CC BY
91Electronic Journal «Technical Acoustics» http://webcenter.ru/~eeaa/ejta/
2004, 17
Allan Kim, Andrey Bykov, Andrey Troshin*
IT Magic Co. Ltd. 4rd WonWoo Bldg, 907-16 Deachi-dong, Gangnam-gu, Seoul Republic of Korea
Active noise control and noise reduction: practical examples
Received 08.11.2004, published 20.11.2004
The practical examples of ANC (active noise control) systems are shown and discussed in the paper. Examples shown for active noise reduction headsets, noise cancellation in the duct system, general noise reduction using adaptive algorithm are discussed. The efficiency of the active noise control is 10-20 dB for the all cases except of transformer where much more control channels have to be implemented in order to obtain the global reduction of three transformer harmonics. It was shown that global reduction for transformer case could be achieved using anti noise loudspeaker and error microphones separated with one half of wavelength.
INTRODUCTION
The idea of active noise control is quite old and referred to Lueg who’s patent was published in 1933. Further theoretical development was done by Nelson and Elliot [1], C. Hansen [2]. The main problem in the active noise control is to duplicate exactly the original sound field using secondary so-called anti-noise sources.
In applications of active noise control (ANC) the cancelling signal is generated electronically and introduced into the system using transducers such a loudspeaker that convert electric signal to the sound. The single control channel usually includes reference sensor, error sensor, preamplifier, digital controller, power amplifier and noise cancellation loudspeaker.
If acoustic field is very complex the huge number of control channels has to be utilized. This circumstance makes commercial implementation of the ANC impossible due to high cost of the total system. That is why most authors declare that effective noise reduction by active means can be obtained if acoustic field is very simple with respect to space. The most attractive case for active noise control is a one-dimensional wave propagating like a sound field in the ducts or headsets.
Another practical limitation is the frequency range where active noise control could be applied with high efficiency. For global noise reduction by active means the cancelling loudspeaker and error sensors have to be separated with one half of the wavelength as it was proved in [1]. It is obvious that in the high frequency range when wavelength becoming shorter the number of cancelling loudspeaker and error sensors substantially increases comparing to low frequency range. The application of active noise control is usually limited
Corresponding author, e-mail: andrey@itmagic.co.kr
to frequencies below 500 Hz. According to [2] there are four basic limits, which can impact the performance of active noise control system: limit determined by control system performance, limit due to quality of reference signal, limit determined by error sensors arrangement and limit determined by placement of anti-noise loudspeakers.
Any kind of sensor for measurement of variable physical values can be implemented as a reference sensor. Most simple and cost effective sensor is electret microphone, however nonacoustic sensors like accelerometer, tacho sensor, strain gauge etc. can be applied also in situ. The implementation of non-acoustic sensors for ANC system eliminates the problem of positive feedback occurred between reference and error microphones. Active noise control technique can provide sometimes very good performances in specific cases when passive noise control measures require more heavy materials to block low frequency noise.
1. ACTIVE NOISE CONTROL: BASICS AND ALGORITHMS
There are two general types of active noise control system feedback and feedforward. The schematic diagram for both systems is depicted in Fig. 1. Feedforward system comprises reference microphone, error microphone, noise cancelling loudspeaker and adaptive electronic controller. The feedback system normally consists of error microphone, noise cancelling loudspeaker and electronic controller.
(a)
(b)
Fig. 1. Feedforward ANC system (a) and feedback ANC system (b) examples given for
duct ANC system
ANC systems may be classified as narrow band and wideband depending on what kind of unwanted noise has to be cancelled: tonal, random or random narrow band.
The LMS (Least Mean Square) adaptive algorithm nowadays is the “working horse” of all commercial ANC applications due to its cost efficiency. Here we briefly consider the FxLMS algorithm, which is more suitable for active noise control by electro acoustic means. More details about digital signal processing (DSP) algorithms for active noise control can be found in [3]. Consider the core of FxLMS algorithm. The schematic diagram of FxLMS algorithm is depicted in Fig. 2.
Fig. 2. Block diagram of FxLMS
x(n) — reference signal from the reference microphone, e(n) — error signal from error microphone, d (n) — primary noise disturbance, x'(n) — filtered reference signal,
_y(n) — cancellation signal, y'(n) — filtered cancellation signal, P( z) — primary acoustic path from reference microphone to error microphone, S(z) — secondary path from controller output to summation point, S'(z) — estimated secondary path obtained during off-line modeling, W (z) — Weiner filter response
Role of reference sensor is to get exact signal, which has to be cancelled, or get the pure signal referred to the noise source. Error sensor role to feed back to controller and monitoring the residual acoustic pressure and adjust controller for optimal results. The optimal frequency response function of the adaptive filter can be expressed in the form:
Sdx' (f)
Wo (f) = dx^’ (1)
S.Af) K)
where: f — frequency; Sdx, (f) — cross spectrum between filtered reference signal and disturbance signal; Sxx (f) — auto spectrum of filtered reference signal.
The equation (1) expresses the frequency response of the optimal controller in least mean
square sense. FxLMS algorithm has advantage for electro acoustic applications because it
takes into account the secondary transfer paths that LMS algorithm does not. Secondary
transfer path can be estimated during so-called off-line modeling. At this stage the impulse response of S (z) using input testing signal is obtained.
The efficiency was calculated using the equation (2) and results given in the Table 1.
Eff (f) = 10lg
f rlef.err(f) ^ 1 -Yleferr (f)
(2)
where:
,(f) — ordinary coherence function between signal from reference and error
microphones.
Table 1. Results of efficiency estimation
Coherence function Potential efficiency, dB
0,6 2
0,7 4
0,8 6
0,91 10
0,93 11
0,94 12
0,95 13
0,96 14
0,97 15
0,98 17
0,99 20
0,999 30
2. ANC HEADSET
The passive headset can suppress effectively the high frequency noise only. The good reduction of low frequency components connected with huge amount of sound absorbing material and big weight of headset air caps. Therefore it is very attractive to implement the active noise control strategy for low frequency noise reduction by headset.
There are two kinds of active headsets - analog and digital. The classification could be more complicated due to various algorithm implementation for digital and difference in circuit arrangement for analog headset. The analog headset can be also considered like a feedback ANC system while digital is a feedforward one. Two kinds of product prototype have been developed. The digital and analog set-ups are shown in Fig. 3 and Fig. 4, respectively.
Fig. 3. Set-up for digital feedforward ANC headset
The performances of both headsets were evaluated and results are shown from Fig. 5 to Fig. 7. The digital ANC headset provide more efficiency around 15 dB in a wide frequency range from 100 Hz to 1000 Hz comparing to analogue one where the 10 dB reduction index in more narrow band can be seen. Another advantage of digitally controlled headset is its ability to make adaptation if noise spectrum changes, for example due to altering the number of revolutions for rotary machinery or propeller.
Fig. 5. Noise reduction efficiency by digital headset (tonal noise) solid line — ANC off, dashed line — ANC on
40 -35 -
30
50 100 150 200 250 300 350
Frequency (Hz)
Fig. 6. Noise reduction efficiency by digital headset (white noise) solid line — ANC off, dashed line — ANC on
SPL, dB
Frequency (Hz)
Fig. 7. Noise reduction efficiency by analogue headset (white noise) solid line — ANC off, dashed line — ANC on
3. NOISE CANCELLATION IN THE DUCT
The most attractive application for the active noise control is duct system. Here the sound field is very simple, only one dimensional wave propagating occurs when duct diameter or largest dimension do not exceed one third of the wavelength. Knowing the speed of sound and dimension of duct cross-section a critical highest frequency can be calculated. Therefore below the critical frequency the single channel ANC system can be applied. An experimental setup for computer fan noise reduction is shown in Fig. 8.
Fig. 8. ANC set-up for reduction of computer fan noise
It can be seen from fan noise spectrum measurement that 295 Hz tonal component is dominant in the fan noise spectrum. Moreover the coherence function between error and reference microphones is close to unity therefore feed forward active noise control system would be very effective for that purpose. The reduction of 18 dB for tonal noise at frequency of 295 Hz was obtained.
4. GENERAL NOISE REDUCTION USING ADAPTIVE ALGORITHM
The urban environmental noise affects the voice intelligibility during communication using cellular phones. There are several problems that should be overcome during design and development of anti noise DSP platform and algorithm. The most essential problem is to save the voice fidelity with implemented DSP algorithm. Another problem is to reduce the power consumption from the cellular phone battery. The third problem referred to the fact that overall sizes of the board should be very small.
A noise reduction solution was developed using DSP platform BS200. The BS200 platform is shown in Fig. 9. Noise reduction effect is shown in Fig. 10.
Fig. 9. BS200 DSP platform
Fig. 10. Noise reduction efficiency by adaptive algorithm
The BS200 noise reduction solution shows a small current consumption from battery less than 7 mA for 12~24 dB noise reduction without serious distortion of the speaker voice. Total harmonic distortion of the BS200 platform does not exceed 0.5%. Obtained performances make this solution quite effective for cellular phone application.
5. NOISE REDUCTION OF INDUSTRIAL TRANSFORMER
The most attractive case for active noise control is a cancellation of low frequency harmonics of industrial transformer. Connover W. B. and Ringlee R. J. have implemented first analog feedback ANC system in order to reduce first harmonic of transformer noise with frequency 120 Hz [4]. The efficiency of that ANC was not stable due to changing the environmental temperature, humidity and speed of sound.
Transformer noise generated due to vibration of the magnetic core with double frequency of the power network and its harmonics. Coherence function between reference signal and error one always very high and feedforward algorithm for noise control seems to be most suitable. The ANC set-up for industrial transformer is shown in the Fig. 11. Here the sub-band FxLMS algorithm was implemented with three frequency bands 120, 240 and 360 Hz. Those frequencies are corresponded to three main harmonics in transformer sound field.
The measurements shown that total reduction effect exists around appropriate error microphone and becomes substantially less if observation point was moved to one half of wavelength far further from error microphone (see Fig. 12). Therefore the number of channels has to be increased and separated with the distance equal to one half of wave length at frequency 360 Hz to obtain a the global noise reduction.
Fig. 11. ANC set-up for industrial transformer
Fig. 12. Quite zone pattern in the front plane of transformer
CONCLUSIONS
Four different cases of active noise control implementation are discussed in the paper. We can conclude that for some specific applications the active noise control can be very effective and deliver at least 20 dB reduction index not only for tonal noise component but in a wide band as well. Example is given for active anti noise headset. Active noise control can be applied also to cancel the tonal component of computer fan with efficiency of 18 dB. Development of general noise reduction algorithm to cellular phones was successfully completed. Usage miniature DSP platform with low power consumption makes possible to reduce wide band noise saving voice intelligibility. The implementation of ANC technology to more complicated field distribution over the space with limited number of control channels does not allow providing the global reduction, just local quit zone can be achieved. Example of sub-band algorithm is given for transformer case study.
Results of the study and development of active noise control for various cases show that there are some specific cases when this technique can be successfully implemented, namely, if acoustic field is quite simple. When acoustic field is three-dimensional like transformer case the implementation of the system becomes too complicate and expensive in commercial applications.
REFERENCES
1. Nelson P. A., Elliott S. J. Active control of sound. Academic press, 1992.
2. Colin H. Hansen. Understanding active noise cancellation. Spon Press, London and New York, 2001.
3. Sen M. Kuo, Dennis R. Morgan. Active noise control: systems algorithms and DSP implementations. John Wiley and Sons Inc, 1996.
4. Connover W. B., Ringlee R. J. Recent contributions to transformer audible noise control. Transactions of the AIEE, Part III, Power Apparatus and Systems, vol. 74, 77-90.
