JPH09134176A - Musical sound signal generating device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain the musical sound signal of high quality by performing filter calculations concerning musical waveform data while using coefficient data and musical waveform data equivalent to sample points to cut unwanted noise components. SOLUTION: An address signal generating means 1 generates an address signal consisting of an integer part IAD and a fraction part FAD with the rate corresponding to the sound pitch of a musical sound to be generated. A musical sound waveform data generating means 2 generates musical sound waveform sample data in accordance with the integer part FAD of the address signal. A filter coefficient supplying means 3 selects and supplies (n) pieces (provided n<m) of filter coefficients among (m) pieces of filter coefficients in accordance with the fraction part FAD of the address signal. A digital filter calculating means 4 performs the filter calculation of an mth order concerning musical waveform data equivalent to (n) sample points by using the (n) pieces of filter coefficients and the musical waveform data equivalent to (n) sample points generated in the musical sound waveform data generating means 2.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a musical tone signal generator which removes aliasing noise and the like and controls tone colors by utilizing digital filter operation, and more particularly, a digital filter having a relatively simple hardware structure and high precision. The present invention relates to a musical tone signal generator capable of performing calculations.

[0002]

2. Description of the Related Art In an electronic musical instrument, a digital filter is used for the purpose of controlling the tone color of a digital musical tone signal or removing noise. For example, in a so-called pitch-asynchronous tone synthesis method that synthesizes a tone signal by always sampling at a constant sampling frequency regardless of the frequency of the tone to be synthesized, in general, the tone frequency and the sampling frequency are non-integer ratios. As is clear from the sampling theorem, there is a possibility that aliasing noise that is inharmonic with the tone frequency may occur, and it is necessary to remove this. As a measure for removing the aliasing noise included in such a pitch-asynchronous tone signal, it has been conventionally considered to pass the tone signal through a digital filter having a characteristic of removing the aliasing noise (Japanese Patent Laid-Open Publication No. Sho. 61-90514).

[0003]

SUMMARY OF THE INVENTION In a digital filter that removes aliasing noise by passing a tone signal through a digital filter, the filter order must be sufficiently high to perform accurate filter operation, which complicates the filter configuration. There is a point. The same applies not only to the digital filter for eliminating aliasing noise, but also for digital filters for tone color control and other applications. The present invention has been made in view of the above points, and when performing tone color control or removing aliasing noise using a digital filter, it is possible to perform accurate filter calculation with a simple configuration. The present invention is intended to provide a musical tone signal generating device.

[0004]

SUMMARY OF THE INVENTION A tone signal generating device according to the present invention is an address signal generating means for generating an address signal consisting of an integer part and a decimal part which change at a rate corresponding to the pitch of a tone to be generated. A tone waveform data generating means for generating tone waveform sample data according to the integer part of the address signal, a control data generating means for generating control data for controlling a tone, and a plurality of filter characteristics showing different characteristics. A filter characteristic corresponding to the control data is selected from among filter coefficient generating means capable of generating a filter coefficient having a predetermined order corresponding to the selected filter characteristic, and a filter coefficient generated from the filter coefficient generating means. The filter coefficient interpolating means capable of generating m-th order filter coefficients larger than the predetermined order by interpolation is provided. By controlling the generation of the filter coefficient in the filter coefficient generating means and the interpolation in the filter coefficient interpolating means according to the fractional part of the address signal, it is possible to correspond to each continuous order for realizing the m-th order filter characteristic. n out of m filter coefficients (where n <m)
Filter coefficient supplying means for selecting and supplying the coefficient data, and the n-th coefficient data and the musical tone waveform data for n sample points generated by the musical tone waveform data generating means are used for the m-th order filter calculation. And digital filter calculation means for performing tone waveform data for n sample points.

The address signal generating means generates an address signal composed of an integer part and a decimal part which change at a rate corresponding to the pitch of a musical tone to be generated. Needless to say, the integer part of the address signal has a coarser resolution than the decimal part. The musical tone waveform data generating means generates musical tone waveform sample data according to the integer part of the address signal. Therefore, the resolution of the musical tone waveform sample data prepared by the musical tone waveform data generating means may be relatively coarse. For example, as in the conventional pitch-synchronized tone signal generator, the tone waveform sample data may only be prepared with an accuracy of about 64 divisions per waveform period. As will be described later, in this case, it does not matter whether the sampling frequency of the musical tone waveform sample data generated by the musical tone waveform data generating means is synchronized with the pitch. If the pitch asynchronous type is used, it is only necessary to prepare the musical tone waveform sample data with a coarse resolution equivalent to that of the pitch synchronous type. The filter coefficient supply means selects and supplies n pieces (where n <m) of coefficient data corresponding to the m-th order filter coefficient according to the decimal part of the address signal. The digital filter arithmetic means uses the n coefficient data and the musical tone waveform data for n sample points generated by the musical tone waveform data generating means to perform an m-th order filter calculation for musical tone waveform data for n sample points. Do about. As a result, an m-th order precise filter calculation (actually, this is not prepared by the musical tone waveform data generating means) with high precision musical tone waveform sample data having the resolution of the decimal part of the address signal (actually, Can perform a filter operation equivalent to performing an operation for n orders). That is, the tone waveform data generated corresponding to the integer part of the address signal is subjected to m-th order filter arithmetic processing in the digital filter,
It is possible to achieve the same processing as waveform interpolation with the resolution of the fractional part of the address signal. This can contribute to the removal of aliasing noise.

Incidentally, for example, in a conventional pitch-asynchronous tone signal generator, in order to eliminate the influence of aliasing noise as much as possible, it is common practice to increase the waveform resolution as much as possible and raise the sampling frequency. For example, waveform data is prepared with an accuracy of 1000 to 16000 divisions per waveform cycle, and therefore,
The waveform memory requires a considerable capacity. In this way, since the waveform memory is required to have a considerable capacity, in the case of the tone signal generating method in which a waveform consisting of a relatively short section such as a one-cycle waveform is stored in the memory and repeatedly read out, the continuous operation is still necessary. This is not suitable for a tone signal generation method in which a waveform composed of a relatively long section such as a typical multiple cycle waveform is stored in a memory and is read out. On the other hand, in the so-called pitch-synchronized tone synthesis method in which the sampling frequency is synchronized with the frequency of the tone to be synthesized, since the tone frequency (pitch) and the sampling frequency are in harmony, the component generated by folding is in harmony with the tone frequency. However, since it does not become noise, it is 6 per waveform cycle.
There is no problem even if the waveform data is prepared with relatively coarse accuracy of about 4 divisions. Therefore, it is also suitable for a tone signal generation method in which a waveform composed of a relatively long section such as a continuous plural-cycle waveform is stored in a memory and is read out.
On the other hand, in the present invention, regardless of whether the sampling frequency is synchronized with the pitch or not, it is only necessary to prepare the musical tone waveform sample data with a coarse resolution equivalent to that of the pitch synchronization type. Therefore, even in the case of a tone signal generating method in which a waveform consisting of a relatively short period such as a one-period waveform is stored in a memory and is repeatedly read out, a relatively long period such as a continuous multiple-period waveform is detected. The present invention is also suitable for the case of a tone signal generation method in which the following waveform is stored in a memory and is read out.

Accordingly, the resolution of the musical tone waveform sample data actually prepared by the musical tone waveform data generating means may be relatively coarse corresponding to the integer part of the address signal, thereby simplifying the circuit configuration. The simplification of the configuration of the digital filter circuit is achieved by the advantage that it can be achieved, and that the filter calculation is actually performed with respect to the limited order corresponding to the tone waveform data of n sample points less than m. While substantially enjoying the advantage of being able to achieve both, the substantial filter operation is performed on the accurate tone waveform sample data having the resolution of the fractional part of the address signal.
It is equivalent to performing the following precise filter operation, which means that various advantages brought about by precise filter operation for high resolution musical tone waveform sample data, for example, surely cutting unnecessary noise components, The advantage of being able to obtain a high-quality musical tone signal can also be enjoyed. Further, according to the present invention, since the filter coefficient supplying means has the filter coefficient interpolating means, it is possible to easily generate m-th order filter coefficients larger than a predetermined order that can be generated by the filter coefficient generating means. Has the effect. Further, the filter coefficient supply means selects a filter characteristic corresponding to control data for controlling a musical sound from a plurality of filter characteristics having different characteristics, and an m-th order filter characteristic corresponding to the selected filter characteristic. In order to realize the above, n (n <m) coefficient data are selected and supplied from among m filter coefficients corresponding to respective consecutive orders according to the fractional part of the address signal. As mentioned above, in addition to being able to obtain the noise removal function by the m-th order precise filter calculation with a simple configuration as described above, it is also possible to realize the tone color control function according to arbitrary tone control data. It has an excellent effect.

[0008]

Embodiments of the present invention will be described below in detail with reference to the accompanying drawings. FIG. 1 is a basic block diagram showing an embodiment of the present invention, in which 1 is an address signal generating means, 2 is a tone waveform data generating means, 3 is a filter coefficient supplying means, and 4 is a digital filter calculating means. . As described above, the address signal generating means 1 generates an address signal consisting of the integer part IAD and the decimal part FAD at a rate corresponding to the pitch of the musical tone to be generated, and the musical tone waveform data generating means 2 The musical tone waveform sample data is generated in accordance with the integer part IAD of the address signal, and the filter coefficient supplying means 3 selects n (though n) out of m-th order filter coefficients in accordance with the decimal part FAD of the address signal. <M) is selected and supplied, and the digital filter arithmetic means 4 produces n by the n filter coefficients and the tone waveform data generating means.
By using the musical tone waveform data for the sample points, the m-th order filter calculation is performed on the musical tone waveform data for the n sample points. In this example, the sampling frequency fs is constant regardless of the pitch of the musical sound to be generated (that is, pitch asynchronous type), and the digital filter calculation means 4 cuts off fs / 2 in order to remove the aliasing noise. Shall realize the low-pass filter characteristic.

In order to explain the principle of the tone signal generator according to the present invention, first, the presupposed digital filter operation will be described with reference to FIGS. FIG.
Is an example of a waveform diagram based on a general sampling frequency conversion theory, and FIG. 3 shows an example of a spectrum envelope thereof. FIG. 2A shows the sampling frequency f.
It is a figure which shows an example of the musical tone waveform sample data sampled by s about some sample points. This figure 2
It is considered that the musical tone waveform sample data of (a) is input to a digital filter that operates at a sampling timing of a frequency (M · fs) that is M times the sampling frequency fs. In that case, one cycle ts = 1 of the sampling frequency fs
Although 1 period ts / M of frequency M · fs is included in / fs, sample data is included at all M filter operation timings during one sample time of the musical tone waveform sample data of FIG. 2A. Is generated, sample data is generated only at one of the M filter operation timings in one period ts of fs, and the sample value is set to "0" at the remaining M-1 timings. . Thus, the frequency M
A sample value of the musical tone waveform sample data of FIG. 2 (a) is generated at a timing of 1 out of M sampling times of fs, and the sample value is "0" at M-1 timings of the remaining M sampling points. Is shown in FIG. 2 (b).

The tone waveform sample data shown in FIG. 2B is input to a digital filter, and a filter operation is performed in accordance with the sampling timing of the frequency M · fs. That is, the musical tone waveform data of each sample point whose one sampling period is ts / M (however, M has a valid sample value at one sampling timing, and the remaining M samples at M-1 timings). The filter coefficient of each order is calculated for the data whose value is "0". Then, as a filter operation output, FIG.
As shown in, it is possible to obtain tone waveform data in which sample values are densely generated corresponding to each of the sampling timings of M · fs. This is because the filter calculation is performed at each sampling timing of the frequency M · fs, and in the filter calculation at each sampling timing, the output signal is obtained by the convolution sum of each filter order and the corresponding sample value. By re-sampling the tone waveform data in which sample values are densely generated as shown in FIG. 2C at a desired sampling frequency, tone waveform sample data converted to a desired sampling frequency can be obtained. FIG. 2D shows an example of musical tone waveform sample data obtained by re-sampling at such a desired sampling frequency. In this case, assuming that the desired sampling frequency to be resampled is M · fs / N, if the waveform data of FIG. 2C is resampled once every N times of the sampling timing of M · fs. Good.

By the above processing, the musical tone waveform data having the sampling frequency fs can be resampled to the sampling frequency of M · fs / N. In the figure, M = 4 and N = 3. 3 (a) is a diagram showing an example of the spectrum envelope of the waveform of FIG. 2 (a), and FIG. 3 (b) is a diagram showing an example of the spectrum envelope of the waveform of FIG. 2 (b). The shape of the spectral envelope is the same in (a) and (b), but the level is shown in FIG.
(B) is 1 / M of (a). This is because the number of substantial sample values (sample values that are not “0”) included in the convolution sum is 1 / M of the original number. FIG. 3C shows an example of digital filter characteristics, and in this case, it is a low-pass filter having a cutoff frequency of fs / 2. FIG. 3D shows the spectrum envelope of the waveform of FIG. 2C obtained by filtering with this low-pass filter characteristic. In this case, the return by the sampling theorem is M
It occurs at a frequency of fs / 2, which is so high that it does not become noise during resampling. FIG.
FIG. 2E is a diagram showing an example of the spectrum envelope of the waveform of FIG. Note that the signal level remains down to 1 / M as described above just by re-sampling the output of the filter. Therefore, when re-sampling, the level of the waveform data in FIG. 2D should be multiplied by M. Therefore, we will return to the original level.

In the above description, the digital filter is used for sampling frequency conversion and is not directly related to the generation of a musical tone signal by setting the musical tone frequency to an arbitrary pitch. On the other hand, the present invention utilizes a digital filter operation based on the above-mentioned principle when generating a tone signal having an arbitrary pitch. That is, when the address signal generating means 1 generates an address signal which changes at a rate corresponding to the pitch of a musical tone to be generated, and musical tone waveform sample data is generated in accordance with this address signal, a digital filter based on the above-mentioned principle. I try to use arithmetic. The address signal is composed of an integer part IAD and a decimal part FAD. In the tone waveform data generating means 2, the integer part IA of this address signal is used.
Musical tone waveform sample data is generated according to D. In other words, the musical tone waveform sample data that can be generated by the musical tone waveform data generating means 2 only corresponds to the resolution of the integer part IAD of the address signal. An example of the musical tone waveform sample data generated from the musical tone waveform data generating means 2 corresponding to the integer part IAD of the address signal is shown in FIG.

The fractional part FAD of the address signal indicates a finer phase between adjacent sample points at the sample points specified by the integer part IAD of the address signal.
For example, the current address value designated by the integer part IAD and the decimal part FAD of the address signal is CA in FIG.
If the phase is indicated by D, the current value of the integer part IAD is, for example, “3”, and the decimal part FAD is CAD and IAD.
Is the difference. That is, CAD = IAD + FAD.
In order to obtain a good-quality musical tone waveform signal without harmonic distortion, subharmonic distortion, and noise, it is desired to obtain musical tone waveform sample data with the resolution of the fractional part FAD of the address signal, that is, corresponding to the phase of CAD. . In the present invention, by utilizing the digital filter operation based on the above-mentioned principle, this is realized with a simple configuration and without particularly increasing the sampling clock frequency. Therefore, in the filter coefficient supply means 3, n pieces of the coefficient data corresponding to the filter coefficient of the mth order (where n <
m) is selected and supplied. Figure 2 above
As is clear from the explanations related to (b) and (c), it is effective at a rate of once per M orders, without needing to supply musical tone waveform sample data corresponding to all filter coefficients of multiple orders. The sample value of the sample data may be set to “0” for the remaining orders,
Even so, since a precise filter operation is performed using the filter coefficients of the multiple orders, a good musical tone signal can be obtained. Based on this, the present invention performs a digital filter operation based on the same idea. However, in the processing based on the above-mentioned general sampling frequency conversion theory, it is necessary to actually perform the filter calculation on the filter coefficients of all orders in accordance with the high-speed filter calculation timing (frequency M · fs). Because the correspondence between the order and the sample data changes with the change of the operation timing, even if the product of the sample data of which the sample value is "0" and a certain filter coefficient is "0". Even if it is clear, it is unclear in which order it holds. Therefore, in the processing based on the conventional general sampling frequency conversion theory, FIG.
As shown in FIG. 2B, the musical tone waveform sample data having the sampling frequency fs as shown in FIG.
It is converted into data of sampling frequency M · fs by inserting “0” as a sample value at sampling timing of M clocks of fs once and M−1 times of M clocks, and this is converted into data of frequency M · fs. It is necessary to input to the operating digital filter, and actually perform the filter operation on the filter coefficients of all orders set by this digital filter.

On the other hand, in the present invention, attention is paid to the fact that the calculation of the sample data of the sample value "0" and the filter coefficient corresponding thereto is substantially unnecessary, and the calculation therefor can be omitted. It is characterized by having done. This makes it possible to simplify the arithmetic circuit while eliminating the need to perform arithmetic at high-speed filter arithmetic timing,
Moreover, by making it possible to perform a precise filter calculation by using a sufficiently large number of filter coefficients, it is possible to realize both simplification of the circuit configuration and high accuracy of the filter calculation all at once. First, similar to the relationship between FIGS. 2A and 2B, a digital filter that operates the musical tone waveform sample data of FIG. 4A at a sampling timing of a frequency M times the sampling frequency fs (M · fs). 1 period ts of the sampling frequency fs
Assuming that 1 cycle ts / M of frequency M · fs is included in = 1 / fs, all M filter calculation timings during one sample time of the musical tone waveform sample data of FIG. Without generating the sample data, the sample data is generated only at one of the M filter operation timings in one period ts of fs, and the sample value is "0" at the remaining M-1 timings. It is assumed that In this way, the sample value of the musical tone waveform sample data of FIG. 4A is generated at one timing for every M of the sampling timings of the frequency M · fs, and at the timing of M−1 for the remaining M timings. FIG. 4B shows that the sample value is "0". Here, the number of divisions of the fractional part FAD of the address signal (1 is divided by the minimum unit of the fractional part FAD, for example, if the minimum unit of the fractional part FAD is 0.1, the number of divisions is 10) If d, then M = d.

Considering that m-th order filter calculation is performed on the musical tone waveform sample data as shown in FIG. 4B (where m is an arbitrary number determined when setting the filter characteristic), FIG. As described with reference to FIG. 2, if the filter operation is performed at the sampling timing of the frequency M · fs by using all the m-th order filter coefficients, the filter output with the fine resolution similar to that shown in FIG. Signals are also obtained for the tone waveform sample data of FIG. But,
Then, as described above, the digital filter must be operated at a high sampling frequency M · fs, and the filter operation must be actually performed on all the filter coefficients of the orders set by the digital filter. However, this is disadvantageous and is not adopted in the present invention. Instead, in the present invention, noting that the calculation of the sample data of the sample value “0” and the filter coefficient corresponding thereto is substantially unnecessary, the calculation for that is omitted, and the filter calculation The calculation is performed without increasing the sampling frequency. According to the present invention, the unnecessary operation can be omitted by dividing the address signal into an integer part IAD and a decimal part FAD and generating musical tone waveform sample data according to the integer part IAD. It is realized by selecting and supplying the filter coefficient of (1) according to the decimal part FAD. In other words, the musical tone waveform sample data is generated with a relatively coarse resolution according to the integer part IAD, but the digital filter calculation is performed with a precise resolution according to the decimal part FAD.
Specifically, by selecting the filter coefficient according to the decimal part FAD of the address signal, the integer part IAD and the decimal part F
Current phase CA of the address signal specified by AD
As if the filter calculation is performed at the fine sampling timing for D, the integer part I
It is possible to determine the order of the filter coefficient corresponding to the musical tone waveform sample data of each sample point at the rough sampling timing corresponding to AD. In other words,
Correspondence between the sample data of each sample point corresponding to the integer part IAD and the order of the filter coefficient is determined according to the decimal part FAD of the address signal.
It is possible to specify the order of the filter coefficient corresponding to the dense sampling timing in which “0” is inserted as the sample value in the case of the assumption as shown in (b). It is possible to process as “0”, and it is only necessary to calculate the filter coefficient corresponding to the discrete order for sample data having a substantial value. Thus, according to the present invention, in the m-th order filter calculation, the number of filter coefficients to be supplied by the filter coefficient supply means 3 is not all the coefficient data corresponding to the m-th order filter coefficient but n (where n <
m) may be selected and supplied according to the fractional part FAD of the address signal.

Here, with respect to M at the sampling frequency M · fs of the assumed digital filter operation, M =
d (the number of divisions of the fractional part FAD), and FIG.
In the sample data of (b), sample data is generated only at one timing of virtual filter calculation timings obtained by dividing one sample interval of the tone waveform sample data of FIG. 4 (a) by M = d, and the remaining d- Since the sample value is set to "0" at one timing, the virtual sample calculation timing obtained by dividing the one sample interval of the musical tone waveform sample data of FIG. It has a value, and the actual calculation may be performed only with respect to the virtual filter calculation timing. Therefore, the above-mentioned n can be determined by the relationship of n = m / d, and it is only necessary to select n orders that are sequentially separated at the interval of d according to the fractional part FAD of the address signal. Further, at the timing of the assumed sampling frequency Mfs = dfs of the digital filter operation, it is only necessary to perform the operation once every d times, and it is not necessary to perform it at other timings, that is, the unit delay of the sample data is It is not necessary to perform sampling frequency M · fs = d · fs, and it is sufficient to perform sampling frequency fs. Therefore, it is possible to perform precise filter calculation with the accuracy of sampling frequency M · fs = d · fs. Instead, the actual filter calculation may be performed at the low sampling frequency fs. Further, in connection with this, even assuming sample data as shown in FIG. 4B, one sampling is actually performed for each d clock of such a sampling frequency M · fs = d · fs. There is no need to perform the process of inserting the sample value “0” d−1 times, and the sampling frequency f
The sample data generated according to the integer part IAD of the address signal according to s may be used as it is.

Explaining this further with reference to the drawings, the filter operation of the present invention is equivalent to performing the filter operation corresponding to the current phase CAD of the address signal for the sample data as shown in FIG. 4 (b). . In this case, an example of the correspondence relationship between each sample data of FIG. 4B and each order of the m-th order filter coefficient (0th order to m−1th order) is shown as follows:
It is as shown in FIG. FIG. 4C shows an m-order FIR.
(Finite impulse response) An example of an impulse response having a low-pass filter characteristic of a filter is shown by an envelope. The frequency domain of this FIR filter is M
-Fs = d-fs, and the cut-off frequency of the low-pass filter characteristic is set to fs / 2 in order to remove aliasing noise related to the sampling frequency fs. In this impulse response, a predetermined reference order (for example, the central order)
Corresponds to the current phase CAD of the address signal. In the filter calculation, the convolution of each sample data of FIG. 4B and the filter coefficient of FIG. 4C is obtained. In that case, the calculation is not performed on the sample data “0” of the sample data of FIG. 4B in the frequency domain of d · fs. That is, only the sample data having a valid sample value in FIG. 4B and the filter coefficient of the corresponding order are calculated. From the relationship of n = m / d, the number of sample data having an effective sample value in the convolution of order m is n. In the illustrated example, m = 96, d = 16, n
= 6. The convolution sum obtained by the filter operation executed corresponding to the current phase CAD of the address signal is shown by the solid line in FIG. Although partly shown by a broken line in FIG. 4 (d), such a convolutional sum, that is, a filter output signal is densely generated in the frequency domain of d · fs as in the case of FIG. 2 (c). Note that, similar to the above, since this convolution is actually performed only for n = 6 sample data, the level of the filter output signal is the original 1 / d = 1 /
Drops to 16. To cope with this, the level of the filter output signal may be multiplied by d = 16. Alternatively, it is not necessary to purposely multiply the level of the filter output signal by d = 16, and the calculation may be performed using a filter coefficient having a level of d = 16 times the original value.

FIG. 4 (e) illustrates the amplitude-frequency characteristic of the FIR low pass filter having the impulse response of FIG. 4 (c). The frequency domain component cut is -80
It has been confirmed that it is attenuated below dB. This is a highly accurate filter characteristic. FIG. 5 is an actual measurement diagram of the amplitude-frequency characteristic of such an FIR low pass filter. FIG. 6 is an actual measurement diagram of a spectrum of a sine wave signal that has passed through the FIR low-pass filter having the characteristic of FIG. As is clear from this, noise components other than the fundamental wave are certainly -8
It is attenuated below 0 dB. 4 (a) to 4 (c) are summarized, each order of the m-th order filter coefficient (0th order to m−
The first order) corresponds to the continuous value of the address signal with the resolution of the fractional part FAD, and the predetermined reference order (for example, the central order k) is the fractional part FA of the current address signal.
It corresponds to the position of D (that is, the position of the current phase CAD). The n orders which are separated from the reference order by an amount corresponding to the distance of each integer part IAD corresponding to n sample points with respect to the fractional part FAD of the current address signal are discretely determined, and thus the current address signal The filter coefficient supply means 3 supplies n filter coefficients corresponding to the n orders determined according to the decimal part. When the reference order corresponding to the current phase CAD of the address signal is the central order k (for example, m = 96, that is, when the total order is 0th to m−1 = 95th order, the central order is k =
47)) and n = 6, “the integer part IAD of the address signal for n sample points” is the integer of n = 6 sample points before and after the integer part IAD of the current address signal. Part, that is, IAD-2, I shown in FIG.
AD-1, IAD, IAD + 1, IAD + 2, IAD +
3 corresponds to this. Each integer part I for this n sample points
AD-2, IAD-1, IAD, IAD + 1, IAD +
2, the reference order (k) by an amount corresponding to the distance of IAD + 3
= 47 orders), n = 6 orders which are separated from each other are generally determined discretely as follows.

Order corresponding to IAD-2: k-FAD-2d Order corresponding to IAD-1: Order corresponding to k-FAD-d IAD: Order corresponding to k-FAD IAD + 1: Corresponding to k-FAD + d IAD + 2 Order: k-FAD + 2d Order corresponding to IAD + 3: k-FAD + 3d Of course, the above is only an example, and another definition method is possible, and even if it is defined as above, k, d, n Depending on how to decide, various singular solutions other than the above may occur. For example, if k = 46 even under the same conditions,
In some cases, the order corresponding to IAD-3 three sample points before the integer part IAD of the current address signal must be determined under the definition of k-FAD-3d. The filter coefficient supply means 3 includes the above table,
Alternatively, an arithmetic circuit for executing the above equation is provided, and n is output according to the fractional part FAD of the current address signal.
The order is determined, and coefficient data corresponding to the determined order n are supplied. An example of the internal configurations of the filter coefficient supplying means 3 and the digital filter calculating means 4 is shown in some detail as shown in FIG. The filter coefficient supplying means 3 is a filter coefficient generating means 3a for generating m-th order filter coefficients (0th order to m-1st order) and n out of the mth order filter coefficients of the fractional part FAD of the address signal. And a selection means 3b for selecting according to the value. The selection means 3b comprises, for example, the above-mentioned table, and each integer part IAD-2, IAD-1, IAD, IAD + for n sample points is provided in accordance with the value of the decimal part FAD of the address signal.
Order k-FAD corresponding to 1, IAD + 2, IAD + 3
-2d, k-FAD-d, k-FAD, k-FAD +
d, k-FAD + 2d, k-FAD + 3d are determined by this table, and filter coefficients h (i-2d), h (id), h (i), h (i + d), which correspond to the respective orders of this skip, h (i + 2
Select d) and h (i + 3d) and output.

In FIG. 7, the digital filter operation means 4
Is composed of an FIR filter and comprises delay means 4a for delaying the musical tone waveform data generated from the musical tone waveform data generating means 2 according to the integer part IAD of the address signal by the clock pulse φ (fs) of the sampling frequency fs. . The delay means 4a includes integer parts IAD-2, IAD-1, IAD, IAD + 1, IAD for n sample points.
+2, IAD + 3, sample data X (Ii−
2), X (Ii-1), X (Ii), X (Ii + 1), X (Ii + 2), X (Ii
i + 3). These sample data X (Ii-2),
X (Ii-1), X (Ii), X (Ii + 1), X (Ii + 2), X (Ii + 3)
Is input to the multipliers 4b1 to 4b6, and the filter coefficients h (i-2d), h (id), h (i), h (i given by the selecting means 3b are inputted.
+ d), h (i + 2d) and h (i + 3d). Multiplier 4b1 ~
The output of 4b6 is added by the adder 4c, and the added output is output as the FIR filter output. The sample data X (Ii) to be handled as the current sample point is the delay means 4
Since it is delayed by a, the filter coefficients h (i-2d) to h (i + 3d) given from the selection means 3b should be appropriately delayed and given to the multipliers 4b1 to 4b6. Here, if the above-mentioned relations of FIGS. 4A to 4C are arranged by another expression, n = m / d according to the division number d (d = 16 in the above example) of the decimal part in the address signal. N relationship
(N = 6 in the above example) is determined, and the n filter coefficients to be determined are those corresponding to the n orders that are sequentially separated by the interval of d, respectively. The n orders are respectively determined according to the values, and the filter coefficient supply means 3 supplies n filter coefficients corresponding to the n orders determined according to the fractional part FAD of the current address signal. Of.

Thus, according to the present invention, in the convolution calculation of the m-th order filter, where n = m, the calculation should be performed for coefficient data of all orders m.
Only the / d coefficient data need be calculated, and the calculation scale can be reduced to 1 / d. Moreover, although the sampling frequency in the actual calculation is fs, a result equivalent to that obtained by performing the digital filter calculation with a high resolution of d · fs can be obtained. Next, it will be described how to improve the accuracy of the filter operation without significantly increasing the circuit configuration scale. In order to carry out a filter operation of order q times, that is, q · m order, using filter coefficients for m-th order, interpolation of resolution q is performed between adjacent ones of filter coefficients for m-th order. The filter coefficients for the mth order may be generated densely. Such q
By the double interpolation, the equivalent sampling frequency in the digital filter calculation becomes a high resolution of q · d · fs and the filter order becomes the q · m order, so that the accuracy of the filter calculation can be considerably improved. Next, a more specific embodiment of the present invention will be described with reference to FIG. In the embodiment of FIG. 8, the decimal part FAD of the address signal is used.
Is set to d = 16, the order of the filter is set to m = 96,
n = 6. Further, the adjacent q-th order filter coefficients are interpolated with the resolution q = 4, respectively, whereby the q * m-th order = 384-th order filter coefficients are densely generated. Therefore, the fractional part FAD of the address signal is basically the division number d = 16 = “2
It is composed of 4-bit data corresponding to "4th power", and the lower 2 bits are further added to this data to instruct the interpolation step of resolution q = 4. Therefore, in this embodiment, the fractional part FAD of the address signal consists of 6-bit data. The sampling frequency is fixed at fs = 50 kHz, and a tone signal is generated asynchronously with the pitch. Also, in this embodiment, the digital filter is configured as a low pass filter characteristic FIR filter for removing the aliasing noise as described above.

The keyboard 10 has a plurality of keys for designating the pitch of a musical tone to be generated. The key pressed by the keyboard 10 is detected by the key assigner 11, and it is assigned to any one of the plurality of tone generation channels that a tone corresponding to the pressed key should be generated. The number of tone generation channels is eight as an example, and each channel is established by time-sharing a common tone generation means. The key assigner 11 outputs the key code KC of the key assigned to each channel, the key-on signal KON, and the key-on pulse KONP in a time division manner corresponding to the channel timing. The address signal generation circuit 12 receives the key code KC and the key-on pulse KONP from the key assigner 11, and time-divisions the address signal that changes at a rate corresponding to the pitch of the key assigned to each channel, corresponding to each channel timing. Occurring in an accident. This address signal is composed of the integer part IAD and the decimal part FAD as described above. The integer part IAD is composed of, for example, 18-bit data, and designates continuous sampling points of a musical tone waveform having a plurality of periods prepared in the musical tone waveform memory 13, and the decimal part FAD is 6 bits as described above. Data. As an example, assuming that the musical tone waveform memory 13 stores data of a plurality of periodic waveforms of the attack portion and data of a plurality of periodic waveforms of the sustain portion, the address signal generation circuit 12
Then, an address signal is generated so that the key-on pulse KONP is used as a trigger to read the multiple-cycle waveform data of the attack portion once, and then repeatedly read the multiple-cycle waveform data of the sustain portion. The address signal generating method in the address signal generating circuit 12 may be any method such as a method of repeatedly calculating a frequency number, a variable frequency dividing method, or a method of counting note clocks.

A touch detection device 14 is provided in association with the keyboard 10 and detects the touch of the pressed key. As an example, the musical tone waveform memory 13 stores sample data of a plurality of periodic waveforms as described above, and a plurality of sets of such waveform sample data are stored corresponding to tone colors selectable by the tone color selection circuit 15. There is. Further, a plurality of sets of waveform sample data may be further stored for the key scaling control of the tone color according to the pitch or the tone color control according to the key touch. Therefore, the tone color selection code TC, the key code KC, and the touch data TD are input to the tone waveform memory 13, the waveform to be read is selected according to these, and this is read according to the integer part IAD of the address signal. The data of the integer part IAD of the address signal generated from the address signal generation circuit 12 is added to the phase address input of the tone waveform memory 13, but this is not directly added, but an arithmetic unit (subtractor 16, adder 17). Join via. The arithmetic units 16 and 17 function as equivalent to the sample data delay means (corresponding to 4a in FIG. 7) in the digital filter. That is, in this embodiment, n (= 6) is obtained by actually delaying the sample data generated from the tone waveform memory 13.
Each integer part IAD-2, IAD-1, IA for sample points
The sample data corresponding to D, IAD + 1, IAD + 2, and IAD + 3 are not obtained, but the data of the integer part IAD of the address signal is time-divided by the arithmetic units 16 and 17 to -2, -1, 0, +1. , +2, +3 are added to each integer part IAD-2 of n (= 6) sample points,
IAD-1, IAD, IAD + 1, IAD + 2, IAD
The address data of +3 is generated in a time division manner, and the memory 13 is read out in response to the generated address data, so that the integer parts IAD-2, IAD-1, IA corresponding to these n (= 6) sample points are generated.
The sample data corresponding to D, IAD + 1, IAD + 2, and IAD + 3 are obtained.

The calculation timing for that purpose is shown in detail in FIG. 9, where CAC indicates a calculation cycle pulse generated in a cycle of the sampling frequency fs = 50 kHz, and this one cycle is divided into 8 time divisions of 8 channels. Timings CH1 to CH8 are formed, and the time-division time slots of each channel are divided into 6 to form 6th-order filter operation time slots. One cycle of the filter calculation time slot is one cycle of the master clock pulse MC, and this master clock pulse MC is modulo 6
By counting with the counter of 6 the 6 filter operation time slots 0 in 1 channel time slot,
Slot count data SLCTR for distinguishing 1, 2, 3, 4, and 5 is obtained. SMC is a filter operation cycle pulse, and one cycle is synchronized with one channel time slot. Assuming that the calculation cycle pulse CAC is 50 kHz, the filter calculation cycle pulse SMC is 40
The master clock pulse MC is 0 MHz and the frequency is 2.4 MHz. Each pulse and count data described above is generated from the master clock generator 22 and the timing signal generating circuit 23. In the subtractor 16, the value IA of the integer part of the sample point two sample points before the integer part IAD of the current address signal.
To obtain D-2, 2 is subtracted from IAD. The data IAD-2 thus obtained is added to the adder 1
7 and the slot count data SLCTR is added. This slot count data SLCTR is
As shown in FIG. 9, since it changes to 0, 1, 2, 3, 4, 5 within one channel time slot,
Corresponding to the six filter calculation time slots 0, 1, 2, 3, 4, 5 in the channel time slot, respective integer parts IAD-2, IAD-1, IAD, for 6 sample points
Address data of IAD + 1, IAD + 2 and IAD + 3 are generated from the adder 17 in a time division manner. Correspondingly, the integer parts I of these 6 sample points from the memory 13
AD-2, IAD-1, IAD, IAD + 1, IAD +
2, sample data corresponding to IAD + 3 are read out in a time division manner.

The sample data read from the memory 13 is input to the multiplier 18 for filter coefficient multiplication. The filter coefficient is supplied from the filter coefficient supply circuit 24 according to the fractional part FAD of the address signal as described later.
The output of the multiplier 18 is input to the accumulator 19 and the convolution sum is obtained. The accumulator 19 performs accumulation at the timing of the master clock pulse MC (that is, at each step of the slot count data SLCTR) and is cleared at the timing of the filter operation cycle pulse SMC. Immediately before clearing the accumulated value, the convolutional sum obtained in this calculation is latched in the latch circuit 20. These subtractor 16, adder 17, multiplier 18, accumulator 19, and latch circuit 2
0 is an FIR type digital filter operation circuit 21
Is equivalent to The filter coefficient supply circuit 24 stores filter coefficient memories 25 and 26 in which m = 96th order filter coefficients (0th order to 95th order) are respectively stored, and n = 6 of the 96th order filter coefficients are decimal fractions of the address signal. It comprises a selection means 27 for selecting according to the value of the section FAD, and an interpolation circuit 28. The two series of filter coefficient memories 25 and 26 are exactly the same, and two series of filter coefficient memories 25 and 26 are provided to read out two adjacent filter coefficients in parallel for interpolation in the interpolation circuit 28. There is. The impulse responses of the filter coefficients stored in the filter coefficient memories 25 and 26 are shown in FIG.
And a filter characteristic realized by this is a low-pass filter characteristic as shown in FIG. 4 (e) or FIG. 5, for example, and fs / 2 = half the sampling frequency fs = 50 kHz. The cutoff frequency is 25 kHz.

The selecting means 27 selects the decimal part F of the address signal.
Each integer part IA for n = 6 sample points according to the value of AD
D-2, IAD-1, IAD, IAD + 1, IAD +
2, the order k-FAD-2d, k- corresponding to IAD + 3
FAD-d, k-FAD, k-FAD + d, k-FAD
+ 2d, k-FAD + 3d is determined, and the determined order is used as an address signal from the filter coefficient memories 25 and 26 to filter coefficient h (i-2d), h (id), h (i), h (i + d), h. (i +
2d) and h (i + 3d) are selectively read out, and in order to make this decision by calculation, a subtracter 29, a multiplier 30,
An adder 31 is provided. The higher-order 4-bit data of the fractional part FAD of the address signal is input to the subtractor 29, and "15-
FAD "is subtracted. The slot count data SLCTR is input to the multiplier 30 to multiply" 16 × SLCTR ". The outputs of the subtractor 29 and the multiplier 30 are added to the adder 3
1 is added, and the above-mentioned orders k-FAD-2d and k-FAD are added.
-D, k-FAD, k-FAD + d, k-FAD + 2
Data for instructing d, k-FAD + 3d is output. The output of the adder 31 corresponding to each value 0 to 5 of the slot count data SLCTR, that is, the determined order is as follows. In the table below, the values IAD-2, IAD-1, IAD, IAD + 1 and
IAD + 2 and IAD + 3 are also shown.

[0027]

[Table 1]

If k = 47 and d = 16, the respective orders k-FAD-2d, k-FAD-d, defined as described above are obtained.
k-FAD, k-FAD + d, k-FAD + 2d, k-
It will be appreciated that FAD + 3d will be as in the table above. Therefore, the arithmetic circuit configuration in the selection means 27 is generally “k−FAD + (SLCTR−2) × d” =
It may be configured to execute the arithmetic expression "47-FAD + (SLCTR-2) × 16". The output of the adder 31 is input to the filter coefficient memory 25 as it is, and is added by 1 in the adder 32 to obtain the filter coefficient memory 26.
Is input to In this way, two filter coefficient data of adjacent orders are read from the filter coefficient memories 25 and 26. These two filter coefficient data are stored in the interpolation circuit 2
8 and is interpolated with 4-step interpolation characteristics (for example, linear interpolation characteristics) according to the lower 2 bits of data of the decimal part FAD of the address signal. In this way, only the m = 96th order filter coefficients are actually stored in the memories 25 and 26, but q · m = 4 × 96 = 384th order filter coefficients are densely prepared by interpolation. Is equivalent to The output of the interpolation circuit 28 is input to the multiplier 18. Note that the order of the filter coefficient output from the interpolation circuit 28 corresponding to each timing of the slot count data SLCTR is substantially changed as shown in the following table by the 4-fold interpolation.

[0029]

[Table 2]

In Table 1, FAD is modulo 16 (d =
16), but in Table 2, FAD is modulo 64 (d = 64)
It is. The filter calculation output signal output from the latch circuit 20 is applied to the multiplier 33, and the envelope generator 3
4 is multiplied by the amplitude envelope signal.
The envelope generator 34 generates an envelope waveform signal controlled according to the key code KC, the tone color selection code TC, and the touch data TD based on the key-on signal KON. The output of the multiplier 33 is input to the accumulator 35, and the sum of the sample data of all channels is obtained. The accumulator 35 performs accumulation at the timing of the filter operation cycle pulse SMC (that is, at each channel timing) and is cleared at the timing of the calculation cycle pulse CAC. Immediately before clearing the accumulated value, the sum of the sample data of all channels obtained this time is latched in the latch circuit 36. The sampling frequency of the tone signal output from the latch circuit 36 is fs = 50 kHz, and aliasing noise is reliably removed by the low-pass filter characteristic filtering with the cutoff frequency fs / 2 = 25 kHz in the digital filter arithmetic circuit 21. There is. Latch circuit 36
Output signal is converted into an analog signal by the digital / analog converter 37 and reaches the sound system 38. The output signal of the latch circuit 36 is input to a digital effect circuit 39 for imparting a musical effect such as reverb, echo, etc., and after the musical effect is imparted, it is converted into an analog signal by a digital / analog converter 37. Is
A sound system 38 is provided.

Incidentally, as described above, in order to cope with the fact that the level of the filter output signal is reduced to the original 1 / d = 1/16, the level of the filter coefficient stored in the filter coefficient memories 25 and 26 is multiplied by 16 in advance. Or the interpolation circuit 28
The coefficient data given to the multiplier 18 from the above may be shifted to the higher order by 4 bits. Next, an embodiment will be described in which the accuracy of the filter calculation is improved without significantly expanding the circuit configuration scale as compared with the configuration of FIG. In the embodiment of FIG. 8, in order to set the musical tone waveform sample data of the sampling frequency fs = 50 kHz as shown in FIG. 4 (a) to the sampling frequency of the domain of apparently d · fs = 16 × 50 = 800 kHz, As shown in FIG. 4B, d · fs =
Processing is performed assuming that the sample value "0" is inserted at a rate of d-1 times for every d times of the clock of 800 kHz, and the filter calculation is omitted for the order corresponding to the sample value "0". On the other hand, in the embodiment of FIG. 11 described below, the sampling frequency fs as shown in FIG.
= 50 kHz of the sound waveform sample data
It is interpreted that the data is the data of the 0th-order hold in the domain of d · fs = 800 kHz as shown in the above, and the above-mentioned embodiment is applied to the sample data having a valid value at all sampling points of the sampling frequency d · fs = 800 kHz. Similarly, the simplified filter operation is performed.

FIG. 10B shows the same m = 96 as in FIG. 4C.
The following is an example of the impulse response of the low-pass filter characteristic. As described above, the phase C of the current address signal is obtained.
Corresponding AD to a predetermined reference order k (for example, the intermediate 47th order), the sample data of FIG. 10A and the sample data of FIG.
Find the convolution of the impulse response of. This Tatami Komi sum is generally

(Equation 1) It is expressed as Here, x (ωs') is a convolutional sum corresponding to the phase CAD of the current address signal, h (95-i) is a filter coefficient, W (i) is a sample value in the domain of d · fs = 800 kHz, that is, FIG. It is sample data of each sample point of (a).

The spectrum envelope of the waveform shown in FIG. 10 (a) is as shown in FIG. 10 (c). This is convenient because unnecessary harmonic components are attenuated. Therefore, even in the filter output signal obtained according to the above equation, unnecessary aliasing components are sufficiently attenuated. By the way, in the general formula as described above, the sum of products is required 96 times, but in the waveform sample data of FIG. 10 (a), the same amplitude continues d = 16 times, so this is collectively calculated. By doing so, the number of sums of products can be reduced to 7, and the calculation scale can be reduced to about the same as 6 times in the embodiment of FIG.
That is, referring to FIG. 10A, the same waveform sample data continues for d = 16 times between the integer parts IAD-2 and IAD-1 of the address signal. IAD-1 and IAD as well as IAD and IAD + 1 as well as IAD + 1
Between IAD + 2 and IAD + 2, and similarly between IAD + 2 and IAD + 3. The same waveform sample data continues between IAD-3 and IAD-2 a number of times less than d = 16, and this can be done as a whole by one coefficient multiplication, and IAD + 3 and IAD + 4
The same is true during the period. Therefore, a convolution equivalent to the above equation can be performed with a total of seven sums of products.

Therefore, in the embodiment of FIG. 11, the filter coefficients of the maximum d = 16 orders corresponding to the same waveform sample data are summed in advance and treated as one filter coefficient data, The product sum is calculated by calculating the coefficient twice. For example, W · h0 + W · h1 + W ·
Instead of performing the calculation of h2 + W · h3 + W · h4 + W · h5 + W · h6 by seven times of multiplication, h0 + h1 + h2 + h3
+ H4 + h5 + h6 is prepared in advance, and W ·
The product sum is obtained by one multiplication of (h0 + h1 + h2 + h3 + h4 + h5 + h6). In FIG.
The same parts as those in the embodiment of FIG. 8 are denoted by the same reference numerals. The changed parts are the subtractor 16 and the master clock generator 2 of FIG.
2, a timing signal generation circuit 23, a subtractor 160 corresponding to the filter coefficient memories 25 and 26, a master clock generator 220, a timing signal generation circuit 230, and filter coefficient memories 250 and 260. As described above, in this embodiment, since the product-sum calculation is performed seven times in the filter calculation at one sample point, seven filter calculation time slots are required within one channel time slot, and the calculation timing is as shown in FIG. Is changed to. FIG.
In 2, the calculation cycle pulse CAC is generated in the cycle of the sampling frequency fs = 50 kHz similarly to the above, and this one cycle is divided into 8 to form the time division timings CH1 to CH8 of 8 channels. However, the time-division time slot of each channel is divided into seven to form seven filter operation time slots. The slot count data SLCTR is changed in this embodiment so as to distinguish the seven filter operation time slots 0, 1, 2, 3, 4, 5, 6 in one channel time slot. Master clock pulse MC
Slot count data SL for distinguishing 7 filter operation time slots 0 to 6 in one channel time slot by counting
CTR is obtained. Therefore, the master clock pulse MC
Frequency is changed to 2.8 MHz. Accordingly, the master clock generator 220 and the timing signal generation circuit 2
30 have been modified.

Further, since the sample data corresponding to the integer part of the address signal for 7 sample points is calculated by one filtering operation, the integer part IAD of the current address signal is obtained.
The extra sample data corresponding to the integer part IAD-3 three sample points before is required. Therefore, the subtracter 160 of FIG. 11 corresponding to the subtractor 16 of FIG. 8 subtracts 3 from the integer part IAD of the current address signal to obtain “IAD-
3 ”is calculated. The filter coefficient memories 250 and 260 have one of m = 96th-order filter coefficients.
Alternatively, a filter coefficient group value corresponding to the sum of a plurality of filter coefficients is stored in advance, and both memories 250 and 260 have the same storage content as described above. Memory 250,
The filter coefficient group value stored in advance in 260 is
As shown in Table 3, the sum of the filter coefficients of the order of d = 16 for all the orders (there are 81 sets, which are stored at addresses 16 to 96 as an example), and FIG.
Sum of less than d = 16 filter coefficients for the left end of the 0 (b) impulse response (this is 15 sets,
As an example, it is stored in addresses 1 to 15) and the sum of filter coefficients of d = less than 16 for the right end of the impulse response (also 15 sets, which are stored in addresses 97 to 111 as an example). , 111 sets of. Although "0" is stored in the address 0, this is merely a design matter caused by the selection means 27 for generating address data having the same configuration as that in FIG. The reference intermediate-order (47th-order) data is the total value of the filter coefficients of 47th to 62nd orders, which is stored in the address 63.

[0036]

[Table 3]

In this configuration, the data output from the adder 31 of the selecting means 27 corresponding to each value 0 to 6 of the slot count data SLCTR does not represent the order itself but is shown in Table 3 above. Addresses of such filter coefficient memories 250 and 260 are represented. Table 4 below shows the values of the coefficient memory address data output from the adder 31 of the selecting means 27 corresponding to the respective values 0 to 6 of the slot count data SLCTR according to the fractional part FAD of the address signal. In the table below, the integer parts IAD-3, IAD-2, I corresponding to 7 sample points respectively corresponding to
AD-1, IAD, IAD + 1, IAD + 2, IAD +
3 is also shown.

The concept of memory reading in this case is the same as in the above-described embodiment, the coefficient data (47 to 6) stored at the address 63 corresponding to the reference intermediate order (47th order).
The sum of the secondary filter coefficients) corresponds to the decimal part FAD of the current address signal, and each integer part IAD- of 7 sample points with respect to the decimal part FAD of the current address signal.
3, IAD-2, IAD-1, IAD, IAD + 1, I
Seven coefficient memory addresses separated from the reference address 63 by an amount corresponding to the distance of AD + 2 and IAD + 3 are randomly determined, and seven sets of filter coefficient group value data are obtained from the thus determined seven coefficient memory addresses. Are read out respectively. Further, similarly to the above, the filter coefficient group value data of adjacent addresses are read out in parallel from both memories 250 and 260, and interpolation is performed.

[0039]

[Table 4]

Referring to Tables 3 and 4, for example, if the fractional part FAD of the address signal is "6", SLCTR is 0.
When the SLCTR is 1, the coefficient memory address is 9, and the total data of the filter coefficients of the 0th to 8th order is read from the filter coefficient memory 250. When the SLCTR is 1, the address is 25 and the total data of the filter coefficients of the 9th to the 24th order. But SLCTR
When is 2, the address 41 is the total data of the filter coefficients of 25th to 40th orders, and when the SLCTR is 3, the total data of the filter coefficients is 57 and the total data of the filter coefficients of the 41st to 56th orders is
When the SLCTR is 4, the total data of the filter coefficients of the 73rd and 57th to 72nd order is 73, when the SLCTR is 5, the total data of the filter coefficients of the 89th and 73rd to 88th order is the SCLTR, and when the SLCTR is 6. At the address 105, the total data of the filter coefficients of the 89th to 95th orders are read out. In this way, the seven filter coefficient group values cover all the filter coefficients of orders 0 to 95.

As described above, according to the embodiment of FIG. 11, 0 is obtained in the domain of d · fs = 800 kHz as shown in FIG.
M = 96 for the sample data in the next hold state
The calculation for calculating the sum of products of all the following filter coefficients can be actually performed by only 7 times. Therefore, it is possible to perform an accurate filter calculation by using the waveform data of the spectrum structure in which the aliasing component is sufficiently attenuated even though the structure is simple. In the embodiments of FIGS. 8 and 11, in the digital filter arithmetic circuit 21, an arithmetic unit for controlling the address of the tone waveform memory 13 is provided instead of actually providing the delay circuit as a means for delaying the sample data. The delay circuit may be actually provided as shown in FIG. Further, although the musical tone waveform memory 13 storing a plurality of period waveforms is used as the musical tone waveform sample data generating means, the musical tone waveform memory 13 is not limited to this, and the musical tone waveform memory simply storing one period waveform or a musical tone waveform sample by frequency modulation calculation is used. Any method such as a method of generating data, a method of generating musical tone waveform sample data by amplitude modulation calculation, a method of converting address data to generate musical tone waveform sample data may be used. Also,
In the above embodiment, the tone generation and filter calculation of each channel are performed by the time division processing method, but this may be parallel processing. Also, not only the double tone generation method,
A single sound generation method may be used. Further, in the embodiment of FIGS. 8 and 11, the tone color selection code TC, the key code KC,
In order to select the waveform corresponding to the touch data TD in the tone waveform memory 13, the address signal generation circuit 1 is used without inputting these data TC, KC, TD into the tone waveform memory 13.
The address signal generation circuit 12 may be configured so that it can be selected according to the upper bits of the address signal.

In the case of an impulse response symmetrical about the intermediate order as shown in FIG. 4 (c) or FIG. 10 (b), it is not necessary to store the filter coefficients for all orders in the memory. Alternatively, only half of them may be stored and the filter coefficients of the same value at symmetrical positions may be shared by different orders. Also, filter coefficient memories 25 and 26
Alternatively, 250 and 260 may be one, and two adjacent coefficient data for interpolation may be read in a time division manner.
In that case, the frequency of the master clock pulse MC is doubled to form two time division time slots for interpolation in one filter operation time slot. 8 and 11
In the embodiment, the filter coefficient read from the filter coefficient memory is interpolated in four steps, but the number of interpolation steps is not limited to this. Further, the interpolation need not be performed.
Also, the filter operation format is not limited to the above-mentioned FIR type,
It may be an IIR (Infinite Impulse Response) type or another type. Further, the use of the digital filter used in the present invention is not limited to the use of aliasing noise removal as in the above embodiment, but may be other uses such as tone color control.
In that case, the filter characteristic is selected according to the tone color selection code TC, the key code KC, the touch data TD, and the like. That is, filter coefficient data corresponding to a plurality of filter characteristics is stored in the filter coefficient memory,
Tone selection code TC, key code KC, touch data T
A set of filter coefficient data that realizes a desired timbre is selected according to D or the like, and this is read according to the decimal part of the address signal.

When performing key scaling control of a tone color according to the key code KC or tone color control according to the touch data TD, the number of filter coefficient data stored in advance in the memory is reduced, and the filter coefficient data is stored. The filter coefficient may be densely generated by interpolating the data according to the key code KC or the touch data TD. Further, a plurality of tone signal generators according to the present invention are provided,
The tone signals generated in each series may be interpolated and synthesized according to the key code KC and the touch data TD. In the above embodiment, the musical tone waveform sample data is generated by the pitch asynchronous method in which the sampling frequency is always constant regardless of the pitch of the musical tone signal, but the musical tone waveform is generated by the pitch synchronous method in which the sampling frequency is synchronized with the pitch of the musical tone signal. The present invention can be applied even when generating sample data. Also, without reading all the waveform sample data stored in the tone waveform memory,
In the high tone range, the address signal may be controlled so as to be thinned out and read out once every two times or once every four times, for example.

[0044]

As described above, according to the present invention, musical tone waveform sample data is generated in accordance with the integer part of the address signal which changes corresponding to the pitch of the musical tone to be generated, and the decimal part of this address signal is generated. N coefficient data corresponding to the m-th order filter coefficient (n <m) are selected, and n sample points corresponding to the n coefficient data and the integer part of the address signal are generated. Since the m-th order filter calculation is performed on the n tone sampling points of the tone waveform data by using the tone waveform data of No. 3, the resolution of the tone waveform sample data actually prepared by the tone waveform data generating means is the same as that of the address signal. The advantage is that the circuit configuration can be simplified by the fact that it may be a relatively coarse one corresponding to the integer part, and in fact it is easy to use for n sample points less than m. Advantage that it is possible to achieve also a simplification of the construction of a digital filter circuit by the fact that it, by performing filter operation with respect to orders limited corresponding to the waveform data, with both enjoy,
Substantially, the filter operation is performed with respect to the accurate musical tone waveform sample data having the resolution of the fractional part of the address signal.
It is equivalent to performing the following precise filter calculation, so that various effects brought about by precise filter calculation for high resolution musical tone waveform sample data, for example, surely cutting unnecessary noise components, This has an excellent effect that the advantage that a high-quality musical tone signal can be obtained can also be enjoyed.

In addition, according to the present invention, since the filter coefficient supplying means has the filter coefficient interpolating means,
It is possible to easily generate m-th order filter coefficients that are larger than a predetermined order that can be generated by the filter coefficient generation means. Further, the filter coefficient supply means selects a filter characteristic corresponding to control data for controlling a musical sound from a plurality of filter characteristics having different characteristics, and an m-th order filter characteristic corresponding to the selected filter characteristic. In order to realize the above, n (n <m) coefficient data are selected and supplied from among m filter coefficients corresponding to respective consecutive orders according to the fractional part of the address signal. As mentioned above, in addition to being able to obtain the noise removal function by the m-th order precise filter calculation with a simple configuration as described above, it is also possible to realize the tone color control function according to arbitrary tone control data. It has an excellent effect.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a basic configuration of an embodiment of a musical tone signal generator according to the present invention.

FIG. 2 is a waveform diagram showing an example of waveform sample data at each stage based on general sampling frequency conversion theory.

3 is a diagram showing an example of a spectral envelope of each waveform sample data of FIG.

FIG. 4 is a diagram for explaining the principle of the digital filter arithmetic operation in the present invention.

FIG. 5 is a graph showing an example of actually measured amplitude-frequency characteristics of the FIR low-pass filter realized according to the present invention.

6 is an actual measurement diagram showing a spectrum of a sine wave signal that has passed through a low-pass filter having the characteristic shown in FIG.

7 is a block diagram showing an example of a filter coefficient supply unit and a digital filter calculation unit in FIG.

FIG. 8 is a block diagram showing a more specific embodiment of the musical tone signal generating apparatus according to the present invention.

9 is a timing chart showing the timing relationship of calculation and other operations in the embodiment of FIG.

FIG. 10 is a diagram for explaining a digital filter arithmetic operation according to another embodiment of the present invention.

11 is a block diagram showing another embodiment of the present invention related to FIG.

FIG. 12 is a timing chart showing the timing relationship of calculation and other operations in the embodiment of FIG.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Address signal generating means, 2 ... Musical sound waveform sample data generating means, 3 ... Filter coefficient supplying means, 4 ... Digital filter calculating means, 3a ... Filter coefficient generating means, 3
b, 27... selection means, 4a... delay means, 4b1 to 4b6
... Multiplier, 4c ... Adder, 12 ... Address signal generation circuit, 13 ... Tone waveform memory, 21 ... Digital filter operation circuit, 24 ... Filter coefficient supply circuit, 25, 26,
250, 260 ... Filter coefficient memory.

Claims (1)

[Claims] 1. An address signal generating means for generating an address signal comprising an integer part and a decimal part which change at a rate corresponding to the pitch of a musical tone to be generated, and a tone waveform sample according to the integer part of the address signal. A tone waveform data generating means for generating data, a control data generating means for generating control data for controlling a tone, and a filter characteristic corresponding to the control data are selected from a plurality of filter characteristics showing different characteristics. , A filter coefficient generating means capable of generating a filter coefficient having a predetermined order corresponding to the selected filter characteristic, and an m-th order filter having more than the predetermined order by interpolating the filter coefficient generated from the filter coefficient generating means A filter coefficient interpolating means capable of generating a coefficient is provided, and the filter coefficient is generated according to the decimal part of the address signal. By controlling the generation of the filter coefficient in the generating means and the interpolation in the filter coefficient interpolating means, n (from the m filter coefficients corresponding to each continuous order for realizing the m-th order filter characteristic ( However, by using the filter coefficient supplying means for selecting and supplying the coefficient data of n <m), and the n pieces of coefficient data and the musical tone waveform data for n sample points generated by the musical tone waveform data generating means, A musical tone signal generating apparatus comprising digital filter arithmetic means for performing m-th order filter arithmetic operation on musical tone waveform data for n sample points.