JP2008067070A - Adaptive noise filter - Google Patents

Abstract

A noise signal is reduced in a high frequency region.
An observation signal d is detected, regarded as an observation error e ′, and a filter coefficient w is calculated so as to minimize the observation error e ′ (step S5). The minimized observation error e ′ or the like is set as a value E (step S6). It is determined whether or not the value E is larger than a preset threshold value (step S7). If YES is determined, an arbitrary phase correction amount Δ ′ is added to or subtracted from the phase correction amount Δ (step S8), and phase correction is performed according to the phase correction amount Δ. A noise removal signal y is applied (step S4). If it is determined NO, the phase correction is prohibited, whereby the noise removal signal y after the phase correction by the immediately previous phase correction amount Δ is continuously applied (step S9).
[Selection] Figure 2

Description

  The present invention relates to an adaptive noise filter that reduces noise signal superposition by a noise removal signal. More specifically, the present invention relates to an adaptive noise filter that cancels the noise signal by a method of generating and applying a noise removal signal having a phase opposite to that of various generated noise signals (noise, electromagnetic noise, etc.).

In recent years, adaptive noise filters based on digital signal processing have been used as countermeasures against various noise signals. The adaptive noise filter reduces noise by generating and applying a noise removal signal having an opposite phase of the generated noise instead of passive noise reduction. So far, the techniques described in Patent Document 1 and Non-Patent Documents 1 to 4 have been proposed.
JP 2006-115401 A Yoshinobu Kajikawa, Yasuo Nomura, "Active noise control system that does not require a secondary path model", IEICE A, vol.J82A, no.2, pp.209-217, Feb.1999 G. Oriti, AL Julian and TALipo, "A new space vector modulation strategy for common mode voltage reduction", Proc. Of 1997 Power Electronics Specialists Conference, pp. 1541-1546, June 1997 S. Senini and PJ Wolfs, "The coupled inductor filter: analysis and design for AC systems," IEEE Trans. On Industrial Electronics, vol.45, issues.4, pp.547-578, Aug. 1998

  However, the conventional adaptive noise filter cannot accurately reproduce the propagation characteristic fluctuation of the observation signal, or an output delay occurs due to the internal processing of the filter, etc., in a high frequency region (for example, about several MHz or more). There were cases where it could not be used. That is, the conventional adaptive noise filter may be limited to use in a low frequency region.

  The present invention has been made in view of the above problems, and an object of the present invention is to provide an adaptive noise filter in which a noise signal is reduced in a high frequency region by a noise removal signal.

  In order to solve the above-described problems, an adaptive noise filter according to the present invention includes a reference signal measurement unit that detects a noise signal variation output from a noise source as a reference signal, and a signal at a position where the noise signal is superimposed as an observation signal. An observed signal measuring means for detecting, a noise removing signal generating means for generating a noise removing signal for reducing superposition of the noise signal based on the detected reference signal, and a noise for applying the generated noise removing signal A removal signal applying means, a filter coefficient calculation means for calculating a filter coefficient for defining the amplitude and phase of the noise removal signal based on the observation error when the detected observation signal is regarded as an observation error, and a preset value Find the measurement error one or more times within the time range, and reduce the noise removal signal level so that the average of the observation errors or the average of the absolute values is minimized. ; And a phase correction control means for controlling a correction amount.

  Further, the adaptive noise filter of the present invention is a filter that defines the amplitude and phase of the noise removal signal based on an analysis error that is a difference between the detected observation signal and the noise removal signal, instead of the filter coefficient calculation means. Filter coefficient calculation means for calculating coefficients is provided.

  Further, the adaptive noise filter of the present invention replaces the phase correction control means with the difference between the observation error before the phase correction and the analysis error after the phase correction once or two times within a preset time range. Phase correction control means for controlling the phase correction amount of the noise removal signal so as to minimize the average of the analysis error or the average of the analysis errors is obtained.

  In the adaptive noise filter of the present invention, the observation signal measuring unit detects a signal at a position where the noise removal signal is superimposed as an observation signal.

  In the adaptive noise filter of the present invention, the observation signal measurement unit detects the reference signal instead of the reference signal measurement unit.

  According to the present invention, a noise signal in a high frequency region can be reduced, and a great effect can be expected for cost reduction and countermeasures against noise.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  FIG. 1 is a configuration diagram of an adaptive noise filter according to the present embodiment.

  In FIG. 1, 1 is an adaptive noise filter, 2 is a noise source, and 3 is a signal source.

  The adaptive noise filter 1 includes a reference signal measurement unit 11, a reference signal measurement sensor unit 111, an observation signal measurement unit 12, an observation signal measurement sensor unit 121, a noise removal signal application unit 13, and a noise removal signal application interface unit 131 (hereinafter, A noise removal signal application IF unit 131) and an optimization processing unit 14.

  The reference signal measurement sensor unit 111 and the observation signal measurement sensor unit 121 include an ISN (Impedance Stabilization Network) terminal, an antenna, a current probe, a voltage probe, a microphone, and the like. These are for detecting signals, and can be arranged outside the adaptive noise filter 1 because higher effects can be obtained if they are arranged near the location where the signals are generated. Moreover, since these may each detect the same signal, they can be shared.

  The noise removal signal application IF unit 131 includes an ISN terminal, an antenna, a current probe, a voltage probe, and the like. These are for applying a signal, and can be configured outside the adaptive noise filter 1 because a higher effect can be obtained if they are arranged near the portion to which the signal is applied.

  The reference signal measurement unit 11 constitutes a reference signal measurement unit of the present invention, and the observation signal measurement unit 12 constitutes an observation signal measurement unit of the present invention.

  The optimization processing unit 14 includes a noise removal signal generation unit 141, a filter coefficient calculation unit 142, and a filter output phase correction unit 143.

  The filter coefficient calculation unit 142 constitutes a filter coefficient calculation unit according to the present invention, and the noise removal signal generation unit 141 constitutes a noise removal signal generation unit according to the present invention.

  FIG. 1 shows that when noise is superimposed on a signal s from a signal source 3 by a noise source 2 such as an electronic device via a transmission line such as a space or a power line, this noise is converted into an adaptive noise filter 1. Corresponding to the image reduced by. Here, a signal at a position where noise is superimposed is referred to as an observation signal d.

  FIG. 2 is a flowchart of an example of the adaptive noise filter 1. The adaptive noise filter 1 is configured with phase correction control means for performing the processing of this flowchart and flowcharts described later.

  First, the reference signal measurement unit 11 detects the noise signal fluctuation output from the noise source 2 as the reference signal x (t) via the reference signal measurement sensor unit 111 (step S1).

  Next, the optimization processing unit 14 sets 0 as the initial value of the phase correction amount Δ when correcting the phase of the noise removal signal y (step S2).

After applying the initial value of the filter coefficient w (n) defined in the filter coefficient calculation unit 142 and the reference signal x (t) detected at time t to the equation (1), the noise removal signal generation unit 141 A noise removal signal y (t) pattern is generated (step S3).

  Next, the filter output phase correction unit 143 performs phase correction on the noise removal signal y according to the phase correction amount Δ, and the noise removal signal application unit 13 converts the noise removal signal y after phase correction to the noise removal signal. The voltage is applied to a predetermined location (for example, a location where noise is superimposed) via the application IF unit 131 (step S4).

  That is, the reference signal x (t) is an input signal of the adaptive noise filter 1. In step S4, even when the phase correction amount Δ is 0 and substantially no phase correction is performed, it is referred to as a noise removal signal y after phase correction.

  In some cases, the noise removal signal y does not have an anti-phase relationship with respect to the noise signal to be removed, and the noise signal may not be removed. Processing is performed.

First, in this state, the observation signal measurement unit 12 detects the observation signal d (t) via the observation signal measurement sensor unit 121, and thereby the filter coefficient calculation unit 142 is defined by Expression (2). The filter coefficient w is calculated so as to minimize the analysis error e (t) which is the difference (step S5).

  Alternatively, the observation signal measuring unit 12 detects the observation signal d, that is, the observation signal d that is the sum of the signal s from the signal source 3 and the noise removal signal y, and regards this as an observation error (hereinafter referred to as an observation error e ′). ), The filter coefficient calculation unit 142 calculates the filter coefficient w so as to minimize the observation error e ′ (step S5).

  As an algorithm for calculating the filter coefficient w, a minimum mean square algorithm, an adaptive lattice algorithm, a sequential least square algorithm, a least square lattice type algorithm, a fast transversal filter (such as an FTF algorithm), or the like can be used. Detailed information is published in Yoji Iiguni, “Adaptive Signal Processing Algorithm”, Baifukan, 2000.

  Next, the observation signal measurement unit 12 calculates the analysis error e (t) of Equation (2) minimized in step S5, its absolute value, its energy amount, observation error e ′, its absolute value, and its energy amount. Either one is set as the value E (step S6).

  Next, it is determined whether or not the value E is larger than a preset threshold value (step S7). If YES is determined, the filter output phase correction unit 143 adds or subtracts an arbitrary phase correction amount Δ ′ to the phase correction amount Δ (step S8), and returns to step S4.

  If NO is determined in step S7, the phase correction is prohibited, and thereby the noise removal signal y after the phase correction by the immediately previous phase correction amount Δ is continuously applied (step S9).

  That is, in FIG. 2, the phase correction amount Δ that minimizes the value E is obtained, and the noise removal signal y after the phase correction by the phase correction amount Δ is applied.

  Since the optimization processing unit 14 requires a certain processing time (for example, about several tens of n to several hundreds of seconds as an example of a current apparatus), the noise signal to be removed and the generated noise removal signal y does not necessarily have an antiphase relationship. That is, it is necessary to prevent the filter effect from being lowered due to the delay time (= phase shift of the output filter). Therefore, as shown in the above flow, in the filter output phase correction unit 143 of the adaptive noise filter 1, the phase correction amount (delay time) of the noise removal signal y based on the fluctuation of the observation signal from the observation signal measurement unit 12. It is possible to optimize the noise removal signal y by controlling the phase correction amount of the noise removal signal y based on the result.

  FIG. 3 is a flowchart of another example of the adaptive noise filter 1.

  The difference from FIG. 2 is that any one of the analysis error e, its absolute value, its energy amount, observation error e ′, its absolute value, and its energy amount is obtained a plurality of times within a preset time range. Is set as the value E (step S6 ′). The time range (time length) and the number of times are constant.

  That is, also in FIG. 3, the phase correction amount Δ that minimizes the value E is obtained, and the noise removal signal y after the phase correction by the phase correction amount Δ is applied.

  FIG. 4 is a flowchart of another example of the adaptive noise filter 1.

  The difference from FIG. 2 is that the difference between the observation error e ′ (t−1) before phase correction and the analysis error e (t) after phase correction or the phase correction amount Δ that minimizes the absolute value thereof is obtained (step S11). ), Applying a noise removal signal y after phase correction by the phase correction amount Δ (step S12).

  FIG. 5 is a flowchart of another example of the adaptive noise filter 1.

  The difference from FIG. 2 is that the difference between the observation error e ′ (t−1) before the phase correction and the analysis error e (t) after the phase correction is obtained a plurality of times within a preset time range, and each time A phase correction amount Δ that minimizes the average of the absolute values or the average of the absolute values at each time is obtained (step S11), and a noise removal signal y after phase correction by the phase correction amount Δ is applied (step S12). It is. The time range (time length) and the number of times are constant.

  2 to 5 show an example in which 0 is set as the initial value of the phase correction amount Δ, but a method of setting an arbitrary value other than 0 is also conceivable.

  6 and 7 are diagrams illustrating application examples of the adaptive noise filter 1. FIG. 6 shows an evaluation system, and FIG. 7 shows an example of an evaluation result.

  As shown in FIG. 6, a noise signal is applied to the communication cable 41 via the ISN 42. A reference signal x is input to the adaptive noise filter 1 from a terminal of the ISN 42 (corresponding to the reference signal measurement sensor unit 111). In addition, the observation signal d generated in the communication cable 41 by the noise signal is input to the adaptive noise filter 1 via the current probe 43 (corresponding to the observation signal measurement sensor unit 121). Then, the noise removal signal y is applied to the communication cable 41 from the adaptive noise filter 1 via the current probe 44 (corresponding to the noise removal signal application IF unit 131).

  In FIG. 7, the observation signal d when the adaptive noise filter 1 is not used is a solid line, the observation signal d when the adaptive noise filter 1 is used is a broken line, and the observation signal d when the LMF algorithm (conventional method) is used. Is indicated by a one-dot chain line. As described above, when the adaptive noise filter 1 is used, an observation signal generated by a noise signal, that is, an unnecessary signal, can be attenuated to 1/10 or less of the conventional method.

  In the above description of step S6, the example in which the value E is set based on the analysis error e and the observation error e ′ has been described. However, in step S6, the filter count update value = μ · e · x or , Filter count update value = μ · e ′ · x (where μ is a step size definition parameter when the filter coefficient w is updated) may be set.

It is a block diagram of the adaptive noise filter which concerns on this Embodiment. 3 is a flowchart of an example of an adaptive noise filter 1. 3 is a flowchart of an example of an adaptive noise filter 1. 3 is a flowchart of an example of an adaptive noise filter 1. 3 is a flowchart of an example of an adaptive noise filter 1. It is a figure which shows the evaluation system in the application example of the adaptive noise filter. It is a figure which shows the evaluation result in the example of application of the adaptive noise filter.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Adaptive noise filter 2 ... Noise source 3 ... Signal source 11 ... Reference signal measurement part 12 ... Observation signal measurement part 13 ... Noise removal signal application part 14 ... Optimization processing part 41 ... Communication cable 42 ... ISN
43, 44 ... current probe 111 ... reference signal measurement sensor unit 121 ... observation signal measurement sensor unit 131 ... noise removal signal application interface unit 141 ... noise removal signal generation unit 142 ... filter coefficient calculation unit 143 ... filter output phase correction unit d ... Observation signal e '... Observation error e ... Analysis error w ... Filter coefficient y ... Noise removal signal x ... Reference signal

Claims (5)

A reference signal measuring means for detecting a noise signal fluctuation output from a noise source as a reference signal;
An observation signal measuring means for detecting a signal of a place where the noise signal is superimposed as an observation signal;
A noise removal signal generating means for generating a noise removal signal for reducing superposition of the noise signal based on the detected reference signal;
A noise removal signal applying means for applying the generated noise removal signal;
Filter coefficient calculation means for calculating a filter coefficient that defines the amplitude and phase of the noise removal signal based on the observation error when the detected observation signal is regarded as an observation error;
Phase correction that controls the phase correction amount of the noise removal signal so as to minimize the simple average or absolute value of the observation error by finding the observation error one or more times within a preset time range An adaptive noise filter comprising: a control means. Instead of the filter coefficient calculation means,
2. A filter coefficient calculation means for calculating a filter coefficient for defining an amplitude and a phase of a noise removal signal based on an analysis error that is a difference between the detected observation signal and the noise removal signal. Adaptive noise filter. Instead of the phase correction control means,
Find the difference between the observation error before the phase correction and the analysis error after the phase correction once or twice in the preset time range, and minimize the simple average of the analysis error or the average of the absolute value The adaptive noise filter according to claim 1, further comprising phase correction control means for controlling a phase correction amount of the noise removal signal. The observation signal measuring means includes
The adaptive noise filter according to any one of claims 1 to 3, wherein a signal at a location where the noise removal signal is superimposed is detected as an observation signal.   5. The adaptive noise filter according to claim 1, wherein the observation signal measurement unit detects a reference signal instead of the reference signal measurement unit.

Images