GB2130651A - Improvements relating to active acoustic attenuators - Google Patents

Abstract

A speaker 17 is connected by a waveguide 19 to an acoustic mixer 15 leading from a duct 13 carrying source sounds to be cancelled by an active acoustic attenuator 11. The coupling component which couples the physical system with the electronic system 11, consists of a microphone array 33 disposed in the duct 13, and a microphone 35 disposed in acoustic mixer 15 downstream from the waveguide 19. Broad band noise is sensed by microphone array 33 which produces a signal which causes an adaptive filter 23 to drive the speaker 17 so as to introduce "cancelling" noise. The microphone 35 senses the degree of cancellation of the source sound by the speaker 17. The signal from microphone 35 produces an error signal, which is also sent to the adaptive filter to adjust its output so as to produce a more effective "cancelling" sound from the speaker 17. <IMAGE>

Description

SPECIFICATION Improvements relating to active acoustic attenuators The invention relates generally to acoustic attenuation, and more particularly concerns apparatus for attenuating a band of lower frequency sound and single frequencies propagating from a given source by the introduction of cancelling sound. In accordance with certain aspects of the present invention, some exemplary forms of an active acoustic attenuator in enclosed rooms or other free space environments shall be described.

In confined enclosures such as ducts, significant reductions in the sound pressure level of sound carried therein has been an unresolved problem for many years. In factories, for example, the noise produced by machinery and various manufacturing operations may be carried from the heating and ventilating ducts in such areas to the ducts which connect to offices and other parts of the plant in which a low level of noise is desired. This is particularly a problem with low frequency noise in the range of infrasound to 800 Hertz, since passive means to attenuate such frequencies are costly, relatively inefficient and physically large in size making them impractical for use in most low frequency applications.

Beginning in 1925 and continuing at an extremely rapid pace today, developments in electronics have made the concept of "active" attenuation of noise to be not only a possible but attractive alternative to passive attenuation of low frequencies. The principle of so-called active attenuation is based on the fact that the speed of sound in air is much less than the speed of electrical signals. In the time it takes for a sound wave to propagate from a location where it is detected to a remote location where it may be attenuated, there is sufficient time to sample the propagating wave, process such information within an electronic circuit and produce a signal to drive a speaker which introduces cancelling sound 1 800 out of phase and equal in amplitude to the propagating sound.Although the process of active attenuation of sound stated above appears to be conceptually simple, a review of the prior art in this area will illustrate the complexity of the problem and the difficulty of obtaining good attenuation over a relatively broad band of lower frequencies.

One of the first efforts in the area of active attenuators is disclosed in U.S. Patent No. 2,043,416 to Lueg. The Lueg system is a monopole consisting of a microphone, amplifier and speaker. The microphone detects the source sound and converts it into an electrical signal which is fed to the amplifier. The loudspeaker, driven by the amplifier, is disposed downstream from the microphone at a location to give the necessary time delay to accomplish a 1 800 phase reversal from the source sound.

The loudspeaker injects a mirror image of the source sound into the duct so that as the source sound passes the location of the loudspeaker, a volume of either high or low pressure air is introduced 1800 out of phase with the corresponding high and low pressure volume of air of the source sound. When the loudspeaker is perfectly synchronized with the passage of the source noise, the pressure of the source noise and that of the loudspeaker averages to O (ambient static pressure) and the noise is "cancelled".

It is apparent from an examination of the Lueg system that attenuation will occur if the distance between the microphone (where the source sound is sampled) and the loudspeaker (where the cancelling sound is introduced) is such that the time delay of the electrical signal sent to the amplifier equals 1800 or an odd multiple of 1800. However, this condition will occur only for a specific stationary acoustic signal which does not vary in time. As a practical matter therefore, the Lueg system is effective only for a single frequency since no means are provided to accommodate phase change.

What this limitation of the Lueg system shows is that there are two parameters which must be met for good attenuation, including delay time, to allow the acoustic wave to move from the point of detection to the point of attenuation, and phase, to assure that attenuation occurs at the point of introduction of the cancelling acoustic wave.

In addition to the limitation of the Lueg system associated with phase detection and accommodation, a problem exists with the generation of standing waves by the loudspeaker in the upstream direction toward the microphone. Because of the standing wave pattern, the pressure of the sound field at the microphone is artificially non-uniform which means that at a given frequency the microphone may be located at a node or antinode of a standing wave. Therefore, the cancellation signal produced by the speaker may be made to be too great or too little. As a result, the sound field may be amplified by; the standing waves in such a way that the resulting propagation downstream from the speaker could be even greater than the sound produced by the source.In addition, the standing wave field could intensify the sound pressure in the duct and more sound could pass through the walls of the duct creating a secondary problem.

In an effort to avoid the above-identified standing wave problem and expand the frequency range of attenuation, several active attenuation systems have been developed subsequent to the Lueg system. One prior art system utilizes the combination of a monopole/dipole arrangement with the dipole being located on one wall of the duct and the monopole being located on the opposite side of the duct as shown. This system was first introduced by M. Jessel and G. Mangiante and is described in a paper entitled, "Active Sound Absorbers in an Air Duct", JSV(1 972) 23 (3,383-390). The dipole and monopole of the Jessel system are phased so that they add in the downstream direction and subtract in the upstream direction allowing a unidirectional propagation to be formed with the sources are balanced.It has been found, however, that the results obtainable with the Jessel system are frequency dependent and related to the half wave length of the dipole separation. In addition, the complexity of this system does not lend itself for use in many practical applications.

As a means of simplification of the Jessel system and to obtain improved performance, further systems were developed. One system is discussed in U.S. Patent No. 4,044,203 to Swinbanks.

Swinbanks removed the monopole found in Jessel and altered the phase characteristics of the dipole so that the propagation of both sources is added in the downstream direction and cancelled in a direction upstream toward the microphone. Another active attenuator, disclosed in U.S. Patent No.

4,109,108 to Coxon et al, locates the microphone between two loudspeakers to produce a minimum level at the microphone position when the proper phasing between the speakers is introduced. While this system is reflective and produces a standing wave pattern upstream of the dipole, the detection system (microphone) is not affected because it is located between the speakers, unlike the Swinbanks system.

In reviewing the performance of the dipole and dipole/monopole systems, it has been found that each of these multisource systems seem to have geometry related limitations. The physical spacing of the loudspeakers and microphones produces a "tuning effect" which sets the frequency of best performance and the bandwidth. Although high levels of attenuation are possible, particularly with the Swinbanks systems, such a performance is obtainable only through a relatively small bandwidth on the order of about 2 1/2 octaves maximum. Accordingly, the most recent approaches to active noise attenuation in ducts have concentrated on improvement of the monopole system first introduced by Leug as discussed above. U.S.Patent No.4,122,303 to Chaplin et al is one example of a monopole system in which relatively sophisticated electronic circuitry is incorporated into the basic microphone/loudspeaker construction found in the Lueg system. In Chaplin et al, a secondary cancelling wave is produced by a first electrical signal which represents the primary sound wave sensed by a microphone. The first electrical signal is convolved with the second signal derived from the system impulse response as a program of operational steps. This process represents the standard operation of an adaptive filter as described in "Feedback and Control Systems', Schaum's Outline Series, 1967, pp. 179-185; "Adaptive Filters", Widrow, B., Aspects of Network and System Theory, (1971); "Principles and Applications of Adaptive Filters: A Tutorial Review", McCool, J.M.; Widrow, B., Institute of Electrical Engineers, (1976). To avoid the problem of standing waves produced by the loudspeaker and propagated upstream to the microphone, Chaplin et al suggests the use of a unidirectional microphone of lining the duct between the speaker and microphone which acoustically absorbant material. In addition, Chaplin et al suggests the possibility of incorporating circuitry to provide a second convolution capable of compensating for the feedback signal produced by the upstream standing waves. However, as discussed in more detail below, the so-called adaptation process of the Chaplin et al system occurs over a period of from 5 to 30 minutes and is thus dependent on the source signal remaining essentially constant over that period.Moreover, the control system functions so slowly that in most practical cases, it must be manually shut off while the first convolution process is still operative to avoid "system hunting" or oscillation between better and poorer results.

It should also be noted that each of the systems identified above share a common problem which in many practical applications will reduce, if not eliminate, their potential use. In each case, the speaker portion of the system is placed directly in the environment of the duct to introduce the cancelling sound. It is anticipated that in many applications, existing speakers will not be able to survive the temperatures, particulates or other foreign materials found in the environment of the duct or in unprotected areas outside of the duct. In a duct environment, an active acoustic attenuator may be employed which is a modified monopole-type active acoustic attenuator which may be thought of as consisting of three separate components including a physical system, electronic circuitry and a coupling component between the physical system and circuitry.The physical system consists of the duct through which sound waves propagate from a given source, an acoustic mixer connected in line with the duct, a speaker disposed in a protective enclosure at a spaced distance from the duct and a waveguide connecting the speaker to the acoustic mixer. The electronic circuitry component of the active attenuator herein consists of three distinct adaptive filters utilizing a modification of the Widrow Hoff LMS algorithm in a true adaptive acoustic cancelling configuration. A microphone array, disposed in the duct at a location upstream from the acoustic mixer, and a microphone positioned within the acoustic mixer downstream from the waveguide form the coupling means between the physical system and electronic circuitry.

As discussed hereinafter, the microphone array senses the source sound in the duct and converts it to an electrical signal which is sent to a modified adaptive filter in the electronic circuitry component herein. The adaptive filter generates a signal to drive the speaker for the production of "cancelling" sound 1 800 out of phase with the source sound, which propagates through the waveguide to the acoustic mixer where it is combined with the source sound. The microphone located in a position downstream from the waveguide within the acoustic mixer, detects the sound resulting from the combination of the source and cancelling sound and produces a signal which represents the "error" or difference between the attenuation achieved in the acoustic mixer and the desired attenuation based on preset levels.This error signal is introduced into the adaptive filter which then adjusts its signal driving the speaker so that the "cancelling" sound propagating into the acoustic mixer more nearly approximates the mirror image of the source sound. High levels of attenuation are thus achieved within the acoustic mixer and everywhere in the far field.

In one preferred form of the active acoustic attenuator, an adaptive compensation filter is provided to assure stable operation of a modified LMS algorithm in producing the cancelling sound.

This adaptive filter provides phase correction and is utilized in lieu of a phase correction filter which would need to be manually calibrated. In a free space environment, it is advantageous to place the cancelling sound generator as near as possible to the source of the sound to be cancelled. In order to aid in accomplishing this desired placement of the cancelling sound source, the adaptive compensation filter should be properly located in the active acoustic attenuator. In accordance with one aspect of the present invention, the adaptive compensation filter is moved out of the cancelling loop of the attenuator, thereby reducing the delays in the cancelling loop. This permits closer placement of the cancelling sound source to the source of sound to be cancelled.

In accordance with a further aspect of the invention, in the employment of the above-described active acoustic attenuator in a free space environment, a plurality of cancelling sound generators are located spaced apart from the source of the sound to be cancelled in different directions of propagation of that sound.

Other objects and advantages of the invention, and the manner of their implementation, will become apparent upon reading the following detailed description and upon reference to the drawings, in which: Figure 1 is a view of a simplified version of the active acoustic attenuator; Figure 2 is a partial block diagram of the digital realization of an adaptive filter with the standard Widrow-Hoff LMS algorithm; Figure 3 is a partial block diagram of the digital realization of an adaptive filter herein with a modified Widrow-Hoff LMS algorithm; Figure 4 is a block diagram of the electrical circuitry component of the adaptive filter; Figure 5 is a block diagram of the adaptive uncoupling filter portion of the adaptive filter; Figure 6 is a block diagram of the adaptive compensation filter portion of the adaptive filter;; Figure 7 is a diagrammatic representation of the relative orientations of a source of sound to be cancelled and cancelling sound sources; Figure 8 is a diagrammatic representation of the relative orientations of sound sources as illustrated in Figure 7 with the addition of sensing means and an active acoustic attenuator; Figure 9 is a block diagram of a modified form of the electrical circuitry of Figure 4 and Figures 10--14 are far field simulated radiation curves for certain particular positions of active acoustic attenuator components relative to a sound source.

While the invention is susceptible to various modifications and alternative forms, certain iliustrative embodiments have been shown by way of example in the drawings and will herein be described in detail. It should be understood, however, that it is not intended to limit the invention to the particular form disclosed, but, on the contrary, the intention is to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the invention, as defined by the appended claims.

Referring now to the drawings, and in particular to Figure 1, illustrating the attenuator in a duct environment, the active acoustic attenuator is labeled generally with the reference 11. The active attenuator 11 may be thought of as consisting of three distinct components, and the discussion in this section will be directed primarily to the physical system component with general references to the other components where necessary. The physical system includes a duct 1 3 through which sound from a given source propagates, an acoustic mixer 1 5 connected to duct 13, a speaker 1 7 which is the source of cancelling sound and a waveguide 1 9 which connects the speaker 1 7 with the acoustic mixer 15.

Although the acoustic mixer 1 5 is shown as having a slightly larger diameter than duct 13, this geometric relationship is not required for proper performance of active attenuator 1 3 and is shown herein for purposes of discussion and illustration only.

One of the immediately apparent advantages of the active attenuator 11 over prior art systems is that the speaker 1 7 is disposed at a remote location from the duct 13, and is connected to acoustic mixer 1 5 by an elongated waveguide 1 9 which may be provided with valve means (not shown) to prevent dust particles, caustic material or other debris flowing through duct 1 3 from damaging speaker 1 7. Prior art active attenuators dispose the speaker directly in the duct where the internal and external conditions which could be encountered in many applications would quickly damage or ruin them. The speaker 1 7 is not only protected from the internal environment of the duct 1 3 and acoustic mixer 1 5 by waveguide 19, but it is contained in an enclosure 21 which protects speaker 17 from the external environment of the duct 13. Enclosure 21 must have a high transmission loss and may also be lined with acoustically absorbant material to further prevent the output of speaker 1 7 from propagating in a direction other than through waveguide 1 9. It should be understood that waveguide 1 9 is completely separate from duct 13 and acoustic mixer 15; that is, waveguide 19 is not a branch duct, and thus no flow from the main duct 13 is carried through the waveguide 19.Waveguide 1 9 is provided solely for the purposes of carrying the cancelling noise from speaker 1 7 to acoustic mixer 1 5 and for isolating and protecting speaker 17 from the environment of duct 13.

The electronic component of the active attenuator 1 is shown in its simplest form in Figure 1 for purposes of the present discussion. A detailed description of the electronics of the subject invention will be provided below including an explanation of the complete circuitry utilized herein. The simplified version of the electronic component of active attenuator 11 includes an adaptive filter 23, an amplifier 25, a phase correction filter 29 and a DC loop labeled generally with the reference 31.The coupling component of the subject invention, which couples the physical system with the electronic system, consists of a microphone array 33 disposed in duct 1 3 in advance or upstream of waveguide 1 9 for sensing the source sound, and a microphone 35 disposed in acoustic mixer 1 5 downstream from waveguide 1 9 for purposes to become apparent below.

Generally, in the cancelling mode of operation, active attenuator 11 operates as follows. Broad band noise propagates down duct 13 and is sensed by microphone array 33 which produces a signal sent to adaptive filter 23. Adaptive filter 23 provides an output to drive speaker 1 7 which introduces "cancelling" noise through waveguide 1 9 into acoustic mixer 1 5. Since a sound wave consists of a sequence of compressions and rarefactions at a given phase and frequency, the pressure of such waves can be reduced or "cancelled" by generating a secondary wave having compressions and rarefactions equal in amplitude and 1800 out of phase with the primary sound waves.The microphone 35 located downstream from waveguide 1 9 in acoustic mixer 1 5 senses the degree of attenuation or cancellation of the source sound after the sound waves produced by speaker 1 7 have been combined with it. The signal from microphone 35 is sent to adaptive filter 23 as an error signal, which, in effect, is an indication of the attenuation achieved within acoustic mixer 1 5. The adaptive filter operates to adjust its output depending on the character produce "cancelling" sound which is more nearly equal in amplitude and 1 800 out of phase with the source sound.

The second primary component of the active acoustic attenuator 11 of the subject invention is the electronic implementation or control of a physical system such as that described above. A brief description of the prior art may be helpful in appreciating the advances made in the electronic control circuitry herein. The linear summation of two equal and opposite signals has iong been recognized as one approach to producing an electronic or acoustic null. Several of the acoustic cancellors developed to date are based on the "equal and opposite" principle wherein a cancelling wave generated by a loudspeaker is introduced into a confined space such as a duct to reduce the pressure variations produced by sound waves propagating through the duct from a given source.In the simplest model, the signal generated by the source is considered to be a pure tone represented by a single rotating vector The cancelling signal must track the source within some maximum permissible error to obtain the desired attenuation. A significant problem with this approach, which is recognized and solved by the present invention, is that the permissible amplitude and phase errors must be held within tolerances which are available only in the very best acoustic devices. For example, to obtain 20 decibels of cancellation, the errors in the microphones, speaker and electronics of such systems must be less than one decibel in amplitude and 60 in phase. As discussed in more detail below, the subject invention overcomes this problem using a tolerance relaxation technique based on feedback principles.

Rather than requiring accuracy in all components, a feedback system concentrates the performance requirements in a few easily controlled devices.

One of the most recent so-called active attenuation systems is found in U.S. Patent No.

4,122,303 to Chaplin et al, which ostensibly uses an "adaptive" process to generate a cancelling wave capable of creating a null when combined with the sound waves from a source. An examination of this system, however, reveals that the electronic circuitry does not rely on a true adaptive process, but involves a trial and error approach in which a series of successive guesses or estimations are made of the error signal, which is the difference between the desired and actual attenuation. Eventually, the guesses of the error signal become closer and closer approximations of the desired error according to preset values. Not only is the trial and error method unduly lengthy (in the range of between 5 and 30 minutes), but after the process is completed, the system must be manually shut down to avoid "system hunting" or oscillation between better and poorer results.In addition, once the trial and error process has begun in the Chaplin et al system, any change in the source sound in the 5 to 30 minutes required to complete the process will also result in "system hunting".

One key aspect of the electronic circuitry herein is that it is a deterministic system utilizing true adaptive filters in contrast to the trial and error approach taught in Chaplin et al. This means that the electronic circuitry of the subject invention automatically adjusts its own parameters and seeks to optimize its performance according to specific criteria. Additionally, by using the feedback principles mentioned above, the electronic circuitry herein is not nearly as dependent on the tolerance requirements of acoustic devices used in many of the known prior art systems. In fact, where the permissible amplitude and phase errors of acoustic devices used in prior art systems are on the order of 1 decibel in amplitude and 60 in phase, the electronic circuitry herein can tolerate amplitude and phase errors in the range of at least 10 decibels and 450 phase. By relaxing the tolerance requirements, the installation and maintenance of this system may be performed by technicians of ordinary skill making the active acoustic attenuator 11 commercially viable in a variety of applications.

The electronic circuitry for the active attenuator 11 utilizes a modified form of the Widrow-Hoff LMS adaptive algorithm in a true adaptive acoustic cancelling configuration. The LMS algorithm was designed for use in signal enhancement systems where the signal to noise improvement, or the noise reduction, was achieved solely in the electronic circuitry. This algorithm was significantly modified in the subject circuitry to permit operation when the vibration or noise reduction and/or cancellation is to be achieved in a physical system having inherent delays such as in an acoustic field. The modifications of the Widrow-Hoff LMS algorithm retain the signal processing advantages inherent in the original algorithm and permit these advantages to be applied to active acoustic cancelling problems.Other adaptive algorithms exist for the solution of the quadratic error function and many of these could be modified for satisfactory operation in an acoustic canceller. The purpose of the adaptive algorithms, as discussed in detail below, is to find an optimum or near optimum solution to the cancelling filter problem. Such other algorithms, modified in accordance with the teachings herein, can accomplish this function and should be considered within the scope of the subject invention.

Before discussing the modification of the LMS algorithm in the adaptive filters of the subject invention, a review of an adaptive filter having a standard LMS algorithm governing its operation will be discussed. Referring to Figure 2, an adaptive filter using the standard Widrow-Hoff LMS algorithm is shown. The basic element of an adaptive filter is known as a transversal filter 53 as indicated in dotted lines in Figure 2. The transversal filter 53 can be visualized as a series of delay elements with the filter output being derived from a weighted summation of the delayed outputs. In Figure 2, a set of n measurements S(t) is sampled to form n sample measurements S(j) where j is the sampled time index.

Each of the points labeled 37 in Figure 2 can be considered as constituting sampled input values, with the Z-' factor representing a delay. Each sample value 37 is multiplied by a corresponding weighting coefficient W(j) in multiplier 39, and the weighted measurements are entered into a summer 41 to form an output yj which is the output of the transversal filter 53. This output yj is compared with a desired response din a summer 43 to form an error signal ei.

The objective of the LMS algorithm which governs the operation of the transversal filter 53, is to deterministically obtain the weighting coefficient in such a way as to minimize the error signal ej and find the weighted sum of the input signals that best matches the desired response. Changes in the weight vector to accomplish this end are made along the direction of the estimated gradient vector based on the method of steepest descent on the quadratic error surface. A detailed treatment of this subject can be found in the article of Prof. Bernard Widrow entitled "Adaptive Filters" from Aspects of Network and System Theory edited by Rudolph E. Kalman and Nicholas DeClaris.A block diagram representation of a digital realization of the LMS algorithm is found on the right-hand portion of Figure 2, which is contained in the dotted lines shown and labeled generally as 55. Although shown in digital form, it should be noted that an analog realization of the LMS algorithm and transversal filter may also be used.

For purposes of illustration, only two sample input values and their corresponding weighted functions are shown in the drawings. The input S(j) is sent to a multiplier 45. The error signal ej, together with a scaling factor u, which controls the rate of convergence and stability of the algorithm, is entered into a multiplier 47. The scaled error signal is then multiplied in multiplier 45 with signal S(j), and that product is introduced to a summer 49 and unit delay 51 or C RAM (Coefficient Random Access Memory). The weight setting W1 (j+l) is sent back to the adjustable weight 39 corresponding to input signal S(j+l) whose product then forms part of the output of the transversal filter 53. The same operatipn is conducted for input signal S(j-I) corresponding to weight W2(j).As mentioned above, the LMS algorithm has proved to be effective in many adaptive signal processing applications wherein the weight setting determined from the error signal can be fed directly back to the adjustable weight corresponding to the input signal for which it was determined with essentially no delay.

In the form shown in Figure 2, the LMS algorithm and transversal filter would not be suitable for use in an acoustic cancellation problem. The cancellation of an acoustic wave propagating down a duct requires an equal and opposite signal to interact with it to reduce the maximum pressure variations generated in the acoustic mixer 1 5. The interaction of the two waves requires a finite length of travel and a corresponding amount of time. Referring now to Figures 3 and 4, it can be observed that in the physical system of active attenuator 11, a finite length of time is required for an acoustic wave to propagate from the microphone array 33 where the input signal is detected.In addition, a delay exists in the time required for the cancelling sound produced by speaker 1 7 to propagate through waveguide 1 9 to acoustic mixer 1 5. These delays may be approximated as the distance in feet through duct 13, acoustic mixer 1 5 and waveguide 19 divided by 1100 feet per second, which is the speed of sound.

The expected delay for most systems will be from a few milliseconds to a few tenths of a second. Using a sampling rate of more than 2000 per second, the delay expressed in sampling intervals will be from several to several hundred.

Expressed in terms of the sampling intervals, the delay may be given as follows: Delay=K (I/sampling rate) Where: K=an integer constant The Widrow-Hoff LMS algorithm was modified to account for the inherent delay in the physical system of active attenuator 11 such that the weighting coefficients determined in the algorithm would be matched with the corresponding signal inputs in the transversal filter for which the weighting coefficients were determined.

For a new input signal sample S(j), the corresponding weighting coefficient may be expressed as follows: W1(j)=W1(j-I)+uS[j-(K+l)]e(j-l) (1) Where: u=scaling factor The next value of the weights can be written in the following form: W(j+l)=WI(j)+uS(jK)e(j) (2) Where: i=tap identification s(j-k)=input sample K intervals past As each new sample value of the input signal is sent to the transversal filter 53, it operates to generate the product of the first weighting coefficient and the last input sample which is added to the product of the second weight and the next to last input sample, and so on untii the last weight times the oldest sample is accumulated. The total accumulation of these terms is the transversal filter 53 output y(j).The weighting coefficients are updated by the corresponding input sample and error signal to account for the inherent delays in the physical system. Figure 3 shows a digital realization of the modified LMS algorithm according to the subject invention (with K=2) in which the delay discussed above is accommodated so that corresponding weighting coefficients, input samples and error signal are combined in the adaptive filter to produce the output signal y(j).

Referring now to Figures 1 and 4, the electronic circuitry of the active attenuator 11 including adaptive filters utilizing a modified Widrow-Hoff LMS algorithm will be discussed. In its simplest form as shown in Figure 1 , the basic operation of the electronic implementation of active attenuator 11 may be described by the following sequence: 1. Sample the source sound wave propagating down duct 13 (signal) 2. Delay, filter and scale the signal 3. Drive speaker 1 7 with the output of step 2, above, to inject the proper replica of the signal 4. Sense the acoustic output of acoustic mixer 1 5 (error) 5. Adjust (2) using the modified LMS algorithm 6. Return to (1) The adaptive filter 23 shown in Figure 1 is a 3-port arrangement, consisting of two inputs and an output.It receives an input from the microphone array 33 which is an electrical signal corresponding to the source sound in duct 13, and produces an output signal to drive speaker 17 for the production of a mirror image replica of the source sound which is introduced in acoustic mixer 1 5 through waveguide 19. The output of acoustic mixer 1 5 sensed by microphone 35 is introduced to adaptive filter 23 as an error signal, or the summation of the source sound and the replica achieved in acoustic mixer 1 5.

The acoustic propagation delay inherent in the physical system of active attenuator 11 must appear as a delay over the full frequency range of interest. The phase tolerance of this delay is approximately +45 degrees at each frequency. In the block diagram schematic of the electronic circuitry shown in Figure 1, a second-order phase-correction filter 29 is included to compensate for the acoustic resonances in the acoustic mixer 15. The characteristics of filter 29 must be determined manually, using appropriate instrumentation, based on the duct characteristics of a particular application. As discussed below, this function may be accomplished with an adaptive filter thus eliminating manual calibration of filter 29.

A DC loop is also provided in the block diagram of active attenuator 11 shown in Figure 1, which is labeled generally with the reference 31. Microphone 35 is not capable of detecting very low frequency components of the output from acoustic mixer 1 5, which, in addition to a DC component, are needed for stable operation of the LMS algorithm of adaptive filter 23. The DC loop 31 is included to provide the DC component, and the LMS algorithm has been modified to accommodate such input.

Referring now to Figures 4-6, an advanced form of the electronic component of active attenuator 11 is shown. Three discrete adaptive filters are included in the electronic component of active attenuator 11, each performing a separate function. Before discussing the adaptive filter processes, the function of the remaining circuit elements should be mentioned. A digital implementation of the adaptive algorithm is used in this embodiment of the electronic circuitry of active attenuator 11. The analog to digital converters (A/D), labeled generally as 63 in Figures 4-6, provide the sample values of the seiected inputs in digital format. The A/D converters 63 are 12 bit successive approximation converters.

Most of the signal processing within the electronic circuitry herein is digital, and the sampling rate of the A/D converters 63 implies an upper limit on the allowable bandwidth of the signal and error inputs. The sampling rate limit requires that the maximum input frequency must be less than half the sampling rate. The low pass filters 65 of the system inputs are required to limit the maximum input frequency to less than one-half of the A/D sampling rate. The output signal from the digital to analog (D/A) converter 67, will also have a frequency aliasing problem. The output signal from the adaptive filters could excite a resonance in the system at any alias, or multiple, of the D/A converter 67 output.

Therefore, the low pass filter 65 on the output of D/A converter 67 is required to limit the maximum output frequency to the band of interest.

The multipliers 39, 45, 47 and 69, and accumulators or adders 41, 43, 49, and 71 shown in Figures 2-6 perform the computations necessary for the adaptive processing herein. It has been found that the TRW Model TDC1 003J 12 bit Parallel Multiplier-Accumulator, or a suitable equivalent, provides the high speed calculation capabilities required.

The first of the three adaptive filters utilized in the electronic circuitry component of active attenuator 11 may be described as the adaptive cancelling filter 23 as shown in Figures 1 and 4. The adaptive cancelling filter 23 includes a.modified LMS algorithm to govern the operation of its transversal filter as discussed above. To expand on the prior discussion, the basic filter functions of the adaptive cancelling filter 23, including phase response, amplitude and delay, are implemented by the transversal filter which is a non-recursive, finite impulse response filter. The filter implementation is digital in nature with both the input signal history and the tap weights being stored in a Random Access Memory (RAM).As mentioned above, the operation of the transversal filter can be described as generating the product of the first weight obtained from the LMS algorithm and the last input sample, plus the second weight from the algorithm times the next to last input sample, and so on until the last weight times the oldest sample is accumulated. In equation form, this may be given as follows:

Where: Sampled input signal corresponding to the ith tap Weight of the ith tap i=tap identification l=number of taps The adaptive cancelling filter 23 may be considered to be a transversal filter when the weight adjustments are stopped.The operation of adaptive cancelling filter 23 will determine the required values for the weights of the desired filter (i.e., optimum filter conditions) by adaptive means and then realize the desired filter function by using the determined weights in the transversal filter. The purpose of the modified LMS algorithm is to govern the transversal filter operation by matching the filter function of adaptive cancelling filter 23 to the desired filter function.

The error signal sensed by microphone 35 and sent to adaptive cancelling filter 23, can be visualized as the summation of the source pressure wave sensed by microphone array 33 delayed by the acoustic travel from that point to microphone 35, plus the cancelling wave produced by; speaker 17 delayed by the travel through waveguide 19 and acoustic mixer 15 to microphone 35. In physically positioning microphone array 33, speaker 17, waveguide 19 and microphone 35 relative to one another in the duct 13 and acoustic mixer 15, the total delay or length from microphone array 33 to microphone 35 must be greater than the total delay or length from speaker 17 to microphone 35 plus the delay associated with the operation of low pass filters 64.In equation form, this relationship may be expressed as follows (See Figure 1): A min L > +d,+d2+Dfc (4) 4 Where: A min=shortest wavelength of interest L=distance between microphone array 33 and microphone 35 c=speed of sound d,=distance from waveguide 19 to microphone 35 d2=distance from acoustic mixer 1 5 (i.e., in alignment with microphone 35) to speaker 17 Delay associated with low pass filters 65 The delay Df of the low pass filters 65 can be expressed in terms of cutoff frequency or Flax. A fourth order low pass filter twill have a delay of 45 degrees for each pole at the cutoff frequency, and a good filter design will approximate a constant delay filter.Therefore, the delay for a fourth order filter can be approximated as 1/2FCutofft and Df may be considered to be approximately equal to 1/FcutofOr the two sets of fourth -order filters in series.

The second adaptive filter found in the electric circuitry component of active attenuator 11 may be termed an uncoupling adaptive filter 75, as shown in Figures 4 and 5. One potential problem associated with most active attenuators is the production of standing waves or mechanical vibrations of the duct as a result of the cancelling sound introduced by the speaker propagating toward the microphone or microphone array which senses the source sound. This coupling will tend to corrupt the signal sensing microphone's estimate of the source sound propagating downstream and reduce the effectiveness of the cancelling system. Although unidirectional microphones have been suggested as a means to avoid this problem, they are not always sufficient acting alone to overcome this inherent system limitation.

The uncoupling adaptive filter 75 herein solves this problem as follows. As shown in Figure 5, a broad band noise source 76 drives the cancelling loudspeaker 1 7 and the uncoupling adaptive filter 75.

Prior to start up of the system, the noise source 76 is automatically adjusted to drive the duct 1 3 at a sound level higher than the source level required to cancel the expected source noise. The adaptive process will reduce the output of the error summer and match the transfer function of the transversal filter in the adaptive uncoupling filter 75 to the acoustic coupling present in duct 13, at which time the noise source 76 will be terminated. At start up of the system, the adaptive uncoupling filter 75 will operate on the drive to the loudspeaker 1 7. The components of the loudspeaker drive that appear in the output from microphone array 33 will be removed by subtracting an equal and opposite image at summer 71. The subtraction process is accomplished electronically in the digital domain of the adaptive uncoupling filter 75 and in summer 71.Although this procedure will not entirely eliminate the effect of the output from speaker 1 7 on microphone array 33, sufficient reduction will be obtainable in most applications.

Referring now to Figures 4 and 6, the third adaptive filter of the electronic circuitry component in active attenuator 11 is shown, which may be called an adaptive compensation filter 79. As mentioned above in connection with the discussion of the basic electronic component shown in Figure 1, compensation means must be provided in series between the output of adaptive cancelling filter 23 and the input of the error signal thereto to assure stable operation of the modified LMS algorithm.

While this compensation was shown as a second-order phase-correction filter 29 in Figure 5, it was noted that the manual calibration of filter 29 could be accomplished by an adaptive filter. The adaptive compensation filter 79 performs this function.

As shown in Figure 6, the adaptive compensation filter 79, loudspeaker 17, waveguide 19, acoustic mixer 1 5 and microphone 35 are in series with each other, and in parallel with the broad band delay circuit consisting of a delay or memory 78. Prior to start up of the system and following the completion of the uncoupling process described above, the noise source 76 will be activated to drive duct 13 and acoustic mixer 1 5. The summation of the parallel signal paths will be the error input into the adaptive compensation filter 79. The adaptive process will match the total transfer function of the series path to the true delay of the noise source 76 by delay 78. In this way, the weights for the desired filter are generated, and the error signal received by adaptive cancelling filter 23 will be within the proper phase tolerance to assure stable operation.

Delay circuit 78 of Figure 6 will delay the signal from noise source 76 by an integer number of input sample intervals (K). The modification of the LMS algorithm in adaptive filter 23 in Figure 4, will provide for a delay of K sample intervals in the calculation of the next value for Wj. The value of K will be greater than the acoustic delay from loudspeaker 17 to the signal input of the adaptive compensation filter 79 of Figure 6. The value of K can be set large so that one value can be used in most applications. The matching of the delay from the output to the error input of the adaptive cancelling filter 23 in Figure 4, to the shift in the sampled signal used in the calculation of the next weight value, will assure the stability of the modified LMS algorithm.

With reference now to the use of the active acoustic attenuator in a free space environment, the general concept of reducing the sound power generated by a sound source, by radiating the sound waves from another source in such a phase relationship with the original source that the sound pressure at the region of interest is substantially reduced, is not new. For example, attempts to reduce the sound levels at the ears of a person working in a noisy environment were made shortly after World War II. Due to the state of the art of electronics at the time, the gains were relatively low and the equipment bulky. As outlined above, various approaches have been made to developing systems to allow the reduction of noise propagation in ducts.For a general discussion of the art in this acoustic attenuation area, including the free space situation, see "Active Attenuation of Noise-The State of the Art" by Glenn E. Warnaka, Noise Control Engineering (May-June 1982).

Prior to a description of the use of the active acoustic attenuator in a free space environment, some theoretical aspects of source interference in free space shall be considered. For a simple monopole sound source in an unbounded space, the sound intensity 1r in a radial direction may be expressed as: Ir=S2pc/2[k/47rr]2 (5) In this expression, r is the distance from the source, k is the wave number (equals co/c), p is the density of the medium, c is the speed of sound, S is the source strength (the average amplitude of surface vibration velocity multiplied by surface area of the source).

For two sources, S,, S2, each of which are equal in amplitude to S, but opposite in sign, the sound intensity may be shown to be approximately I,-S2pc/2 [k/27Gr]2sin2(kd cosO) (6) As illustrated in Figure 7, the two sound sources are separated by a distance 2d, and the angle O is measured from the line connecting the two sources. In the derived expression for sound intensity for the two monopoles, r is much greater than d (that is, the expression is for the far field). It can be seen from the two-monopole expression that cancellation of the sound is effective along the axis of symmetry between the two sources, where O is equal to 900.

It can be shown that zeroes (cancellations) are produced where cos 0=nA/2d n=0, 1, 2... (7) In this expression, A is the wavelength of each frequency of sound, and cos 0 is, of course, constrained to be less than or equal to one. Because cos 0 is less than or equal to one, values of n for which nA/2d is less than or equal to one are permissible. Therefore, for large d compared to A(A/d 1), the maximum permissible values of n can be high, which means that the radiation pattern consists of many lobes. The zero for 0=0 will be relatively narrow and the area of reduced sound pressure level will be small.So even if the radiation into the side lobes was reduced by absorption, this source arrangement is fairly inefficient and impractical. The closer the sources are, the broader is the zero (cancellation) in the vicinity of 0=:r/2. If d/R, or kd, becomes small, the two monopoles may be treated as a dipole. For a dipole, the approximate sound intensity is I,=[S2pc/2] [2dk2/47Gr]2Cos28 (8) As can be seen from the formula (8), to achieve an efficient noise reduction by an additional source radiating sound in opposite phase, the position of this source must be as close as possible to the original source. The distance between the sources 2d must be such that the condition for a dipole, k(2d)=27r(2d/A) < 1, must be satisfied. The higher the frequency, the smaller the distance must be.

Because of the physical size of the source, it is more difficult to satisfy these conditions at high frequencies. However, in this higher frequency region, any enclosures surrounding the free space environment usually offer a good insertion loss. On the other hand, the low frequency region is usually more difficult to attenuate passively, and here the active noise control systems have a good application.

The total radiated sound power may be determined as the double integral of I,. dS over a spherical surface surrounding the source or sources. Relative to the sound power emanating from a single monopole, the sound power for a double monopole is twice as great if they are far apart.

However, in the case of the dipole, the total radiated power is less than that for two monopoles. The exact amount of total radiated power depends, of course, upon the spacing between the sound sources and the frequency of the sound produced thereby.

Methods exist for computing the far field radiation characteristic using near field techniques. See, for example, "Near Field, Underwater Measurement System" by Robert D. Marciniak, J. Acoust Soc.

Am., 66(4), October 1979. The Marciniak article describes a near field acoustic measurement techniques for accurately computing the far field radiation characteristics of sound transducers. The analysis was based on evaluation of a form of the Helmholtz integral which utilizes a Green's function that vanished over the surface of integration, thereby requiring only knowledge of the near field acoustic pressure for computing the far field radiation.

From the theoretical considerations of far field calculations, and with regard to the source array, for example, of Figure 7, it can be seen that a first sound source and a second sound source, which may be a "mirror image" cancelling sound source, produce a far field sound intensity pattern which can be substantially calculated. As discussed, as the two sound sources are moved closer and closer together, more and more complete cancellation occurs. Using further techniques, such as those outlined in the Marciniak article, the far field effects of more complex near field arrays may also be calculated. The active acoustic attenuator earlier described herein is particularly suitable for use in such near field arrays in order to produce good sound cancellation and/or reduction in some or all areas of the far field.

Proper near field positioning for a plurality of cancelling sound sources can be determined either empirically or by utilizing the above-mentioned calculation techniques.

For example, as shown in Figure 8, an active acoustic attenuator 11 includes a first cancelling sound generator such as a speaker Scl and a second cancelling sound generator Sc2. Both of the cancelling sound generators ScX, Sc2, are driven by a common output of the active acoustic attenuator circuitry. The attenuator 11 also includes first sensing means such as a microphone Ma for detecting the vibrations from a source S of sound to be cancelled. A second sensor such as a microphone M2 detects the summation of the vibration from the source Sand the cancelling vibrations SC1 Sc2, in order to produce an error input signal for the circuitry of the attenuator 11.The single cancelling loop of the attenuator may, for example, accommodate the average acoustic delay from the generators to the error sensor. In the illustrated arrangement, the microphone M1 is substantially coincident with the source S of the sound to be cancelled, and the three sound sources and the microphone M2 are arranged in a straight line.

The electrical circuitry section of the active acoustic attenuator 11 shown in Figure 8 may be substantially as shown in Figure 4. However, it has now been determined that certain modifications to that circuit, as shown in Figure 9, result in improved performance of the attenuator. With reference to Figure 9, the adaptive compensation filter 79' has been removed from the cancelling loop and placed in series with the error signal path from the error microphone to the adaptive cancelling filter 23'. The adaptive compensation filter 79' is initialized as indicated in Figure 6, and then retained in that location in the circuit during the operation of the attenuator circuit. Moving the adaptive compensation filter out of the cancelling loop removes some of the delay introduced between the pick-up microphone 33 and the cancelling speaker 1 7.Therefore, the cancelling speaker may be placed closer to the pick-up microphone, and hence, to the source sound.

Referring now to Figures 1 0-14, a series of far field radiation curves represent simulations of dual cancelling sources operating on a single point source of noise. In each case, the frequencies are in hertz and all distances are in feet. The curves represent a point source of sound to be cancelled located at the center of the grid system. The strength of the source is arbitrarily +30 dB. Each 10 dB is marked by dotted circles concentric about the noise source, the center of the grid. The +30 dB circle is the outermost circle of the three and is slightly "heavier" to indicate the strength of the source. The orientation of the cancelling sources and detectors in each simulation corresponds substantially to the orientations shown in Figure 8.Two matched cancelling sources are symmetrically spaced on either side of the noise source on the horizontal axis, which is also common to the error sensor. The source sensor is assumed to be located substantially at the location of the original sound source. The system is implemented using a single adaptive acoustic cancelling loop as disclosed for the acoustic attenuator 11.

The solid curve is the far field radiation pattern of the source and two cancelling generators combined. The plot represents only the far field noise reduction in a 3600 sweep centered about the noise source. There is no distance information on the plot. Where the solid pattern exceeds the heavy dotted circle, the combined field strength is greater than the original source; or, in other words, the cancelling arrangement is not decreasing but increasing the radiated sound. The advantages of greatly decreasing the radiated sound level in a large "cone" in many cases overcomes any problems caused by the increase of sound in other directions.

A comparison of Figures 10 and 12, and also a comparison of Figures 11 and 13, shows the improved performance at the lower frequency for the same source to cancelling generator spacing. As will be recalled, as a general proposition, the closer the spacing of the sound sources, or the longer the wavelength (lower the frequency), the better the performance in the two-source arrangement. The same appears to hold true for the three linearly arranged sources of the simulated radiation curves. As can be further seen in Figure 14, as the distance between the source of sound to be cancelled and the cancelling sound source becomes significantly larger, the far field radiation curve becomes more complex. Again, as shown in Figure 14, even in the illustrated configuration, a substantial "cone" of cancelled, or substantially attenuated, sound is produced.

In the illustrated radiation plots, the error microphone is located spaced apart from the source of sound to be cancelled in the direction of desired cancellation. In the plots, this is to the right of the source along the horizontal axis. Although the radiation plots illustrated are for single frequencies, the active acoustic attenuator of the present invention provides a broad band reduction of acoustic vibrations, or noise. The far field radiation pattern produced by the attenuator would, of course, vary across the effective spectrum of attenuation, dependent upon the spectral components of the noise.

Claims (13)

Claims

1.An apparatus for the cancellation of vibrations from a source comprising: First sensing means for detecting said vibrations, said first sensing means producing first electrical signals representing the amplitude and phase characteristics of said vibrations; cancelling means for generating cancelling vibrations for combination with said vibrations from said source including a first cancelling vibration generator spaced from said source in a first direction of propagation of said vibration and a second cancelling vibration generator spaced from said source in a second direction of propagation of said vibration;; second sensing means disposed at a location spaced from said first sensing means in a direction of propagation of said vibrations, said second sensing means detecting the summation of said vibrations from said source and from said cancelling means for producing second electrical signals representing the amplitude and phase characteristics of said summation;; and an adaptive filter for driving said cancelling means to produce said cancelling vibrations, said adaptive filter including a transversal filter employing an LMS algorithm modified to accommodate the delay inherent in the propagation of said vibrations from said first sensing means to said second sensing means, and the delay inherent in the propagation of said cancelling vibrations from said first and second cancelling vibration generators to the point of combination of said cancelling vibrations with said vibrations from said source, said adaptive filter being operable to successively delay, filter and scale said first electrical signals from said first sensing means to produce corresponding outputs for driving said cancelling means, and then to deterministically adjust said delay, filtering and sealing operation based on said second electrical signals received from said second sensing means to provide corresponding revised outputs for driving said cancelling means to produce cancelling vibrations having the mirror image amplitude and phase characteristics of said vibration from said source within present limits.

2. The apparatus of claim 1 in which the second sensing means is disposed at a location spaced from said first sensing means in a direction of propagation of said vibrations in which cancellation is desired.

3. The apparatus of claim 1 in which the first cancelling vibration generator, the second cancelling vibration generator, and the second sensing means are aligned in a linear array.

4. The apparatus of claim 3 in which the first sensing means is located substantially at the vibration source.

5. The apparatus of claim 4 in which the second sensing means is disposed at a location spaced from said first sensing means in a direction of propagation of said vibration in which cancellation is desired.

6. An apparatus for the cancellation of vibrations from a source comprising: first sensing means for detecting said vibrations, said first sensing means producing first electrical signals representing the amplitude and phase characteristics of said vibrations; cancelling means for generating cancelling vibrations for combination with said vibrations from said source; second sensing means disposed at a location spaced from said first sensing means in a direction of propagation of said vibration, said second sensing means detecting the summation of said vibrations from said source and from said cancelling means for producing second electrical signals representing the amplitude and phase characteristics of said summation; and an adaptive filterfor driving said cancelling means to produce said cancelling vibrations, said adaptive filter including a transversal filter employing an LMS algorithm modified to accommodate the delay inherent in the propagation of said vibration from said first sensing means to said second sensing means, and the delay inherent in the propagation of said cancelling vibrations from said cancelling means to the point of combination of said cancelling vibrations with said vibration from said source, said adaptive filter being operable to successively delay, filter and seals said first electrical signals from said first sensing means to produce corresponding outputs for driving said cancelling means, and then to deterministically adjust said delay, filtering and scaling operation based on said second electrical signals received from said second sensing means to provide corresponding revised outputs for driving said cancelling means to produce cancelling vibrations having the mirror image amplitude and phase characteristics of said vibration from said source within preset limits, said adaptive filter further including an adaptive compensation filter having means for deterministically adjusting the phase characteristics of said second electrical signals produced by said second sensing means for introduction to said adaptive cancelling filter to assure stable operation thereof, said adaptive compensation filter being coupled in series between the second sensing means and an input of the adaptive cancelling filter.

7. A controller for an apparatus for the cancellation of vibrations from a source comprising an adaptive cancelling filter including a transversal filter employing an LMS algorithm modified to accommodate acoustic delays, the adaptive cancelling filter having an input coupled to first sensing means for detecting vibrations from a source of vibrations to be cancelled and having an output coupled to a cancelling means for generating cancelling vibrations for combination with the vibrations from said source and having an error input, and an adaptive compensation filter, coupled between second sensing means disposed at a location spaced from said first sensing means and a direction of propagation of said vibrations to be cancelled and the error input of the adaptive cancelling filter, the adaptive compensation filter being operable to deterministically adjust the phase characteristics of the output of the second sensing means to assure stable operation of the adaptive cancelling filter.

8. An apparatus for the cancellation of vibration in a physical system comprising: first sensing means for detecting said vibration, said first sensing means producing first electrical signals representing the amplitude and phase characteristics of said vibration; cancelling means for generating cancelling vibration for combination with said vibration from said physical system; second sensing means disposed at a location spaced from said first sensing means in the direction of propagation of said vibration, said physical system and said cancelling vibration and producing second electrical signals representing the amplitude and phase characteristics of said summation; and an adaptive filter, said adaptive filter including a modified LMS algorithm and a transversal filter, said adaptive filter being operable to successively delay, filter and scale said first electrical signals from said first sensing means to produce corresponding outputs for driving said cancelling means, and then to deterministically adjust said delay, filtering and scaling operation based on said second electrical signals received from said second sensing means to provide corresponding revised outputs for driving said cancelling means to produce cancelling vibration having the mirror image amplitude and phase characteristics of said vibration from said physical system within preset limits.

9. The apparatus of claim 8 wherein said modified LMS algorithm includes means for accommodating the delays inherent in the propagation of said vibration from said first sensing means to said second sensing means, and the propagation of said cancelling vibration from said cancelling means to the point of combination of said cancelling vibration with said vibration from said physical system.

1 0. An apparatus for the cancellation of vibration in a physical system comprising: first sensing means for detecting said vibration, said first sensing means producing first electrical signals representing the amplitude and phase characteristics of said vibration; cancelling means for generating cancelling vibration for combination with said vibration from said physical system; second sensing means disposed at a location spaced from said first sensing means in the direction of propagation of said vibration, said second sensing means detecting the summation of said vibration from said physical system and said cancelling vibration and producing second electrical signals representing the amplitude and phase characteristics of said summation; and an adaptive filter for driving said cancelling means to produce said cancelling vibration, said adaptive filter including a transversal filter and an LMS algorithm modified to accommodate the delay inherent in the propagation of said vibration from said first sensing means to said second sensing means, and the delay inherent in the propagation of said cancelling vibration from said cancelling means to the point of combination of said cancelling vibration with said vibration from said physical system, said adaptive filter being operable to successively delay, filter and scale said first electrical signals from said first sensing means to produce corresponding outputs for driving said cancelling means, and then to deterministically adjust said delay, filtering and scaling operation based on said second electrical signals received from said second sensing means to provide corresponding revised outputs for driving said cancelling means to produce cancelling vibration having the mirror image amplitude and phase characteristics of said vibration from said physical system within preset limits.

11. An apparatus for the cancellation of vibration in a physical system comprising: first sensing means for detecting said vibration, said first sensing means producing first electrical signals representing the amplitude and phase characteristics of said vibration; cancelling means for generating cancelling vibration for combination with said vibration from said physical system; second sensing means disposed at a location spaced from said first sensing means in the direction of propagation of said vibration, said second sensing means detecting the summation of said vibration from said physical system and said cancelling vibration and producing second electrical signals representing the amplitude and phase characteristics of said summation; and electronic circuitry means including an adaptive cancelling filter and an adaptive compensation filter, said adaptive compensation filter being operable to deterministically adjust the phase characteristics of said second electrical signals produced by said second sensing means for introduction into said adaptive cancelling filter to assure stable operation thereof, said adaptive cancelling filter having a modified LMS algorithm and a transversal filter, said adaptive filter being operable to successively delay, filter and scale said first electrical signals from said first sensing means to produce corresponding outputs for driving said cancelling means, and then to deterministically adjust said delay, filtering and scaling operation based on said phase compensated second electrical signals received from said second sensing means to provide corresponding revised outputs from driving said cancelling means to produce cancelling vibration having the mirror image amplitude and phase characteristics of said vibration from said physical system within preset limits.

12. The apparatus of claim 11 wherein said modified LMS algorithm includes means for accommodating the delays inherent in the propagation of said vibration from said first sensing means to said second sensing means, and the propagation of said cancelling vibration from said cancelling means to the point of combination of said cancelling vibration with said vibration from said physical system.

13. An apparatus for the cancellation of vibration in a physical system comprising: first sensing means for detecting said vibration, said first sensing means producing first electrical signals representing the amplitude and phase characteristics of said vibration; cancelling means for generating cancelling vibration for combination with said vibration from said physical system; second sensing means disposed at a location spaced from said first sensing means in the direction of propagation of said vibration, said second sensing means detecting the summation of said vibration from said physical system and said cancelling vibration and producing second electrical signals representing the amplitude and phase characteristics of said summation; and electronic circuitry means including an adaptive cancelling filter for driving said cancelling means to produce said cancelling vibration, and an adaptive compensation means operable to deterministically adjust the phase characteristics of said second electrical signals produced by said second sensing means for introduction into said adaptive cancelling filter to assure stable operation thereof, said adaptive cancelling filter including a transversal filter and an LMS algorithm modified to accommodate the delay inherent in the propagation of said vibration from said first sensing means to said second sensing means, and the delay inherent in the propagation of said cancelling vibration from said cancelling means to the point of combination of said cancelling vibration with said vibration from said physical system, said adaptive filter being operable to successively delay, filter and scale said first electrical signals from said first sensing means to produce corresponding outputs for driving said cancelling means, and then to deterministically adjust said delay, filtering and scaling operation based on said second electrical signals received from said second sensing means to provide corresponding revised outputs for driving said cancelling means to produce cancelling vibration having the mirror image amplitude and phase characteristics of said vibration from said physical system within preset limits.