Content uploaded by Keith Ballagh
Author content
All content in this area was uploaded by Keith Ballagh on Mar 14, 2018
Content may be subject to copyright.
.
,
AMIiet/ atolls!it's 31 (1990) 245264
Noise of Simulated Rainfall on Roofs
K. 0. Ballagh
, Marshall Day Associates, PO Box 5811.
Wellesley Street, Auckland. New Zealand
(Reccivcd 9 October. 1989; revised
VCrsion reccived 21 February 1990;
accepted 28 February 1990)
A BSTRrl CT
N('11' Zculd, Id11u. , PIOb/e!lis 1111/1 noise 111 buildiiigs CUNTed bi' I'dintir/. 7/11. , If
d!, c 10 111g/I Jul'"^1/1 Idics undd11^hill'ei:g/lisii, /e o1'bt, 11drng. Expeli'111eiiid/in^of
Iheo!'eiicd/ 1101'A lid. , he'll c'd!'1'1ed owl io exdiiiiiie Ihe 417t'ci q1 "di'lolls 1'00/
<'011,111'tit'!lolls o11 ,101'., e 11'diiJiii!'/led 10 1'00nis hened!/I Ihe 1001 uiid to
11/11c. ingnic Ihe 111cc'hdii!'., Ills of 1101. ,e gcnc!'niloii. 7/1ei'c i's d xi 111ficd/11
1,111diioii 111 1101. ,c gc!lei'unoii bi' cliffei'e, 11 PIOfi/CT q1100/ c'/, Idd!'Jig, diffei'eiii
tinddriig 171dicii(!/s, diid (/^ffbi'e, 11 c'elmig ('on $11'11c'!ions. He CJD!'ei c/( of I' d
d(/at 11b, I'll JimS. , ill Ihe 1'00/ ., 11'11('1111'e genei'rilli' I'ethice 1101'se, 11'111i OOF, Ie
('o111/1b!, 1101i 11'01ii .soulit/ fibsoJp/I'd 111fi//.,. A whip/e plediciion 111eihod is
PI'c. Willed
,
,
INTRODUCTION
The noise produced inside buildings by rainfall impactin on their To f '
significant problem in New Zealand. New Zealand is a land of hi h rainf H
rates-in some areas average maximum rainfall rates over 10 min can
exceed 80 mm/h, and peak rainfall rates are even higher. Coupled with these
high rainfall rates, New Zealand favours a lightweight st to of b 11d'
Td't' Ib'Id' g-
Traditional building techniques utilise a timber frame with a Ii ht ' ht
steel sheet cladding on the roof. These two factors combine to roduce
buildings In which speech intelligibility is severely reduced dunn moder. t
to heavy rainfall.
App/^t, / At oilyits 0003-682X/'901S0350 C 1990 Elsevier Science Publishers Ltd, E I d
Printed in Great Britain
245
246
300
K. 0. 8,111ngh
BCHn
Galv anise d tr ough s ection
steel cladding <0.62mm)
160, H
Galva nis e d tr ough section
steel cladding <0.63mm or 047mm>
Corrugated galv anised st e eI
cladding <0.48mm>
From about 1970, long run trough section steel claddings, as distinct from
the traditional corrugated steel section (see Fig. I) became available. These
could be used with pitches of only a few degrees. A large number of
commercial and public buildings began to use these trough section claddings
for lightweight skillion roof construction. Skillion roofs effectiveIy eliminate
the attic space in traditional designs, and attach the ceiling directly to the
underside of the roof structural members. A widespread problem with rain
noise was experienced, particularly in school buildings where good speech
intelligibility is vital, but economical construction is sought.
Against this background, practical research was carried out in 1978 by the
Ministry of Works and Development, in conjunction with the Department
of Scientific and Industrial Research. This consisted of producing simulated
rain on a test hut and measuring the noise inside the nut with four different
rooflceiling constructions. After a lapse of several years further experi-
mental work was carried out during 1987, together with some consideration
of the theoretical aspects of noise generation of the impact of water droplets
on lightweight claddings.
This paper presents the results of the experimental work with simulated
rainfall on a number of roof and ceiling constructions, and conclusions are
drawn as to how rain noise can be reduced and insulated against. The basic
mechanisms of noise generation and the relationship between noise and
Golf ug a te d Pilaf o. 5 cement
sheets <6mm)
L
,
125. H
Fig. I. PIOfiles or roof claddings
*
Butyl rubber membrane
plywood
on
"
"
rainfall are also investigated, and some comparisons between theory and
experimental results are made.
it was decided at a very early stage to use simulated rainfall. Because
comparisons were to be made between various constructions It was
important that conditions could be held as constant as possible between
tests. This was possible only with simulated rainfall. it was realised that this
introduced some uncertainty in using the results for real rainfall, nowever
reasonable correspondence with natural rainfall was obtained (see below).
Noise qf siniu/died Juln/all o11 I o0/3
TEST ARRANGEMENT
Apparatus for rainfall simulation
The simulated rainfall was produced by a 25 mm (internal pipe diameter)
square spray, full jet, wide angle nozzle which was located 2.8 in above the
centre of the roof. The water supply to the nozzle was taken from the water
mains through a pressure reducing valve. The pressure was set to about
55 kPa. The water spray was directed upwards so that water drops fell from
about 3 to 35 in under the action of gravity, having zero vertical velocity at
the top of their arc. In the first series of measurements rainfallrates between
17 and 95 mm/h were simulated. In the second series of measurements, a
single rainfall rate of about 80 mm/h was simulated.
247
Rainfall measurement
Rain falling on the known area of roof was channelled into a guttering
system, the run off collected for 90 s and then accurately weighed. Under
normal conditions the rainfall rate was stable though it varied slightly from
day to day. Corrections were made to the noise levels to account for the
slight differences in rainfall rates between tests (see below).
Test hut
IdealIy the test roof should have radiated the sound energy into a
reverberant room qualified under ISO procedures for measuring sound
power. Economic and time considerations prevented the construction of a
specialised facility with sufficient volume and other required acoustic
characteristics. Although the results are considered internally consistent,
comparison with other situations may be maccurate at lower frequencies.
A test hut of floor plan 2.4 in x 3.0 in (internal dimension) was erected on a
POOF 11ne
-------- ---
co
ni
ni
Q
A
I>._
01
Q
o
Q
^
dlrectlon OF Fall
L
gutter
12nm
I^^
BituFoam
2400
2600
Plywood
POOF claddlng
. "g ~\~
.
concrete
Fig. 2. Plan and section of test hut
Floor
12nm Sop'tboard
18nm Customwood
o
.
Q
co
19
4:..
co
Section A-A
.
o
CD
^
co
co
co
.*
>^
o
CS
^
;^
...
.~
>
.
.
*
concrete floor. The walls of the hut were of 100 mm x 50 mm timber framing
clad internally with 18 mm thick medium density wood fibre board and
externalIy with 12 mm thick plywood. The roof pitch was 3' (see Fig. 2)
There was a vibration break between the roof and the walls consisting of
30 mm thick bitumen impregnated plastic foam.
Because of the small size of the room the sound field lacked diffusion
below 500 Hz (Schroeder's large room frequency) although in other contexts
it has been found that rooms can stillbe considered diffuse down to half this
frequency. ' Below 250 Hz the results should be viewed with caution if
comparisons are to be made with other test setups. Fortunately the
frequencies below 250 Hz were unimportant to the loudness or speech
interference of any of the roof constructions tested.
The uniformity of rainfall was measured over the roof by putting
containers of known volume at various points on the roof and timing how
long it took to fill them. The rainfall in tensities at various positions are
shown in Fig. 3. it can be seen that there is a wide variation in intensity,
certainly more than is desirable. However, by the time this was discovered
half way through the second series of tests, considerable data had already
been collected which would have been invalidated if the rainfall simulation
Non, e of 511"Minie, / 1,111fti// on loon
249
40
60
80
100
120
100
80
100
80
60
60
,,,,, in Ih, *
Fig. 3. Variation of rainfall intensity across the roof (ISO-rainfall contours)
120
80
140
100
120
140
60
,
,
100
80
60
40
20
250
had been altered. The reasonable agreement between simulated rainfall and
natural rainfall (see below) suggests that the variation in rainfall rate did not
have a large effect on overall radiated sound intensity.
Acoustic measurements
The space and time averaged sound pressure level in the test hut was
measured with a 12.5 mm diameter microphone mounted on the end of a
rotating boom which swept out an inclined arc through the room of about
1.5 in diameter. The measurements were recorded on a tape recorder and
later analysed with a 1/3 octave band real time analyser and computer.
It was estimated that the overall noiselevel for each roof was reproducible
to +/- IdB, and spectrum levels to +/- 2 dB. Only variations between
roofs of more than these values are significant.
K. 0. Bundgh
Correction for rainfall rate
The first series of measurements was done at four rainfall rates between 17
and 95 mm/h. From these measurements the relationship between simulated
rainfall rate and sound intensity for this apparatus was found to be
approximately
(1)
LitR) = L, (R, ) + 20 log, , R/R,
where L, is the sound intensity radiated by a given roof and is a function of
the rainfall rate R, and R is a reference rainfall rate. The second series of
measurements (1987) were done at only one rainfall rate (80mm/h). Small
variations in rainfall rate between tests were corrected using expression (1).
All results in this paper are for a nominal rainfall rate of 80 mm/h.
The factor relating the logarithm of rainfall to sound intensity for the
simulated rainfall was found to be approximately 20, as compared to that for
natural rainfall of 17.3 (Ref. 2) and that predicted from theory of 154 (see
below). This indicates that the simulation of the rainfall in terms of range of
drop sizes and the variation of this range with rainfall rate, although not
perfect, did provide characteristics similar to natural rainfall.
Description of roofing systems tested
A total of nine different roof systems with several variations were measured.
In the first series of measurements (1978) the emphasis was on solving an
immediate problem with the standard school roof construction. One type of
trough section steel cladding profile was used with four different under-roof
constructions.
Details of the roofs are shown in Fig. 4.
co
System I&5&6
___----400mn wide steel troughs
Building paper
150 x 50 tinbe" Joists
12nm Softboard
System a
400mm wide steel troughs
Building rag Felt
38nn Woodtex
150 x 50 Joists
12nm Softboa"d
~~~~,.,, ^^11din^, rag, Felt k n
System 3
400mn wide steel troughs
150 x 50 timber Joists
~~'~"- 12nm Softboard
400nn wide steel troughs
-,_,,, Building rag Felt
System 4
___,_--- 12nn Soptboa"d
150 x 50 timber Joists
12nn SDFtboa"d
System 7
corrugated steel
Bulldlng paper
150 x 50 timber Joists
.
Fig. 4. Sections or roofs ICstcd
System 8
6nn corruga ted Fibrous
cement sheets
Build!rig paper
150 x 50 timber Joists
12.1m Softboa"d
System 9
___,----~ 1.0nn Butyl rubber
~~~~~' 12nn PI wood
150 x 50 timber Joists
12nn Softboard
:a:
Q
,,
r*
.
~\
*
^
^
^
.~
^
~
=:..
.~
a
\,,
^
.~.
o
=
\
C
Q
>
I\,
1/1
^
252
System I had 400 mm profile, galvanised trough section cladding (0.63 mm
thick) on 150mm x 50mm purlins at 600mm centres with 50mm x 50mm
nogs at 600 mm centres (referred to as frame A) with a 12 mm softboard
ceiling.
System 2 was similar but had a layer of building rag felt supported on
38 mm Woodtex Woodwool and 50 mm dwangs at 600 mm centres.
System 3 was similar to I, but had alayer of building rag felt supported on
25 mm x 50 mm boards laid on their flat, with 50 mm gaps between
adjacent boards, and solid blocking at 600 mm centres.
System 4 was similar to I, but had alayer of building rag felt supported on
12 mm plywood with heavy duty malthoid (bitumin impregnated paper)
drapped over the purlins and cut around the solid blocking.
The second series of measurements was designed to explore more
variables in the roof design such as the cladding profile, cladding thickness,
roof structure, and different roof materials. it was intended that these tests
would produce a wider data base to be able to predict rain noise generation
by more types of roof and would also shed light on the mechanisms of rain
noise generation.
System 5 was similar to I, but had 400 mm profile galvanised trough
section cladding (0.47 min thick) on frame A
System 6 was similiar to System I, but had a 300 mm profile trough section
instead of the 400 mm profile. TITree dinei. Grit frames were tested with
System 6; purlins and nogs at 1200 mm centres (frame C), purlins at 600 mm
centres and nogs at 1200 mm centres (frame B), purlins and nogs at 600 mm
centres (frame A).
System 7 had 0 48 mm thick corrugated galvanised steel cladding on
frame A.
System 8 had 6 mm thick corrugated fibrous cement sheets on frame
System 9 had a 1.0 min thick Butynol membrane roof on 12.5 mm thick
plywood on frame A. System 9 was also tested with the softboard ceiling
removed.
K. 0. Baling/I
A
,
The overall results are given in Table I and in Figs 5-8. Table I gives the A
weighted sound intensity level, the Noise Criteria rating (NC) and the Speech
Interference Level (SIL) of the sound intensity level. Although NC and SIL
are normally applied to sound pressure levels, it is useful in this context to
apply them to the sound intensity levels so that the effect of roof area and
room absorption can be removed.
RESULTS
*
TABLE I
Sound intensity Levels Radiated into the Test Hut at 80 mm/h Simulated
Rainfall
Noise of simulated rail!full on roofs
Desci. ip!1011 of 1001
System I
System 2
System 3
System 4
System 5
System 6 roame A)
System 6 drame B)
System 6 drame Q
System 7
System 8
System 9
System 9 minus softboard
A 11'e!:gh!ed
sound in lensi!j.
(dB)
59
54
52
50
58
56
55
55
53
43
51
62
Speec/I
mrei/ei. ence
fobe/ (dB)
253
80
53
48
46
43
52
50
50
50
49
36
44
56
Noise
t'rifei'Iu
(NC)
ai
E
\
=
D.
,,
70
"
L
co
.
60
55
50
48
46
54
52
50
50
47
39
47
56
-
"
>
o
.
50
>
+I
.r,
in
=
"
+,
=
H
40
.
=
=
o
co
30
^,^ ^
20
Fig. 5. Sound intensity radiated by roofs due to simulated rainfall of 80mm/h. (^,
System I; -0--0--, System 2; -C----C. -, System 3.1
10
\^^
\
o
31.5
63
125
250
500 1000 2000 4000 8000 16000
Froq, nrey Hz
254
80
co
E
\
=
a
,,
co
L
co
.
70
K. 0. Bundg/I
60
F1
01
>
co
.
>
+,
H
co
C
co
+I
=
H
.
.
=
o
co
50
40
a
30
20
Fig. 6. Sound intensity radiated by roofs due 10 simulated rainfall of 80 mm/h. (-^,
Systeml; -0---a, System 5; -C. ---, 3--, System 6. )
10
o
\
31.5
63
Comparison among roofs
System I is a basic roof construction. High levels of sound are produced by
simulated rainfall on this roof-in a typical room the noise levels would be
70 dB(A) or more, obviously making speech communication very difficult.
All other systems are quieter than this, and will be discussed in comparison
with this base construction.
In System 2, alayer of building rag felt under the roof cladding reduces the
vibration energy transmitted into the purlins and ceiling. Sound trans-
mission across the ceiling cavity is reduced by the absorption of the Woodtex
Woodwool. Overall, a 5 dB reduction is obtained.
System 3 also has the building felt, and substantially increases the mass of
the roof with the 25 mm x 50 mm boarding. A 7 dB reduction from System I
is obtained.
,
\
125
\^
250
\
500 1000 2000 4000 8000 16000
Frequer"y Hz
DISCUSSION
BO
Noise of sini!,/alert rni'"/all on roq/3
01
E
\
=
a
,I
ai
L
co
.
70
60
F1
co
>
co
.
>
~
H
co
C
co
+,
=
H
D
C
=
o
co
50
40
30
a
a q
'6 '''6
^,,
20
Fig. 7. Sound intensity radialed by roofs due to simulatcd rainfall of 80 mm/h. (^.
SystenT I; 0---0-. System 7; -C. - -0-. System 8.1
10
Q
255
^.
o
System 4, with solid 12 mm plywood underneath the cladding, increases
the mass of the roof, and some reduction in vibration is achieved by the use
of building felt under the roof cladding. A 9 dB reduction is obtained
System 5 is identical to System I, but with thinner steel cladding (0.47 mm
versus 0.63 mm). The difference in mass would be expected to produce a
2.5 dB increase in level for System 5 if the roof cladding behaved as a simple
limp mass. However a slight reduction was measured. This may have been
because the expected variation is small, and is about the same as the
estimated repeatability. Alternatively it may be because the claddings are
quite stiff and undamped
System 6, with a slightly different profile, was 3 dB quieter than System I,
while System 7, with a corrugated profile, was 6 dB quieter than System I.
The exact profile of the cladding obviously has an effect on the noise
generation, but for reasons that could not be explained by the simple model
developed below
The tests on three different frames with System 6 revealed no significant
difference in noise radiation. it was concluded that the spacing of the ceiling
Q
31. S
'a
63
e e-Q
125
\
a
250
a
500 1000 2000 4000 8000 1600o
F, ~us, rev Hz
a
@
256
80
tv
E
\
=
a
^
co
L
co
.
70
K. 0. Balingh
60
F1
co
>
co
.
>
+,
*,
un
=
a,
+,
C
H
D
=
=
o
co
50
40
30
,
0.6
20
,.
a
Fig. 8. Sound intensity indiaied by roofs duc 10 simulated rainfall of 80mm/h. (-----,
System I; -0---0-, Sysieni4; -C----C. -, System 9; -A----A--, System 9b. )
members, and hence structural stiffness of the roof, cannot be used to
control rain noise.
System 8 was the quietest of the systems tested. The mass and damping of
the fibre cement board contribute to a 16 dB reduction over System I.
System 9, using a Butynol membrane over plywood achieved an 8 dB
reduction from System I. it is also interesting to compare System 9 with
System 4, the main difference being the replacement of the steel cladding of
System 4 with a Butynol rubber membrane. Very similar results were
obtained, indicating that the extra damping provided by the Butynol rubber,
and possible cushioning of the impact by the rubber, had little effect on the
noise generation and transmission.
System 8, with the softboard removed, was about 11 dB noisier than
System 8 with the softboard ceiling.
Comparison with natural rainfall
it is interesting to examine the difference between the simulated rainfall and
natural rainfall by comparing System I with Dubout's measurements. The
10
A ." A
o
31.5
,
b.
.
63
,
A
125
A
250
500 1000 2000 4000 8000 16000
Froqu. rev Hz
,
*
BO
Noise of $1mu/died ruin/all on roofs
CU
E
\
=
a
,,
70
co
L
co
.
60
.-
"
>
co
.
>
+,
.H
un
C
co
+,
C
H
.
=
.
a
co
50
@
I
40
II
- _,. _ -a,
30
20
\
Fig. 9. Comparison or soundintensity produced by natural and simulated rainfall. (^-,
natural rainfall (Dubout); -0- - -0-, simulated rainfall (see text).)
10
\
257
\
o
results for System I were corrected by subtracting off the measured insertion
loss of the softboard. This was then coinparable to Dubout's experimental
set-up, although the roofing profiles were somewhat different. it can be seen
from Fig. 9 that the simulated rainfall is about I-5 dB less than the natural
rainfall. Some of this variation could be due to the differencein profile, or to
other differences between test set-ups, but some could also be due to
inadequacies in the simulation of rainfall. Although the agreement is not
perfect, it is probably adequate for the purposes of the experimental work,
which were primarily comparisons between constructions.
Applicability to other situations
While the roofs tested were typical New Zealand constructions, it is
interesting to consider whether the results are generally applicable to other
countries using different building materials and with different rainfall
characteristics. In many countries rain noise is not a problem because
traditionally, heavy materials such as clay tile or slate are used for the
\
31.5
\\
b
63
125
250
500 1000 2000 4000 8000 16000
Froqoorrey Hz
258
roofing material, and/or because a loft space separates the roof from rooms
beneath. However in many newly developed countries and countries in
which winters are not severe lightweight building styles are common. in
these countries and particularly where thin steel claddings are common the
results in this paper should be directly applicable.
The simple theory developed below should also allow some extrapolation
to similar but non-identical materials and constructions. For each country
some data on the expected occurrence of various rainfallratesis required so
that the required level of protection against rain noise can be assessed.
K. 0. Bandgh
Although the principal direction of the work reported here was
experimental, it was considered important to gain solne understanding of
the mechanism of rain noise generation. A simple theoretical model was
developed which explains the major features of rain noise.
MECHANISM OF NOISE GENERATION
Impact force
Consider a water drop of radius R falling at terminal velocity Pandimpact-
ing on an infinite limp sheet of surface mass M. The assumption of limpness
means that the coincidence frequency is above the frequency range of
interest, and therefore the noise radiated can be found simply from the point
force applied by the impact of the drop.
When the water drop hits the sheet it will be rapidly decelerated until the
forces within the drop exceed the surface tension of the drop. Once the drop
breaks, the water will flow outwards from the the point of impact, with the
downwards velocity of the drop being constant. The force on the roof is
given by the rate of change of momentum of the drop, first as it is
decelerated, and then as its downward velocity is directed radially outward. '
Because of theftattened shape of a raindrop, themitialdeceleration phase
will be very short. Peterson' assumes that the second phase (flow phase)
be o1ns immediately, and that the velocity of the drop as it flows down and
sideways is the same as the falling velocity at impact. The force then is
.
,
F(I) ' pitR2 ,,,
- ---,- + - - -I
2R
I :^ ^
~3P
2R 5R6
^::;I:i^^
3P ~ ~3P
(2)
.
.
Noise radiation
The responseofthe roof to this pointimpact can be found by considering the
point impedance of the roof. Cremer ei a/.' give the point impedance of an
infinite fiat sheet in the frequency region below the concidence frequency as
Z = 81^'17
(3)
where Z is the mechanical point impedance, B' is the bending stiffness and
M is the mass per unit area of the sheet. The velocity of the plate can then be
found, and the sound power radiated calculated
PI'
P-
47rcn?~
where Pts the sound power radiated on one side of the sheet, c is the speed of
sound in air, p is the air density and F the point force.
This is a surprisingIy simple expression. Note that the only significant
property of the sheet is its surface mass M, no other property of the roof
cladding appears. However for finite sized roofs damping is also a significant
factor. 5
To reduce the generation of noise by rain, either the surface mass of the
roof cladding must be increased (though there are obvious practical
limitations) or the force of impact of the water drops must be decreased or
the transmission of vibration to the ceiling underneath must be reduced.
Norse of siniu/u!ed mill^/ on fool\
259
Raindrop characteristics
Because the magnitude and duration of the force is dependent on the size of
the water drop, it is necessary to know the range of drop sizes and relative
distribution in natural rainfall. If some degree of temporal and spatial
averaging is involved (as it always will be in any practical situation), then a
reasonable approach to the statistical characteristics of natural rainfallis an
exponential distribution of raindrop size, a well known model being the
Marshall-Palmer distribution. 6
This takes the form
N =N e~AD
(5)
where D is the diameter, N, 60 is the number of drops in unit volume of
space and N, is the value of N, for D = 0. Marshall and Palmer give the
values of parameters as N, = 8000 in ~ ' mm ~ ' and A = 4-1R ~ 0'2 I mm ~ I,
where R is the rainfall in mm/h.
(4)
260
K. 0. Bundgh
A simple empirical formula for the terminal velocity of raindrops in still
air is given by Best7
P. = 9.58 a - exp I-(D/0,855)"'"l)
(6)
where I", is the terminal velocity in in/s.
Although individual rain showers may differ significantly from the
Marshall-Palmer distribution, and short-lived drops may not have had time
to reach terminal velocity at impact, these characteristics in expressions (5)
and (6) are reasonable for an investigation of the mechanisms of rain noise
generation.
A computer program was written to evaluate the contribution of each
drop size to the overall radiated sound power, taking into account the
change in distribution of drop size with rainfall rate. The predicted radiated
sound power level varied with 15.4 times the logarithm of the rainfall rate.
This can be compared with the coefficient measured by Dubout for natural
rainfall of 17.3 and for the simulated rain of this paper of 20.
it is interesting in passing to note that the major contribution to rain
noise is from larger sized drops even though the number of drops/in'
decreases sharply with drop size (Fig. 10). Not only do larger drops have
more mass, but their terminal velocity is higher and the duration of impactis
longer. (The sound power is proportional to about the fifth power of the
drop size. ) Although much of the noiseis produced by large size drops, most
.
,
10 '
10
I\
^
' 10
E
E
in
E
.-.
z 10
10
4'
<:3
4> 4, ~
e,
3
a
Drop size D Cmm>
Fig. 10. Distribution or numbers or drops with drop size
10 ~'
\
,;a
6*
t.
I
?
'3
,:,,
,^.,
4
5
,
*
,
.
Noise offinIu/alert rain/all on roofs
of the water volume in rainfall is made up by the smaller drop sizes because
of the much greater number of smaller drops. At 80 mm/h most of the noise
is generated by drops in the range from 2 to 6 mm though most of the volume
of rainfallis from dropsless than 4 mm (half the volume from dropsless than
2 mm).
Comparison with measurements
The sound intensity level was predicted for two different roofs at a rainfall
rate of 80 mm/h. The roof claddings were 4.7 kg/in' and 9 kg/in'. These were
chosen to approximate Systems I and 8 respectively. For Systems I and 8,
the measured insertion loss of the softboard ceiling was subtracted. The
results are shown in Figsll and 12n can be seen that the agreement is
excellent for System 8 (corrugated fibrous cement), but poor for System I
(trough section steel).
It is surmised that the fibrous cement most closely fullils the assumptions
of the simple theory, in that the damping is sufficiently high so that the
forced vibration is more important than the resonant vibration. With the
261
BO
CU
E
\
=
a
-
a,
L
co
.
70
60
F1
U
>
ai
.
>
+,
+,
co
C
co
+,
C
H
.
=
=
a
co
50
^
' ^" G. 0- 0. ' e-- @- a
a- d
40
ag
a
30
20
Fig. 11. Sound intensity produced by 80mm/h rainfall. (-a--0-, measured systeml
(without ceiling); ---, predicted for 47 kg/in' roof. )
10
\
o
Q
31.5
^
63
125
Q
250
b
500
1000 2000 4000 8000 16000
Froq","y Hz
262
80
ai
E
\
=
a
.,
70
K. 0. Bd//ag/I
co
L
co
.
60
F,
co
>
co
.
>
+,
.H
un
=
"
,
=
H
.
=
=
a
co
50
40
30
V
20
Fig. 12. Sound inICnsity PIOduccd by 801nm/11 lainfall. (-C----C--, nieasured systciii8
(without ceiling); ^, 131'cdicied for 9 kg/in' 1'00f. )
steel cladding, the damping is assumed to be less, and resonant vibration
which is not accounted for by the simple theory is a significant contribution.
The theory could be extended to account for resonant vibration and the
effect of damping. However, damping factors for the roofs were not
available. This would be a fruitful areain which to do more work to see if the
measured variation in damping between roofs could be used to improve the
agreement between theory and experiment.
it should be noted that the simple model described above is successful in
predicting the overall spectrum shape of both roofs.
10
o
.
31.5
"
63
,
.25
250
500 1000 2000 4000 8000 16000
Froqu. rev Hz
Q
PREDICTION METHOD
For evaluating potential rain noise problems, a simple prediction method
has been developed from the measured data. The Noise Criteria produced in
a room by rainfall on its roof is given by
(7)
'Nc ' F1 + F2 + F3 + '4
*
~\
.
Norse o131mu/died ruin/all on loon
Skillion roofs
Steel cladding (thickness 0 4741 62 mini
on 150 x Soloists
to) 400 mm profile trough section
(b) 300 mm profile trough section
(c) corrugated profile
Butynol membrane on 12 mm plywood
Corrugated fibrous cement 6 mm
TABLE 2
Rain Noise Constants
(for use in eqn (7))
Descrip!1011 of roof tolls!ruclion
where
F, = constant for each roof/ceiling construction (see Table 2)
F2 = 10 log, 4 where A is the area of the roof (in')
F3 = - 10 log S, where S is the equivalent sound absorption area in the
room (in2)
F4 = 1710g R where R is the rainfall rate (mm/h)
LNc = the Noise Criteria rating within the room due to rainfall.
Desirable maximum levels of background noise depend on the purpose of
a room, size of room and acoustical characteristics (e. g. reverberation time).
Recommended levels can be found for instance in Be ranek '
Notes: Addition of (a) 12 mm softboard ceiling under
Joists, - 11; (b) Fibreglass in ceiling cavity, -4; (c) 12 mm
plywood sarking under steel cladding, -9
263
F1
36
32
30
27
16
The generation of rain noise has been investigated experimentalI and
theoretically. A number of different roof claddings and ceiling structures
have been investigated. Lightweight roof claddings can potentially produce
high levels of noise in rooms underneath dunno rainfall. Increased mass in
the roof cladding, sound absorption in the ceiling cavity, vibration isolation
of the ceiling and variations in cladding profile have all been shown to reduce
the transmitted noise.
A simple method has been developed to predict the level of noise due to
rainfall for a range of roof constructions.
A theoretical model has been developed which was reasonably successful
in explaining a number of features of rain noise. Theseincluded the variation
CONCLUSIONS
264
of noise with rainfall rate, the overall spectrum shape, and for well dampe
claddings, the absolute sound intensity levels.
ACKNOWLEDGEMENTS
The author would like to extend his grateful thanks to the following people
who contributed to the measurements and information: R. S. Carter, H.
James, L. S. Wong, S. Potter, C. WassiliefT and H. Larsen.
K. 0. Baling/I
REFERENCES
I. Davy, J. L. , Dunn, I. P. & Dubout, P. , The variance of decay rates measured in
reverberation rooms. Arusiica, 43 (1979) 12-25.
2. Dubout, P. , The sound of rain on a steel roof. I. Sound Fib. , 10 (1969) 144-50.
3. Peterson, B. , On the structural acoustic effects of liquid drop impacts. Lurid
Institute of Technology Report TVBA-3020, November 1985.
4. Cremer, L. , Heckl, M. & Ungar, E. E. , Silli('/wJ'e-Bullie Soiliid. Springer-Veilag,
Berlin, 1973.
Fahy, F. , Soliiid tint/ SII'uciiiiw/ rift', 1110/7. Academic Press, New York, 1985.
5
Marshall, J. S. & Palmer, W. MCK. , The distribution of raindrops with size
6
1011/11u1 o1' MCIeoi'o10g. I', 5 (1948) 165-6.
7. Best, A. C. , The size distribution or raindrops. Qud!'jelly. JouJ'rid/ q/' Ihe ROJ, d/
MCIcoi'o10gi't'd/ SOC'leiJ, , 76 (1950) 16-36.
BeTanek, L. L. , Noi'se ,!, Id Fibi. diio, I ColliJ. o1. McGraw-Hill, New York, 1971.
8
,
A