ISSN 0005−1144

ATKAAF 42(1−2), 5−12 (2001)

Branko Somek, Martin Dadi}. Mladen Maleti}

Active Noise Control in Ducts

UDK 621.371.56:534.8
IFAC IA 4.7.1;2.1.8
Original scientific paper

After reviewing the development of active noise control principles, we analyzed the elements of the active noise
control system in ducts. On the basis of this analysis, we created a model of these elements. Especially, we brought
a model of the loudspeaker in the z-domain, suitable for description of systems containing analog and digital parts.
Such model enabled us to analyze work and convergence of the adaptive signal processing algorithms applied to
active noise control. As an example, we analyzed performance of FXLMS algorithm on simplified model of active
noise control system in ventilation duct, and have shown a strong influence of the loudspeaker's transfer function
on the power spectrum of the error signal.
Key words: active noise control, adaptive filters, electroacoustical transducers, loudspeakers, ventilation ducts

1. INTRODUCTION the influence of the acoustical feedback between

loudspeaker and input microphone must be either
Large ventilation systems are often very noisy. minimized or compensated in order to avoid unsta-
Because of their wide application in the human ble operation. Efficacious dealing with these two
neighbourhood, there is need for decreasing its noi- problems became possible only with application of
se level. Passive noise control can not give the satis- the adaptive signal processing.
fying result for low frequencies – it is too costly and
it is not quit efficient.
Active noise control uses principle of destructive
interference of the sound waves, e. g. in order to
cancel undesired noise a sound wave with inverse
sound pressure is generated. To achieve a large
amount of cancellation, the anti-noise source must
generate an inverted signal of original noise signal
with great accuracy.
The principle of the active noise control was
firstly introduced in Paul Lueg's patent in the
1930's [1]. A microphone detects unwanted sound Fig. 1. Lueg's system
(noise) and gives this input signal to the electronic
system that drives the loudspeaker. The distance be-
tween microphone and loudspeaker, as well as elec- After Lueg, in 1953 Olson and May carried their
tronic system itself are adjusted in this way that a researches on active absorber [2]. In 1956, Conover
destructive interference between original and gene- described an active system for reducing transfor-
rated sound waves is achieved (Figure 1). The en- mer's noise [3]. Conover applied manual adjusting
tire process depends on relatively slow propagation of antinoise's phase and amplitude. Onoda and Kido
of the sound waves compared to the fast processing developed an automated system for transformer's
of the electrical signals. The amplitude and phase noise reducing in 1968 [4]. In 1981 Burgess sugges-
response of the electronic system must be adjusted ted application of adaptive digital filters in active
with a great accuracy to get a satisfying perfor- noise control [5]. In 1983 Chaplin developed a sys-
mance. There are two big problems in this appro- tem for the cancellation of periodic noise with use
ach. First, the electronic system must be capable to of waveform generator [6]. Roure described an ac-
compensate nonuniform amplitude and phase res- tive noise control system for ventilation ducts based
ponse of transducers, filters and amplifiers. Second, on frequency-domain analysis in 1984 [7].

AUTOMATIKA 42(2001) 1−2, 5−12 5

Active Noise Control in Ducts B. Somek, M. Dadi}, M. Maleti}

The influence of acoustical feedback was com- 2. MODELING OF SOUND FIELD IN INFINITE
pensated in ducts with use of Chelsea dipole source DUCTS
[8], (Figure 2), Swinbanks two-elements unidirec-
tional source [9], (Figure 3) and use of Jessel-Man- If the sound wave propagates through the infi-
giante-Canévet tripole source [10]. The Chelsea di- nite duct, and if its wavelength is much longer than
pole uses a pair of loudspeakers driven out of pha- duct's diameter then we can assume that we have a
se and spaced one-half wavelength apart. Jessel- plane acoustical wave. In the cases when the wave-
-Mangiante-Canévet tripole uses a monopole source length is less of the doubled diameter of the duct,
to cancel upstream propagation from the dipole we have high-order modes in the soundwaves pro-
source. There was also proposed application of pagation. Although most work on active noise con-
FURLMS adaptive signal processing algorithm to trol systems in ducts has focused on plane-wave
solve this problem [11]. propagation, higher order modes can become pro-
blem for large duct dimensions even for low fre-
quencies. This problem may be solved through the
use of a partitioned duct, as described in [16].
There are efficient passive noise control measures
for high frequencies with high-order propagation mo-
des, and we will discuss ahead only low frequency
soundwaves' propagation, e. g. plane acoustical
waves.
Plane wave propagation of the sound in the +x
axis direction of the Cartesian coordinate system is
Fig. 2. Chelsea dipole
described by the differential equation [1]

¶2 p 1 ¶2 p
2
- 2 × 2 =0 (1)
¶x c0 ¶ t
where p denotes sound pressure, t is the time and c0
is the velocity of the sound propagation.
The simplest antinoise source for the infinite
duct is one loudspeaker fixed on the wall of the
duct. For the frequencies low enough, neglecting
the area close to the loudspeaker, we can assume
the antinoise sound wave to be a plane wave, and
Fig. 3. Swinbanks system we can model the loudspeaker as a plane acoustical
monopole in the infinite duct [1].
A plane acoustical monopole can be imagined as
From a geometric point of view, active noise a pair of two massless pistons on the infinitesimal
control systems may be divided on duct noise con- distance, moving on the opposite directions. Their
trol systems, interior noise control systems, free- moving is directed by the time-variant volume ve-
-space control systems and personal hearing protec- locity of the monopole q(t). For the soundwave's
tion systems. Interior noise was actively controlled lengths long enough compared to the physical di-
in passenger cars [12], and in the propeller-driven mensions of the loudspeaker, the monopole's and
aircrafts [13]. The occupational noise can be redu- the loudspeaker's volume velocities will be the same.
ced by use of active acoustic barriers [14]. Active
hearing protectors were proposed in [15]. Active If we have the noise source in the xP point on
control of sound has some close associations with the x-axis, and the secondary source positioned in
the active control of vibration. In many cases active
control of sound and vibration can be treated as
one problem, especially when dealing with struc-
ture-borne sound.
In this paper a model of active noise control sys-
tem for infinite ducts will be developed, useful for
modeling of the active noise control systems in the
ventilation systems. On the basis of this model, the
efficiency of the adaptive signal processing algo-
rithms will be analyzed. Fig. 4. Two plane acoustical monopoles in an infinite duct

6 AUTOMATIKA 42(2001) 1−2, 5−12

B. Somek, M. Dadi}, M. Maleti} Active Noise Control in Ducts

Fig. 5. The elements of active noise control system

the xS point (Figure 4), and if their volume veloci- Good microphones could have very uniform fre-
ties are qP an qS, respectively, the sound pressure quency response.
will be canceled for x > xS if following equation is
satisfied: Piezoelectric accelerometers have the best proper-
ties, and this type of accelerometer is almost only
FG IJ
qS = − q P t − D , (2)
in use. Lower frequency limit in vibration measure-
H K
c0 ments is defined by the lower frequency limit of
the preamplifier and the lower frequency limit of
where D denotes the distance between two sources the accelerometer itself. The lower frequency limit
[1]. of the preamplifier usually can be below 1 Hz.
Tempe-rature sensitivity of the accelerometer define
its lower frequency limit, and using shear type one
3. THE ELEMENTS OF THE ACTIVE NOISE
can bring this limit below 1 Hz too.
CONTROL SYSTEM
In the middle frequencies range, a power ampli-
If we want to satisfy (2), we have to assure time
fier can be modeled as a voltage source with adjust-
delay between qS and qP equal to the time of sound
able clipping distortion. Dynamic range of the po-
waves' propagation between two sources. In the real
wer amplifier is extremely important for the proper
conditions, it means that we first have to detect ori-
work of the system.
ginal noise signal using an appropriate sensor, and
transduce acoustic signal to its electrical equivalent.
The electric signal has to be processed for ensuring 5. LOUDSPEAKERS
time delay condition, and after amplifying, brought
to the actuator (loudspeaker). As a part of the active noise control system, the
loudspeaker influences in great measure on overall
This chain can be described by the Figure 5. All work and efficiency of such a system. The loud-
elements of an active noise control (ANC) system speaker is a complex system consisted of electric,
have their own transfer functions. P(z) is transfer mechanical and acoustical parts. In this analysis an
function of the primary channel, ideally represen- assumption that the cone moves as uniform piston
ting by a pure time delay. The microphone, amplifi- is taken. Such assumption can be made for the low-
er and the loudspeaker are introducing amplitude -frequency range.
and phase changes in the system. The purpose of
the signal processor is to compensate these chan- The terminal voltage E on the voice coil of the
ges, and to ensure overall time delay for satisfying loudspeaker is given by:
condition (2) by its transfer function W(z). E = Z E I + Bl v . (3)

4. SENSORS AND AMPLIFIERS ZE is electrical impedance of the voice coil

As the noise sensor in an active control of sound Z E = R + sL (4)
system, one can use acoustic or non-acoustic sensor.
Non-acoustic sensor, for instance the accelerometer, where R is the resistance of the voice coil, L is in-
can be used in cases when noise is produced by ductance of the voice coil and s denotes complex
mechanical vibrations of the noise source. For pe- frequency. Also l denotes wire length of voice coil,
riodical noise signals, other non-acoustic sensors, for B is flux density in region of voice coil, and v is
instance tachometers can be used too. When using cone velocity amplitude. The relationship between
acoustical sensors, there is a problem of unwanted terminal voltage and cone movement is given by
acoustic feedback. [21]:
Overall characteristics of a microphone are influ- v = Bl (5)
E Z Z 2
enced by its acoustical and electrical properties too. E MT + ( Bl )

AUTOMATIKA 42(2001) 1−2, 5−12 7

Active Noise Control in Ducts B. Somek, M. Dadi}, M. Maleti}

where ZMT denotes total mechanical impedance. impedance was transformed using area ratio, as in
ZMT is defined by (7) on the loudspeaker loading, and the modified
total mechanical impedance ZMT was calculated for
Z MT = sm + D + K (6) the each frequency. Effective loading of the air
s mass on the back of the loudspeaker was estimated
where m denotes total effective mass of moving sy- by mass loading of a semi infinite space on a equi-
stem (i.e. cone, voice coil and effective air mass), D valent piston. From the calculated piston velocity
denotes damping, including mechanical damping in (5) and known load impedance, the driving point
moving system and radiation resistance, and K de- sound pressure was determined, and finally, sound
notes effective stiffness of suspension and back en- pressure on the more distant end. Resistance of the
closure, if any. voice coil was increased for 50 Ω output impedance
of the source. Figure 6 presents calculated and mea-
Generally, effective air mass and radiation resis- sured response. The discrepancies between two cur-
tance are frequency-varying, and are depending on ves can be explained with the losses in the duct
mounting conditions. For loudspeaker fixed on one and directivity characteristic of the microphone.
wall of the infinite duct, it is assumed, following re-
sults given in [18], that its radiation impedance
(mechanical) can be approximated at low frequen-
cies (below cut-on frequency of the first higher or-
der mode) with
S L2
RMA = ρ0 c0 (7)
2 SD
where ρ0 is density of air, c0 is velocity of sound in
air, SL is cone area, and SD is duct area. This ap-
proximation is verified with an experiment on the
short duct. A broadband loudspeaker was laterally
mounted on a 70 cm long rigid duct with circular
cross-section. Radius of the duct was 38 mm. Phy-
sical constants of the loudspeaker were measured,
and they are listed in the Table 1. Loudspeaker was
mounted 14 cm away from the one side of the duct.

Table 1. Physical constants of the broadband loudspeaker

Fig. 6. Sound pressure on the end of the opened short duct
Magnetic coupling factor Bl = 2.306 Tm
Resistance of voice coil 7.33 Ω
Using relations (3–7), a transfer function of the
Inductance of voice coil 0.2 mH loudspeaker in s-domain can be derived:
Mechanical stiffness of suspension 1962 N/m
v( s )
Mass of voice coil, cone, etc. 4.031⋅10 −4 kg H L ( s) = , (8)
E( s )
Mechanical damping 0.102 Ns/m
so we have finally
Piston diameter 77 mm
H L ( s) = s , (9)
K 3 s3 + K 2 s2 + K1 s + K 0
The transfer function between sound pressure on with coefficients:
the more distant end of the duct (56 cm away from
the loudspeaker) and terminal voltage was mea- K 3 = Lm
sured. The both ends of the duct were opened, and Bl
a miniature Brüel &Kjaer microphone was used in
these measurements. The transfer function was also K 2 = LD + Rm
Bl
calculated. The radiation impedance on the ends was (10)
taken as for the pulsation sphere, and driving point K 1 = LK + DR + Bl
impedance was calculated using transmission lines Bl
theory [19]. The losses in the duct were not taken
into account. The calculated overall driving point K 0 = KR .
Bl

8 AUTOMATIKA 42(2001) 1−2, 5−12

B. Somek, M. Dadi}, M. Maleti} Active Noise Control in Ducts

Applying the bilinear transformation [20]

−1
s = 2 ⋅ 1 − z −1 , (11)
T 1+ z

a z-domain transfer function can be achieved:

A3 z −3 + A2 z −2 + A1 z −1 + A0
H L ( z) = (12)
B3 z −3 + B2 z −2 + B1 z −1 + 1

with coefficients:
M = 8 K 3 + 4TK 2 + 2T 2 K 1 + T 3 K 0
2
A3 = −2T
M
2
A2 = −2T
M Fig. 7. Frequency response of the loudspeaker's model – magnitude
2
A1 = 2T
M
2 (13)
A0 = 2T
M
−8 K 3 + 4TK 2 − 2T 2 K 1 + T 3 K 0
B3 =
M
24 K 3 − 4TK 2 − 2T 2 K 1 + 3T 3 K 0
B2 =
M
−24 K 3 − 4TK 2 + 2T 2 K 1 + 3T 3 K 0
B1 = ,
M
where T is sampling period.
Following the results from section 2, sound pres-
sure in duct is defined for plane waves by:
v ⋅ SL
p = ρ 0 c0 . (14) Fig. 8. Frequency response of the loudspeaker's model – phase
2 ⋅ SD

To check the feasibility of the entire analysis, we

compared our results with more sophisticated nu-
merical analysis given by Shepherd et al. [21], as
well as with their measurements. In their study,
acoustic load on the face of the piston was deter-
mined numerically, following analysis given by Doak
Table 2. Physical constants of the Plessy-Foster loudspeaker

Magnetic coupling factor Bl = 12.8 Tm

Resistance of voice coil 6.7 Ω
Inductance of voice coil 1 mH
Mechanical stiffness of suspension 1400 N/m
Mass of voice coil, cone, etc. 0.028 kg
Mechanical damping coefficient 4.0
Piston diameter 200 mm
Fig. 9. Unit pulse response of the loudspeaker's model

AUTOMATIKA 42(2001) 1−2, 5−12 9

Active Noise Control in Ducts B. Somek, M. Dadi}, M. Maleti}

[22], and taking into account both propagating and error signal is
evanescent modes. They used 200 mm diameter e( n) = d( n) − y ′( n), (13)
Plessy Foster 200FO5 mounted in an unvented en-
closure of 20 litres. The attenuator was set up on a and coefficients updating is
rigid walled duct of 244 mm square section. Physi-
cal constants of that loudspeaker are listed in Table W ( n + 1) = W ( n) + 2 µX ′( n) e( n) (14)
2. The effective air mass was estimated by mass
where W is weight vector of the adaptive FIR filter,
loading of a semi infinite space on a piston, and
µ is step size that satisfies stability condition [23], X
suspension stiffness was not increased to account
is vector of the input samples and X′ is vector of
for the enclosure [21].
input samples filtered by secondary path estimate
With sampling frequency 1/T = 40 kHz, covering C(z). Generally, C(z) is achieved by some method of
audio frequencies range, we obtained loudspeaker's system identification.
response in examined duct presented with figures For simulation purposes, it is assumed that P(z)
7–9. There are also presented predictions, given by includes only time delay:
Shepherd et al. Frequency response is presented
only for the frequencies below the cut-on frequency P( z) = z −60 . (15)
of the first higher order mode in this duct. Compa-
ring their results with our predictions, one can see
very good matching.

6. SIGNAL PROCESSING

To ensure proper work of the system, signal pro-

cessing has to be applied. It can be done by fixed
or adaptive filters. The need for a very accurate
phase and amplitude response leads toward applica-
tion of the adaptive signal processing. It is caused
by the fact that properties of transducers, as well as
the properties of electronic elements are time and
temperature varying. The air temperature also in-
fluences on the sound propagation speed.
If we idealize properties of microphones and am-
plifiers, as well as acoustic path properties between
noise source and secondary source, and if we apply
loudspeaker model described in section 5, the ap-
Fig. 11. Power spectra of the residual noise (mean of 50 indepen-
plication of the classical FXLMS adaptive algorithm dent trials) before adaptive processing (ANC off) and during adap-
[23, 24] can be represented with Figure 10, where tive procesing (i.e. 1300 iterations, ANC on)
e(n) denotes error signal, and d(n) is unwanted noi-
se signal that is acoustically summing with antinoise
signal y ′(n).

Fig. 10. FXLMS algorithm

Filter output is defined by

y( n) = X T ( n) W ( n), (12) Fig. 12. Residual noise in time-domain with H(z) = C(z) = z−1

10 AUTOMATIKA 42(2001) 1−2, 5−12

B. Somek, M. Dadi}, M. Maleti} Active Noise Control in Ducts

For the secondary path transfer function is as- [5] J. C. Burgess, Active Adaptive Sound Control in a Duct: A
sumed: Computer Simulation. J. Acoust. Soc. Am., 70 (3), pp.
715–725, 1981.
H ( z ) = H L ( z ), (16)
[6] G. B. B. Chaplin, R. A. Smith, Waveform Synthesis-The Es-
with loudspeaker and duct values presented in sec- sex Solution to Repetitive Noise and Vibration. Proc. Inter-
-Noise 83, pp. 399–402, 1983.
tion 5.
[7] A. Roure, Self Adaptive Broadband Active Sound Control
In ideal case it is also System. Journal of Sound and Vibration, 101, pp. 429–441,
1985.
C( z ) = H ( z), (17) [8] K. Eghtesadi, H. G. Leventhall, Active Attenuation of Noise:
The Chelsea Dipole. Journal of Sound and Vibration, 75, pp.
so we are capable by applying these idealizations to 127–134, 1981.
investigate loudspeaker's transfer function influence [9] M. A. Swinbanks, The Active Control of Sound Propagation
on active noise control system performance. For in Long Ducts. Journal of Sound and Vibration, 27, pp.
Gaussian white noise as the noise signal, sampling 411–436, 1973.
frequency 1/T = 2 kHz and filter length of 100 coef- [10] G. Mangiante, Active Noise Control in a Duct: the JMC
ficients, we get residual noise spectrum presented Method. Proc. Inter-Noise, pp. 1297–1300, 1990.
with figure 11. As a comparison, modeling C(z) = [11] L. J. Eriksson, M. C. Allie, R. A. Greiner, The Selection and
= H(z) = z−1 gives as result almost absolute canceling Application of an IIR Adaptive Filter for Use in Active
of noise signal (figure 12). Sound Attenuation. IEEE Transactions on Acoustics,
Speech, and Signal Processing, Vol. ASSP-35, No. 4, pp.
433–437, 1987.
7. CONCLUSION [12] S. J. Elliott, I. M. Stothers, P. A. Nelson, A. M. McDonald,
An active noise control system in small area D. C. Quinn, T. Saunders, The Active Control of Engine
Noise Inside Cars. Proc. Inter-Noise, pp. 987–990, 1988.
ducts is analyzed. Acoustically, we problem estab-
lished as the active noise control in the infinite [13] M. A. Simpson, T. M. Luong, M. A. Swinbanks, M. A. Rus-
sell, H. G. Leventhall, Full Scale Demonstration Tests of
duct. Such system is consisted of acoustic paths, Cabin Noise Reduction Using Active Noise Control. Proc.
sensors, signal processor, amplifier and actuator Inter-Noise, pp. 459–462, 1989.
(loudspeaker). The need for very accurate ampli- [14] S. Ise, H. Yano, H. Tachibana, Application of Active Con-
tude and phase relations between noise source sig- trol to Noise Barrier. Proc. Int. Symp. Active Control of
nal and generated anti-noise signal, as well as the Sound Vib., pp. 309–314, 1991.
time-varying characteristics of the system elements, [15] C. Carme, A New Filtering Method by Feedback for ANC
leads toward application of the adaptive digital fil- at the Ear, Proc. Inter-Noise, pp. 1083–1086, 1988.
ters. In order to verify performance and conver- [16] M. Takahashi, K. Matsumoto, R. Gotohda, H. Hamada, Ac-
gence speed of the adaptive signal processing algo- tive Noise Control in Large Ventilation Ducts by Using
rithms, we presented elements of these systems, and Single-Channel Adaptive Controllers. J. Acoust. Soc. Jpn.
a z-domain model of the loudspeaker mounted on (E), Vol. 19, pp. 413–416, 1998.
the wall of a small-area duct is developed. Compa- [17] B. Somek, M. Dadi}, Kundt's Tube As a Model for Appli-
ring frequency response of our models with much cation of Active Noise Control. MIPRO '99 Proceedings,
Volume 1, Opatija, Croatia, pp. 78–81, May 1999.
more sophisticated numerical calculations as well as
with measurements, we achieved a very good mat- [18] S. Irrgang, Optimisation of Active Absorbers in Rectangu-
lar Ducts. Active '97 Proceedings, Budapest 1997.
ching. Our model allows time-domain simulations,
and using this model, we analyzed influence of the [19] I. B. Crandall, Theory of Vibrating Systems and Sound.
loudspeaker's transfer function on FXLMS algo- Van Nostrand, New Jersey, 1954.
rithm performance. A great influence of the loud- [20] A. V. Oppenheim, R. W. Schafer, Digital Signal Processing.
speaker on the overall system performance and on Prentice-Hall International (UK), London, 1975.
the power spectrum of the error signal is shown. [21] I. C. Shepherd, A. Cabelli, R. F. LaFontaine, Characteristics
Such a model allows us to analyze efficiency of par- of Loudspeakers Operating in an Active Noise Attenuator.
Journal of Sound and Vibration, 110(3), pp. 471–481, 1986.
ticular signal processing algorithms for active noise
control purposes. [22] P. E. Doak, Excitation, Transmission and Radiation of
Sound from Source Distributions in Hard Walled Ducts of
Finite Length (1): Effects of Duct Cross-section Geometry
REFERENCES and Source Distribution Space-Time Pattern. Journal of
[1] P. A. Nelson, S. J. Elliott, Active Control of Sound. Acade- Sound and Vibration, vol. 31, pp. 1–72, 1973.
mic Press, London, 1992. [23] B. Widrow, S. D. Stearns, Adaptive Signal Processing. Pren-
[2] H. F. Olson, E. G. May, Electronic Sound Absorber. J. Aco- tice Hall, Englewood Cliffs, 1985.
ust. Soc. Am., 25, pp. 1130–1136, November 1953. [24] S. M. Kuo, D. R. Morgan, Active Noise Control Systems:
[3] W. B. Conover, W. F. M. Gray, Noise Reducing System for Algorithms and DSP Implementations. New York, John
Transformers, U. S. Patent 2,776,020, January 1, 1957. Wiley, 1996.
[4] S. Onoda, K. Kido, Automatic Control of Stationary Noise [25] L. J. Eriksson, M. C. Alie, A Practical System for Active
by Means of Directivity Synthesis. Paper presented at the Attenuation in Ducts. Sound&Vibration, vol. 22, no. 2, pp.
Sixth International Congress on Acoustics, 1968. 30–4, Feb. 1988.

AUTOMATIKA 42(2001) 1−2, 5−12 11

Active Noise Control in Ducts B. Somek, M. Dadi}, M. Maleti}

Aktivni sustavi za za{titu od buke u cijevima. Uz dani prikaz razvoja aktivne za{tite od buke, analizirani su ele-
menti sustava aktivne za{tite od buke u cijevima, a na temelju te analize napravljen je model elemenata. Posebno
je provedena analiza zvu~nika te je napravljen model zvu~nika u z-domeni, prikladan za analizu sustava koji imaju
analogne i digitalne dijelove. Primjenom takvog modela mo`emo u vremenskoj domeni analizirati rad i konvergen-
ciju pojedinih adaptivnih algoritama obrade signala. Kao primjer provedena je analiza rada sustava aktivne za{tite
na pojednostavljenom modelu ventilacijskog kanala primjenom FXLMS algoritma, te je pokazan jak utjecaj prije-
nosne karakteristike zvu~nika na spektar snage zvu~nog signala preostale buke.

Klju~ne rije~i: adaptivni filtri, aktivna za{tita od buke, elektroakusti~ki pretvornici, ventilacijski kanali, zvu~nici

AUTHORS’ ADDRESSES:
Prof. dr. sc. Branko Somek
Department of electroacoustics
Faculty of Electrical Engineering and Computing
University of Zagreb
Unska 3, HR-10000 Zagreb, Croatia
Dr. sc. Martin Dadi}
Department of electrical engineering fundamentals and
measurements, Faculty of Electrical Engineering and Computing
University of Zagreb
Unska 3, HR-10000 Zagreb, Croatia

Doc. dr. sc. Mladen Maleti}

Department of electroacoustic
Faculty of Electrical Engineering and Computing
University of Zagreb
Unska 3, HR-10000 Zagreb, Croatia
Received: 2001−10−05

12 AUTOMATIKA 42(2001) 1−2, 5−12