JPH05303388A - Muffler device - Google Patents

Abstract

PURPOSE:To reduce the computing time so as to muffle the sound whose fundamental period is long by detecting a sound period, dividing the period into plural sections and setting a sampling period for each section. CONSTITUTION:A period detection dividing circuit 1 of a controller 104 detects a fundamental period T of a noise sound source and the period T is divided into two different length sections T1 and T2. The ratio between lengths T1 and T2 is beforehand set depending on the kinds of noises. A clock 2 generates clock pulse signals S2 for every beginning of each section T1 and T2 and on the other hand, a sampling frequency is set for every section and the clock 3 generates clock signals which are synchronized with the sampling period set for every section. Therefore, sounds which have different frequency ranges for every section are properly processed, the coefficient length of the filter coefficient for each section becomes short and computations are executed in a short sampling time.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a device for suppressing noise in a place where periodic noise is generated, such as in a factory, a construction site, or a vehicle, by producing a sound having an opposite phase and the same amplitude as the noise. Regarding

[0002]

2. Description of the Related Art Conventionally, a so-called "active muffling" method has been known, which aims to muffle sound by generating sounds having the same amplitude and opposite phase to noise.

The so-called "Filtered X LMS" is an algorithm applied to active muffling.
Is generally used, and in this filter based on Filtered X LMS, the coefficient of the digital filter is sequentially updated for each sampling period in order to create a sound deadening waveform. However, the calculation of the coefficient of this digital filter requires a complicated convolution calculation when it is based on a random noise waveform. This complicated calculation must be executed within a minute sampling period, which requires an increase in memory capacity and an increase in the calculation capacity of the processor, as well as a high cost. This is all the more so when the noise becomes high frequency and the sampling frequency is increased accordingly. For this reason, there is a practical limit to silencing at high frequencies, and it is also an obstacle to increasing the convergence speed.

Therefore, a method of limiting the noise reduction target to periodic noise instead of random noise, generating a pulse signal synchronized with this period, and updating the coefficient of the digital filter based on this pulse signal has been proposed before this application. ("Synchronous adaptive filter and its application to active noise and vibration control" / Haruo Hamada et al / Proceedings of the Acoustical Society of Japan / Heisei 4
Published in March, 2013). In addition, a patent application has already been filed for its content.

According to the method according to the above-mentioned documents, since a pulse signal is used, a complicated convolution operation is not required at the time of filtering, so that it is possible to solve the problem that it is difficult to muffle high frequencies.

[0006]

However, since the coefficient length of the digital filter is made equal to the number of samples according to the noise period, the coefficient length becomes very large. Especially when the sampling frequency is increased. Therefore, the calculation time is extremely long due to the increase in the coefficient length.
If the noise source is the engine, the cycle corresponding to the engine speed is low and it does not matter, but the noise of the press machine, which has a longer cycle than the engine speed, requires a large amount of computation time and a huge memory capacity. Therefore, it is practically difficult to implement the device.

The present invention has been made in view of the above circumstances, and an object thereof is to provide an apparatus for realizing the silencing of a periodic sound without causing the above problems.

[0008]

Therefore, according to the present invention,
A sound source that generates a periodic sound, a cycle detection unit that detects the cycle of the sound source, and a sound wave having the same amplitude as the sound wave propagating from the sound source but having a phase opposite to that of the sound wave propagated to cause the sound waves to interfere with each other at a predetermined position. A sound wave generating actuator for silencing, an interference sound wave detecting means arranged at the predetermined position for detecting an interference sound wave of the both sound waves, and a pulse generating a pulse signal synchronized with the cycle detected by the cycle detecting means. A reference signal is generated by performing an operation of filtering the pulse signal with a first filter coefficient indicating a transfer function between the actuator and the interfering sound wave detecting means. And the detection signal of the interference sound wave detection means, the second filter coefficient is sequentially updated at predetermined sampling intervals, and the second filter coefficient is updated. In a silencer that creates a signal for driving the actuator by performing an operation of filtering the pulse signal with the generated second filter coefficient, the cycle detected by the cycle detecting means is divided into a plurality of sections, and The sampling period is set for each section, the number of samples is calculated for each section, and the numbers of the first and second filter coefficients in each section are made to match the number of samples in each section.

[0009]

According to this structure, the sound cycle is detected, this cycle is divided into a plurality of sections, and the sampling cycle is set for each section. Therefore, it is possible to flexibly deal with sounds having different frequency ranges in each section. Then, the number of samples is obtained for each section, but this number of samples is smaller than the number of samples obtained for the entire period. Therefore, if the coefficient length of the second filter coefficient in each section is set to a length corresponding to the number of samples in each section, the coefficient length becomes small. For this reason, the coefficient length of the reference signal generated by performing the calculation for filtering the pulse signal by the first filter coefficient indicating the transfer function between the actuator and the interfering sound wave detecting means for each section also becomes small. Then, when sequentially updating the second filter coefficient for each predetermined sampling cycle based on the reference signal and the detection signal of the interfering sound wave detecting means, the coefficient length becomes small and calculation can be executed during a short sampling time. It will be possible.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of a silencer according to the present invention will be described below with reference to the drawings. FIG. 2 is a block diagram showing the configuration of the embodiment.

As shown in the figure, the noise source 101 is assumed to be a press machine and generates a tapping sound with a longer cycle than that of an engine or the like. Note that an engine, a motor, a fan, or the like can be arbitrarily applied as long as it emits a periodic sound.
The synchronization signal sensor 102 is provided to detect the fundamental period of the noise source 101, and a microphone is used, for example. The period may be directly detected by using a rotary encoder attached to the rotating machine or an accelerometer that detects a sine wave synchronized with the sound period. The signal detected by the sync signal sensor 102 is the sync pulse generator 10
The basic period T is detected from the input signal added to 3 and the pulse signal S1 synchronized with this period T is output (see FIG. 3A).

The controller 104 outputs the output signal S1 of the sync pulse generator 103 and an error sensor 106 described later.
A calculation to be described later is executed based on the output error signal e of the above, and a signal y for silencing the propagation noise from the noise source 101 is generated in the silencing area 107 centered on the error sensor 106 and added to the actuator 105.

The actuator 105 is, for example, a speaker and is driven according to the signal y. As a result,
A sound wave for noise reduction is radiated and propagated toward the noise reduction area 107, and the noise reduction in the noise reduction area 107 is achieved. In addition,
The actuator 105 is not limited to a speaker, and an electric or hydraulic vibrator that can indirectly emit a sound wave by vibrating a structure can be used. The error sensor 106 detects an interfering sound wave of a sound wave propagating from the noise source 101 and a radio wave propagating from the actuator 105, that is, a sound deadening error at the center position of the sound deadening area 107, and an error signal e indicating this sound deadening error.
To the controller 104.

The calculation executed by the controller 104 will be described below with reference to the block diagram shown in FIG.

In the figure, the period detection division circuit 1 receives the external synchronizing pulse signal S1 and detects the fundamental period T of the sound of the noise source 101. This basic period T is, as shown in FIG. 3B, two first sections having different lengths from T1 and T2,
It is divided into a second section. The ratio of the lengths T1 and T2 of the first and second sections is, for example, 2: 3, which is set in advance according to the type of noise. Thereby, it is possible to deal with a variation in the length of the cycle T.
Note that the lengths of the sections may be the same. In addition, the division ratio may not be fixed but may be changed according to the operating condition of the press machine. The number of divisions is not limited to two, and may be more than that. The first clock 2 generates the clock pulse signal S2 at each start of the first and second sections (see FIG. 3B).

On the other hand, the sampling frequency f for each section
k (k = 1, 2) is set. This setting is made according to the type of noise. For example, in a press machine, the cycle T
The frequency is relatively high in the first section and relatively low in the last section. Therefore, the sampling frequency f2 of the second section is set to the sampling frequency f2 of the first section.
It can be set smaller than 1. It should be noted that the sampling frequency in each section may be a preset value, or may be appropriately changed according to the fluctuation of the basic period T and other circumstances. Therefore, the second clock 3 generates a clock signal synchronized with the sampling period set for each section (FIG. 3).
(See (c)). Here, from the sampling period for each section and the lengths T1 and T2 of each section, the number of samples N for each section
k (k = 1, 2) is obtained. Conversely, the number of samples Nk may be given in each section, and the sampling frequency may be determined from this and the length of each section.

Next, a reference signal (reference signal) rk
The calculation of (n) will be described. Here, the reference signal is a signal created by arithmetic processing equivalent to filtering the pulse signal with a filter coefficient indicating the transfer function C between the actuator 105 and the error sensor 106. Note that the calculation of this signal is well known from the above-mentioned documents and the like, and since it is not directly related to the gist of the invention of the present application, the characteristic part of the invention of the present application will be mainly described. Further, hereinafter, the subscript k added to the symbol r or the like is a symbol indicating a divided section, and n is a symbol indicating the n-th sampling in each section.

The transfer function C between the actuator 105 and the error sensor 106 is modeled, and the filter coefficient sequence C ^ of the transfer function C is stored in the C ^ register 5. The identification of C needs to be performed as a preprocessing before the control is executed, but it is also possible to appropriately change it to the latest data during the control.

FIG. 4A shows a filter coefficient sequence C ^ having a coefficient length (tap length) J at the sampling frequency f. When it is determined that the control enters the section k by the first clock 2, the filter coefficient sequence C ^ k of the coefficient length Jk when sampling at the sampling frequency fk is shown in FIG. Created as shown.
In this case, the sampling frequency fk of each section is set so that p can be represented by f / p where p is a positive integer. Figure 4
(B) shows the case of f = 2fk.

Next, a filter coefficient sequence Ck is created as shown in the following equations (1) to (3) depending on whether the coefficient length Jk is larger or smaller than the length Nk corresponding to the number of samples Nk in the section k. , Ck register 4 is stored. In addition,
As the length of the section k and the sampling frequency fk, that of the section k one cycle before is actually used. Therefore, if there is no change from the previous cycle, the previous filter coefficient sequence Ck can be used as it is and the creation of Ck can be omitted.
In this case, the Ck register 4 is provided for the number of sections K, and the first
The registers may be sequentially switched according to the output signal S1 of the clock 2 of FIG. Of course, the filter coefficient sequence Ck may be recreated for each period T for all sections k.

The calculation of the filter coefficient sequence Ck is performed by the following equations (1) to
It is calculated by (3).

Ck = {Ck0, Ck1, Ck2 ..., CkNk-1} (1) Ckn = ΣC ^ kn-aNk (however, a is added from -infinity to + infinity) (2) C ^ k = {C ^ k0, C ^ k1, C ^ k2 ..., C ^ kJk-1} (3) Therefore, as shown in FIG. 4C, Nk = 10 and coefficient length Jk = 8. If Nk is larger than Jk, the contents are set to zero at two locations in the filter coefficient sequence Ck, and the coefficient length of the filter coefficient sequence Ck is N samples.
The length is set according to k. On the other hand, when Nk is smaller than Jk as in the case where Nk = 5 and the coefficient length Jk = 8 as shown in FIG. 4D, the filter coefficients C ^ are separated from each other at three positions in the filter coefficient sequence Ck. By adding, the coefficient length of the filter coefficient sequence Ck is set to a length corresponding to the number of samples Nk. Now, the filter coefficient is sequentially updated for each sampling period based on the reference signal and the detection signal of the error sensor 106, and the signal for driving the actuator 105 is created by filtering the pulse signal with the updated filter coefficient. The formula for updating the above filter coefficients is shown below.

Wkn + 1 (i) = Wkn (i) + αe (n) rk (n−i) (4) where α is a convergence coefficient.

Here, the coefficient length Ik of the filter coefficient string Wkn
As for the filter coefficient Ck and the coefficient length Jk, the length is set according to the number of samples Nk. Therefore, the subscript i is 0, 1, 2, ... Ik-1.

In the above equation (4), the filter coefficient sequence Wkn
In order to obtain, the reference signal rk (ni) must be obtained, and this reference signal is read from the Ck register 4 according to the following equation (5) (FIG. 4 (c),
(See (d)).

Rk (n−i) = Ckn−1 (i = 0 ... Ik−1) (5) Since the reference signal is obtained based on the pulse signal, the Ck register 4 is simply used without performing the convolution operation. It can be seen that it is only necessary to read out the contents of (1) according to (ni).

The Wk register 11 stores a filter coefficient sequence Wkn at the current sampling time n for each section k.

To update the filter coefficient sequence Wkn to Wkn + 1, first, Wkn is read from the Wk register 11 and added to the adder 9. On the other hand, α is read from the memory 6 that stores the convergence coefficient α, and is multiplied by the error signal e (n) in the multiplier 7. The multiplication result is added to the multiplier 8 and further multiplied by the reference signal rk (n-i) generated by the Ck register 4. The result of this multiplication is added to the adder 9 to become the right side of the above equation (4), which becomes the filter coefficient sequence Wkn + 1 at the next sampling time n + 1. This passes through the delay circuit 10 and the content of the register in the section k is updated. The switching of the sampling time n → n + 1 is executed according to the output signal S3 of the second clock 3.

The k registers of the Wk register 11 are switched for each section k in accordance with the output signal S2 of the first clock 2, and only one register operates. Therefore, when paying attention to one register corresponding to the section k, only the noise waveform in the section k repeatedly operates in each cycle T, and it is in the rest in the other sections.

The signal ykn for driving the actuator 105 at the sampling time n in the section k is generated according to the following equation (6).

Yk (n) = Wkn (Pk) (6) where Pk is nmod Nk.

As is clear from the equation (6), the drive signal y is generated based on the pulse signal, so that it is understood that a complicated convolution operation is unnecessary.

The drive signal yk (n) is converted into noise d ^ (n) by the transfer function C and added to the noise reduction area 107, and the noise d (n) from the noise source 101 is interfered with to effectively muffle noise. To be done.

For an arbitrary section (for example, a section with a low noise level), the control in that section is turned off, and the time during which it is off is subjected to another processing (update calculation of another section with a high noise level or some kind of processing). It is also possible to carry out the processing that is prioritized).

[0035]

As described above, according to the present invention, the period of a sound is detected, this period is divided into a plurality of sections, and the sampling cycle is set for each section. It becomes possible to flexibly cope with sounds having different frequency ranges. Then, the number of samples is obtained for each section, and the coefficient length of the filter coefficient in each section is set to a length according to the number of samples in each section, so that the coefficient length can be made small and even when a sound with a long basic period is silenced. The calculation time can be significantly reduced.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a controller in a device of an embodiment of a silencer according to the present invention.

FIG. 2 is a block diagram showing the overall configuration of the embodiment.

3 is a time chart showing a pulse signal applied to the controller of FIG. 1 and a clock signal generated by a clock of the controller.

FIG. 4 is a conceptual diagram used for explaining arithmetic contents of the controller shown in FIG. 1.

[Explanation of symbols]

 1 synchronization detection division circuit 2 clock 3 clock 4 register 5 register 11 register

Claims (1)

[Claims] 1. A sound source that generates a periodic sound, a cycle detection unit that detects a cycle of the sound source, and a sound wave having a phase opposite to that of a sound wave propagating from the sound source and having the same amplitude and radiated at a predetermined position. A sound wave generating actuator for silencing sound waves by interfering with each other, and arranged at the predetermined position,
An interference sound wave detecting means for detecting an interference sound wave of the both sound waves,
Pulse signal generating means for generating a pulse signal in synchronism with the cycle detected by the cycle detecting means, and the pulse according to a first filter coefficient indicating a transfer function between the actuator and the interference sound wave detecting means. A reference signal is created by performing an operation of filtering the signal, and the second filter coefficient is sequentially updated at predetermined sampling intervals based on the reference signal and the detection signal of the interference sound wave detection means, and the updated In a silencer that creates a signal for driving the actuator by performing a calculation for filtering the pulse signal with a second filter coefficient, the cycle detected by the cycle detection means is divided into a plurality of sections, and each section is divided. Set the sampling period to, calculate the number of samples for each section, A silencer adapted to match the number of the first and second filter coefficients in a section with the number of samples in each section.