View metadata, citation and similar papers at core.ac.

uk brought to you by CORE

provided by Purdue E-Pubs

Purdue University
Purdue e-Pubs
International Compressor Engineering Conference School of Mechanical Engineering

1976

Refrigerant Muffler Analysis

T. C. Hundley

Follow this and additional works at: https://docs.lib.purdue.edu/icec

Hundley, T. C., "Refrigerant Muffler Analysis" (1976). International Compressor Engineering Conference. Paper 218.
https://docs.lib.purdue.edu/icec/218

This document has been made available through Purdue e-Pubs, a service of the Purdue University Libraries. Please contact [email protected] for
additional information.
Complete proceedings may be acquired in print and on CD-ROM directly from the Ray W. Herrick Laboratories at https://engineering.purdue.edu/
Herrick/Events/orderlit.html
 REFRIGERANT MUFFLER ANALYSIS

T.C. Hundley, Senior Physicist

Major Appliance Laboratories
General Electric Co., Louisville, Ky.

INTRODUCTION
Most refrigeration systems require mufflers to rapid assembly of a computer program to analyze
reduce noise due to gas pulsations in compressor most refrigerant muffler configurations. The
suction and discharge lines. In the past, the muffler and associated tubing are broken down into
design and analysis of refrigerant mufflers has sections, each of which can be represented by a
been largely by cut-and-try methods, due to the matrix relating input and output quantities.
complexity of the numerical calculations necessary Matrix elements are functions of the physical para-
for even relatively simple muffler configurations. meters of the section. Only a very limited number
As in other areas of design, the widespread of different section matrices are required. After
availability of digital computers has made prac- the matrices have been defined, multiplication of
tical the use of more comprehensive design and the matrices by a computer gives the muffler trans-
analysis methods. The purpose of this paper is to fer function, from which other desired quantities
describe one such method which yields information can be obtained.
of value to the designer, and to describe also
transmission loss measurements made in air which In this paper, a general discussion of muffler
were used to validate the computer model. section types will first be given, followed by
application to a particular configuration. Next,
The usual purpose of muffler analysis is to obtain the measurement method will be described and
a plot of transmission loss as a function of fre- results of computation and measurement compared.
quency. The designer does not need extremely Finally, some advantages and disadvantages of the
precise information of this sort, but he does need method will be outlined.
to know the locations in frequency at which the
muffler amplifies rather than attenuates, and he COMPUTER PROGRAM
needs to know the approximate attenuation which is
attained in the muffler stop bands. Further, it is Almost any reactive muffler having one input and
of great value to be able to relate singularities one output port can be modeled by using four dif-
in the muffler response to specific elements of the ferent network elements. These are a transmission
muffler. line section, a shunt element, a lumped capaci-
tance, and a termination. Other elements can, of
A variety of methods have recently been developed course, be conceived but have not yet been needed.
to obtain this information. A few of these are
described in References 1 through 5. Methods which Figure 1-A is the representation of a transmission
have been used range from highly simplified models line section. This would be used to model a muf-
using lumped circuit theory to sophisticated fler element having approximately constant cross
methods which provide numerical solutions to the sectional area. The area can be of any shape, but
wave equation. The lumped circuit methods, at the the largest transverse dimension must be less than
one extreme, are open to the objection that they one half wavelength at the highest frequency of
are limited to low frequencies only, and that they interest. If the cross section changes slowly, it
fail to account for finite wavelength effects which may be possible to represent the section as two or
are of much practical importance. The more elabo- more separate parts. This network element is used
rate computations, on the other hand, require much to represent tubes, for example, and also holes in
time on a powerful computer and often yield con- baffles. These holes, even in thin baffles, have
siderably more information than can be utilized by a finite effective length. The important para-
the practical muffler designer. meters of the section are the area, length, velo-
city of sound, and density of refrigerant.
The method to be described here is intermediate in
complexity and accuracy to the extremes mentioned Figure 1-B shows the electrical analog, a trans-
above. Based on the prior work of Miller and mission line. Here, 1f, is the acoustical volume
Hatten 1 , it utilizes a modular approach to enable current into the section and~ is the alternating

382
component of pressure at the input. As shown in program using the identifying number as a sub-
Figure 1-C, these quantities are complex, having script and a beginning frequency assigned. Thus,
both real (subscript R) and imaginary parts the length of the 6th section is L6. Next, the
(subscript I). elements of each matrix are defined, using the
identifying letter together with row and column
The relationshi p between input and output quanti- number, as, for example, A(l,l) for the element in
ties can be written in electrical network foxm as the first row and first column of the A matrix.
sho~n in Figure 2-A. The complex constants A,~. After this process is completed, the program is
t, D for a lossless transmission line are evalu- directed to multiply each matrix by the preceeding
ated by assuming two waves on the line, with one one, beginning with the termination matrix. When
wave traveling in the forward direction and one this is completed, the result is a 4xl matrix.
in the backward direction. The result of this The first element in this matrix is the real part
evaluation is shown in Figure 2-B, with an addi- of the input pressure and the second element is
tional intervening step omitted. The omitted step the imaginary part. Thus, since a unit real out-
separated the two equations with complex coeffi- put pressure was assumed, the transmissio n loss is
cients and variables into four equations having obtained by calculating ten times the logarithm
only real coefficient s and variables. This step is to the base 10 of the sum of the squares of the
necessary because the computer language used, BASIC, two input pressure components. This process is
does not allow use of complex numbers. Figure 2-B then repeated for each frequency of interest.
also gives the definitions of two intermediat e
functions, I and J, used for convenience in com- An example will perhaps make this somewhat more
putation. Finally, Figure 2-C gives the matrix understandable. Figure 6 is the schematic of a
formulation of the equations. The input quantities, typical refrigerant muffler. The cylinder head
separated into real and imaginary parts, form a volume is connected to the muffler by a short
matrix with one column and four rows as do the tube. The muffler itself has two expansion cham-
output quantities. The coefficient matrix is 4x4, ber sections separated by a thin baffle having
with elements which are functions of cross sec- several holes. The downstream side of the muffler
tional area and length of the muffler section and feeds into a long tube. Position 7 is arbi-
of refrigerant density and velocity of sound in trarily chosen as the output point. For purposes
the refrigerant . Any consistent set of units can of calculation , the muffler termination impedance
be used. is chosen to be purely resistive, of magnitude
equal to the characteris tic impedance of the tube.
The matrix representat ion for a shunt section is
shown in Figure 3. This is derived by requiring As shown on Figure 6, each section of the muffler
continuity of pressure and volume current at the is assigned a letter and a number and the type of
junction point and treating the shunt as a section section determined. Section 1 is modeled as a
of transmissio n line with shorted termination . lumped capacitance and hence, the elements A(i,j)
That is, the closed end has volume current set to of matrix A are those for the lumped capacitance
zero. on Figure 4. The dimensions of this section are
Ll and Sl and refrigerant properties are Cl and
The lumped circuit representat ion for a capaci- pl.
tance is shown on Figure 4. This is used primarily
for representin g the head volume. Geometry is Section 2 through 6 are all transmissio n line
usually complex and therefore, it is necessary to sections and the elements of matrices B, C, D, E,
use a highly simplified model. A lumped capaci- and F are defined by Figure 2-C. Lengths are L2
tance has been found to be a reasonable model. through L6, and so for other parameters. Section
7 is the termination and its elements are as shown
Finally, the column matrix used for a resistive on Figure 5. This particular muffler does not
termination is shown on Figure 5. This is a good include a shunt section.
approximation for a muffler which is connected to
a long length of tubing, as a discharge muffler on In the actual program, it is first necessary to
a refrigerant compressor often is. Other termina- provide dimensions for all matrices. Next, the
tions have been used, however, and any desired one velocity of sound in each section, lengths, areas,
can be programmed in. One useful termination some- and density of refrigerant are read in. The
times found in practice in suction mufflers is a intermediat e functions I and J are defined for
hole in a plate. The first two terms of the each section. The program then asks for the fre-
infinite series solution for an orifice were used quency at which the computation is to start, the
in one analysis of this type with good results. ending frequency, and the frequency increment
For any termination , the procedure is to assume a size. For the starting frequency, the program
real unit output pressure and to ~alculate the then computes the I and J functions and the ele-
resulting output volume current, U, which may then ments for all matrices. Then the matrix multi-
be either real or complex. plications are done.

A computer program to analyze a particular muffler For the example shown here, the first multiplica-
is assembled by first dividing the muffler into tion is that of the G or termination matrix by
sections, each of which can be represented by one the F matrix. The program is written in BASIC
of the four matrices described above. Each matrix language which provides a single statement for
is assigned an identifying letter and number. The matrix multiplicat ion. The 4xl matrix resulting
data describing each section is then read into the from the first multiplicat ion is then multiplisd

383
 by the E matrix and so on. It is necessary to phone fitted into the output line so as to provide
specify in the program names for the intermedi ate as little acoustica l mismatch as possible.
natrices. Thus, the matrix formed by the product
of G by F might be designated matrix P. Matrix P Outputs of the microphone are amplified and con-
would then be multiplied by Matrix E and the pro- nected through a switch to a tuned narrow band
duct designated Q. voltmeter . Use of a sharply tuned filter on the
output is necessary to reject extraneous noise and
After all multiplic ations are completed, the tran~ harmonics of the driving frequency. Without this
mission loss for that frequency is computed and precautio n, highly incorrect results can be ob-
stored. The starting frequency is then incremented tained due to amplifica tion of harmonics of the
by the specified amount and the calculatio ns driving frequency by muffler resonances.
repeated. This continues until the desired stop-
ping frequency is reached and the data is printed Measurements are performed by successiv ely tuning
out. the oscillato r in small frequency increments
through the range of interest and recording the
It has been found desirable to use relatively sound pressure level in decibels at each micro-
large frequency steps for the first calculatio ns. phone. The differenc e between input and output
After results of calculatio ns are printed out, it is the transmission loss. The voltmeter must be
is usually found necessary to repeat calculatio ns tuned to the driving oscillato r frequency for
in selected frequency ranges using smaller fre- each measurement. As for the computed transmis-
quency steps to detail attenuatio n in the vicinity sion loss curves, frequency increments can be
of transmiss ion loss singulari ties such as large in regions not near resonances but near
resonances. resonances, the increments are made smaller.
Results of computations for this example will be Figure 8 shows a comparison between measured and
discussed after the measurement arrangement is calculated results. For this figure, the fre-
described . quencies measured in air have been scaled by the
ratio of the velocity of sound in refrigera nt at
MEASUREMENT ARRANGEMENT operating condition s to the velocity in air.
This factor usually ranges between 0.5 and 0.6.
It was considered highly desirable that the com-
puter model be validated by comparison with ex- The dashed line in Figure 7 is the transmission
perimental data. Measurements were made in air, loss measured in air and the solid line, the
since measurements in an operating refrigera nt calculated curve. Note that positive attenuatio n
system present considera ble experimental diffi- is plotted downward. Thus, regions where the
culties. Even measurements in air require con- muffler amplifies are above the zero line. What
siderable care if reproduci ble data is to be is desired for good noise suppression is that the
obtained. attenuatio n be positive (downward) and as large as
possible. The negative (upward) peaks need to be
Some factors which must be considered in muffler located as far from harmonics of the running speed
measurements are the establishm ent of realistic as possible.
source and terminatin g impedances, avoidance of
vibration and flanking paths, and eliminatio n of It will be noted that agreement between the two
electrica l distortion products. Means to avoid curves is good. One differenc e which can be seen
these problems are shown on Figure 7, which shows is that peaks in the calculated curve are sharp,
the equipment arrangement used for measurements. while measured peaks are rounded. ·This is because
effects of dissipatio n are not included in the
An oscillato r is used to feed a power amplifier computer program. Dissipatio n could be included
which in turn is connected to a horn drive~ unit. but for the purpose for which the analysis is
The driving frequency is measured with an elec- intended, the increased complexity was not
tronic counter. The horn driver unit is enclosed believed warranted.
in a soundproof enclosure to prevent flanking
paths for acoustic energy from the horn driver It is possible to correlate the singulari ties in
directly to the output microphone, bypassing the the transmission loss curves with the element or
muffler. elements which cause them. Each tube will, for
example, contribut e a peak at each frequency at
Energy from the driver is fed into the cylinder which it is a multiple of one half wavelength long.
head volume by means of a capillary tube filled One way to connect elements with singulari ties is
with fine wires. This provides an approximately to vary the length of a section and observe the
constant acoustica l current drive into the closed effect on the transmission loss.
head volume. Sound pressure level in the head is
measured with a l/4 inch condenser microphone. Better agreement between computed and measured
response curves can be obtained by adjusting the
The muffler is placed on a soft foam pad to pre- effective lengths and areas of sections. Some
vent transmiss ion of room vibration into the judgment is required because muffler sections
measurement system. A long (about 50 feet) copper often have rounded ends and thus, it is not clear
tube with the open end filled with loose cotton what the effective length is. A dimension which
serves as the anechoic terminatio n. Output sound requires considera ble care is the effective length
pressure is measured with a second 1/4 inch micro- of tube assigned to represent holes in a thin

384
baffle. Adequate theory is available for a baffle Sound and Vibration, Vol. 29, No.3, 1973,
with only one hole, but the effective length de- pp. 263-280.
creases as the number of holes increases.
3) Miwaj, T., and Igarashi, J., "Fundamentals of
CONCLUSION Acoustical Silencers", Aeronautical Research
Institute, Univ. of Tokyo, Report No. 344,
The method of muffler analysis outlined here has May, 1959.
proved to be of considerable value in providing
guidelines for muffler design. It has some ad- 4) Gatley, W.S., and Cohen, R., "Development and
vantages as well as disadvantages and it would be Evaluation of a General Method for Design of
well to discuss some of these. Small Acoustic Filters", ASHRAE Transaction s,
Vol. 76, 1971.
One limitation is that the method assumes plane
wave propagation and hence, the upper frequency 5) Cummings, A., "Sound Transmission in 180°
range is limited to about 1.22 c/d for a tube Duct Bends of Rectangular Section", Journal of
feeding into the center of a cavity and about Sound and Vibration, Vol. 41, No.3, 1975,
0.59 c/d for a tube feeding off center into a pp. 321-334.
cavity. Here, c is the velocity of sound and d is
the cavity diameter. For R22 at typical discharge
conditions and for a muffler 3 inches in diameter,
the upper limits are about 3,200 hertz and 1,600
hertz, respectivel y. For smaller mufflers, the
limits are higher. The program is still useful
above those limits,·but other effects not pre-
dicted by the computation will occur causing
significant effects on the transmission loss. / "'
(
\
The effects of flow and of dissipation are not
A

included. These could be added if desired but the u.~j

r
\
s I
I

additional complication has not been felt justi- P. \

/
I

fiable. A more important disadvantage is that

interaction s between compressor and muffler are
not accounted for. TUBE SECTIO N
In contrast to the disadvantages, it is very easy 1-A
and straightforw ard task to assemble a program to
analyze almost any single input-singl e output
muffler. The programs are very flexible and cost A
little to run. Different refrigerant properties
--1"~
A

can be used in each section of the muffler, if u, -u2

desired, and any source and load impedance can be
used. It can be modified to analyze mufflers for
t~
~----------------------~
multicylind er compressors but some different
methods must be used. It yields results which are
sufficientl y accurate and detailed for design pur- TRANS MISSION LIN~
poses and it is easy to see results of varying 1-8
muffler parameters.
Additional data, such as pressure at input or
output of any section and muffler input impedance
A

P, ~ P,P. + j P.:r:
" .
P:~."' P.:~~~. "'J P2r
can easily be read out. /1 •
U, oU,,., -~"J U,:x U,_ o: {4.A. +J U;;.z
In summary, it has been found that this approxi-
mate method of muffler analysis is a good compro- COMPLF'X VARIABLES
mise between simple methods which provide insuf-
ficient output data and more sophisticat ed 1-C
methods which are difficult to use, expensive to
run, and may provide more output data than a
muffler designer can utilize.
TRANS/'1/SS/0/V LIN£ StCT!O N
REFERENCE-S
1) Miller, D.F. and B.W. Hatten, "~1uff1er
F!CUA £ I
Analysis by Digital Computer", ASHRAE Trans-
actions, Vol. 66, 1960, pp. 202-202
2) Munjal, M.L., Narasimhan, M.V., and Sreenath,
A.V., "A Rational Approach to the Synthesis of
One-Dimensional Acoustic Filters", Journal of

385
 A "-A "- "-
P, ::: A fir 8 U;.,
.1\ A A II. Jl
U, :: Cf.,.. .- D U._

C£NFRAL NETWORK EQUA TION S

IJJ COM PLEX FORM
- PP,:r,~~.l ~
l[~~]
2 -A 0

l
I 0 0
0 I 0 0
u,~~. "' , o -(-};.) TAN{I2 ) 0
DEFIN E; Io:2-r rrLje- U, 1: (;;-;_) TA N(I2) 0 0

J= f7 c;q SHUNT SECT ION

po: GAS DENSI TY F!CU RE 3
C~ VELOC ITY OF SOUND A
U, - " ~ Uz
A

F c: F R E Q U £ N C Y (1-1 £ R T Z) 1 _P,_ _ _-,~J~C...;...o_ _ _j Pz

S'O"CR OSSSE CTfON AREA
V=VO LUME =LS
w ~2'Tr'f'

Ptt~.~~~~.. P,J.:P.;.x
~A =' AzR COS (I)- U".:r J SIN (I)
U,,. ::: Uu - w Go P, ~

-l
A:r ~ P,n:: COS (1) 1- U:u. J SIN (I) U,x = UH -t- 6J C0 ~ P.
U,A =--(Pz.'jr) SIN (I)+ U.2.11.. COS (I)
0 0
U,:t = (~'j:j) 5/N(I)+ U~:r COS (I) Pr"'] [
P,1: 0I 0I 0 0
U,~~. = o -w FC 0 I 0
U,:r: w(C6 0 0
NET\11/0R/{ [QUA T!ON S IN P.FAL F0Rfr1
2-B LUMPED CAPA CITAN C£ MATRIX
{!CUR £ 4

ftP.l
P 1:r
[COS(I)
Uu,. :::
[U11:
0
0
(:f)SIN(I)
0 0
COS(I) JSIN(I )
-(-})$1/U(l) CDS(I)
0 0
-J"SIN
0
0
(J)] [P.v.]
COS (.I)
Pz-r
U11 11.
Uu
0-:o pI ZT= ~&
'/§ IN
LINE TERMI NATED
Cf/ARA GTFRI ST!C
I fl.1 P£DA IJC £
UP.'= Pp, (ffc) 1\
MATR IX FORM U:z_ ~ P:z. (?"~)
CHOOSE P= I+Jo

2-C

D[V[LOPM£NT OF MATRIX [QUAT!ONS

F/GUP.E 2
TFRM INATI ON MATRIX FIG. 5

386
 A 8 c D £ F G
I 2 3 '{ 5 b 7
I
I
I

e;:;...,.6_::_,......J I
~-
...
VOLUME. t:i~'--PISTON Pour
...
Uour

2 3 '-1 5 (.,. 7
S £ CTl ON AI 0. I

LETT£ A A B c D r.: F G

TP,ANSNI SSIO/J
LUHPED TRAAJSMISSION TP. A 1J SM t5£ JON TRAAISM IBI()N TRANSMI SSION TERMINA TION
TYPE CAP L/ IJ £ L WE. LINE' LIIJ£ L!AJE"

p/I.Jll.
I
Puu. 0
= A B c D £ F
Uu111. 57;
u /NI
l/rc7
0

TYPICAL MUF FL[R AND C0/1PUT£P-. fv10D£L

FIGURE 6

387
 MICROPHONE
AMPLIFIER

TUNED
VOLTMETER
OSCILLATOR
MICROPHONE
AMPLIFIER

FREQUENCY OSCILLOSCOPE
COUNTER (4 CHANNEL)

POWER
AMPLIFIER
B&K lt"
MICROPHONES
MICROPHONE

IN SOUNDPROOF
BOX TUBING
MUFF L£P. M £AS UR £/VIE NT A RR A AJ G £ M £NT
f:!G UR£ 7

-J..o ~~~~ittltlli-r~-~·:rj
1 ."'
. I •.
_:_j +

!_I
I
~--------++----'----;----~-'>--!-+-+++->-+ -"-H::htittrmtt
I

l-1 wtl~act:lrtl!MII~ittttmtt;r--H:;trt:ttttl\tttlt·.'r\l:~:
· ,

30

/00 200 300 '100 500 700 2000

FR!:QU£NC Y (1-I~R T.2)

FIGURE 8

388