WO1994029848A1

WO1994029848A1 - Error path transfer function modelling in active noise cancellation

Info

Publication number: WO1994029848A1
Application number: PCT/US1994/005153
Authority: WO
Inventors: Sen M. Kuo; Mary K. Christensen
Original assignee: Caterpillar Inc.
Priority date: 1993-06-11
Filing date: 1994-05-11
Publication date: 1994-12-22
Also published as: AU6786394A

Abstract

A signal containing the full range of frequencies to be cancelled is employed in establishing an initial estimate of the error path loss in an active noise cancellation system. The signal is from about 80 cycles per second to about 500 cycles per second and sounds like a 'chirp'. The estimate is in the form of coefficients used in the adaptive filter (13) of the algorithm (15) used in generating the cancelling signal.

Description

ERROR PATH TRANSFER FUNCTION MODELLING IN ACTIVE NOISE CANCELLATION

Technical Field

The invention relates to the control of sound using active noise cancellation and in particular to the development of an estimate representing the losses in the error path for use in an adaptive cancelling signal algorithm.

Background and Relation to the Prior Art

Active noise cancellation involves superimposing on a noise acoustic wave an opposite acoustic wave that destructively interferes with and cancels the noise wave. The cancelling acoustic wave is of equal amplitude but of opposite phase to the noise acoustic wave. The generation of the proper interference signal to produce cancellation, in the proper position at the right time requires taking into consideration a number of variables resulting in elaborate signal processing. The active noise cancellation principle is most useful at frequencies below 500 cycles per second (Hz) . Above that frequency range, noise attenuating materials applied to surfaces are more effective.

The implementation of the principle of active noise cancellation generally involves sensing of the characteristics of the noise acoustic wave, generating the cancelling acoustic wave and through monitoring of the combined waves, developing a feedback signal that keeps the cancelling wave in adjustment. The monitoring signal is frequently called an "error" signal. Most implementations of the active noise cancellation principle also accommodate changes in the frequency and intensity characteristics of the noise. This involves incorporating adaptability into a feedback from the monitoring microphone that provides information used in adjusting the cancelling wave. The computations involved in determining the adjustments to the cancelling wave are performed under a procedure known in the art as an algorithm.

A number of algorithms with adaptability have evolved in the art. A survey article by J.C. Stevens entitled "An Experimental Evaluation of Adaptive Filtering Algorithms for Active Noise Control", Georgia Institute of Technology, GHTI/AERO, Atlanta, GA 1992 Pages 1-10, provides an illustrative description of the current capabilities in the art. The implementation of the algorithms has been achieved in the art using digital signal processing and fabricated in the form of single semiconductor chip active noise controller devices.

In the computations employed in the algorithms, value estimates are used in iterative incremental comparisons until a minimum difference between the cancelling signal and the monitoring or error signal is achieved. The accuracy of the estimate in turn affects the efficiency of the adaptability of the algorithm. In developing the estimate, a factor that is involved is a loss that occurs in the translation of the cancelling signal to an acoustic signal, the transport of that signal through the air and the sensing and reconversion of that signal back to an electric signal by the monitoring or error microphone. In other words the loss is the difference between the cancelling signal and the monitor signal of it. That difference is known in the art as the error path loss. The response sensitivity of adaptive algorithms to the accuracy of the estimate of the error path loss is addressed in the text in the art "Active Control of Sound" by P.A. Nelson and S.J. Elliott, Academic Press, Harcourt, Brace Jovanovich Publishers, in Section 6.14 Pages 195-198; and in Section 6.15 Pages 202-203 a description is given of the use of a white noise signal in developing the estimate.

Summary of the Invention

The invention is the use of a signal that includes all the frequencies in the range being cancelled in the development of the estimate of the value representing the error path loss or the loss in the transfer between the cancelling signal and the monitor or error signal. The signal varies in frequency from about 80 to about 500 (Hz) and being in the audio range, it sounds like a "chirp."

Brief Description of the Drawings

Figure 1 is a schematic representation of the error path loss in an active noise cancellation system. Figure 2 is a schematic of the elements involved in an active noise cancellation system.

Figure 3 is a diagram of a functional block representation of an adaptive algorithm used in the processing of an active noise cancellation signal. Figure 4 is a schematic representation of the use of the signal of the invention in developing an estimate of the error path loss values. Description of the Invention

Referring to Figure 1 a schematic representation is provided of the error path loss in an active noise cancellation system. In Figure 1 a cancellation signal is graphically depicted and labelled as element 1. The signal 1 is delivered to an electroacoustical transducer or speaker labelled element 2 through a conductor 3. An acoustic replica of the signal 1 shown as dashed lines 4 travels through the air and is sensed by a monitoring or error microphone 5. The acoustic replica 4 is again converted to an electrical signal in an acoustoelectrical transducer or microphone 5 and delivered as the monitor or error signal 6 through a conductor 7. The signal 6 has some of the characteristics, that were present in signal 1, missing, and the difference is known in the active noise cancellation art as the error path loss. The loss results from variations in the performance of the components 2, 3, 5 and 7 and the transmission properties of the acoustic signal 4 due to for example heat, humidity and component aging. The error path loss, while a major unknown in the active noise control system must be estimated accurately or the algorithm computational effectiveness will be reduced or operability of the active noise cancellation system may be compromised.

Referring next to Figure 2, a schematic representation is provided of the elements involved in an active noise cancellation system. In Figure 2 wherein like reference numerals are employed for like elements as in Figure 1 the overall operation takes in sound sensed at a remote microphone 10 as one input to a controller 11 for generation of the sound cancelling signal on conductor 3. During operation, in an individual iteration, the monitor or error signal on conductor 7 entering controller 11 reflects both the error path loss shown by the difference between signals 1 and 6 and also any difference with respect to the previous iteration's cancelling signal. The algorithm, in each iteration, in an incremental step, is to compare the last iteration's cancelling signal and the present monitoring or error signal and generates a new cancelling signal modified in a direction to minimize any difference between the previous cancelling and the present monitoring signals. In order to do this there must be supplied to the controller 11 an estimate of the error path loss in order that the difference between the previous iteration cancelling signal and the present iteration monitoring signal can be employed in the incremental step that moves the next cancelling signal toward minimizing the difference. Referring next to Figure 3, a diagram of a functional block representation of an adaptive algorithm such as is used in the controller 11 is provided. In the diagram of Fig. 3 the elements are representations of the functions of variables that influence the signals. The algorithm is the filtered X least mean squares type known in the art. The algorithm operates by calculating an error correction for the cancelling signal, applying the correction and repeating till a minimum variation is achieved. The more accurate the estimate of the error path loss the fewer iterations to convergence.

In Fig. 3 the input X travels the main path and through the adaptive filter. The cancelling signal is determined by the adaptive filter to which is added the error path loss which when acoustically superimposed with main the path becomes the error signal. The error signal in turn enters the controller for another iteration with the dotted arrow indicating a change in the adaptive filter.

In accordance with the invention there is a much improved estimate for the error path loss when a signal that contains all the frequencies in the range being cancelled is used in establishing the estimate.

That range is from about 80 cycles per second (Hz) to about 500 cycles per second (Hz) . The signal being audio sounds like a chirp. The lower number 80 is the response limit of most off the shelf equipment. The chirp signal is provided in a standard variable frequency generator and the value to be used for the error path loss is established in a separate operation and entered between X_n and the controller at the location labelled "error path estimate".

Referring to Figure 4 an illustration is provided of the use of the chirp signal containing all the frequencies in the range being cancelled in establishing an estimated value for the error path loss for use in the adaptive filter of the algorithm. In Fig. 4 the chirp signal generator 12 has the output connected to the speaker 2 and the adaptive filter 13 at Y . The output of the adaptive filter is summed in summing amplifier 14. From the output, labelled e' , through the LMS Algorithm 15 connected to both Y and the adaptive filter 13, the adaptive filter 13 is updated at each iteration. The chirp generator 12 is connected through a disconnecting element 16, shown as a switch. When the adaptive filter 13 converges, that is when e' is minimized, the disconnecting element is opened and separates the chirp generator 12. The coefficient representing the estimate of the error path loss is thus established and the value is transferred to the error path estimate in Fig. 3. In a preferred embodiment the controller uses a commercially available semiconductor integrated circuit. A satisfactory model for controller 11 is the TMS 320C30 Floating point DSP manufactured by Texas Instruments, Inc., Dallas, Texas; an adaptive filter portion of which may also be employed as the adaptive filter 13. A satisfactory model of a microphone for a monitoring microphone 5 and an input microphone 10 is a model SM98-A made by the SHURE Co. A satisfactory model of a cancelling speaker for the speaker 2 is the Rockford Fosgate PRO-128 12" Sub Woofer.

What has been described is the use of a signal containing the full range of frequencies that are to be cancelled as a signal used to establish an initial and more accurate estimate of the error path loss in an active noise cancellation system.

Claims

1. An active noise cancellation system comprising: a cancelling signal error path serially through a cancelling signal speaker (2) , the air and a monitoring microphone (5) , and an adaptive filter type algorithm (15) operable in providing cancelling signals to said cancelling speaker (2) based on iterative minimum difference seeking comparison measurements of signals from said monitoring microphone (5) with signals for said cancelling speaker (2) , said adaptive filter of said algorithm (15) having an initial estimate coefficient of loss in said error path derived through iterative computations with a signal containing all frequencies to be cancelled.

2. The active noise cancellation system of claim 1 wherein said signal containing all frequencies to be cancelled is from about 80 to about 500 cycles per second.

3. The active noise cancellation system of claim 1 wherein said coefficient is developed in a separate adaptive filter (13) for use in said adaptive filter type algorithm (15) .

4. Apparatus for establishing a coefficient representing an initial value for error path loss for use in an adaptive filter type algorithm (15) in active noise cancellation comprising a cancelling signal error path serially through a cancelling signal speaker (2) , the air and a monitoring microphone (5) , an adaptive filter (13) having the input connected to said speaker (12) and having the output summed in a summing amplifier (14) with the output of said monitoring microphone (5) and having at least one of incrementing and decrementing means for each iteration in seeking a minimum value of said summing amplifier output, and a variable frequency generator (12) having a signal output containing all frequencies to be cancelled in said active noise cancellation, said variable frequency generator (12) being detachably connected to the input of said adaptive filter (13).

5. The apparatus of claim 4 wherein said frequency generator (12) provides frequencies from about 80 to about 500 cycles per second.

6. The apparatus of claim 4 wherein said adaptive filter (13) is changed by a least mean squares type algorithm (15) .

7. In an active noise cancellation system the improvement comprising: the use of a signal containing the full range of frequencies to be cancelled in said active noise cancellation system in an initial determination of the error path loss to be used in computations in an adaptive filter algorithm (15) in said active noise cancellation system.

8. The improvement of claim 7 wherein said initial determination is made in a separate apparatus and transferred as a coefficient value in said adaptive filter algorithm (15) .

9. In a process of active noise cancellation including the cancellation of sound with a cancellation signal, the monitoring of the combined sound and cancellation signals and iteratively adjusting the cancellation signal in response to comparisons in a minimum seeking filtered adaptive algorithm (15) the improvement comprising the steps of: initially establishing a value for said adaptive algorithm (15) representing the loss between said cancellation signal and said monitoring signal (6) through the use of a signal containing the full range of frequencies to be cancelled.

10. The process of claim 9 wherein said step of initially establishing a value is performed in a separate apparatus and transferred as an initial coefficient setting in said filtered adaptive algorithm (15) .

11. The process of claim 10 wherein said full range of frequencies is from about 80 to about 500 cycles per second.