See discussions, stats, and author profiles for this publication at: [Link]

net/publication/289741635

Validating insertion loss predictions for HVAC silencers

Article · January 2009

CITATIONS READS

3 2,078

3 authors, including:

Ray Kirby
RMIT University
102 PUBLICATIONS 1,741 CITATIONS

SEE PROFILE

All content following this page was uploaded by Ray Kirby on 14 October 2016.

The user has requested enhancement of the downloaded file.

 VALIDATING INSERTION LOSS PREDICTIONS FOR
HVAC SILENCERS
Ray Kirby
Brunel University, School of Engineering and Design, Mechanical Engineering, Uxbridge,
Middlesex. UB8 3PH, United Kingdom
e-mail: [Link]@[Link]
David Herries and Ken Amott
CAICE Acoustic Air Movement Ltd., Riverside House, 3 Winnerish Fields, Gazelle Close,
Winnerish, Wokingham. RG41 5QS, United Kingdom

Ventilation systems normally contain large dissipative splitter silencers that are used to at-
tenuate fan noise. Recently, numerical models have been developed that seek to predict the
insertion loss of these silencers; however, there is very little experimental data available in
the literature that is suitable for validating these insertion loss predictions and so questions
marks still remain regarding the accuracy of the models. This issue is investigated here by
comparing theory with experiment for a range of splitter silencers. Here, the insertion loss
predictions are compared against experimental measurements obtained on a test rig that con-
forms to ISO 7235. Measurements are presented for silencers of differing geometry, al-
though each silencer is restricted to one central splitter with two baffles lining the opposite
walls of the duct. Furthermore, validation of the numerical models is based on measurements
taken both with and without flow. Comparison between prediction and experiment is re-
ported in one-third octave frequency bands up to a frequency of 8 kHz. Results indicate that
the theoretical model is capable of providing accurate predictions for silencers of high per-
centage open area, but when the percentage open area is low significant discrepancies appear
between prediction and experiment at higher frequencies.

1. Introduction
Dissipative silencers are widely used in heating, ventilation and air conditioning (HVAC) sys-
tems to attenuate broad band noise emanating from fans or other air-moving devices. Large ducting
systems are common, which normally require the use of large silencers and for rectangular duct-
work silencers typically consist of vertical rectangular baffles arranged in parallel with one an-
other1, whereas for circular ductwork so-called pod silencers are common2. It is normal for industry
to attempt to specify the acoustic performance of these silencers in octave bands up to a frequency
of 8kHz, and here the acoustic insertion loss of the silencer is typically specified (either with or
without flow). Clearly, quantifying the performance of potentially large dissipative silencers over a
frequency range than extends up to 8 kHz presents a significant challenge, both from a theoretical
and experimental point of view. Accordingly, this article will present some new measurements that

ICSV16, Kraków, Poland, 5-9 July 2009 1

16th International Congress on Sound and Vibration, Kraków, Poland, 5–9 July 2009

have been obtained for relatively large silencers, which are suitable for use in rectangular ductwork
and with the baffles arranged in a so-called splitter formation so that one central baffle is placed
parallel to two baffles that line opposite duct walls. The aim here is to validate theoretical predic-
tions that have recently appeared in the literature3-5 by comparing insertion loss predictions with
experimental measurements with a view to developing more accurate models of HVAC silencers
suitable for use in a design environment.
Very little experimental data has been published for large HVAC silencers. Here, the major-
ity of measured data comes from the Fraunhofer-Institut für Bauphysik (IBP) in Germany, see for
example Refs. 6 and 7, although data was also reported by Ramakrishnan and Stevens8. In fact,
very few facilities are currently available for testing large HVAC silencers and this is probably be-
cause of the cost involved in commissioning and running such a test rig. Accordingly, there is a
need for more published experimental data with which to validate current theoretical models.
Moreover, the experimental methodology should conform to the most up-to-date standards (for ex-
ample Ref. 9) and the silencer fabrication and sound pressure measurements need to be carefully
controlled if one is to fully validate the new theoretical models. Therefore, this article will present a
sample of some insertion loss measurements taken for splitter silencers on a new bespoke silencer
test rig that has been designed and commission according to ISO 72359. Measurements are reported
for a range of splitter silencer geometries in order to fully validate the predictions and conclusions
will be drawn on how well the theory performs.

2. Experimental apparatus
The experimental facility to measure silencer insertion loss is shown in Fig. 1. The test facil-
ity is a closed loop ductwork system constructed from galvanized steel and is designed to measure
the insertion loss of splitter silencers suitable for use in circular and rectangular ductwork. The fa-
cility can measure so-called static insertion loss, when no mean flow is present, and so-called dy-
namic insertion loss where mean flow is generated by the fan (which may be moved through 180°
in order to change the orientation of the flow with respect to the sound propagation). This test facil-
ity has been designed in accordance with ISO 7235 and has undergone extensive preliminary testing
and commissioning in order to ensure that all the requirements of the ISO standard have been met.
Furthermore, the ability of the test rig to meet ISO 7235 has been further independently verified and
certified by the UK Accreditation Service10.

Primary fan
Volume flow rate Silencers
Auxiliary measurement
Silencer

Fan

Microphone
Modal filter Anechoic
locations
Test Silencer termination

Auxiliary
Loudspeaker
Silencer
Approx. 20m

Figure 1. Experimental Apparatus

2
 16th International Congress on Sound and Vibration, Kraków, Poland, 5–9 July 2009

Here, the test facility is accredited to measure static and dynamic insertion loss in one third octave
bands over a frequency range of 50 Hz to 10 kHz for silencers between 280 and 1200 mm square
with a maximum length of 4800 mm. The test facility is also accredited to measure circular (or pod)
silencers from 300 mm to 1200 mm in diameter, which is accomplished by changing the ducting for
the test silencer.
In Fig. 1 the silencer to be tested (hereafter the “test silencer”) is placed between a sound
source (a 500 W loudspeaker) and an anechoic termination that is placed downstream of the micro-
phone positions. In this article we shall focus on testing rectangular splitter silencers and so the test
silencer is placed within a rectangular duct, which is joined via a transition section to a duct of cir-
cular cross-section that runs upstream to the loudspeaker enclosure and downstream to the anechoic
termination. Between the loudspeaker enclosure and the test silencer a straight through circular
dissipative silencer is placed that dampens down higher order modes excited by the loudspeaker.
This silencer, or modal filter, is specified in the ISO standard and is designed to deliver a sound
pressure field incident on the test section that is approximately planar. Accordingly, the modal fil-
ter must be designed to dampen down the energy contained within the higher order modes but at the
same time it must not adversely affect the signal to noise ratio at the microphones. Clearly, when
mean flow is present this places limits on the air speed it is possible to measure and so those results
presented here are limited to those in which the signal to noise ratio is acceptable. It is interesting
to note that the modal filter is not specified in the equivalent U.S. standards11 and so this filter was
investigated prior to commissioning the test facility. To do this sound pressure measurements were
taken in one third octave bands over the cross section of the duct downstream of the modal filter
and the sound pressure distribution was analyzed. It was seen that the modal filter generally per-
forms well and delivers a reasonably consistent plane wave up to about 4 kHz, beyond which some
modal contamination is present. Accordingly, it was judged that a modal filter can work well pro-
vided it is properly designed, which has important ramifications for the theoretical predictions that
follow.
Downstream of the test silencer is a duct of circular cross-section in which a standard ½ inch
microphone (and windshield for flow measurements) feeds data into an analyzer. Here, the micro-
phone is supported by a track, which allows the microphone to be moved automatically across the
duct using a stepper motor and for the position of the microphone to be carefully controlled. Down-
stream of the microphone is an anechoic termination that has been carefully designed to deliver the
sound attenuation specified in the standards so that minimal sound energy is reflected back towards
the microphone. This anechoic termination is designed to accommodate mean flow and so it is of
the “straight-through” variety whereby the depth of the surrounding material is gradually increased
in order to smooth out changes in impedance. Beyond the anechoic chamber, and upstream of the
loudspeaker enclosure, the ducting system is designed to accommodate a fan. The fan is designed
to provide air volume flow rates of up to 2 m 3 /s and can be rotated through 180°. In order to meas-
ure the volume flow rate in the system a Wilson Flow Grid is used and further air static pressure
measurements are also taken around the test rig. In order to remove unwanted fan noise primary
and auxiliary rectangular splitter silencers are installed on both sides of the fan, as well as before the
loudspeaker enclosure and after the anechoic termination. This latter auxiliary silencer further adds
to the performance of the anechoic termination and so enables the test rig to meet, and sometimes
exceed, the ISO requirements.

2.1 Experimental methodology

The test facility measures silencer performance by first measuring the sound pressure field
with a test silencer present, and then removing the test silencer and measuring the sound pressure
field with an empty substitution duct in place of the silencer. This delivers the so-called silencer
insertion loss, which is simply calculated by subtracting the sound pressure level (in dB) measured
when the silencer is present from the value measured when the silencer is absent. This is carried
out using pink noise and is processed in one third octave bands, with and without flow. Here, the

3
16th International Congress on Sound and Vibration, Kraków, Poland, 5–9 July 2009

sound pressure level for each measurement is calculated by averaging the data measured at five
locations that span diagonally across the duct in order to minimize the influence of higher order
modes when estimating the equivalent planar sound pressure level. By using five locations errors
associated with quantifying silencer performance are minimized.
The testing programme that is reported in this article is based on testing rectangular splitter si-
lencers only, and is further restricted to those silencer containing one baffle placed centrally in the
duct with two baffles lining the opposite walls. For these silencers numerous commissioning pro-
cedures were undertaken in order to address problems such as, for example, break-out/break-in
noise, sound reflection from the loudspeaker enclosure and the minimization of background noise.
A methodology for testing silencer insertion loss was then arrived at that delivered data that was
reproducible with an accuracy better than that quoted by ISO 7235. As a guide to the accuracy that
is achievable from this test facility, a typical measurement uncertainty (± dB) for a rectangular split-
ter silencer is given in Table 1, for octave bands.

Table 1. Estimation of measurement uncertainty for splitter silencers.

±dB in each Octave Band centre frequency (Hz)

63 125 250 500 1k 2k 4k 8k
Static insertion loss 2 2 2 2 3 3 3 3
Dynamic insertion loss 2 2 2 2 3 3 3 3

3. Theory
The experimental measurements are to be compared here against theoretical predictions gen-
erated using the point collocation method. This method is reported extensively elsewhere, see for
example Refs. 3-5, and so only a brief summary is reported here. The point collocation method is a
numerical method that employs a modal representation of the sound pressure field for a uniform
duct section and then enforces continuity conditions over impedance discontinuities in the duct us-
ing discrete collocation points. Here, the modal eigenfunctions and wavenumbers for the empty
duct and silencer sections are found using the finite element method and, after applying continuity
of pressure and displacement over axial discontinuities, the modal amplitudes may be computed; it
is then straightforward to find the silencer insertion loss. In the previous section it was noted that
the current ISO 7235 standard specifies an incident plane wave and this allows for a considerable
simplification of the theoretical model4. This is because plane wave excitation can be shown to
excite only symmetric modes in the silencer section and so one only needs a one dimensional model
for calculating the eigenmodes in each duct section. Obviously this will significantly reduce the
degrees of freedom required for the point collocation approach and it is this that makes it feasible to
study a typical splitter silencer up to a frequency of 8 kHz. Accordingly, the model of Kirby5 is
used here with a one-dimensional eigensolution for each duct section; moreover, the rectangular
silencer section is assumed to be joined to two uniform rectangular ducts and so the converging and
diverging duct sections seen in the experimental work are omitted.
For a harmonic time dependence of eiωt ( i = − 1 , ω is radian frequency, t is time) the
acoustic wave equation for region q is written as
2
1 D p′q
− ∇ 2 pq′ = 0 . (1)
cq2 Dt 2

where cq is the speed of sound and pq′ is the acoustic pressure.

4
 16th International Congress on Sound and Vibration, Kraków, Poland, 5–9 July 2009

Outlet Duct
Silencer section Region 4 b
Region C z
a2
Baffle L a1
fairings
z
x

y
y
Porous Perforate
Inlet Duct, material sheet
Region 1
Figure 2. Geometry of rectangular splitter silencer

The walls of each duct, and the baffle fairings at each end of the silencer, are assumed to be
rigid and impervious to sound, which gives
∇pq′ ⋅ n q = 0 (2)

where n q is the outward unit normal vector in region q. The acoustic field in each region is
expanded as an infinite sum over the duct eigenmodes to give
∞
p1′ ( x, y, z ) = F 0Φ 0i ( y, z )e -ik0λi x + ∑ An Φ nr ( y, z )e-ik0λr x
0 n
(3)
n =0

∞ ∞
pc′ ( x, y, z ) = ∑ B n Ψin ( y, z )e -ik0ki x + ∑ C n Ψrn ( y, z )e -ik0kr x
n n
(4)
n=0 n=0

∞
p4′ ( x′, y, z ) = ∑ D n Φ in ( y, z )e -ik0λi x .
n
′
(5)
n =0

Here, An, Bn, Cn and Dn and are the (unknown) modal amplitudes; λin are the incident and λnr
the reflected axial wavenumbers in regions 1 and 4, kin are the incident and krn the reflected axial
wavenumbers in the silencer section, Φ in are the incident and Φ nr the reflected eigenfunctions in
regions 1 and 4, and Ψin are the incident and Ψrn the reflected eigenfunctions in the silencer sec-
tion. In addition k0 = ω c0 , where c0 is the isentropic speed of sound in air, and for an incident
plane wave in region 1, F 0 = 1 . The eigenfunctions and eigenvectors for each region are found
using the finite element method5. Point collocation is then used to match continuity conditions over
the inlet and outlet planes of the silencer. Here, Eq. (2) is enforced over the baffle fairings; over
each airway continuity of pressure and displacement are enforced12. Thus, at the silencer entrance
p1′ ( x, y, z ) = pc′ ( x, y, z ) (6)
∂p1′ ∂p′
[1 − M 1λn ]−2 = [1 − M c k n ]−2 c , (7)
∂x ∂x
and at the silencer exit
pc′ ( x′, y, z ) = p′4 ( x′, y, z ) (8)
∂pc′ ∂p′
[1 − M c k n ]−2 = [1 − M 4λn ]−2 4 . (9)
∂x ∂x

5
16th International Congress on Sound and Vibration, Kraków, Poland, 5–9 July 2009

Here, Mq is the Mach number of the mean flow in region q. After applying these axial match-
ing conditions the problem may be solved for the unknown modal amplitudes in Eqs. (3)-(5). The
insertion loss IL is then given as
NI γ n I n Dn2
IL = −10 log10 ∑ , (10)
n =0 I0
2
where I n = ∫Φ
R4
n ( y, z ) dydz .

Note that the analysis here is for a general three-dimensional (uniform) waveguide; however,
in the following section a two-dimensional (x, y plane) representation is used. Furthermore, provi-
sion is made in Eq. (10) to include higher order modes propagating in the outlet duct.

4. Results and Discussion

Static and dynamic IL predictions are presented here for four different splitter silencers.
These results represent a sample of a large number of measurements carried out over a range of
silencer geometries. Those results presented here were chosen to represent the different trends
found when comparing prediction to experiment over the whole range of silencers tested. Each
silencer contains rock-wool with an estimated flow resistivity of 1650 Pa s/m2; values used for the
bulk acoustic properties and perforated sheet are the same as those reported in Ref. 6. In Fig. 3 the
measured and predicted static IL is presented for Silencer A, which has a percentage open area of
50%. It is evident in Fig. 3 that it is possible to obtain good agreement between predicted and
measured data for splitter silencers over a wide frequency range. Moreover, this quality of agree-
ment can generally be replicated for those silencers of similar percentage open area.

Figure 3. Static IL Silencer A , theory; ♦, experiment.

( a1 = 0.1 m , a2 = 0.3 m , b = 0.4 m , L = 1.8 m )
However, if one lowers the percentage open area then the agreement between prediction and
experiment is less satisfactory. For example, in Fig. 4 Silencer B has a 20% open area and it is evi-
dent that at higher frequencies the IL is over predicted (similar behaviour is also seen in Ref. 5). It
should, however, be noted here that there are limits to the level of IL it is possible to measure. This
limit must be found using tests outlined in ISO 7235 and for this test rig the limiting IL was found
to be approximately 70 dB above 1 kHz. Accordingly, it is not possible solely to explain the dis-
crepancies observed in Fig. 4 on the basis of the limits of the test rig; moreover, significant discrep-
ancies are still observed above, say, 2 kHz where the IL is lower. This behaviour is also generally
reflected in other silencers of low percentage open area measured with this test rig.

6
 16th International Congress on Sound and Vibration, Kraków, Poland, 5–9 July 2009

Figure 4. Static IL Silencer B: , theory; ♦, experiment.

( a1 = 0.1 m , a2 = 0.15 m , b = 0.25 m , L = 1.8 m )

Figure 5. Static IL Silencer C: , theory; ♦, experiment.

( a1 = 0.1 m , a2 = 0.225 m , b = 0.325 m , L = 1.2 m )
In Fig. 5 comparisons are shown for a silencer with an open area of 38% and it is clear that, as
the open area increases, the agreement between prediction and experiment improves. It is notice-
able, however, that the experimental measurements do not pick up the predicted peaks in IL. This
may partly be explained by the averaging over the octave bands; however, the most likely cause is
the suppression of higher order modes in the experimental measurements (by averaging over 5 loca-
tions) whereas these peaks appear in the predictions as a result of higher order modes cutting on.
These peaks can also be seen in Fig. 6, which also includes dynamic IL (with a silencer face veloc-
ity of 2.6 m/s). This silencer also has a low percentage open area so again the agreement between
prediction and experiment at higher frequencies is limited. One can, however, see that the influence
of mean flow is generally restricted to the middle frequency range and here the IL is seen to reduce.
This lowering of IL is in line with expectation and is also reflected in the theoretical predictions,
although here the effect is not so pronounced at lower frequencies. It is also noticeable that when
mean flow is present the predicted peaks in IL at higher frequencies are seen to shift.

7
 16th International Congress on Sound and Vibration, Kraków, Poland, 5–9 July 2009

Figure 6. Silencer D: , SIL theory; ♦, SIL experiment; , DIL theory;
×, DIL experiment. ( a1 = 0.1 m , a2 = 0.15 m , b = 0.25 m , L = 0.6 m )

5. Conclusions
A representative sample of IL measurements and predictions for rectangular splitter silencers
is presented here. Agreement between measurement and prediction is generally good at low fre-
quencies and when the percentage open area of the silencer is large. However, as the open area
reduces the theory over predicts the IL at higher frequencies. This behaviour is similar to that found
elsewhere and cannot be explained by experimental limitations. Accordingly, further work is re-
quired in order to understand the reasons for these discrepancies and to improve correlation between
prediction and experiment for those silencers with a low percentage open area.

REFERENCES
1
A. Cummings, “Sound attenuation in ducts lined on two opposite walls with porous material, with
some application to splitters,” Journal of Sound and Vibration, 49, 1976, 9–35.
2
R. Kirby, “Multi-mode sound propagation in pod silencer”, Proceedings of the Thirteenth Interna-
tional Congress on Sound and Vibration, Vienna, Austria, July 2006.
3
R. Kirby and J.B. Lawrie “A point collocation approach to modelling large dissipative silencers”,
Journal of Sound and Vibration, 286, 2005, 313-319.
4
J.B. Lawrie and R. Kirby, “Mode matching without root finding: Application to a dissipative silencer,”
Journal of the Acoustical Society of America, 119, 2006, 2050-2061.
5
R. Kirby, “The influence of baffle fairings on the acoustic performance of rectangular splitter silenc-
ers,” Journal of the Acoustical Society of America, 118, 2005, 2302-2312.
6
F.P. Mechel, “Numerical results to the theory of baffle-type silencers”, Acustica, 72, 1990, 7-20.
7
P. Brandstätt, W. Frommhold and M.J. Fisher, “Program for the computation of absorptive silencers in
straight ducts”, Applied Acoustics, 43, 1994, 19-38.
8
R. Ramakrishnan and R. Stevens, “Improving the accuracy of duct insertion loss predictions”, Journal
of Sound and Vibration, 169, 1994, 423-427.
9
European Standard EN ISO 7235, “Measurement procedures for ducted silencers – insertion loss, flow
noise and total pressure loss”, 2003.
10
United Kingdom Accreditation Service, [Link]
11
ASTM E477, “Standard Test Method for Measuring Acoustical and Airflow Performance of Duct
Liner Materials and Prefabricated Silencers,” DOI. 10.1520/E0477-06A, [Link]
12
R. Kirby and F.D. Denia, “Analytic mode matching for a circular dissipative silencer containing mean
flow and a perforated pipe,” Journal of the Acoustical Society of America, 112, 2007, 3471-3482.

View publication stats