Mar 07, 2016
Acoustical Instruments and Measurements May 2015, Argentina
SCATTERING COEFFICIENT
FEDERICO DAMIS1, NAHUEL CACAVELOS
2
Universidad Nacional de Tres de Febrero (UNTREF), Buenos Aires, Argentina.
[email protected], [email protected]
2
Abstract A one-dimensional (1D) diffuser surface was designed, built and measured
according to ISO 17497-1 standard. Basic theory behind diffusion and scattering was analyzed
and applied in the design phase. The design included the use of randomly generated number
series with low autocorrelation in order to improve the diffuser efficiency. The diffuser was
scaled down by a factor of 2, in order to make the measurement process easier and fit the
requirements imposed by the standard. Measurement methodology was fully described and the
scattering coefficient was calculated, as well as its corresponding standard deviation and
measurement uncertainty.
1. INTRODUCTION
Diffusion, in acoustics and architectural
engineering, is the efficacy by which
sound energy is spread evenly in a given
environment. A perfectly diffusive sound
space is the one that has certain key
acoustic properties which are the same
anywhere in space. A non-diffuse sound
space would have considerably different
reverberation time as the listener moved
around the room. Virtually all spaces are
non-diffuse. Spaces which are highly non-
diffuse are ones where the acoustic
absorption is unevenly distributed around
the space, or where two different acoustic
volumes are coupled [1].
Diffusors (or diffusers) are used to treat
sound phenomena in rooms such as
echoes. They are an excellent alternative or
complement to sound absorption because
they do not remove sound energy, but can
be used to effectively reduce distinct
echoes and reflections while still leaving a
live sounding space. Compared to a
reflective surface, which will cause most
of the energy to be reflected off at an angle
equal to the angle of incidence, a diffusor
will cause the sound energy to be radiated
in many directions in time as well as
spatially (as seen in Fig.. 5).
There is a growing trend away from the
traditional use of absorbers on the rear wall
of auditoria towards the use of diffusers.
[2] Fig. 1 shows diffusers applied to the
rear wall of Carnegie Hall in New York.
This form of reflection phase grating was
the starting catalyst for modern diffuser
research. The diffusers were installed in
Carnegie Hall because a long delayed
reflection from the rear wall caused an
echo to be heard on the stage, making it
difficult for musicians to play in time with
each other. Adding diffusers dispersed the
reflection, reducing the reflection level
arriving on the stage and consequently
making the echo inaudible. The diffusers
also improved spaciousness on the main
floor by uniformly diffusing rear wall
reflections and masking echoes from the
boxes in this case. [3]
Figure 1. Schroedder diffusers applied to
the rear wall of Carnegie Hall to prevent
echoes.
The visual and aesthetic aspect of
diffusers is also a relevant matter in
Acoustical Instruments and Measurements May 2015, Argentina
diffuser design. The acoustical treatment
needs to complement the visual appearance
of the room to be acceptable to architects.
If the visual aesthetic is not agreeable, the
treatment will have to be hidden behind
fabric; however, this would turn the side
wall diffusers into absorbers. [2]
In order to show how diffusers disperse
reflections finite difference time domain
(FDTD) models can be used. Fig. 2 shows
a cylindrical wave reflected from a planar
hard surface. An impulse is generated, and
so a single cylindrical wavefront is seen in
frame 0, which is traveling from right to
left towards the surface. The wave simply
changes direction on reflection, traveling
back in the specular reflection direction,
where the angle of incidence equals the
angle of reflection. The reflected
wavefront is spatially unaltered from the
incident sound. Consequently, the sound
from the source reflects straight back, and
is unchanged and not dispersed.
Figure 2. Cylindrical wave reflected from a
flat surface.
Fig. 3 shows the effect of changing a
surface to help disperse reflections. In this
case, part of an ellipse is used. It can be
seen that the reflected wavefront is more
bowed. The change is not great, because
the curve of the ellipse was quite gentle;
even so, in the far field the sound will be
more spatially dispersed. The wavefront
generated is still very ordered, however, so
although single semicylinders or ellipses
are good at spatial dispersion, they are not
the best diffusers, because temporal
dispersion is not achieved.
Figure 3. Cylindrical wave reflected from a
curved surface.
Fig. 4 shows the effect of using a
Schroeder diffuser; the reflected wavefront
is much more complex than the previous
examples. Inspection of the different
frames shows why this complexity arises.
Sound can be seen taking time to
propagate in and out of wells, causing parts
of the reflected wavefronts to be delayed.
The different depts of the wells cause
different delay times, and the resulting
interference between the reflected waves
forms a complex pattern.
Acoustical Instruments and Measurements May 2015, Argentina
Figure 4. Cylindrical wave reflected from a
Schroeder diffuser.
Figure 5. Spatial and temporal dispersion
generated by a Schroeder diffuser.
One of the most significant occurrences
in diffuser design, if not the most
important event, was the invention of
phase grating diffusers by Schroeder.
Apart from very simple constructions,
previous diffusers had not dispersed sound
in a predictable manner. The Schroeder
diffuser offered the possibility of
producing optimum diffusion, and also
required only a small number of simple
design equations. Single plane or 1D
Schroeder diffusers consist of a series of
wells of the same width and different
depths. The wells are separated by thin
fins. The depths of the wells are
determined by a mathematical number
sequence, such as the quadratic residue
sequence. Single plane diffusers cause
scattering in one plane, in the other
direction, the extruded nature of the
surface makes it behave like a plane
surface. Because of this, it is normal to just
consider the plane of maximum dispersion.
Plane waves are reflected from the bottom
of the wells and eventually re-radiate into
the space. All these waves have the same
magnitude but a different phase because of
the phase change due to the time it takes
the sound to go down and up each well.
Consequently, the polar distribution of the
reflected pressure from the whole surface
is determined by the choice of well depths.
Schroeder showed that by choosing a
quadratic residue, the energy reflected into
each diffraction lobe direction is the same.
Nevertheless, it is important to note that
lobes can be generated in the energy polar
patterns due to surface periodicity.
Figure 6. QRD diffuser cross-section.
For the design theory to be correct,
plane wave propagation within the wells
must dominate. Consequently, an upper
frequency for the diffusion to follow the
simple design principles can be found
from:
(1)
where is the minimum wavelength
before cross-modes in the wells appear,
and w is the well width. Above this
Acoustical Instruments and Measurements May 2015, Argentina
limit dispersion will continue to occur
because these are complicated
structures. Consequently, this is just a
limit of applicability of a theory, and
not necessarily an upper limit for the
diffusion quality. Schroeder diffusers
work at integer multiples of a design
frequency, f0. The design frequency is
normally set as the lower frequency
limit, given by:
(2)
where is the maximum number of
the sequence used, c is the speed of
sound, N is the amount of wells and
is the maximum well depth.
Similarly to fmax, this design frequency
is not the lowest frequency at which the
surface produces more dispersion than
a plane surface, it is just the first
frequency at which even energy
diffraction lobes can be achieved. It has
been shown that Schroeder diffusers
reflect differently from a plane hard
surface an octave or two below the
design frequency [4].
2. DIFFUSER DESIGN
An optimized diffuser was designed
and built, taking into account the principles
of diffusion and scattering.
(Mathematically) optimized diffusers
differ from non-optimized diffusers
because the latter offer only spatial energy
dispersion whereas optimized ones offer
dispersion both in space and time.
Figure 7. 1D optimized diffuser.
Figure 8. Examples of non-optimized
diffusers.
Due to costs, transportation and
constructive limitations, a 1D (one
dimensional Fig. 7) diffuser surface of 1
m2 was proposed (1m x 1m). The material
used was expanded polystyrene (EPS) of
20 kg/m3 density. As mentioned earlier, in
order to achieve a great amount of
diffusion, we need to avoid surface
periodicity. If we consider a numerical
series, in direct relation with the surface
shape, we need to achieve maximum
randomness of the series, randomizing the
surface accordingly. This numerical series
will be then considered as a discrete signal,
and its Fourier Transform will give
information about the amount of pressure
in the far field, as a function of the output
angle of the sound wave propagated,
according to Equation 3[5]:
| ( )| | ( ) ( ) ( ) | (3)
Acoustical Instruments and Measurements May 2015, Argentina
The Autocorrelation function is a great
mathematical tool to analyze the
periodicity of a signal; it shows the
similarity between observations of a signal
as a function of the time lag between them.
It is defined in the discrete domain as:
( )
(4)
The WienerKhinchin theorem states
that the autocorrelation function of wide-
sense-stationary random process has a
spectral decomposition given by the power
spectrum of that process [6]. By applying
this theorem, it can be deduced that in
order to obtain a plain power spectrum
(which equals to uniform distribution of
energy in space), the autocorrelation of a
given signal must be as similar as possible
to a Delta Dirac function (Fig. 9).
Figure 9. Fourier Transform of a Delta Dirac
function.
Consequently, if the numerical series
autocorrelation approaches a Delta Dirac
function, the diffusion will be higher for
the given surface. For that matter, we
define an Autocorrelation Quality Factor
(ACQF) by means of the following:
(5)
where is the autocorrelation function
of the numerical series used. As ACQF
reaches 0 greater decorrelation is achieved
in the signal, causing greater pseudo-
randomness of the diffuser shape and
better diffusion of the sound energy, since
its autocorrelation approaches a Delta
Dirac function (Fig. 10).
Figure 10. Autocorrelation for different
sequences.
In view of the fact that a 1D diffuser
was proposed, the most important thing to
be determined is the well widths and
depths. By using a fixed well width of 5
cm, a series of 20 well depths were
required. For the sake of simplicity, only 6
different well depths were considered, each
one differing 2 cm from each other. The
maximum depth was then 10 cm. In order
to adapt to measurement requirements, the
diffuser built was scaled by a factor of 2
(section 3.2 Scaling) so bandwidth was
calculated according to full scale
dimensions (Eq. 6-7). The numerical series
to establish the diffuser surface shape was
then required to contain a series of 20
numbers from 0 to 5. An algorithm was
coded in MATLAB, so as to find the
optimum numerical series with the lowest
ACQF. After several hours of continuously
running the algorithm, it found that the
optimum sequence was: [2 2 2 2 1 2 3 0 1
0 5 0 1 1 0 3 4 3 3 2] with an ACQF value
of 0,1784.
Once set the sequence values,
maximum and minimum frequencies of
operation were determined according to
Eq. 1 and 2 (even though the equations
apply to Schroeder diffusers only):
(6)
Acoustical Instruments and Measurements May 2015, Argentina
(7)
The theoretical bandwidth of the
diffuser was 3200 Hz approximately.
However, it must be noted that due to its
low amount of mass, the diffuser will
hardly work at such low frequencies, since
the acoustical impedance presented will be
very low. Diffusing behavior will only be
expected at mid and mid-high frequencies.
Figure 11. 3D model of the designed diffuser.
The sample was first modeled in
Google Sketchup and then constructed
(Fig. 11-13). The sample consisted of a flat
base of 4 cm thick of EPS, and several bars
of 2 cm thick according to the calculated
sequence. Materials were glued with
Unipox glue and then reinforced with duct
tape. The footprint surface was 1 m2, total
volume was 0,077 m3 and weight was 1,54
kg.
Figure 12. Final construction.
Figure 13. Final construction.
3. MEASUREMENT PROCEDURE
3.1 PRINCIPLE
Measurements were made to the
diffuser sample according to ISO 17497-
1[7]. This International Standard is used to
determine the sound-scattering properties
of a surface. The degree of acoustic
scattering from surfaces is very important
in all aspects of room acoustics, and is
used mostly by acoustical simulation
software. Insufficient scattering may cause
strong deviations from exponential sound
pressure decay. On the other hand, an
approximately diffuse sound field may be
obtained with highly scattering surfaces in
a room. It is of great relevance to
differentiate between scattering and
diffusion coefficients. While the scattering
coefficient is a rough measure that
describes the degree of scattered sound, the
diffusion coefficient describes the
directional uniformity of the scattering; i.e.
the quality of the diffusing surface.
Therefore, there is a need for both concepts
and they have different applications. To
sum up, results obtained with this standard
can be used to describe how much the
sound reflection from a surface deviates
from a specular reflection.
The general principle of the method can
best be explained by looking at the effect
of reflection and scattering in the time
domain. Fig. 14 shows three bandpass-
filtered pulses which were reflected from a
corrugated surface for different
Acoustical Instruments and Measurements May 2015, Argentina
orientations of the test sample in the free
field.
Fig 14. Examples of band-pass filtered impulse
responses measured at three different positions
of the test sample.
The initial parts of the reflections are
highly correlated. This coherent part is
identical with the specular component of
the reflection. In contrast, the later parts
are not in phase and depend strongly on the
specific orientation. The energy in the
tail of the reflected pulse contains the
scattered part. The principle of the
measurement method is to extract the
specular energy from the reflected pulses.
This is done by means of averaging the
impulse responses obtained for different
sample orientations. In addition to
conventional measurements of absorption
coefficient (based on ISO 354), the sample
is placed on a turntable and impulse
responses are obtained for different sample
orientations. By synchronized averaging of
the pressure impulse responses, the
specular components add up in phase,
whereas the scattered sound interferes
destructively. Assuming statistical
independence between scattered
components, it can be shown (Fig. 14) that
after synchronized addition of n room
impulse responses, the initial decay is
related to the combined effects of
absorption and an apparent energy loss due
to sound scattered from the sample.
3.2 SCALING
In the interest of ease of measurement,
scaling can be applied to both the test
sample and the reverberation room used. A
physical scale ratio of 1:N can be used,
that is, the ratio of any linear dimension in
a physical scale model to the same linear
dimension in full scale. However, it must
be noted that the wavelength of the sound
used in a scale model for acoustic
measurements obeys the same physical
scale ratio. So, if the speed of sound is the
same in the model as in full scale, the
frequencies used for the model
measurements will be a factor of N times
than those in full scale. A scale factor of 2
was used in this case. Since the frequency
range should be measured from 100 Hz to
5000 Hz, in third-octave bands (full scale),
taking into account the scale factor, the
scaled frequency range will be from 200
Hz to 10000 Hz, in third octave bands also.
3.3 REVERBERATION ROOM
The reverberation room must comply
with the ISO 354 [8] standard, but with
certain differences. For example, diffusing
elements within the room must be fixed;
moving diffusers like rotating vanes shall
not be used. The room and its contents
should be invariant, as far as possible. The
temperature and humidity have a very
significant effect. Any devices such as
circulation systems that cause movement
or change the properties of the air in the
room should not be operated. For the
current test room, this was not an issue,
since the room was empty during
measurements, except for the equipment
used and its operators.
The dimensions of the test room are
shown in Fig. 15. This gives a total volume
of 62,2 m3 (for the extended volume
analysis section 3.4 of the present work
must be checked) and a total surface of
98,38 m2. The total volume, in cubic
meters, shall be at least:
Acoustical Instruments and Measurements May 2015, Argentina
(8)
(9)
Figure 15. Test room dimensions
Reverberation time (T20) was measured
in the room to verify maximum absorption
area. Results are shown in Table 1.
Absorption area was calculated according
to 8.1.2 section of ISO 354 as:
(10)
where V is the room volume, c is the speed
of sound and T20 is the reverberation time.
Frequency T20 (s) A1 (m2)
100 1,53 6,55
125 1,74 5,75
160 1,97 5,10
200 1,60 6,26
250 1,70 5,91
315 1,56 6,42
400 1,70 5,91
500 1,70 5,90
630 1,69 5,92
800 1,57 6,39
1000 1,60 6,29
1250 1,53 6,54
1600 1,47 6,82
2000 1,44 6,96
2500 1,42 7,05
3150 1,31 7,65
4000 1,20 8,35
5000 1,12 8,95
6300 1,03 9,75
8000 0,91 11,04
10000 0,77 13,00 Table 1. Reverberation time and absorption
area for the test room.
The equivalent absorption area of the
empty room, A1, including the air
attenuation should not exceed:
(11)
(12)
This was not achieved due to the
limitations of the test room dimensions and
room volume, which lowered the
reverberation time, even though the
surfaces were very reflective.
As regards shape of the room, if the
room is not rectangular, none of its
surfaces should be parallel. If the room is
rectangular, its proportions should be
selected so that the ratio of any two
dimensions does not equal or closely
approximate an integer. The proportions
1:21/3
:41/3
are frequently used. Other room
dimension ratios that have been found to
be satisfactory are given in Table 2 [9].
ly/lx lz/lx
0,83 0,47
0,83 0,65
0,79 0,63
0,68 0,42
0,7 0,59
NOTE The symbols lx, ly and lz represent the room dimensions.
Table 2. Recommended test room
proportions.
Room proportion determined was 1:
1.1: 2 since ly/lx=1,81 and lz/lx=0,9
(considering lx=3,36m; ly= 6,09m; lz=
3,04m). Consequently the test room will
not guarantee an even modal distribution in
the low frequencies. However, since the
lowest frequency of analysis was 200 Hz,
modal distribution issues were not that
significant. Also, according to ISO 354 the
length of the longest straight line (Imax)
which fits within the boundary of the room
should obey the following equation:
Acoustical Instruments and Measurements May 2015, Argentina
(13)
(14)
This conditions could not be met either, for
the same reasons already established.
3.4 DIFFERENCES IN ROOM
VOLUME
The reverberation room had a
suspended ceiling built in PVC (6mm
thick). Above this surface, there is a
distance of 0,3 m up to a concrete ceiling.
It should be clarified then, that the actual
volume of the room will be a function of
frequency. For low frequencies, the roof
has a very high rate of transmissibility, so
all the sound radiated on its surface is
easily transmitted to the air gap between
ceilings, thereby increasing the volume to
be considered. But at high frequency, the
transmissibility rate is sufficient to prevent
sound propagation. By means of
calculating Transmission Loss (TL) of the
ceiling material this effect can be
considered. TL simulation was performed
with a MATLAB software algorithm to
estimate this parameter.
Figure 16: Tranmission loss calculated for the
suspended ceiling.
Two relevant ranges of frequency can
be noted (Fig. 16). Above 1 kHz, the TL is
high enough to consider a constant room
volume. Below 1 kHz the TL is less than
30 dB, so volumetric increase must be
considered in this range. Fig 17 shows the
volume as a function of frequency, where
volume linearly decreases up to 1 kHz, and
then it remains constant.
Figure 17: Corrected volume of the room.
3.5 TEST SAMPLE
The area of the test sample should be as
large as possible in order to obtain good
measurement accuracy. The test sample
should be ideally circular, but the standard
allows the use of square samples (flush-
mounting the sample in the turntable). The
sample shall have a minimum edge size of
. In this case this gives an amount
of 1,5 m, which was not achieved. Flush
mounting was not done either. Structural
depth of the test sample should be
(Fig. 18).
Figure 18. Structural depth of test sample.
where d is the diameter of the test sample
and h is the structural depth. In this case,
this condition could not be achieved either,
since h = 0.1 m and d=1m. As a result,
scattering coefficients obtained with this
sample could be above 1 and below 0 for
particular third-octave frequency bands,
because edge effects can occur due to
variations in the height of the sample along
the edge of the test sample. The surface of
the perimeter of the test sample should be
as smooth as possible and rigid as possible.
The perimeter was not covered during
measurements. The absorption coefficient
of the test sample should not exceed a
value of s = 0,50 because the
measurement method will not produce
Acoustical Instruments and Measurements May 2015, Argentina
reliable results for samples with high
absorption coefficient.
3.6 BACKGROUND NOISE
When measuring impulse responses in a
room it is important to have a good amount
of signal-to-noise ratio (S/N) for each
frequency band in order to obtain accurate
values of reverberation times. When
measuring T20 for instance, a S/N ratio of
25 dB is the minimum required in order to
linearize the decay recorded. The reason
for this is described in the requirements of
decay signals, found at section 3.5 of the
present work. In the interest of reaching
higher values of S/N ratio, the source
sound pressure level was set notably high
and perceptually above the background
noise. To complement this, all impulse
responses measured were checked for
appropriate S/N ratio using Aurora
software (Fig. 19). It is worth noting that
the log sine sweep method was used,
which improves the actual S/N ratio
performance of the measurement.
Figure 19. S/N ratio evaluation for an
impulse response measured.
3.7 MEASUREMENT OF IMPULSE
RESPONSES
Impulse responses were measured
without and with the test sample following
ISO 354. Two source positions and three
microphone positions were used, giving a
total of six measurements for each
reverberation time. The reverberation time
was then estimated as the arithmetic
average of the individual reverberation
times determined in each position.
Measurement overview can be seen in
Table 3.
Reverberation Time
Test sample
Turntable
T1 not
present not
rotating
T2 present not
rotating
T3 not
present rotating
T4 present rotating
Table 3. ISO 17497-1 Measurement overview.
A turntable was required in order to
rotate the sample. The turntable shall be
provided with a rigid base. The base plate
shall be symmetrical with respect to the
axis of rotation. The size of the base plate
shall correspond to the maximum
dimension of the test sample. In this case,
an Outline ET250-3D electronic turntable
was used. It has a circular base with a
diameter of 350 mm. Dimension of the
turntable can be seen in Fig. 20.
Figure 20. Dimensions of Outline ET250-3D
turntable.
No part of the turntable may be closer
than N-1
. 1,0 m to the walls of the room (in
this case 0,5 m). The scattering coefficient
for the base plate itself shall be measured
to check the quality of the arrangement.
For each combination of source and
receiver positions, (in some cases
Acoustical Instruments and Measurements May 2015, Argentina
according to Table 3) the test signal was
continuously radiated and received while
the turntable was rotating. The total
measurement duration was equal to the
time of one revolution of the turntable. The
test signal used to measure impulse
responses shall be deterministic since the
evaluation requires a coherent averaging.
The integrated impulse response method
shall be applied. In this case, the signal
was a sine sweep from 50 Hz to 16000 Hz
which lasted 180 seconds. In order to avoid
measurement error due to air movements
or other unstable conditions in the room,
the measurements were not started until 15
minutes after closing the door.
Figure 21. Experimental setup
Audio samples were recorded using a
Tascam US-1641 usb audio interface, with
16 bit resolution and 44.1 kHz sampling
frequency. Measurements were made using
three simultaneous measurement
microphones: two Earthworks M50 and
one DPA 4007; all of them being
omnidirectional with great impulse and
frequency response. Test signal was
generated in Aurora 4.4 software and then
reproduced in an Outline Globe Source
Radiator and Subwoofer, as seen in Figs.
21 and 23. Audio samples were then
deconvolved with inverse filters using
Aurora to obtain impulse responses.
Impulse response decay was analyzed in
third-octave bands using Easera 1.0
software. According to ISO 17497-1, the
backward integration shall be restricted to
the linear slope of the impulse response
level. The decays for T1, T2 and T3 should
be linear down to the background noise
level, whereas the decay for T4 consists of
two superposed decay curves, and only the
first decay should be evaluated. This can
clearly be seen in Fig. 22.
Figure 22. Example of reverberation times
slopes for the different setups.
Integration limit must be set at -30 dB
and the reverberation time must be
evaluated in the range between -5 dB and
-20 dB provided that the first decay is
within the range. In order to comply with
the standard as much as possible, T20 were
used for these particular reasons.
Figure 23. Experimental setup.
For the specification and positioning of
microphones and sources ISO 354 was
referred, but taking into consideration the
scale factor applied. As a result, the
minimum distance between microphones
was 0,75 m; the minimum distance from
microphones to source was 1m and the
Acoustical Instruments and Measurements May 2015, Argentina
minimum distance to any room surface
was 0,5 m. Both sound source positions
had to be at least 1,5 m apart.
Figure 24. Microphone and source positions for
the first measurement setup.
All this distance conditions were met
accordingly; distance was measured with a
Bosch GLR225 laser distance meter. In
Fig. 24-25 the two measurement setups
used can be seen.
Figure 25. Microphone and source positions for
the second measurement setup.
As regards temperature and relative
humidity, changes during the course of
measurement can have a large effect on the
measurement results, especially at high
frequencies. Reducing the air attenuation
improves the measuring accuracy.
Therefore, temperature and relative
humidity shall be measured in the room
before and after each of the four
measurement situations. This was partially
accomplished, since only temperature
measurements were made before and after
all measurements. Overall changes were
minimal, a difference of 0,7C. The initial
temperature was 18,6 C whereas the final
temperature was 19,3 C. Measurements
were made using a Luft digital
thermometer with a precision of 0,1C. No
relative humidity measurements were
made due to lack of measurement
equipment for that matter.
4. RESULTS
In pursuit of calculating the scattering
coefficient of the diffuser, others
coefficients had to be calculated first. One
of them was the random-incidence
absorption coefficient s. According to ISO
17497-1, it shall be calculated for third-
octave bands using the formula:
(
)
( ) (15)
where V is the volume of the reverberation
room in cubic meters (m3); S is the area of
the test sample in square meters (m2); T1 is
the reverberation time obtained without
sample but with the plate present, in
seconds (s); T2 is the reverberation time
obtained for the test sample in seconds (s);
c1 is the speed of sound in air, in meters
per second (m/s), during measurements of
T1; c2 is the speed of sound in air, in meters
per second (m/s) during the measurement
of T2; m1 is the energy coefficient of air, in
reciprocal meters (m-1
), calculated
according to ISO 9613-1, using the
temperature and relative humidity during
the measurement of T1; m2 is the energy
attenuation coefficient of air, in reciprocal
meters (m-1
), during the measurement of
T2. The reverberation times T1 and T2 are
measured without rotation of the turntable.
Acoustical Instruments and Measurements May 2015, Argentina
The speed of sound in atmospheric air
can be calculated according to ISO 9613-1
as:
(16)
In this case, since the test room volume
was not high enough to consider air
absorption, m coefficients were zeroed.
Also, the specular absorption coefficient
spec was calculated similarly to s:
(
)
( ) (17)
where V is the volume of the reverberation
room in cubic meters (m3); S is the area of
the test sample in square meters (m2); T3 is
the reverberation time obtained for the
rotating base plate without sample, in
seconds (s); T4 is the reverberation time
obtained for the test sample on a rotating
turntable in seconds (s); c3 is the speed of
sound in air, in meters per second (m/s),
during measurements of T3; c4 is the speed
of sound in air, in meters per second (m/s)
during the measurement of T4; m3 is the
energy coefficient of air, in reciprocal
meters (m-1
), during the measurement of
T3; m4 is the energy attenuation coefficient
of air, in reciprocal meters (m-1
), during the
measurement of T4. Again, m coefficients
were zeroed in this equation.
Finally, the determination of s and spec leads to the calculation of the random-
incidence scattering coefficient using the
following formula:
(18)
Spatially averaged results obtained for
reverberation time (T20) in all cases can be
seen in Table 4 and Fig. 26. It must be
noted that all values were transposed to
full scale measurement: for example,
values for 100 Hz were measured at 200
Hz, and so on. From this data, it can be
observed that in low frequency there are
some variations due to uneven modal
density, but then the reverberation time
tends to decrease, as expected. Also, T2
and T4 values show a determined amount
of decrease respect to T1 and T3, given the
fact that in these cases the test sample was
present and provided some absorption.
Frequency T1 T2 T3 T4
100 1,60 1,44 1,54 1,39
125 1,70 1,52 1,72 1,48
160 1,56 1,46 1,51 1,46
200 1,70 1,53 1,72 1,55
250 1,70 1,52 1,67 1,56
315 1,69 1,54 1,71 1,59
400 1,57 1,50 1,63 1,48
500 1,60 1,48 1,60 1,51
630 1,53 1,38 1,51 1,41
800 1,47 1,34 1,49 1,32
1000 1,44 1,29 1,45 1,30
1250 1,42 1,28 1,38 1,28
1600 1,31 1,19 1,32 1,19
2000 1,21 1,10 1,20 1,09
2500 1,10 1,02 1,13 1,04
3150 1,05 0,95 1,04 0,94
4000 0,92 0,85 0,92 0,85
5000 0,78 0,74 0,78 0,73 Table 4. Results obtained for reverberation
time (T20).
Figure 26. Results obtained for reverberation
time (T20).
Acoustical Instruments and Measurements May 2015, Argentina
Standard deviation of the reverberation
time was estimated according to Annex A
of the ISO 17497-1 as:
( )
( ) (19)
where N is the number of measurements of
the reverberation time, and the spatial
average of the reverberation chamber is:
(20)
The 95 % confidence limit for T20
values was estimated as two times the
standard deviation in all cases. Table 5
shows measurements as well as their
corresponding expanded uncertainty.
Frequency T1 T2 T3 T4
100 1,6 0,1 1,44 0,15 1,54 0,12 1,39 0,14
125 1,7 0,13 1,52 0,12 1,72 0,18 1,48 0,14
160 1,56 0,08 1,46 0,12 1,51 0,1 1,46 0,09
200 1,7 0,13 1,53 0,07 1,72 0,09 1,55 0,14
250 1,7 0,09 1,52 0,08 1,67 0,1 1,56 0,11
315 1,69 0,09 1,54 0,06 1,71 0,15 1,59 0,09
400 1,57 0,09 1,5 0,06 1,63 0,04 1,48 0,06
500 1,6 0,06 1,48 0,03 1,6 0,02 1,51 0,04
630 1,53 0,04 1,38 0,04 1,51 0,04 1,41 0,05
800 1,47 0,04 1,34 0,03 1,49 0,05 1,32 0,06
1000 1,44 0,05 1,29 0,03 1,45 0,04 1,3 0,08
1250 1,42 0,04 1,28 0,02 1,38 0,04 1,28 0,06
1600 1,31 0,03 1,2 0,01 1,31 0,02 1,18 0,03
2000 1,2 0,02 1,09 0,04 1,22 0,03 1,09 0,02
2500 1,12 0,04 1,03 0,01 1,12 0,02 1,04 0,02
3150 1,03 0,03 0,95 0,03 1,04 0,02 0,94 0,03
4000 0,91 0,02 0,86 0,02 0,93 0,01 0,85 0,03
5000 0,77 0,01 0,73 0,02 0,78 0,01 0,73 0,02
Table 5. T20 results with expanded uncertainty.
Figs. 27-28 show graphically the expanded
uncertainty for T1 and T2 in order to
compare both measurements. T3 and T4 are
not displayed because the values are very
similar to the latter. The data dispersion is
greater for low frequencies due to room
resonances, and it decreases as frequency
increases.
Figure 27. Measurement uncertainty for T1.
Figure 28. Measurement uncertainty for T2.
Final values obtained for scattering are
showed in Table 6, and obtained according
to Eq. 15-18. The first thing to analyze is
the fact that scattering values calculated
are not trustworthy below 1250 Hz for
various reasons. The most crucial factor is
assumed to be the test sample surface,
which was not enough to fit the standard.
Other factors could have been the low
amount of diffusion of the test room (the
second most relevant factor), design
limitations and considerations such as
weight and spatial dimension of diffusion
(1D diffusers offer lower scattering than
2D diffussers), the edge absorption of the
Acoustical Instruments and Measurements May 2015, Argentina
test sample due to the fact that the sample
was not flush mounted, low amount of
decay samples (source and microphone
positions), instrumental error and other
secondary requirements of the ISO 17497-
1 which were not met and specified during
this product specification. However, for
the frequencies above 1600 Hz, results
obtained were much more reliable, partly
because diffraction at higher frequencies
(due to the short wavelengths) is a
phenomenon with a higher probability to
occur.
Frequency alfa s alfa spec s
100 0,72 0,71 -0,02
125 0,69 0,93 0,78
160 0,46 0,22 -0,45
200 0,66 0,67 0,03
250 0,69 0,46 -0,76
315 0,59 0,45 -0,33
400 0,30 0,63 0,48
500 0,51 0,39 -0,25
630 0,74 0,50 -0,97
800 0,69 0,87 0,59
1000 0,82 0,75 -0,35
1250 0,77 0,59 -0,81
1600 0,78 0,83 0,22
2000 0,83 0,87 0,25
2500 0,72 0,81 0,35
3150 1,02 1,01 0,39
4000 0,90 0,94 0,44
5000 0,70 0,88 0,61 Table 6. Results calculated for s, spec and
scattering coefficient (s).
The uncertainties in the absorption
coefficients are obtained by means of the
following equations:
(
) (
) (21)
(
) (
) (22)
Finally, the standard deviation in the
scattering coefficient is:
|
| (
) (
) (23)
Frequency s spec s
100 0,84 0,90 4,43
125 0,69 0,86 2,79
160 0,68 0,61 2,17
200 0,53 0,65 2,43
250 0,46 0,58 3,22
315 0,43 0,64 2,08
400 0,47 0,32 0,57
500 0,28 0,22 0,84
630 0,29 0,29 2,52
800 0,27 0,43 1,43
1000 0,31 0,51 3,63
1250 0,24 0,41 2,59
1600 0,21 0,27 1,46
2000 0,33 0,26 2,12
2500 0,31 0,26 1,17
3150 0,38 0,35 1,76
4000 0,34 0,37 1,58
5000 0,36 0,36 1,28 Table 7. Standard deviation for s, spec and s
coefficients
Standard deviations for all coefficients
are shown in Table 7. High values of
standard deviation were introduced, due to
the uncertainties already mentioned.
Current results for scattering were
compared and contrasted to four diffuser
solutions offered by the company RPG,
only in the frequencies above 1600 Hz.
The diffusers compared were 1D and can
be seen in Fig. 29. Fig. 30 shows data
comparison. As it can be seen, scattering
of the designed diffusor at these
frequencies is not as good as other
products scattering, but it shows similar
performance as the FlutterFree diffuser.
However, it must be noted that the
comparison is only made at high
frequencies; most of the RPG diffusers
Acoustical Instruments and Measurements May 2015, Argentina
shown offer scattering at lower frequencies
also.
Figure 29. 1D commercial diffusers:
FlutterFree, QRD 734, FlutterFree-T,
Formedffusor (from top left to bottom right).
Figure 30. Comparison between the designed
diffuser and commercially available products.
5. CONCLUSION
A diffuser design was proposed and
constructed, taking into consideration the
background theory associated with diffuse
sound fields and acoustics. Measurements
of the diffuser were carried out according
to ISO 17497-1 and it was determined that
a very important factor was the sample
size; since we are calculating the scattering
coefficient indirectly, by means of
reverberation time, if the sample footprint
has a low amount of surface, the
reverberation time considered will not vary
accordingly and error will be introduced.
As stated, diffusion is a very desirable
effect on room acoustics and even though
the quantity of diffusion surface is hard to
define, it is empirically demonstrated and
well known that in order to obtain good
sounding rooms there are some obligatory
locations for diffusers. For example, in
control rooms for recording/mixing
studios, a preferable place for diffusers is
the rear wall, because sound incidence is
critical and highly probable for that
surface. The election of 1D or 2D diffusers
will depend on the overall geometry of the
room; 2D will tend to distribute the energy
along the horizontal and vertical axis of
incidence whereas 1D will only sparse the
energy in a unique axis depending on the
diffuser mounting. In the case of control
rooms, if sound reflections from the rear
wall to the ceiling are to be avoided, 1D
diffusers mounted vertically will be
preferable, and they will only scatter the
sound energy laterally; this could be the
case when the height of the control room is
not very large and would avoid strong
second order reflections at the sweet spot.
In the case of concert halls, the
situation will be different since these
rooms have greater volume and different
geometry in general. Here, the compulsory
location for diffusers will be the front side
of the balcony surface; scattering on this
surface can be accomplished by means of
optimized diffusers or non-optimized ones
such as ornaments which are very common
in concert halls. The scattering on the
balcony will avoid reflections from the
stage causing unwanted echoes which
could perceptually affect the performers or
the audience located at the front rows.
Also, diffusion can be introduced on the
rear walls for the same reasons as in
control rooms, but considering the distance
between the rear walls and the last row of
seats. Diffusers tend to behave like
resonators in the short distance, and its
Acoustical Instruments and Measurements May 2015, Argentina
frequency of resonance will be determined
by well depth and width. Therefore,
listeners at the last rows will perceive
variations in the frequency response that
will not be applied to the rest of the
audience.
In the end, diffusion and scattering are
very complex phenomena in the room
acoustics field and even though objective
influence of these parameters in actual
enclosures and sound fields are not well
established, they can improve acoustical
preference of rooms when used properly,
and the design of surfaces which introduce
diffusion and scattering behavior such as
the present in this specification is of great
relevance.
6. REFERENCES
[1] Strube, H. W. (1981). More on the
diffraction theory of Schroeder
diffusors. The Journal of the Acoustical
Society of America, 70(2), 633-635.
[2] Cox, T. J., & D'antonio, P. (2009).
Acoustic absorbers and diffusers: theory,
design and application. CRC Press.
[3] L. L. Beranek, Concert and Opera
Halls: How They Sound. AIP Press, 6974 (1996).
[4] T. J. Cox and Y. W. Lam. Prediction
and evaluation of the scattering from
quadratic residue Diffusors. J. Acoust.
Soc. Am., 95(1), 297305 (1994).
[5] Schroeder, M. (1997) Number Theory in science and Communication. Chapters 13 & 15.Third edition. Springer. 1997.
[6] C. Chatfield (1989). The Analysis of
Time Series An Introduction (Fourth Ed.). Chapman and Hall, London. pp. 9495.ISBN 0-412-31820-2.
[7] ISO 17497-1. Acoustics - Sound-
scattering properties of surfaces - Part 1:
Measurement of the random-incidence
scattering coefficient in a reverberation
room.
[8] ISO 354. Acoustics - Measurement of
sound absorption in a reverberation room.
[9] ISO 3741. Acoustics - Determination of
sound power levels of noise sources using
sound pressure - Precision methods for
reverberation rooms.