Dr.
Denenberg has over 20 years of experience in the electronics, communications,
and computer industries. He has worked
for Motorola, Bell Laboratories, ITT, and Prodigy Services Co. prior to joining
Noise Cancellation Technologies in 1990 as Vice President, R&D and Chief
Technology Officer. Dr. Denenberg
received his BS. from Northwestern University in 1966 and both a M. S. and a
Ph. D. from the Illinois Institute of Technology in 1968 and 1970 respectively,
all in Electrical Engineering. He is a
Senior Member of the IEEE and holds 11 U. S. patents.
An
Electronic Silencer For Industrial Machines:
|
Environmental noise has been a problem since the industrial revolution. Noise affects our health and safety, interferes with communication, and reduces our ability to enjoy life at its fullest. High frequency noise and vibration have been amenable to passive controls. However, it has proven difficult to passively control low frequency (500 HZ and below) noise and vibration. The application of Active Noise Cancellation (the use of an equal, but opposite, anti-noise to eliminate noise) to the control of low frequency noise was proposed over 50 years ago. But at that time it was quite impractical due to the complexity of analog electronics required.
Since that time, Digital Signal Processing technology has had the same rapid advances as those in general purpose computer technology. As a consequence, Active Noise Cancellation is a practical solution to many previously difficult problems in environmental noise. A demonstration of this advance is the recent successful application of Active Noise Cancellation techniques at several CSX Transportation facilities where bulk material is transferred between railroad cars and trucks by the MasterVac bulk loaders.
The use of Active Noise Cancellation technology on the MasterVac resulted in a 24 dBA reduction of the noise level at the equipment operator's position. A surprising benefit of this technology is a significant improvement in energy efficiency and productivity. Fuel usage of the MasterVac was reduced by 24% and the time to transfer a rail car of materials was reduced by 20%.
Active Noise Cancellation is not a new idea. Creating a copy of the noise and using it to cancel the original dates back to the early part of this century. The first systems used a simple "delay and invert" approach and showed some promise, but the variability of real world components limited their effectiveness.
In the mid 1970's a major step forward took place with the application of adaptive filters to generate the anti-noise. This greatly enhanced the effectiveness of the systems because they could continuously adapt to changes in their external world as well as changes in their own components. A second breakthrough in the mid 1970's was the recognition that many noise sources, particularly those produced by man-made machines, exhibit periodic (tonal) noise. This tonal noise allows a more effective solution as each repetition of the noise is similar to the last and the predictability of the noise allows creation of an accurate anti-noise signal.
Practical application of this technology still had to wait since the electronic technology available at that time was not mature enough for implementation of Active Noise Cancellation systems. Now digital computer technology has evolved to the point where cost effective Digital Signal Processing (DSP) microcomputers can perform the complex calculations involved in Active Noise Cancellation. This technology advance has made it feasible to apply Active Noise Cancellation to previously insolvable problems in low frequency environmental noise at a reasonable cost.
The
MasterVac
Custom designed for CSX, a large rail transportation company on the east coast of the United States, the MasterVac is a vacuum transfer system. It is used to load and unload dry bulk materials (such as sugar, grain, and plastic pellets) between railroad cars and trucks at the CSX Bulk Intermodal Distribution Services (BIDS) terminals. The MasterVac features a 100 horsepower diesel engine that drives a powerful, positive displacement, vacuum pump attached to a long, flexible, 6 inch diameter, stainless steel pipe. The air flow propels the material into a separation bucket that is periodically emptied into the target vehicle.
The system, as manufactured, has excessive noise. At the operator's position the noise was measured at 115 dBA. Even with hearing protection, OSHA safety rules limited the operator exposure to the machine to 15 minutes a day. It takes about an hour to fill or empty a railroad car.
Industrial noise sources like the CSX MasterVac tend to simultaneously generate noise with two distinct characteristics. One characteristic noise is due to turbulence or grinding. This noise is distributed evenly across the frequency bands and is referred to as "Broadband Noise".
The other characteristic noise is different. "Narrowband Noise" concentrates most of its noise energy at specific frequencies. When the source of this noise is a rotating or repetitive machine, the noise frequencies are all multiples of a basic "Noise Cycle" and the noise is approximately periodic and consists of a set of tones. This "Tonal Noise" is common in the environment as man made machinery tends to generate it (along with a smaller amount of broadband noise) at increasingly high levels.
FIGURE 1: A
Typical Noise Spectrum
The original MasterVac noise is due to two sources:
The diesel engine produces a noise like that in Figure 1, but with many tonal components at multiples of the rotational rate ( ~40 Hz). This noise was radiated from the exhaust as well as from the engine itself.
The vacuum pump exhaust is the principal noise source in the machine. With the original passive silencer removed, the noise was 133 dB one meter from the exhaust outlet. Most of this noise energy is contained in three tonal components at 180 Hz, 360 Hz and 540 Hz. The vacuum inlet is not a problem since it is typically routed into a rail car or truck which muffles the noise.
The first line of defense against noise is good design. Machines should be well balanced. Symmetry in design, careful manufacturing, and good maintenance can significantly reduce vibration and noise. Turbulence can be reduced by good aerodynamics. High "Q" (low loss) resonances in structures and gas flows should be avoided.
There were opportunities to improve the MasterVac using passive techniques. Since a good noise control system is usually a combination of active and passive techniques, passive treatments were applied before considering active noise cancellation..
The diesel engine was originally enclosed in a sheet metal housing. This housing was rebuilt using sound absorbing materials.
The muffler on the diesel engine exhaust was ineffective and was replaced with a high performance, low back pressure unit.
With these changes the diesel engine noise was reduced to a manageable level, leaving the primary noise source: the vacuum outlet.
A standard passive silencer was used on the vacuum exhaust in the original design in an attempt to control the vacuum outlet noise. It was a baffled unit like those commonly found on engine exhausts (but significantly larger). Baffler silencers are not effective enough at low frequencies and create additional turbulent noise which makes active noise cancellation more difficult, so we looked at other possibilities.
The best alternative was a straight through or "glass pack" muffler. It is similar to those used in the "Hot Rods" of the 50's. This absorptive silencer consists of a length of pipe with holes wrapped by fiber glass (or other sound absorbent material) which is then enclosed by a larger diameter pipe. Absorptive silencers are very effective at high frequencies, reduce turbulence in the exhaust, and produce little or no back pressure. They have little effect, however, on the low frequency tonal noise in these applications unless the muffler is made impractically large and heavy.
We therefore chose a high quality absorptive silencer to handle the high frequency and turbulent noise in the outlet air stream and left the three major tonal components for the Active Cancellation System.
The idea of creating a copy of the noise and using it as "Anti-Noise" to cancel the original dates back to the early part of this century. Figure 2 shows the relationship in time of a noise signal, an anti-noise signal and the residual noise that results when they meet.
FIGURE
2: Noise Cancellation
Note that Active Noise Cancellation does not mask the noise; it removes a significant portion of the noise energy from the environment.
The Synchronous Feedback noise cancellation technique, developed by G. B. B. Chaplin in the mid 1970's and recently improved and productized by Noise Cancellation Technologies, is very effective on repetitive noise and does not rely on causality. In this approach a tachometer signal is used to provide information on the rate of the noise. Since all of the repetitive noise energy is at harmonics (or multiples) of the machine's basic rotational rate, the DSP microcomputer can dedicate its resources to canceling these known noise frequencies.
THE ACTIVE NOISE CANCELLATION
CONTROLLER The cancellation algorithm is executed in a modern DSP computer that fits on one 250 cm2 printed circuit board. Included on this board are: A digital signal processing chip (such as the AD2101) capable of 10 million operations per second. Two Low Pass filters set at 500 Hz to
avoid aliasing, the confusion of high frequency signals with An A/D converter to measure the noise remaining after cancellation. A D/A converter to output the Anti-Noise. |
Figure 3 shows an active muffler system as applied to the vacuum exhaust. The passive element is a simple, but large, straight through glass pack silencer (absorptive) that controls high frequency noise (above 500 Hz). The active silencer (see photo above) is a speaker cabinet (three 12 inch high power speakers each driven by a 100 watt amplifier to produce the required anti-noise level) that is concentric to the exhaust pipe and outputs the anti-noise in a ring around the end of the exhaust. The symmetry of the noise and anti-noise sources in this arrangement provides for global cancellation of the low frequency noise (500 Hz and below). A microphone in the exhaust noise sound field feeds back the RESIDUAL noise (after cancellation) so that the ADAPTER (usually a LMS adaptation algorithm -- see below) can continuously adjust the cancellation to drive the RESIDUAL noise toward zero at the noise frequencies. The tachometer signal drives a harmonic generator to internally provide pure tones at the harmonics of the vacuum's basic cycle (two full revolutions of the pump). This sets up the whole system to concentrate its efforts on the noise from the engine.
FIGURE
3: Synchronous Cancellation
The
MasterVac Active Silencer |
Active Noise Cancellation Systems employ a variant of the LMS algorithm known as Filtered-X. The basic LMS algorithm correlates an error signal (the Residual Noise in this case) with a reference signal (called "X" by Widrow and Stearns). The result is then multiplied by an adaptation rate constant and then used to adjust the relevant parameter of the adaptive filter. This is done repeatedly for each filter parameter with the objective being convergence to an operation that minimizes the average power in the error signal.
In real world systems, however, the LMS algorithm does not converge due to the delay and gain effects of the physical path taken by the anti-noise signal. Using a compensating filter on the reference signal (hence the name "Filtered-X") restores stability and produces a well-behaved system. Figure 4 shows a single tone cancellation system
A Single Tone Cancellation System
The Equalizer is a digital filter that adjusts the gain and phase of the reference signal (a rotating unit vector in the complex plane). Its design is determined by measuring the actual gain and phase shift of the anti-noise path at the operating frequencies, inverting the resulting complex impedance, and using these results to set the filter parameters.
The procedure for updating the weights at time k+1 are:
Wc,k+1 = Wk + 2mekXc,k
Ws,k+1 = Wk + 2mekXs,k
Where:
m < 1 is the adaptation constant
ek is the residual or error signal at time k
Xc,k is the in-phase component of the equalized reference signal
Xs,k is the quadrature component of the equalized reference signal
The MasterVac Active Silencer reduces the three major noise tones by over 30 dB. The overall results as measured at the operator's position (near the diesel engine) are shown in Table 1.
TABLE
1: CSX MasterVac Active
Silencer Performance
CONDITION |
NOISE |
BACK PRESSURE |
NOTES |
Original System |
115 dBA |
3 inches HG |
Passive silencer on vacuum exhaust |
Active Muffler System |
91 dBA |
None |
24% fuel savings |
This solution to the CSX noise problem was first demonstrated in the Fall of 1990 at CSX Transportation's Bulk Intermodal Distribution Services terminal in Baltimore Maryland. The use of minimal hearing protection now allows a worker to accomplish a full shift without the threat of hearing damage. An operator's enhanced ability to hear other sounds also makes the work site safer. Roger Posey, Operations Manager for CSXT/BIDS says, "The noise from the original MasterVac can be likened to a jet plane during takeoff ... now, with the NCT Active Industrial Silencer, the MasterVac sounds no louder than an average living room air conditioner. Workers are happier and we have eliminated the possibility of noise complaints from the surrounding community." Active Industrial Silencers are now installed on 20 MasterVacs and are planned for all 80 units owned by CSX Transportation.
Active noise cancellation is now an effective approach to solving many environmental noise problems created by man-made machinery. Where standard passive approaches are ineffective Active Noise Cancellation can significantly reduce noise and vibration while, in many cases, improving the energy efficiency of the machine.
The rapid advances in computer technology during the last decade have now made it feasible to do the sophisticated calculations required to implement effective noise cancellation using a low cost digital signal processing computer. Many applications are now practical. In addition to the Active Industrial Silencer described here there are:
active mufflers for transportation
active isolation mounts for industry and transportation
active headsets to protect hearing and enhance safety
quiet zones to provide comfort and productivity
Active noise cancellation products for all of these applications are now moving from the laboratory into commercial products.
J. N. Denenberg, " Anti-Noise, Quieting the Environment with Active Noise Cancellation Technology", IEEE Potentials, Volume 11, No. 2, pp. 36-40 (April, 1992)
Kh. Eghtesadi, M. McLoughlin, and E. W. Ziegler, "Development of the Simulation Model of the Multiple Interacting Sensors and Actuators (MISACT) for an Active Control System", Recent Advances in Active Control of Sound and Vibration, Edited by C. A. Rogers and C. R. Fuller, Virginia Polytechnic Institute, pp. 246-257, (15-17 April, 1991).
Colin Ross, "The Control of Noise Inside Passenger Vehicles", Recent Advances in Active Control of Sound and Vibration, Edited by C. A. Rogers and C. R. Fuller, Virginia Polytechnic Institute, pp. 671-681, (15-17 April, 1991).
Proceedings of NOISE-CON 90, "Reducing the Annoyance of Noise", University of Texas at Austin (October, 1990).
Kh. Eghtesadi and J. W. Gardner, "Experimental Results of the Non-Restrictive Electronic Muffler on Internal Combustion Engines", Acoustics Letters, Volume 4, Number 4, pp. 70-73 (1990).
P. A. Nelson, "Causal Constraints in the Active Control of Sound", IEEE Conference on Acoustic Speech and Signal Processing, Volume 1 (April, 1987).
B. Widrow and S. D. Stearns, Adaptive Signal Processing, Prentiss-Hall (1985).
G. B. B. Chaplin, "Anti-Noise - the Essex Breakthrough", Chartered Mechanical Engineering, Volume 30, pp. 41-47 (1983).