|
Active Noise Control
Introduction
Passive suppression of low-frequency noise or vibration has
numerous disadvantages, mainly because of the volume of the absorbers. By the
usage of signal processors it has became possible to utilize different active
noise control methods. These are based on the phenomenon of the destructive
interference. A "secondary" noise has to be generated, which suppresses the
"primary" (i.e. the original) noise at the properly situated microphones.
Signals of the microphones are used to control the secondary noise.
Fig. 1. The basic idea of active noise control
The whole problem has acoustical, signal processing and control
aspects. In some cases, if the secondary source can be installed near to the
primary source (e.g. a ventilation duct), a single loudspeaker-microphone pair
is enough. In other cases, if the noise source is not concentrated, and the
suppression in an enclosure has to be achieved (e.g. an airplane), more
loudspeakers and microphones are utilized.
Active noise control is a new research and development
direction at our department. However, the knowledge-base in measurement, system
identification and signal processing helps us to obtain new results in this
field. The performance of the noise control system is usually evaluated by the
spectra of the microphone signals measured with and without control. Typical
spectra can be seen in the figure below.
Fig. 2. The effect of active noise control on the
spectrum of the noise
Technical Background
Active noise control requires multifarious knowledge in the
field of measurement, analog and digital signal processing and acoustics. Design
of active noise control systems needs measurements on the plant to be
controlled. The records derived during the measurements play an essential role
in plant identification and system simulation. Algorithm design and
evaluation has a central role. It should be in close ties with the other
parts of the design process. The main task is to implement or improve digital
signal processing methods which can be used for this purpose. Although the
design needs the first measurement results, involving the a priori knowledge in
the field (literature, experiences) the development can be started at the
beginning of the design. This development work includes theoretical derivations,
simulations and laboratory experiments. Due to the special features (e.g. high
temperature) of the possible acoustic systems, the active noise control system
needs special hardware tools. Sensor and actuator design is an important
part of the system design. The final phase requires industrial
experiments giving feedback to the above parts of the design. This
experimental phase is also for revealing the scope of active noise control for
the actual purpose.
Research Activities
Conventional noise controllers are usually adaptive filters,
with the parameters updated mainly on LMS basis. The first systems for this task
were feedback systems. The most successful applications of them are single
channel systems in which the loudspeaker is closely positioned to the error
microphone (e.g. an ear defender). The next step was the implementation of
feedforward controllers. These systems need an additional reference signal which
is correlated with the primary noise. Feedforward control was successfully
implemented in various active noise systems. Such adaptive systems utilize the
filtered-X LMS (XLMS) algorithm, therefore they need a built-in model on the
acoustic path between the secondary source and the microphone. The accuracy of
this model influences the success of the control. Inaccurate model leads to loss
in the performance, or in serious cases the system will be unstable. Adaptive
systems are able to suppress both broadband and harmonic noise.
Periodic noise control
If the noise to be suppressed is periodic, adaptive feedforward
control is straightforward, because of the easily available reference signal.
This control task is rather simple, therefore the controller should be simple,
as well. However, the structure of the usual adaptive controllers is similar to
that for broadband noise control. The aim of our first research work was to
design an adequate controller, which is fitted to periodic noise cancelation
providing simpler structure and better control results than the conventional
adaptive controllers.
The theoretical background of such a controller design is the
adaptive Fourier analysis. The adaptive
Fourier analyzer (AFA) is a structurally adaptive system for exact
measurement of band-limited periodic signals of arbitrary fundamental frequency.
It is an extension of the resonator based observers developed earlier to perform
the recursive discrete Fourier transform (RDFT). In these observers the
resonators work in a common feedback loop providing zero steady-state feedback
error at the resonator frequencies. The AFA adapts the resonator frequencies to
coincide with those in the input signal. The proposed noise controller can be
considered as an extension of the AFA. The block diagram of the system can be
seen in the following figure.
Fig. 3. Periodic noise control
More details can be found in the paper
presented at the Active'97 conference.
Broadband Noise Control
The experience with periodic noise controller design helped us
to extend the ideas to the field of broadband noise control. In the scope of the
research was the convergence of the LMS based adaptive filter structure.
Adaptive filters updated by the least mean square (LMS)
algorithm are successfully utilized for both identification and control purposes
in active noise control systems. In the case of noise control, the output of the
adaptive filter drives the secondary loudspeaker, and the error signal is
derived by the microphone only at the end of the secondary path. In such cases
the filtered reference LMS (XLMS) algorithm guarantees the stable adaptation.
Multiple channel systems utilize the multiple error LMS algorithm. However, the
convergence of these algorithms can be very slow, depending on the transfer
function of the secondary path. We have introduced a novel algorithm which
provides much higher convergence rate than the XLMS algorithm. The proposed
structure is a modification of the XLMS structure: in addition to the filter in
the reference signal path, a secondary filter is applied, and the same filter is
applied in the error signal path. This secondary filter is designed so that the
resulted magnitude response in the adaptation loop oscillates around the unity.
The proposed algorithm was extended also for multiple channel active noise
control. Numerical and practical examples verify that the proposed method
improves the convergence rate significantly without high additional
computational demand. The block diagram can be seen in the next
figure.
Fig. 4. Broadband noise control with the application
of the extended XLMS algorithm
More details can be found in the paper
presented at the Active'99 conference.
On-line Identification of the Secondary Path
The transfer function of the secondary path can change
considerably that the built-in model cannot ensure the stability of the system.
In this case the system is unstable, it should be turned off, and the
identification should be carried out again. In certain applications such a
change is quite common, so the off-line identification is not a proper solution
for active noise control. Such applications are the outdoor noise control tasks,
where the change of the temperature or the humidity of the air causes meaningful
change in the secondary path. Another possibility is the sudden change in rooms
(e.g. door opening).
The research of this topic has just recently started, with the
application of the adaptive filter based methods available in the literature.
The resonator-based controller developed for periodic noise control could be
improved so that it is able to retain its stability even if the secondary path
changes.
Application of Sensor Networks
Nowadays sensor networks are used more frequently for data
acquisition. It is a straightforward idea that the signal of the microphones can
be pre-processed and transmitted by the elements of the networks (these are the
so-called motes). The great advantage of the solution is that the installation
of the microphones is much simpler, and the geometry of the system is easily
reconfigurable. The block diagram of our first system can be seen in the
following figure.
Fig. 5. Active noise control with sensor
network
One of the motes has not microphone, it has a wired connection
to a DSP board, where the algorithm is implemented. The loudspeakers are
connected also by wires.
Applications
Active noise control poses many theoretical problems, the
research results can be used by other engineers working in this field. On the
other hand the implementation has an essential role. The development of certain
applications answers whether a new algorithm is really viable.
Active Noise Control for Industrial Phones
It is often unavoidable that phones in a factory operate in a
very noisy environment. Noise of different machines and equipment is also
transmitted in the phone line. There are passive methods to suppress this noise,
but they cannot solve the problem. Active noise control is a possibility for
efficient noise suppression, even if the spectra of the speech and the noise are
overlapped in the frequency domain. The solution is depicted in the figure
below.
Fig. 6. Active noise control for industrial
phones
Active Noise Control for Airplanes
Propeller-driven airplanes generate a high-level noise which is
disturbing for the passengers. The problem was posed by the airplane
manufacturer Fokker, and the experiments were made at the Technische Physische
Dienst in Delft, the Netherlands. Unfortunately, the successful experiments were
interrupted because of the breakdown of Fokker. The resonator-based structure
for periodic noise control mentioned above was tested in a laboratory
environment. The main characteristics of the system were the following:
|
- 2 fundamental frequencies
- 4 controlled harmonics
- 5 microphones
- 4 loudspeakers
- 1300 Hz sampling frequency
- TMS320C30 DSP
- Adaptive Fourier Analyzer
- Resonator Based
Controller
|
Active Noise Control for Furnaces
The aim is development of active control systems which suppress
the noise appearing in furnaces and boiler rooms. The noise generated by the
combustion chamber of the furnace has different injurious influences: in extreme
situations it can destroy the equipment, but even at normal conditions the noise
is harmful for the personnel working in the boiler room. The noise leaving the
boiler room is a heavy environmental load. Conventional methods of suppressing
this noise using sound absorbers generally do not work well at such
circumstances. On the one hand, the power of this noise is concentrated in the
low-frequency range (below 500 Hz), on the other hand traditional noise
absorbers cannot be installed on furnaces. Active cancelation of the
furnace-noise seems to be a promising solution. This project was granted by the
Hungarian Fund for Scientific Research OTKA T-023868.
Noise Control of Transformers
The noise of high-power (in the range of 1 megawatts)
transformers causes high environmental load, especially at night. The noise
radiation cannot be reduced properly by the design of the transformers, this is
why active noise control is a real alternative. The task is a nice example of
outdoor noise control, where the noise radiation should be reduced in a certain
direction. Since the noise to be suppressed is periodic, the resonator-based
controller has been suggested.
Other Applications
The principle of active noise control can be applied not only
for vibroacoustic problems. It has other physical content, but very similar
problems are to be solved if active magnetic shielding is to be produced. Our
experimental system works well. The theoretical connection is more complicated
but also a similar approach can be used if a high-voltage or high-current sine
wave generator is to be designed. Usual electronic devices generate a distorted
waveform, and the undesired higher harmonics can be canceled actively.
Recommended publications:
S. M. Kuo, D. R. Morgan, "Active noise control: a tutorial review",
Proceedings of the IEEE, vol. 87, No. 6, June, 1999, pp.
943-973. |
Review of the research field. The most important algorithms and
applications are investigated, and more than 150 references
cited. |
L. Sujbert, G. Péceli, "Noise cancelation using resonator based
controller", presented on the Active '97 - The International
EAA Symposium on Active Control of Sound and Vibration,
Aug. 21-23, 1997, Budapest, Hungary, in proceedings pages 905-916. |
Introduction of the resonator-based controller, developed at the DSP
laboratory. |
L. Sujbert, "A new filtered LMS algorithm for active
noise control", presented on the Active '99 - The
International EAA Symposium on Active Control of Sound and Vibration,
Dec. 2-4, 1999, Fort Lauderdale, Florida, USA, in proceedings pages
1101-1110. |
Improvement of the classical XLMS algorithm. |
L. Sujbert, B. Vargha, "Active distortion cancelation of sinusoidal
sources", presented on the IEEE Instrumentation and
Measurement Technology Conference, May. 18-20, 2004, Como, Italy, in
proceedings pages 322-326. |
Application of the theoretical results in a different
field. |
Other related publications of the DSP laboratory can be found on
this page.
Useful Links:
ISVR Demonstartions |
Acoustic applications of signal processing, from physical basics to
modern algortithms. The "isvr" (University of Southampton,
Institute of Sound and Vibration Research) is one of the most
known research center of the topic. |
Active noise control at the
isvr |
Introduction to active noise control, from the basics to
interesting real applications. |
Further information: László Sujbert
|