sound Vib. 13, 27-35 (1970).
uation of aircraft noise for aircraft on or near the sur- ?H. G. Jonasson,"SoundReductionby Barriers onthe Ground,"
face. These predictions are required out to ranges of 5 J. Sound Vib. 22, 113-126 (1972).
or 6 km for use in developing Noise Exposure Forecasts ST. W. F. Embleton,J. E. Piercy, andN. Olson, "Outdoor
Sound Propagation Over Ground of Finite Impedance," J.
around major airports. However, the results described
Acoust. Soc. Am. 59, 267-277 (1976).
here, particularly the measurements of Parkin and
9M. E. DelanyandE. N. Bazley, "A Noteonthe Effect of
Scholes•5at long ranges, indicatethat groundabsorption Ground Absorption in the Measurement of Aircraft Noise,"
effects and the effects of atmospheric turbulence and re- J. Sound Vib. 16, 315-322 (1971).
fraction are closely linked and cannot be treated in iso- løM. E. DelanyandE. N. Bazley, "AcousticalPropertiesof
lation for long range propagation. Thus in applying this Fibrous Absorbent Materials," Appl. Acoust. 3, 105-116
model to the aircraft ground attenuation problem a com- (1970).

bined approach in which due account is taken of the me- llW. L. NyborgandD. Mintzer, "Reviewof Sound
Propagation
in the Lower Atmosphere," WADC Report 54-602 (1955)
teorological effects wouldbe required. A further factor (ADO67880).
which has not been considered here is the effect of an
12T.W. Putnam,"Propagation
of WavesAlonga FiniteImped-
uneven terrain. Incoherent scattering from a rough ance Boundary"; NASA TM X-56033 (September 1975) (N75-
boundary may contribute to the ground absorption atten- 32119).
uation, particularly at higher frequencies, and the im- 13L.A. Chernov,WavePropagationin a RandomMedium
portance of this contribution also requires investigation. translated by R. A. Silverman (McGraw-Hill, New York,
1960).

ACKNOWLEDGMENTS 14U.IngardandG. C. Maling,"OntheEffectof Atmospheric

Turbulence on Sound Propagated over Ground," J. Acoust.
The reverse thrust noise measurements described in Soc. Am. 35, 1056-1058 (1963).
this paper were made with the assistance of personnel of 15p.H. ParkinandW. E. Scholes,"TheHorizontalPropaga-
the Department of Transport, in particular Mr. B. G. tion of Sound From a Jet Engine Close to the Ground, at
Radlett," J. Sound Vib. 1, 1-13 (1964).
Harris and Mr. J. Davey, and their contribution is
16p.H. ParkinandW. E. Scholes,"TheHorizontalPropaga-
gratefully acknowledged. tion of Sound From a Jet Engine Close to the Ground, at
Hatfield," J. Sound Vib. 2, 353-374 (1965).
•TW.E. ScholesandP. H. Parkin, "The Effect of Small
Changes in Source Height on the Propagation of Sound over
l"StandardValues of Atmospheric Absorption as a Function of Grassland," J. Sound Vib. 6, 424-442 (1967).
Temperature and Humidity for Use in Evaluating Aircraft •SA.R. Wenzel,"Propagation
of WavesAlongan Impedance
Flyover Noise," Society of Automotive Engineers, Inc., Boundary," J. Acoust. Soc. Am. 55, 956-963 (1974).

J. Acoust. Soc. Am., Vol. 62, No. 4, October 1977

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 99.244.138.70 On: Sun, 18 May 2014 03:09:18