JPH05118906A - Method and device for acoustic measurement - Google Patents

Abstract

PURPOSE:To effectively measure sounds of an object to be measured which is influenced by dark noise for increasing the low frequency spectrum of the object by using signals LSS(Logarith Swept Sine) as non-inpulse measuring signals. CONSTITUTION:An LSS generator 10 produces LSS by address controlling the LSS generation data stored in a waveform memory 11 in a digital signal form, from a waveform reading circuit 12. The data is converted into analog signals by D/A converter 20 and input to an object to be measured 30, then the output is converted to digital signals by A/D converter 40 and input to a LSS reverse filter 50 of a deconvolution means. The filter 50 divides input signals on the axis of frequency by means of high speed Fourier conversion means to obtain transmission functions in the range of frequency axis, and it is converted into inpulse response in the range of time axis and outputted. By the use of measuring signals excellent in S/N at the low range and with high signal power, sounds of object to be measured is measured effectively.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an acoustic measurement method and apparatus for measuring acoustic characteristics such as impulse response and transfer function, and more particularly to a low-frequency acoustic measurement method using a measurement signal obtained by sweeping a sine wave in logarithmic proportion with frequency. TECHNICAL FIELD The present invention relates to an acoustic measurement method and apparatus for improving S / N in a frequency range.

[0002]

2. Description of the Related Art When measuring acoustic characteristics such as impulse response and transfer function of a linear, time-invariant, and causal system to be measured,
In general, (1) impulses of extremely short duration,
(2) A flat chirp signal in the time domain, (3) a flat TSP (time stretching pulse) in the frequency domain, etc. are used as the measurement signal. FIG. 20 is a waveform diagram of the chirp signal, where the horizontal axis represents time (Time) and the vertical axis represents response (Respo).
nse). FIG. 21 is a spectrum of the chirp signal, where the horizontal axis represents frequency (Frequency) and the vertical axis represents amplitude (Magnitude). Also, FIG. 22 shows TSP.
In the waveform diagram of, the horizontal axis represents time and the vertical axis represents response. FIG. 23 shows the spectrum of TSP, where the horizontal axis is frequency and the vertical axis is amplitude. FIG. 24 shows the inverse filter characteristic of TSP, where the horizontal axis represents time and the vertical axis represents response. As an example of the TSP system, as shown in Japanese Patent Laid-Open No. 3-6467, the spectrum has a flat spectrum at all discrete frequencies, the phase characteristics are continuous and proportional to the square of the discrete frequency, and the number of samples is 1 /
Some use a measurement signal having a waveform such that the discrete Fourier transform of the input signal at two discrete frequencies is 1 + j0.

FIG. 11 shows the impulse response of the impulse method of the above (1). The horizontal axis of this characteristic is time, and the vertical axis is response. FIG. 12 is an enlarged view of the tail of this impulse response. FIG. 13 shows the spectrum of this impulse method. The horizontal axis of this spectrum is frequency, and the vertical axis is amplitude. FIG. 14 shows the TSP of (3) above.
The impulse response of the method is shown. The horizontal axis of this characteristic is time, and the vertical axis is response. FIG. 15 is an enlarged view of the tail of this impulse response. FIG. 16 shows the spectrum of this TSP method. The horizontal axis of this spectrum is frequency, and the vertical axis is amplitude.

[0004]

[Problems to be Solved by the Invention]

The method using the impulse of (1) has an advantage that it can be measured with a simple configuration, but has a small signal power. Therefore, as shown in FIG. 13, the S / N in the low range is not good (the influence of dark noise is large in the low range). In order to improve this point, a considerable amount of synchronous addition is required, which has the drawback of increasing the measurement time. On the other hand, the chirp signal of (2) and the TSP of (3) have high signal power, so S
/ N is good and therefore the measurement time is short.
However, considering that the noise at the time of acoustic measurement is mainly dark noise, and the spectrum has a characteristic of increasing as it goes to lower frequencies as shown in FIG. 10, it is possible to obtain sufficient S / N in the middle and high frequencies. However, it is often impossible to sufficiently remove low-frequency noise. In the spectrum of the TSP method shown in FIG. 16, the influence of dark noise still remains in the low range (the amplitude is larger than that in the middle and high ranges).

According to the present invention, by using a measuring signal having a large signal power and a good S / N in the low frequency range,
It is an object of the present invention to provide an effective acoustic measurement method and a measurement device for a system under measurement which is affected by dark noise whose spectrum increases as the frequency becomes lower.

[0006]

In order to achieve the above object, according to the present invention, a non-impulse measurement signal and an impulse for this measurement signal are applied to a system to be measured having linear, time-invariant, and causality. In an acoustic measurement method for measuring an impulse response or a transfer function of the system under measurement by using an inverse convolution process, a signal obtained by sweeping a sine wave in logarithmic proportion to frequency is used as the non-impulse measurement signal. This is the first feature.

In the present invention, the non-impulse measurement signal and the deconvolution processing for making the measurement signal into impulse are applied to the system under measurement having linear, time-invariant, and causality. An acoustic measuring device for measuring an impulse response or a transfer function of a system, which comprises a measuring signal generating means for digitally creating and outputting a measuring signal obtained by sweeping a sine wave in a logarithmic proportion to a frequency, and a measuring signal generating means. A signal to be measured which is converted into an analog signal and supplied to the system to be measured, and the response of which is taken and converted into a digital signal before being output, and a device which is obtained from this system to be measured I / O And an inverse convolution means for performing an inverse convolution process for converting the measurement signal into an impulse with respect to the digital response signal. The output of the inverse convolution means is used to measure the measured object. That it has to obtain the impulse response or transfer function of the system is the second feature.

[0008]

As shown in FIG. 2, an acoustic measurement signal used in the present invention is a signal (Logarithm Sweep Sine; LS) obtained by sweeping a sine wave in logarithmic proportion to frequency.
(Abbreviated as S), the chirp signal and T
Similar to SP, it has a large signal power. Therefore, the S / N in the middle and high frequencies is originally good. In addition, as shown in FIG. 3, LSS is 3 dB / oct from the high range to the low range.
Since it has a spectrum that rises in, and the power in the low range is larger than that in the high range, the S / N in the low range is also good. Therefore, it is effective when the system under measurement, which is affected by dark noise whose spectrum in the low frequency range is high, is an acoustic measurement target.
Since acoustic measurement signals are digitally created and measurements are made in a digital configuration, all outputs, inputs and circuit operations are stable and reproducible, and perfect synchronization between signals required for deconvolution is easy. The synchronous state can be easily controlled with one clock, and an accurate impulse response or transfer function can be obtained very easily. Further, it is possible to use it in combination with the S / N improving method which has been used conventionally such as synchronous addition.

[0009]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram of an acoustic measuring device showing an embodiment of the present invention. In the figure, 10 is an LSS generator, 20 is a D / A converter, 30 is an object to be measured, and 40 is A.
The / D converter, 50 is an LSS inverse filter. The LSS generator 10 is a measuring signal generating means including a waveform memory 11 and a waveform reading circuit 12, and the LSS generating data stored in the waveform memory 11 in the form of a digital signal is address-controlled by the waveform reading circuit 12. Then LS
Create S. For example, all sample data of a desired LSS are calculated in advance and stored in the waveform memory 11, and when LSS is desired to be generated, the waveform reading circuit 12 outputs an address which is sequentially incremented from 0 at a constant speed. The LSS is created based on the value read from the waveform memory 11 given to the waveform memory 11. If the LSS is digitally created in this way, the stability and reproducibility of the waveform will be good, and it will be convenient when synchronizing the signals in the subsequent processing. D /
The A converter 20 and the A / D converter 40 constitute the measured system input / output means, and the D / A converter 20 converts the output of the LSS generator 10 into an analog signal and inputs it to the DUT 30. Further, the A / D converter 40 converts the output of the DUT 30 into a digital signal and outputs it. The LSS inverse filter 50 is an inverse convolution means, and in this example, is configured by using a digital signal processor. The input of this filter 50 is A /
The output of the D converter 40 may be directly input in real time, or may be temporarily stored in a storage means 60 such as a memory, a hard disk, or a digital audio tape and then read and input at appropriate times. good.

When the LSS generated by the LSS generator 10 is Fourier transformed, it is found that the LSS has a spectrum that rises from the high band to the low band at 3 dB / oct, as shown in FIG. In addition, when the 1 / N octave analysis is performed, it can be seen that the level is constant at any 1 / N octave band. Furthermore, when viewed in a time window of a constant width, it can be seen that the spectrum has a substantially constant shape on the logarithmic frequency axis, and the center frequency thereof moves with time. A signal having such characteristics can be calculated, for example, by the following formula 1.

[0011]

[Equation 1]

FIG. 2 is a waveform diagram of the LSS calculated by the above formula 1, and FIG. 3 is a spectrum obtained by Fourier transforming this LSS waveform by an FFT (Fast Fourier Transform). As is clear from FIG. 2, since the LSS waveform has a constant amplitude, A / A in FIG.
As the D converter 20, an inexpensive one having a relatively small word length can be used, and thus sufficient accuracy can be secured. Further, as is clear from FIG. 3, the spectrum is rising at 3 dB / oct toward the low range,
It has sufficient power in the low-frequency component that is likely to be buried in dark noise, etc., and it can be seen that the noise suppression capability in the low frequency range is high.

FIG. 25 shows another example of the configuration of the LSS generator 10 shown in FIG. 1, which is particularly suitable for generating the LSS waveform calculated by the above formula 1. That is, the formula of θ (i) in the above formula 1 is modified as follows.

[0014]

[Equation 2]

A sine wave table 111 and an exponent (exp) table 112 are prepared as the waveform memory 11, and the start frequency f ₀ , the stop frequency f ₁ ,
By calculating the values of A and a from the length N samples and giving them to the waveform reading circuit 12, the exponent (EXP) table 112 is used and θ (i) = A {exp (ai) −
The value of 1} is obtained, and the sine wave table 111 is read by using the value as an address to obtain the final LSS waveform.

When a non-impulse signal is used as a measurement signal, the measurement signal is deconvoluted (Deconvolutio) in order to obtain an impulse response or a transfer function.
n) is required. The following method can be considered for this reverse convolution processing. (A) Obtain an inverse filter analytically and fold it. (B) Division on the frequency axis by FFT. (C) Approximately obtain an inverse filter by FFT and fold it. The above method (A) can be applied to the above-described conventional TSP signal, which is a great advantage when the TSP signal is used. On the other hand, the conventional chirp signal and the LSS wave of the present invention can be subjected to the deconvolution processing by the method (B) or (C).

FIG. 4 shows a procedure for performing the deconvolution processing by the method (B) in the LSS inverse filter 50 of FIG. Since this method uses FFT, it is difficult to process in real time, but it is a reliable method. That is, the input x (t) and the output y of the DUT 30.
(T) is in the time domain, and noise n (t) is added to the output y (t). Therefore, if the impulse response of the DUT 30 is h (t), y (t) = h ( t) * x (t) + n (t). The FFTs 51 and 52 respectively convert these into complex numbers X (f) and Y (f) in the frequency domain. In this case, Y (f) is expressed as follows. Y (f) = H (f) · X (f) + N (f)

The complex conjugate calculator 53 inputs the complex number X on the input side.
The conjugate complex number X ^* (f) is calculated from (f). And
One of the multipliers 54 calculates X (f) · X ^* (f), and the other multiplier 55 calculates Y (f) · X ^* (f). Y (f) · X ^* (f) = H (f) · X (f) · X ^* (f)
+ N (f) · X ^* (f) Further, the divider 56 obtains the transfer function H (f). In this case, n (t) and x (t) are uncorrelated, and X (f) · X ^*
Since (f) is large, the second term on the right side can be ignored in the following equation. Y (f) · X ^* (f) / X (f) · X ^* (f) = H (f)
+ N (f) · X ^* (f) / X (f) · X ^* (f) Therefore, approximately H (f) = Y (f) · X ^* (f) / X (f) · X ^* ( f) is represented. Finally, the transfer function H (f) in the frequency domain is converted into an impulse response h (t) in the time domain by the inverse FFT 57 and output.

FIG. 5 shows an example in which an LSS wave inverse filter is created by the method (C). This is normalized into a time length twice as long as the original LSS wave, and is converted into 32 bits. This inverse filter allows real-time processing if the convolutional calculation word length is secured.
The processing result can be obtained in a shorter time than the method (B). For example, to perform real-time deconvolution in the case of 2 channels, 1
6k sample LSS wave can be used (inverse filter is 32
Similarly, in case of 4 channels, LSS wave of 8k samples can be used (inverse filter is 16kta).
ps).

When an LSS inverse filter is created with a double length and 32 bits, the error of the deconvolution is 9 × on the time axis.
It is about 10 ⁻⁷ (corresponding to 20-bit precision) and 0.001 dB or less on the frequency axis. Figure 6 LSS deconvolution error (data 18 bits, coefficient 32 bits)
Is a characteristic diagram showing on the time axis, and FIG. 7 shows this on the frequency axis. As described above, the characteristics of the present invention are all flat, which is extremely ideal.

When a similar error is measured by the conventional TSP,
About 6.4 × 10 ^{-7 on the} time axis (equivalent to 20-bit precision)
And 0.004 to 0.006 dB on the frequency axis
It is below. FIG. 8 is a characteristic diagram showing the deconvolution error (data 18 bits, coefficient 32 bits) of TSP on the time axis, and FIG. 9 shows this on the frequency axis.

FIG. 17 is a characteristic diagram showing an impulse response of the LSS system according to the present invention, and FIG. 18 is an enlarged view of the impulse response tail thereof. Furthermore, FIG. 19 shows this L
The SS system spectrum is shown. As is clear from FIG. 19, it can be confirmed that the characteristics close to the speaker response in the low frequency range are realized, and the influence of dark noise in the low frequency range is considerably removed. In the above description, as the LSS, when the sine wave is logarithmically (exponentially) proportional to the frequency and is swept, the low frequency is changed to the high frequency, but in principle, the high frequency is changed to the low frequency. You can change it. However, in general, low-frequency sound has a long residual time in the sound field, and therefore the deconvolution process in the subsequent process is often complicated.

[0023]

As described above, according to the present invention, by using a measurement signal having a large signal power and a good S / N in the low frequency range, the dark noise in which the spectrum increases as the frequency becomes lower. It is possible to provide an acoustic measuring method and a measuring apparatus that are effective for a system under measurement that is affected by. In addition, since the acoustic measurement signal is created digitally and the measurement is done in a digital configuration, all outputs, inputs and circuit operations are stable and reproducible, and perfect synchronization between the signals required for deconvolution is achieved. Can be easily obtained, these synchronous states can be easily controlled with one clock, and an accurate impulse response or transfer function can be obtained very easily, and synchronous addition, which is a conventionally used S / N improvement method, can be obtained. Can be used together.

[Brief description of drawings]

FIG. 1 is a configuration diagram showing an embodiment of the present invention.

FIG. 2 is a waveform diagram of an LSS wave of the present invention.

FIG. 3 is a spectrum of the LSS wave of the present invention.

FIG. 4 is a block diagram of the deconvolution processing of the present invention.

FIG. 5 is an LSS wave inverse filter characteristic diagram of the present invention.

FIG. 6 is a characteristic diagram showing the deconvolution error of LSS on the time axis.

FIG. 7 is a characteristic diagram showing the deconvolution error of LSS on the frequency axis.

FIG. 8 is a characteristic diagram showing a TSP deconvolution error on a time axis.

FIG. 9 is a characteristic diagram showing the deconvolution error of TSP on the frequency axis.

FIG. 10 is a frequency characteristic diagram of background noise in a studio.

FIG. 11 is an impulse response characteristic diagram of an impulse method.

FIG. 12 is an enlarged view of an impulse response tail of the impulse method.

FIG. 13 is an impulse-type spectrum.

FIG. 14 is a TSP type impulse response characteristic diagram.

FIG. 15 is an enlarged view of a TSP type impulse response tail.

FIG. 16 is a spectrum of the TSP method.

FIG. 17 is an impulse response characteristic diagram of the LSS method.

FIG. 18 is an enlarged view of an LSS impulse response tail.

FIG. 19 is a spectrum of the LSS method.

FIG. 20 is a waveform diagram of a chirp signal.

FIG. 21 is a spectrum of a chirp signal.

FIG. 22 is a waveform diagram of TSP.

FIG. 23 is a spectrum of TSP.

FIG. 24 is a characteristic diagram of an inverse filter of TSP.

FIG. 25 is a configuration diagram showing another example of the LSS generator of the present invention.

[Explanation of symbols]

10 ... LSS generator, 11 ... Waveform memory, 12 ... Waveform data reading circuit, 20 ... D / A converter, 30 ... DUT, 40 ... A / D converter, 50 ... LSS inverse filter, 5 ...
1, 52 ... FFT, 53 ... Complex conjugate operation unit, 54, 55
... Multiplication section, 56 ... Division section, 57 ... Inverse FFT.

Claims (2)

[Claims] 1. An impulse of the system under measurement using a non-impulse measurement signal for the system under measurement having linearity, time invariance, and causality, and deconvolution processing for making the measurement signal into an impulse. An acoustic measurement method for measuring a response or a transfer function, wherein a signal obtained by sweeping a sine wave in logarithmically proportional frequency is used as the non-impulse measurement signal. 2. An impulse of the system under measurement using a non-impulse measurement signal for the system under measurement that is linear, time-invariant, and causal, and deconvolution processing for converting the measurement signal into impulses. An acoustic measurement for measuring a response or transfer function, which is a measurement signal generating means for digitally creating and outputting a measurement signal obtained by sweeping a sine wave in logarithmic proportion to the frequency, and an analog signal for this measurement signal. And input to the system to be measured, and the response of the system to be measured, which receives the response, converts it to a digital signal, and outputs the digital signal, and a digital response signal obtained from the system input / output unit to be measured. And an inverse convolution means for performing an inverse convolution process for converting the measurement signal into an impulse signal, the output of the deconvolution means being used for impulse response of the measured system. Or acoustic measurement device which is characterized in that so as to obtain the transfer function.