JP2708037B2 - Music signal generator - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [0001] The present invention relates to a digital filter.
Elimination of aliasing noise and tone control using filter operation
In particular, a relatively simple c
High-precision digital filter operation with hardware configuration
A tone signal generator capable of performing
You. [0002] 2. Description of the Related Art In an electronic musical instrument, a digital musical tone signal is used.
Purposes such as controlling the tone of a signal or removing noise
Uses digital filters. For example, synthesis
Constant sump regardless of the frequency of the musical tone
By sampling at the ring frequency
In a so-called pitch-asynchronous tone synthesis method for synthesis
Generally means that the tone frequency and the sampling frequency are
It is a numerical ratio, and as is clear from the sampling theorem,
There is a possibility that aliasing noise that is not
Need to be removed. Such pitch unequal
In order to remove aliasing noise contained in the period type tone signal
As a countermeasure, a digital
To pass the tone signal through the filter
(JP-A-61-90514). [0003] SUMMARY OF THE INVENTION For digital filters
For removing aliasing noise through musical signal
Is an accurate filter with sufficient filter order
Operation must be performed, which complicates the filter configuration.
There is a problem that becomes. In addition, for aliasing noise removal
Not limited to digital filters, other applications for tone control
The same applies to the digital filter described above. The invention
Has been made in view of the above points, and
When performing tone control and aliasing noise removal using
In addition, a simple configuration, yet accurate filter
A music signal generator that can perform calculations
It is something to offer. [0004] [MEANS FOR SOLVING THE PROBLEMS] A musical tone signal according to the present invention
The signal generator has a rate corresponding to the pitch of the musical tone to be generated.
Generates an address signal consisting of an integer part and a decimal part
Address signal generating means for generating the address signal;
A musical sound wave that generates musical sound waveform sample data according to several parts
Shape data generation means and control data for controlling musical sounds
Control data generating means for generating
Number of filter characteristicsFor a filter of predetermined order
Numbers, each of which is stored in the filterof
Select the filter characteristics corresponding to the control data from among
Corresponding to the selected filter characteristicSaidPredetermined order
Filter coefficientsFrom the decimal part of the address signal
Reads out multiple filter coefficients according to the predetermined upper bits of
YouFilter coefficient generating means and filter coefficient generating means
FromMultiple readFilter coefficientsThe address signal
According to the predetermined lower bits of the fractional part of the signalBy interpolating
M-order filter coefficients that are larger than the predetermined order
Filter coefficient interpolation means, and
Filter in the filter coefficient generating means according to the decimal part
Coefficient ofreadingAnd the interpolation by the filter coefficient interpolation means
Control to achieve m-th order filter characteristics
Among m filter coefficients corresponding to each successive order of
Select and supply n (where n <m) coefficient data
Filter coefficient supply means, and the n coefficient data
For n sample points generated by the musical sound waveform data generating means
Using the tone waveform data of
Digital sound waveform data for sample points
Filter operation means. The address signal generating means generates a musical tone to be generated.
Integers and fractions that change at a rate corresponding to the pitch of the sound
Generates an address signal. Needless to say, ads
The integer part of the address signal has lower resolution than the decimal part. Musical sound wave
The shape data generating means responds to the integer part of the address signal.
The tone waveform sample data is generated. Therefore, the musical tone
Musical sound waveform sample data prepared by waveform data generation means
May have a relatively coarse resolution. For example,
In conventional pitch-synchronous tone signal generators.
As shown in FIG.
Only the pull data may be prepared. Note that later
In this case, the tone waveform data
The sampling frequency of the sampled waveform data
It does not matter whether or not it is synchronized with the pitch. Suppose, pitch
Assuming that it is an asynchronous type, a coarse resolution similar to the pitch synchronous type
It is only necessary to prepare musical sound waveform sample data with Noh
is there. In the filter coefficient supply means,Duplicates with different characteristics
Filter function of a predetermined order for a number of filter characteristics
Number of each of the filter characteristics.
Select the filter characteristics corresponding to the control data from among
The predetermined order corresponding to the selected filter characteristic
The fractional part of the address signal from the filter coefficients
Reads out multiple filter coefficients according to the predetermined upper bits of
Filter coefficient generating means and filter coefficient generating means
The plurality of filter coefficients read from the
By interpolating according to the predetermined lower bits of the decimal part of the signal
M-order filter coefficients that are larger than the predetermined order
Function filter coefficient interpolation means,Of the above address signal
According to the decimal part,coefficient data corresponding to the m-th order filter coefficient
Data (where n <m) is selected and supplied.
In the digital filter operation means, the n coefficient data
And a sump generated by the musical tone waveform data generating means.
M-th order filter performance using the tone waveform data for
The calculation is performed on the tone waveform data for n sample points. This
As a result, the precision of the resolution with the fractional part of the address signal
Good tone waveform sample data (this is actually a tone waveform
Although not prepared by the data generation means), m order
Filter operation (actually for n orders)
Filter operation which is equivalent to performing
It can be performed. That is, the integer part of the address signal
The tone waveform data generated in response to
Filter processing at the filter
Therefore, waveform interpolation is performed at the resolution of the decimal part of the address signal.
It is possible to achieve the same processing as. this
Can contribute to the elimination of aliasing noise
You. Incidentally, for example, the conventional pitch asynchronous type
In a tone signal generator, the effects of aliasing
In order to eliminate as much as possible, make the resolution of the waveform as high as possible.
Therefore, it is common practice to increase the sampling frequency.
You. For example, with an accuracy of 1000-16000 divisions per waveform cycle
Preparation of waveform data is being performed,
The waveform memory requires a considerable capacity. in this way
Since a considerable capacity is required for the waveform memory, one cycle
Stores waveforms consisting of relatively short sections, such as waveforms, in memory
In the case of a tone signal generation method in which
If it is still relatively long, such as a continuous multi-period waveform,
Store the waveform consisting of the interval in the memory and read it out.
It is not suitable for a simple tone signal generation system. to this
On the other hand, sampling to the frequency of the musical tone to be synthesized
A so-called pitch-synchronous tone synthesis method that synchronizes frequencies
The tone frequency (pitch) and the sampling frequency
The component generated by folding to be harmonious is the musical frequency
And does not become noise.
Prepare waveform data with relatively coarse precision of about 4 divisions
No problem. Therefore, the ratio of continuous multi-period waveforms
Store a relatively long section of the waveform in memory and read it.
It is also suitable for a tone signal generation system that protrudes.
In contrast, in the present invention, the sampling frequency is
Pitch-synchronous, whether or not synchronized with the pitch
Prepares sample data of musical tone waveform with coarse resolution
Is good. Therefore, a relatively short section such as a one-cycle waveform
Memorize the waveform consisting of the interval in memory and read it out repeatedly
In the case of such a tone signal generation method,
Stores waveforms consisting of relatively long sections, such as multiple cycle waveforms
In a tone signal generation system where
In any case, the invention is suitable. As a result, tone waveform data is actually generated.
The resolution of the tone waveform sample data prepared by
Even if it is relatively coarse, corresponding to the integer part of the
Good means that the circuit configuration can be simplified
Advantages of being able to, and Actually, the tone waveform data for n sample points less than m
Filter operation on the limited order corresponding to the
That the digital filter circuit
Both advantages that the configuration can be simplified
While enjoying Effective filter operation is the decomposition of the fractional part of the address signal.
M for accurate sound waveform sample data with high performance
It is equivalent to performing the following precise filter operation.
This makes it possible to deal with high-resolution musical sound waveform sample data.
The various benefits provided by the precise filter operation
Points, for example, to ensure that unnecessary noise components are
The advantage of being able to obtain the tone signal of
Can be Further, according to the present invention, the filter
The number supply means has a filter coefficient interpolation means.
And the filter coefficient generation meansRememberFrom the specified order
M-order filter coefficients, The filter coefficient generating means
Without increasing the storage capacity of the filter coefficients inSimple
This has the effect of being able to occur in Also,
In the filter coefficient supply means, a plurality of filters having different characteristics are provided.
From the filter characteristics to the control data for controlling the tone
Select the corresponding filter characteristics and select the selected filter characteristics.
To realize the m-th order filter characteristic corresponding to the characteristic
Out of m filter coefficients corresponding to each order
According to the decimal part of the address signal, n (where n <m)
The coefficient data is selected and supplied.
M-order precise filter performance with a simple configuration
In addition to being able to obtain the noise removal function by arithmetic
Also implements tone control functions corresponding to arbitrary tone control data.
It has an excellent effect that it can be realized.Furthermore,
The tone signal generating apparatus according to the present invention provides
From the integer part and the decimal part that change at the rate corresponding to the pitch
Address signal generating means for generating an address signal,
The tone waveform sample data is determined according to the integer part of the address signal.
Sound waveform data generating means for generating data,
The filter coefficients are stored before Decimal of address signal
Multiple different order fills according to the predetermined upper bits of the
Filter coefficient generating means for reading out filter coefficients;
A plurality of different orders read from the coefficient generator.
The filter coefficient is determined by a predetermined lower bit of the decimal part of the address signal.
Interpolated according to the order, and filters with orders higher than the predetermined order
Filter coefficient interpolation means for generating filter coefficients;
According to the filter coefficient generated by the filter coefficient interpolation means
The tone waveform data generated by the tone waveform data generating means.
Digital filter operation that performs filter operation on data
Means. According to this, the filter
Since it has a number interpolation means, the filter coefficient generation means
Filters of orders greater than the number of stored filter coefficients
Filter coefficients in the filter coefficient generation means.
Can easily occur without increasing the storage capacity of numbers.
It has the effect of cutting. [0008] BRIEF DESCRIPTION OF THE DRAWINGS FIG.
The embodiments of the present invention will be described in detail. FIG. 1 shows one embodiment of the present invention.
FIG. 1 is a basic block diagram showing an embodiment, where 1 is an address signal.
Signal generating means, 2 means musical tone waveform data generating means, 3 means fill
And 4 are digital filter operation means.
is there. As described above, the address signal generating means 1 has an integer part.
Generates an address signal consisting of IAD and decimal part FAD
Generated at a rate corresponding to the pitch of the
The tone waveform data generating means 2 outputs the address signal
Generates tone waveform sample data according to the integer part IAD
And the filter coefficient supply means 3
Filter coefficient of the m-th order according to the fractional part FAD of the
In addition, n units (where n <m) are selected and supplied.
The digital filter operation means 4 uses the n filters.
Luther coefficient and n generated by the tone waveform data generating means.
Using the tone waveform data for the sample points,
Performs a filter operation on musical sound waveform data for n sample points.
Is Umono. Note that in this example, the sampling frequency
fs is constant regardless of the pitch of the tone to be generated
(That is, pitch asynchronous type), digital filter operator
In stage 4, fs / 2 is used to remove aliasing noise.
Realizes low-pass filter characteristics with cut-off frequency
Shall be. The principle of the tone signal generator according to the present invention
In order to explain, first, the digital filter
The data operation will be described with reference to FIGS. FIG.
Is a waveform diagram based on the general sampling frequency conversion theory
FIG. 3 shows an example of the spectrum envelope.
This is an example. FIG. 2A shows the sampling frequency f.
An example of musical tone waveform sample data sampled by s
It is a figure showing about some sample points. This figure 2
The tone waveform sample data of FIG.
Sampling timing of M times frequency (M · fs) of several fs
Input to a digital filter that operates
You. In that case, one cycle ts = 1 of the sampling frequency fs
/ Fs contains M times one cycle ts / M of frequency M · fs.
However, one sample of the tone waveform sample data shown in FIG.
All M filter operation timings during the sample time.
M data in one cycle ts of fs without generating sample data
Only one of the filter operation timings
Generates sample data and the remaining M-1 timings
Let the sample value be "0". Thus, the frequency M
・ One of the sampling timings of fs
At the timing, the sample of the tone waveform sample data of FIG.
Generate a pull value, and M-1 timings for the remaining M
FIG. 2 (b) shows the sample value set to "0".
You. The sound waveform sample data shown in FIG.
Is input to a digital filter, and a sample of a frequency M
A filter operation is performed according to the ring timing. Toes
Each sampling point whose sampling period is ts / M
Musical sound waveform data (However, one sample
Has a valid sample value at the timing of
At M-1 timings, the sample value is "0".
To calculate filter coefficients of each order for
It is. Then, as a filter operation output, FIG.
As shown in the figure, each of the sampling timings of M · fs
The tone waveform data for which sample values are densely generated corresponding to
Obtainable. This means that each sample of frequency M
Filter operation is performed at the sampling timing,
For each filter timing,
The convolution of the number and its corresponding sample value
This is because an output signal is obtained. As shown in FIG.
The desired sound waveform data with dense sample values
Re-sampling at the pulling frequency
Tone waveform sample data converted to the sampling frequency of
Data can be obtained. FIG. 2 (d) shows such a desired
By re-sampling at the sampling frequency.
An example of the obtained musical tone waveform sample data is shown. This place
The desired sampling frequency to resample
Is M · fs / N, the waveform data of FIG.
Is calculated once every N times of the sampling timing of M · fs.
What is necessary is just to resample at a ratio. By the above processing, the sampling cycle
Music waveform data with wave number fs is sampled by M-fs / N
Sampling frequency. The figure
In the drawing, M = 4 and N = 3 are shown. FIG. 3A is a diagram.
An example of the spectrum envelope of the waveform 2 (a) is shown.
Fig. 3 (b) shows the spectrum envelope of the waveform of Fig. 2 (b).
FIGS. 3A and 3B show an example of a loop.
The shape of the spectral envelope is the same but the level is
(B) is 1 / M of (a). This was
Substantial sample values included in the convolution sum ("0"
Number of sample values) is 1 / M of the original number
It is. FIG. 3C shows an example of a digital filter characteristic.
Where fs / 2 is the cutoff frequency and
This is a low-pass filter. FIG. 3 (d) shows this roper.
Filter characteristics.
Shows the spectral envelope of the waveform of FIG. 2 (c).
It is. In this case, the return by the sampling theorem is M
It occurs at a frequency of fs / 2, which is quite high
Therefore, noise does not occur at the time of resampling. FIG.
(e) is one of the spectrum envelopes of the waveform of FIG.
It is a figure showing an example. Resample the output of the filter.
Just by ringing, the signal level is reduced to 1 / M as described above.
Since the level remains down, when resampling
By multiplying the level of the waveform data in FIG.
And return to the original level. In the above, the digital filter is used as a sampler.
It is used for converting the tone frequency
Setting a desired pitch and generating a tone signal directly
Has nothing to do with. In contrast, the present invention
When generating a tone signal with a pitch of
Digital filter operation is used.
is there. That is, the pitch corresponding to the pitch of the musical tone to be generated
Address signal which changes with the
Generated according to the address signal.
Digital data based on the above principle
The filter filter operation is used. Address information
The signal consists of an integer part IAD and a decimal part FAD.
In the data generating means 2, the integer part IA of the address signal
The tone waveform sample data is generated according to D. Paraphrase
Then, the musical tone that can be generated by the musical tone waveform data generating means 2
The waveform sample data corresponds to the integer part IAD of the address signal.
It only corresponds to the resolution. Address signal integer
Generated from musical sound waveform data generating means 2 corresponding to section IAD
FIG. 4 (a) shows an example of the sample data of the musical tone waveform to be generated.
You. The decimal part FAD of the address signal is
At the sample point specified by the integer part IAD of the signal
2 shows a finer phase between adjacent sample points.
For example, in the integer part IAD and the decimal part FAD of the address signal,
Therefore, the current address value indicated by 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.
High quality sound wave signal free of harmonic distortion, subharmonic distortion and noise
Resolution of the fractional part FAD of the address signal to obtain the signal
In other words, corresponding to the CAD phase, the musical sound waveform sample
It is desirable to seek data. In the present invention, the aforementioned
Using digital filter operation based on the principle
This can be done with a simple configuration and with a sampling clock.
Clock frequency without special speed up.
I have to. Therefore, the filter coefficient supply means 3
Is an mth-order filter according to the decimal part FAD of the address signal.
Of the coefficient data corresponding to the data coefficients (where n <
m) is selected and supplied. Figure 2 above
As is clear from the explanations related to (b) and (c),
Sample data corresponding to all of the filter coefficients
One time per M orders without needing to supply data
Supply valid sample data at and for the remaining orders
May set the sample value of the sample data to “0”,
Even so, precise filtering is performed using multiple-order filter coefficients.
Filter operation is performed, so that a good tone signal can be obtained.
Can be. Based on this, a similar idea
In this invention, a digital filter operation based on
I have to. However, the general sampling frequency mentioned above
In processing based on number conversion theory, high-speed filtering
According to the arithmetic timing (frequency M · fs),
Must actually perform the filter operation on the filter coefficients.
Absent. Because the correspondence between order and sample data
Changes with time due to the change of operation timing.
Therefore, even if there is sample data with a sample value of "0"
It is clear that the product with the filter coefficient is "0"
It is unclear what order it holds
Because it is. Therefore, conventional general sampling
In the processing based on the frequency conversion theory, FIG.
A tone waveform sample data with a sampling frequency fs
As shown in FIG. 2 (b), the data
sampling once per M clocks of fs and M clocks
At sampling timing of M-1 times per sample,
Sampling by inserting “0” as pull value
Is converted to data of frequency M · fs, and this is converted to frequency M · fs.
Input to the operating digital filter, this digital
Filter coefficients of all orders set in the filter
The filter operation must be actually performed. On the other hand, in the present invention, the sample value
“0” sample data and corresponding filter coefficients
Noting that the operation of is virtually unnecessary,
The feature is that it is possible to omit the operation for
You. As a result, calculations can be performed at high-speed filter
The operation circuit is simplified while eliminating the need to perform
In addition, using a sufficiently large number of
By enabling the
Simultaneous simplification of circuit configuration and high-precision filter operation
To be realized. First, the functions shown in FIGS.
Similarly, the tone waveform sample data of FIG.
A sample with a frequency (M · fs) that is M times the sampling frequency fs
Input to a digital filter that operates at the
Assuming that one cycle ts of the sampling frequency fs
= 1 / fs contains M cycles of one cycle ts / M of frequency M · fs
Assuming that the tone waveform sample data shown in FIG.
All M filter operation timings in one sample time
Without generating sample data during one cycle ts of fs.
One of the M filter operation timings
Only sample data is generated, and the remaining M-1
It is assumed that the sampling value is set to “0”. This
, The sampling timing of the frequency M · fs
4 (a) at one timing per M pieces.
Generate sample values of sample data, and
At M−1 timings, the sample value is set to “0”.
This is shown in FIG. Where the decimal part of the address signal
FAD division number (1 divided by the minimum unit of decimal FAD
Indicates the number, for example, the minimum unit of the decimal part FAD is
If 0.1, the number of divisions is 10), and d is M.
You. FIG. 4B shows a musical sound waveform sample data as shown in FIG.
Consider performing m-th order filter operation on
(M is an arbitrary number determined when setting the filter characteristics
A), and as described with reference to FIG.
Sampling timing of frequency M · fs using all data coefficients
When the filter operation is performed by the filtering, it is as shown in FIG.
A filter output signal having a similar precision resolution is shown in FIG.
It can also be obtained for musical sound waveform sample data. But,
Then, as mentioned above, the digital filter
Must be operated at the sampling frequency M
In addition, all of the following
Actually perform the filter operation on the number of filter coefficients
Is disadvantageous because it must be adopted in this invention
do not do. Instead, we use the sample values
“0” sample data and corresponding filter coefficients
Noting that the operation of is virtually unnecessary,
Omit the calculation for
The calculation is performed without increasing the ring frequency.
It is. The elimination of such unnecessary operations is in accordance with the present invention.
Then, the address signal is divided into an integer part IAD and a decimal part FAD.
And divide the tone waveform sample data according to the integer part IAD.
Generated, but a filter for digital filter operation
Select and supply coefficients according to the fractional part FAD
It is realized by doing. In other words, the musical sound waveform sample
Data generation is relatively coarse resolution according to the integer part IAD
The digital filter operation is performed in accordance with the decimal part FAD.
The feature is that it is performed at a precise resolution.
For details, fill in according to the fractional part FAD of the address signal.
By selecting the data coefficient, the integer part IAD and the decimal part F
Current phase CA of the address signal specified by AD
At fine sampling timing related to D
The integer part I as if a filter operation had been performed.
Each sample of coarse sampling timing corresponding to AD
Filter coefficients corresponding to the tone waveform sample data
Can be determined. In other words,
In accordance with the fractional part FAD of the address signal, the integer part IAD
Sample data and filter coefficients for each corresponding sample point
4 is determined, and as a result, FIG.
As a sample value under the assumption as shown in (b)
Corresponds to dense sampling timing with “0” inserted
The order of the filter coefficient can be specified, and the order
Regarding, the product is set to “0” without performing any operation.
For sample data with substantial values
Calculate filter coefficients corresponding to discrete orders
It will be good. Thus, according to the present invention, the m-th order
In the filter operation of (1), the filter coefficient supply means 3
The filter coefficients to be supplied correspond to the m-th order filter coefficients.
Not all coefficient data, but n of them (where n <
m) according to the fractional part FAD of the address signal.
You just have to pay. Here, the assumed digital filter operation
M at the sampling frequency M · fs of
d (the number of divisions of the fractional part FAD), and FIG.
The sample data shown in FIG. 4B is a tone waveform sample shown in FIG.
Virtual file obtained by dividing one sample interval of file data by M = d
Sampling at only one of the data operation timings
Data is generated, and at the remaining d-1 timings,
Since the sample value is “0”, the ease of FIG.
Provisional data obtained by dividing one sample interval of sound waveform sample data by d
In one of the virtual filter operation timings
Only have virtual sample values and their virtual filter
Actual calculations only need to be performed for
You. Therefore, the above n is determined by a relationship of n = m / d.
Can be added, and n orders sequentially separated by an interval of d can be added.
What is necessary is just to select according to the decimal part FAD of the address signal. Ma
Also, the sampling frequency of the assumed digital filter operation
Once every d times at the timing of the wave number M · fs = d · fs
Only needs to be performed, and must be performed at other times.
No, that is, the sample data unit delay is
The sampling frequency does not need to be set at M · fs = d · fs,
At the frequency fs.
Precise filtering with the sampling frequency M-fs = d-fs
Despite performing the filter operation, the actual filter operation is
It should be done at a low sampling frequency fs.
You. In connection with this, a sump as shown in FIG.
Data, but actually
Once per d clock with the pulling frequency M-fs = d-fs
Sampling and insertion of the sample value “0” d-1 times
There is no need to perform any processing, and the sampling frequency f
s generated according to the integer part IAD of the address signal according to
The sample data obtained may be used as it is. This will be further described with reference to the drawings.
The bright filter operation is performed on the sample data as shown in FIG.
Data corresponding to the current phase CAD of the address signal.
This is equivalent to performing a filter operation. In this case, the figure
4 (b) and each of the m-th order filter coefficients
An example of a correspondence relationship with the order (0th order to m-1th order) is as follows.
It is as shown in FIG. FIG. 4C shows an m-order FIR.
(Finite impulse response) filter low-pass filter
Envelope shows an example of an impulse response
Things. The frequency domain of this FIR filter is M
-Fs = d-fs, and cut-off of low-pass filter characteristics
Is the return noise related to the sampling frequency fs.
Set to fs / 2 to remove noise. This impulse
A predetermined reference order (eg, center order)
Corresponds to the current address signal phase CAD
And In the filter operation, each sample shown in FIG.
Convolution of the pull data with the filter coefficients of FIG.
Is required. In that case, the frequency domain of d
Of the sample data shown in FIG.
No calculation is performed for the element with the default value "0". Sand
In FIG. 4B, a sample having a valid sample value
Of the data and the corresponding order filter coefficients
Just do. From the relationship of n = m / d, m-order convolution
Sample data with valid sample values in
Is n. In the illustrated example, m = 96, d = 16, n
= 6. Current address signal phase
It is obtained by the filter operation executed corresponding to CAD.
The convolution is shown by a solid line in FIG.
You. In FIG. 4D, a part is indicated by a broken line.
The convolution sum, that is, the filter output signal has a frequency of d
In several domains, it occurs as densely as shown in FIG.
Is done. Note that, as before, this convolution is actually
Performed only on n = 6 sample data
Therefore, the level of the filter output signal becomes 1 / d = 1 /
Drops to 16. To address this, the filter output signal
May be multiplied by d = 16. Alternatively,
Even if you increase the level of the filter output signal by d = 16
And a filter section with d = 16 times the original value
The operation may be performed using a number. FIG. 4 (e) shows the impulse response of FIG. 4 (c).
Example of amplitude-frequency characteristics of FIR low-pass filter with
It is shown. The frequency domain component cut is -80
It has been confirmed that it is attenuated below dB. this is
The filter characteristics are quite high precision. Figure 5 shows such
In the actual measurement diagram of the amplitude-frequency characteristics of the FIR low-pass filter
is there. FIG. 6 shows the result of passing through the FIR low-pass filter having the characteristic shown in FIG.
FIG. 7 is an actual measurement diagram of the spectrum of the sine wave signal obtained. from here
As is apparent, noise components other than the fundamental wave are reliably -8.
It is attenuated below 0 dB. 4 (a) to 4 (c)
In summary, each order of the m-th order filter coefficient (0th order to m−
(Primary) is the resolution of the fractional part FAD.
Corresponding to a given reference order (eg,
For example, the central order k) is the fractional part FA of the current address signal.
Corresponding to the position of D (that is, the position of the current phase CAD)
I have. For the fractional part FAD of this current address signal
A quantity corresponding to the distance between each integer part IAD for n sample points
Each of the n orders separated from the reference order by
The address part of the current address signal.
N filters corresponding to n orders determined according to
The filter coefficient is supplied by the filter coefficient supply means 3.
You. The reference next to the current phase CAD of the address signal
Let the number be the central order k (for example, when m = 96,
When the order is 0th to m-1 = 95th, the center order is k =
47) and n = 6, then “n sample points
Each integer part IAD of the minute address signal "
N = 6 samples before and after the integer part IAD of the address signal
The integer part of the point, ie, 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), the n = 6 orders separated from the general
Are determined in the following manner. [0019] Order corresponding to IAD-2: k-FAD-2d Order corresponding to IAD-1: k-FAD-d Order corresponding to IAD: k-FAD Order corresponding to IAD + 1: k-FAD + d Order corresponding to IAD + 2: k-FAD + 2d Order corresponding to IAD + 3: k-FAD + 3d Of course, the above is only an example, and other definitions are possible.
In addition, even if it is defined as above, how to determine k, d, n
Depending on the situation, various other specific solutions may occur.
is there. For example, if k = 46 even under the same conditions,
3 sample points before the integer part IAD of the current address signal
The order corresponding to IAD-3 is defined as k-FAD-3d.
May need to be decided under filter
The coefficient supply means 3 includes a table as described above,
Alternatively, do not provide an arithmetic circuit that executes the above expression.
Thus, according to the decimal part FAD of the current address signal, n
Are determined, and coefficients corresponding to the determined n orders are determined.
Supply data respectively. This filter coefficient supply means 3
Some examples of the internal configuration of the digital filter operation means 4
This is shown in detail in FIG. Filter coefficient supplier
Stage 3 generates filter coefficients of order m (order 0 to m-1)
Filter coefficient generating means 3a and the m-th order filter
N of the coefficients are the values of the fractional part FAD of the address signal.
And selecting means 3b for selecting according to. Hand of choice
Stage 3b comprises, for example, the table described above,
N sample points according to the value of the fractional part FAD of the address signal.
Each integer part IAD-2, IAD-1, IAD, IAD +
Order k-FAD corresponding to 1, IAD + 2, IAD + 3
-2d, k-FAD-d, k-FAD, k-FAD +
d, k-FAD + 2d and k-FAD + 3d
And the filter corresponding to each discrete order
H (i-2d), h (i-d), h (i), h (i + d), h (i + 2
d) and h (i + 3d) are selected and output. In FIG. 7, the digital filter operation means 4
Is composed of FIR filter and generates musical tone waveform data.
Generated from stage 2 in response to the integer part IAD of the address signal
Clock of sampling frequency fs
With delay means 4a for delaying by pulse φ (fs)
I have. This delay means 4a is provided for each integer part for n sample points.
IAD-2, IAD-1, IAD, IAD + 1, IAD
+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)
Are input to the multipliers 4b1 to 4b6, and given from the selection means 3b.
The obtained filter coefficients h (i-2d), h (i-d), h (i), h (i
+ d), h (i + 2d) and h (i + 3d). Multiplier 4b1 ~
The output of 4b6 is added by the adder 4c.
Output as FIR filter output. The current sump
Sample data X (Ii) to be handled as
a, the file provided from the selection means 3b
The filter coefficients h (i-2d) to h (i + 3d) are appropriately delayed accordingly.
To the multipliers 4b1 to 4b6. here
Then, the above-described relationships of FIGS. 4A to 4C are arranged in another expression.
Then, the division number d of the decimal part in the address signal (the above-described number d)
In the example, n = m / d according to d = 16).
(N = 6 in the above example), and the n
The filter coefficients are based on n orders that are sequentially separated at intervals of d.
Corresponding to the decimal part of the current address signal
Each of the n orders is determined according to the value of FAD.
Is determined according to the fractional part FAD of the current address signal.
N filter coefficients corresponding to the n orders
It is supplied by the coefficient supplying means 3. Thus, according to the present invention, the m-th order fill
In the convolution operation of the data
Where the operation must be performed on the data, n = m
It is only necessary to perform an operation on / d coefficient data.
The calculation scale can be reduced to 1 / d. And actually
The sampling frequency in the calculation of fs is fs
Digital filter operation with high resolution of d · fs
The result is equivalent to Next, check the circuit configuration scale.
Improve the accuracy of the filter operation without enlarging
Will be described. Using the mth order filter coefficients,
To perform a filter operation of q times the order of
Is the difference between adjacent filter coefficients of the m-th order.
Interpolation of resolution q is performed.
The coefficients may be generated densely. Such q
Equivalent in digital filter operation by double interpolation
The typical sampling frequency is high resolution of qdfs
And the filter order is q · m order,
The accuracy of the calculation can be considerably improved. Then this
A more specific embodiment of the invention will be described with reference to FIG.
I do. 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. And the m-th order
Between adjacent ones of the filter coefficients, each having a resolution of q = 4.
The filter coefficients for the qm order = 384 order
Are generated densely. Therefore, the address signal
The decimal part FAD of the signal is basically the number of divisions d = 16 = “2
4 bits of data corresponding to
Furthermore, the lower 2 bits are added to the interpolation step of resolution q = 4.
To instruct the user. Therefore, in this embodiment,
If the decimal part FAD of the address signal is 6-bit data,
Become. The sampling frequency is fs = 50 kHz.
It is fixed and generates tone signals asynchronously with pitch.
Swelling. In this embodiment, the digital filter
Filter is used to remove aliasing noise as before.
It is configured as an FIR filter with pass filter characteristics.
You. The keyboard 10 specifies the pitch of a musical tone to be generated.
With multiple keys for Pressed on keyboard 10
Key is detected by key assigner 11 and corresponds to pressed key
Tones that should generate multiple tone generation channels
. An example of the number of tone generation channels
And each channel has a common musical tone generating means.
Established by time sharing
You. The key assigner 11 stores the key assigned to each channel.
Key code KC, key-on signal KON and key-on pal
KONP is time-divisional according to channel timing
Output to The address signal generating circuit 12 is a key assigner
Key code KC and key-on pulse KONP from 11
To the pitch of the key assigned to each channel.
Address signal that changes at a
Occur in a time-sharing manner in response to the This address signal
Consists of the integer part IAD and the decimal part FAD as described above.
ing. The integer part IAD is, for example, 18-bit data
Or more than one period prepared in the tone waveform memory 13
Specifies continuous sample points of a musical sound waveform consisting of
Yes, the decimal part FAD is 6-bit data as described above.
You. As an example, the musical tone waveform memory 13
Multi-period waveform data and multi-period waveform data for the sustain part
Is stored, the address signal generation circuit 12
Now, the attack is triggered by the key-on pulse KONP
Part of the multi-period waveform data is read once,
Address so that multiple period waveform data of
Signal. Note that the address signal generation circuit 12
Address signal generation method repeats frequency number
Calculation method, variable frequency division method, or note clock
Any method, such as a method for counting
No. In connection with the keyboard 10, the touch detection device 14
And detects a touch of a pressed key. Musical sound waveform
As an example, the memory 13 has a multi-period waveform as described above.
The sample data of
Supports pull data that can be selected by the tone selection circuit 15
And a plurality of sets are stored. In addition, the key of the tone corresponding to the pitch
ー Scaling control or tone control according to key touch
In addition, multiple sets of waveform sample data are stored for
May be. For that, the tone selection code TC, key
The code KC and the touch data TD are stored in the tone waveform memory 13.
The waveform to be read is selected according to these
This is read according to the integer part IAD of the address signal.
Is spilled out. Generated from address signal generation circuit 12
The data of the integer part IAD of the address signal is stored in the tone waveform memory.
Thirteen phase address inputs, which directly participate
Not through the computing unit (subtractor 16, adder 17)
Join. The arithmetic units 16 and 17 are provided with digital filters.
Sample data delay means (see 4a in FIG. 7)
This is equivalent to the above. Sand
In this embodiment, the tone generated from the tone waveform memory 13 is used.
By actually delaying the sample data, n (= 6)
Each integer part IAD-2, IAD-1, IA for sample points
D, IAD + 1, IAD + 2, IAD + 3
Sample data, but not address signals.
Arithmetic units 16 and 17 perform time division on data of several IADs
-2, -1, 0, +1, +2, +3
, Each integer part IAD-2 for n (= 6) sample points,
IAD-1, IAD, IAD + 1, IAD + 2, IAD
+3 address data is generated in a time-sharing manner.
By reading the memory 13 by using the n (= 6)
Each integer part IAD-2, IAD-1, IA for each sample point
D, IAD + 1, IAD + 2, IAD + 3
They have obtained sample data. The calculation timing for this will be described in detail.
This is as shown in FIG. 9, where CAC is the sampling frequency.
Calculation cycle pulse generated at a period of several fs = 50 kHz
This period is divided into eight and time division of eight channels is performed.
Timings CH1 to CH8 are formed.
Each time-division time slot is divided into six parts to fill the sixth order
A data operation time slot is formed. Filter operation tie
One cycle of the memory slot is one cycle of the master clock pulse MC.
And the master clock pulse MC is modulated modulo 6
1 channel tie by counting with
Filter operation time slots 0,
Slot count data that distinguishes 1, 2, 3, 4, and 5
SLCTR is obtained. SMC is the filter operation cycle
Pulse, one cycle equals one channel time slot
Synchronized. When the calculation cycle pulse CAC is 50 kHz
If there is, the filter operation cycle pulse SMC is 40
0 kHz and the master clock pulse MC is 2.4 MHz.
You. Each pulse and count data described above are
Generated from the clock generator 22 and the timing signal generation circuit 23
Is done. The subtractor 16 calculates the integer part of the current address signal.
The value IA of the integer part of the sample point two sample points before the IAD
Subtract 2 from IAD to find D-2
is there. The data IAD-2 thus obtained is added to the adder 1
7 and the slot count data SLCTR is added.
Is calculated. This slot count data SLCTR is
As shown in FIG. 9, within one channel time slot
0, 1, 2, 3, 4, 5
6 filter operation types in channel time slot
6 slots corresponding to 0, 1, 2, 3, 4, 5
Each integer part IAD-2, IAD-1, IAD, for the pull point
Address data of IAD + 1, IAD + 2, IAD + 3
Is generated from the adder 17 in a time division manner. According to this
Thus, each integer part I for these six sample points is stored in the memory 13.
AD-2, IAD-1, IAD, IAD + 1, IAD +
2, the sample data corresponding to IAD + 3
Is read. The sample data read from the memory 13
The data is input to a multiplier 18 for multiplying the filter coefficient. H
The filter coefficient is filtered according to the fractional part FAD of the address signal.
It is supplied from the filter coefficient supply circuit 24 as described later.
The output of the multiplier 18 is input to the accumulator 19,
The convolution is required. This accumulator 19 is
At the timing of the star clock pulse MC (that is,
Accu) for each step of the count data SLCTR)
The filter operation cycle pulse SMC
Cleared at the timing. Clear accumulated value
Immediately before the operation, the convolution sum obtained in this operation is latched.
Latched by circuit 20. These subtracter 16, adder
17, multiplier 18, accumulator 19, latch circuit 2
0 is an FIR type digital filter operation circuit 21
Is equivalent to The filter coefficient supply circuit 24 has m = 96 order
Filter coefficients (0th order to 95th order)
Data coefficient memories 25 and 26 and the 96th-order filter coefficient
Of the address signals correspond to the value of the fractional part FAD of the address signal.
Selection means 27 and an interpolation circuit 28
I have it. The two series of filter coefficient memories 25 and 26
Are exactly the same,
To read in parallel two filter coefficients adjacent to
Are provided with two series of filter coefficient memories 25 and 26.
I have. The filters stored in the filter coefficient memories 25 and 26
The impulse response of the filter coefficient is, for example, as shown in FIG.
It is like the one shown in
The filter characteristics are shown in FIG. 4 (e) or FIG.
Low-pass filter characteristics, and the sampling frequency
Cut off fs / 2 = 25 kHz, half of fs = 50 kHz
Frequency. 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, and the determined order
From the filter coefficient memories 25 and 26 as dress signals,
Filter coefficients h (i-2d), h (i-d), h (i), h (i + d), h (i +
2d) and h (i + 3d) are selectively read out.
The subtracter 29, multiplier 30,
An adder 31 is provided. The address signal of the
Input the upper 4-bit data of the decimal part FAD and enter "15-
FAD "is subtracted.
Input the data SLCTR, and enter “16 × SLCTR”
Perform multiplication. The outputs of the subtracter 29 and the multiplier 30 are added to the adder 3
1 and the above-described order k-FAD-2d, k-FAD
-D, k-FAD, k-FAD + d, k-FAD + 2
Data indicating d, k-FAD + 3d is output. S
Corresponding to each value 0-5 of lot count data SLCTR
The output of the adder 31, that is, the determined order is as follows.
You. The following table shows the integers for each of the six sample points.
Part values IAD-2, IAD-1, IAD, IAD + 1,
IAD + 2 and IAD + 3 are also shown. [0027] [Table 1] Assuming that k = 47 and d = 16,
K-FAD-2d, k-FAD-d,
k-FAD, k-FAD + d, k-FAD + 2d, k-
It is understood that FAD + 3d is as shown in the above table.
U. Therefore, the arithmetic circuit configuration in the selection means 27 is
Generally, “k−FAD + (SLCTR−2) × d” =
Calculation of “47−FAD + (SLCTR−2) × 16”
What is necessary is just to comprise so that an expression may be performed. Output of adder 31
Is input to the filter coefficient memory 25 as it is
The adder 32 adds 1 to the filter coefficient memory 26
Is input to Thus, two fills of adjacent orders
Data is read from the filter coefficient memories 25 and 26.
Will be issued. These two filter coefficient data are stored in the interpolation circuit 2
8 and the lower 2 bits of the fractional part FAD of the address signal.
4-step interpolation characteristics (for example, a straight line)
(Interpolation characteristic). Thus, the memories 25 and 26
Actually stores only m = 96th order filter coefficients.
Not available, but q · m = 4 × 96 = 384th order by interpolation
This is equivalent to densely preparing the filter coefficients of. Supplement
The output of the inter-circuit 28 is input to the multiplier 18. What
It should be noted that, by quadruple interpolation, the interpolation circuit 28
Output corresponding to each timing of the
The order of the filter coefficients used varies substantially as shown in the table below.
Has been updated. [0029] [Table 2] In Table 1, FAD is modulo 16 (d =
16), but in Table 2, FAD is modulo 64 (d = 64)
It is. Filter operation output from latch circuit 20
The force signal is supplied to a multiplier 33, and the envelope generator 3
4 is multiplied by the amplitude envelope signal.
Envelope generator 34 has key code KC and tone selection
Encoder controlled according to code TC and touch data TD
Generates a envelope waveform signal based on the key-on signal KON
You. The output of the multiplier 33 is input to the accumulator 35.
And sum the sample data of all channels.
You. This accumulator 35 is a filter operation cycle
At the timing of the SMC
Accumulate and calculate cycle pulse)
It is cleared at the timing of CAC. Accumulate value
Just before clearing
The total of the data is latched by the latch circuit 36. Luck
Sampling frequency of the tone signal output from the
The number is fs = 50 kHz, and the digital filter
The cutoff frequency is fs / 2 = 25 kHz on the road 21
By filtering the low-pass filter characteristics
Aliasing noise is reliably removed. Latch circuit 36
Output signal is analog-converted by the digital / analog converter 37.
The signal is converted to a sound signal and reaches the sound system 38. Ma
The output signal of the latch circuit 36 is, for example, reverb,
Digital effects to add kor and other musical sound effects
After being input to the circuit 39 and being given a tone effect,
Is converted into an analog signal by the
A sound system 38 is provided. As described above, the level of the filter output signal is
Dealing with bell drop to 1 / d = 1/16
Is the filter stored in the filter coefficient memories 25 and 26
The level of the coefficient is increased 16 times in advance, or the interpolation circuit 28
The coefficient data given to the multiplier 18 by 4 bits higher
I just need to shift to Next, as compared with the configuration of FIG.
Filter operation without significantly expanding the circuit configuration scale
An embodiment for improving the accuracy of the above will be described. Fig. 8
In the embodiment, the sampling frequency as shown in FIG.
Sample of sound waveform sample data with wave number fs = 50kHz
Upper d.fs = 16 x 50 = 800 kHz domain sampling
In order to set the frequency, as shown in FIG.
At a rate of d-1 times per d times of 800 kHz clock,
Processing is performed assuming that the pull value “0” has been inserted, and sampling is performed.
Filter operation is omitted for the order corresponding to the default value “0”.
Abbreviated. In contrast, the implementation of FIG.
In the example, the sampling frequency fs as shown in FIG.
= 50 kHz of the sound waveform sample data
0th-order hole in the domain of d.fs = 800 kHz as shown in
The data is interpreted as the
A sample having a valid value at all sample points of several d · fs = 800 kHz
Sample data was simplified as in the previous example.
The filter operation is performed. FIG. 10 (b) shows the same m = 96 as in FIG. 4 (c).
Illustrate the impulse response of the following low-pass filter characteristics
Things. As described above, the phase C of the current address signal is obtained.
AD corresponds to a predetermined reference order k (for example, intermediate 47th order)
Then, the sample data of FIG.
Find the convolution of the impulse response of. This fold
Komiwa is generally (Equation 1) It is expressed as Where x (ωs') is the current address signal
Is the convolution sum corresponding to the phase CAD, and h (95-i) is the
Luther coefficient, W (i) is the sample in the domain of d · fs = 800 kHz.
The pull value, that is, the sample data of each sample point in FIG.
It is. The spectrum envelope of the waveform shown in FIG.
The rope is as shown in FIG. Unnecessary harmonic components
It is attenuated and convenient. Therefore, according to the above equation
Unwanted aliasing components in the filtered output signal
It will be sufficiently attenuated. By the way, one of the above
The general formula requires 96 sums of products, but the wave of FIG.
The same amplitude continues for d = 16 times in shape sample data
So, if you perform this operation as a whole,
Can be reduced to seven times, and six times in the embodiment of FIG.
The operation scale can be reduced to a value that is not so different from the above.
That is, referring to FIG. 10A, the integer of the address signal
Same waveform sample data between IAD-2 and IAD-1
D = 16 times, and this is collected once
Multiplication of the IAD-1 and IAD-1
Similarly, between IAD and IAD + 1, IAD + 1
And between IAD + 2 and IAD + 3
Similarly, And between IAD-3 and IAD-2
Now, the same number of waveform sample data
Data, which are grouped together and multiplied by one coefficient.
And IAD + 3 and IAD + 4
The same is true during the period. Therefore, the above equation is obtained with a total of seven product sums
Can be performed. Therefore, in the embodiment shown in FIG.
D = 16 order filters corresponding to the shape sample data
Filter coefficients are summed in advance, and this is
Treat as numeric data and find sum of products by one coefficient operation
I am trying to. For example, W · h0 + W · h1 + W ·
Calculation of h2 + Wh3 + Wh4 + Wh5 + Wh6
Is performed by seven multiplications, h0 + h1 + h2 + h3
+ H4 + h5 + h6 is prepared in advance, and W ·
(H0 + h1 + h2 + h3 + h4 + h5 + h6)
The sum of products is obtained by one multiplication. In FIG.
The same parts as those in the embodiment of FIG. 8 are denoted by the same reference numerals. Strange
Additional points are the subtractor 16 and the master clock generator 2 shown in FIG.
2, timing signal generation circuit 23, filter coefficient memory
A subtractor 160 corresponding to 25 and 26 generates a master clock.
Generator 220, timing signal generating circuit 230, filter
This is a part of the coefficient memories 250 and 260. As mentioned above
In this embodiment, in the filter operation of one sample point,
1 time slot to perform 7 times sum of product calculation
7 filter operation time slots in the
Thus, the operation timing is changed as shown in FIG. FIG.
2, the calculation cycle pulse CAC is the same as described above.
Generated at a sampling frequency fs = 50 kHz,
Divides one cycle into 8 and time division timing of 8 channels
CH1 to CH8 are formed in the same manner as described above.
However, the time division time slot of each channel is divided into 7
To form seven filter operation time slots.
You. The slot count data SLCTR is used in this embodiment.
In the case of 7 channels in one channel time slot,
Time slot 0,1,2,3,4,5,6
Changed to distinguish. Master clock pulse MC
Is counted by a modulo 7 counter,
7 filter operation times in channel time slot
Slot count data SL for distinguishing slots 0 to 6
CTR is obtained. Therefore, the master clock pulse MC
Is changed to 2.8 MHz. According to this
Clock generator 220 and timing signal generation circuit 2
30 have been modified. Further, seven sampling points can be obtained by one filtering operation.
To the sample data corresponding to the integer part of the minute address signal
, The integer part IAD of the current address signal is calculated.
Corresponding to the integer part IAD-3 three sample points before
Extra data is needed. Therefore, the pull in FIG.
In the subtractor 160 of FIG.
3 is subtracted from the integer part IAD of the address signal to obtain "IAD-
3 "has been changed. Filter coefficient memo
The filters 250 and 260 are one of the filter coefficients of m = 96 order.
Or a filter coefficient corresponding to the sum of a plurality of filter coefficients
Group values are stored in advance, and both
250 and 260 have the same storage contents. Memory 250,
The filter coefficient group value stored in advance in 260 is
As shown in Table 3, d = 16
The sum of the filter coefficients (81 sets, one example
1 are stored at addresses 16 to 96).
D = 16 for the left end of the impulse response of 0 (b)
The sum of the full filter coefficients (there are 15 sets,
(For example, stored in addresses 1 to 15).
D = less than 16 fills for the right end of the impulse response
Sum of the data coefficients (there are also 15 sets, as an example
(Stored at addresses 97 to 111)
Consists of Note that “0” is stored in the address 0.
However, this is equivalent to the selection means 27 for generating address data.
This is a design issue caused by the same configuration as 8
Only. The data of the intermediate order (47th order) is used as the reference.
Is the sum of the 47th to 62nd order filter coefficients.
Is stored in the address 63. [0036] [Table 3] In this configuration, the slot count data
Of the selection means 27 corresponding to each value 0 to 6 of the data SLCTR.
The data output from the arithmetic unit 31 represents the order itself.
Instead of filtering, the filter
The addresses of the number memories 250 and 260 are shown. A
Slot count data according to the fractional part FAD of the dress signal
Of the selection means 27 corresponding to each value 0 to 6 of the data SLCTR.
Of the coefficient memory address data output from the adder 31
The values are as shown in Table 4 below. The following table corresponds to each
Each integer part IAD-3, IAD-2, I for 7 sample points
AD-1, IAD, IAD + 1, IAD + 2, IAD +
3 is also shown. The concept of memory reading in this case is as described above.
Similarly to the above-described embodiment, the reference intermediate order (47th order)
The coefficient data (47 to 6) stored in the corresponding address 63
The sum of the secondary filter coefficients) to the current address signal
Corresponding to the decimal part FAD,
Each integer part IAD for 7 sample points for several parts FAD-
3, IAD-2, IAD-1, IAD, IAD + 1, I
AD + 2, IAD + 3
Seven coefficient memories separated from the primary address 63
7 dresses decided each and every one of the dresses
7 filter coefficient groups from the coefficient memory address of
The value data is read out respectively. Also before
As before, filter coefficient group of adjacent address
Value data is read from both memories 250 and 260 in parallel.
In addition, interpolation is performed. [0039] [Table 4] Referring to Tables 3 and 4, for example, the address
If the fractional part FAD of the source signal is "6", SLCTR becomes 0
In this case, the coefficient memory address is 9, and the 0th to 8th order
The total data of the filter coefficients is obtained from the filter coefficient memory 250.
When SLCTR is 1, the address is 25 and
The total data of the 9th to 24th filter coefficients is SLCTR
If the value is 2, the address is 41 and the filter is 25th to 40th order
When the total data of the coefficient is 3 in SLCTR, the address
Is 57 and the total data of the 41st to 56th order filter coefficients is
When SLCTR is 4, the address is 73 and 57th to 72nd order
When the total data of the filter coefficients of SLCTR is 5
Is the sum of 73-88th order filter coefficients with address 89
When the data is SLCTR = 6, the address is 105
The total data of the 89th to 95th order filter coefficients is
Will be issued. Thus, 7 sets of filter coefficient group values
Therefore, all filter coefficients of the 0th to 95th orders are covered.
You. As described above, according to the embodiment of FIG.
In the domain of d.fs = 800 kHz as shown in FIG.
M = 96 for the sample data in the next hold state
The operation to find the product sum of all the following filter coefficients is actually
Can be performed with only seven operations. Follow
And simple structure, but with sufficient attenuation of aliasing components
Using the spell vector waveform data
A good filter operation can be performed. 8 and FIG.
In the eleventh embodiment, the digital filter operation circuit 21
A delay circuit is used as a means to delay sample data.
Address of the tone waveform memory 13 instead of actually providing
Is provided, which is controlled as shown in FIG.
A delay circuit may be actually provided. Also, music
Multi-period waveforms are recorded as waveform sample data generation means.
Uses the stored tone waveform memory 13, but is not limited to this.
Instead, a tone waveform memory that simply stores a one-cycle waveform, or
Generates tone waveform sample data by frequency modulation
Tone waveforms based on the
Method of generating sample data, or address data
To generate tone waveform sample data by converting data
Any method such as a formula may be used. Also,
In the above embodiment, the tone generation and the filter of each channel are performed.
The calculation is performed by the time-division processing method.
Column processing may be used. Also, not only the double tone generation method,
A single sound generation method may be used. 8 and FIG.
In the embodiment, the tone color selection code TC, the key code KC,
The waveform corresponding to the touch data TD is stored in the tone waveform memory 13.
To select, these data TC, KC, TD
Without input to the waveform memory 13, the address signal generation circuit 1
2 and can be selected by the upper bits of the address signal.
The address signal generation circuit 12 may be configured so that
No. Note that FIG. 4 (c) or FIG. 10 (b)
Symmetrically around the middle order
Response, the filter coefficients are stored in memory for all orders.
There is no need to keep it, just remember half and
A filter coefficient with the same value can be shared between different orders.
You may do it. Also, filter coefficient memories 25 and 26
Or 250 and 260 as one and two adjacent for interpolation
The coefficient data to be read may be read out in a time-division manner.
In that case, double the frequency of the master clock pulse MC.
Then, within one filter operation time slot,
Two time division time slots are formed. 8 and 11
In the embodiment, the filter read from the filter coefficient memory is used.
The interpolation of the filter coefficient is performed in 4 steps.
The number 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,
IIR (infinite impulse response) and other formats
You may. The digital filter used in the present invention
Is used for removing aliasing noise as in the above embodiment.
The present invention is not limited to this, and may be used for other purposes such as tone control.
In that case, the tone selection code TC, key code KC,
To select the filter characteristics according to the
I do. In other words, multiple filter characteristics are stored in the filter coefficient memory.
The filter coefficient data corresponding to the characteristics are stored respectively,
Tone selection code TC, key code KC, touch data T
Filter coefficient data for realizing a desired tone according to D etc.
And read them according to the decimal part of the address signal.
It is to come out. Also, a key having a tone color corresponding to the key code KC
Tone control according to scaling control and touch data TD
Filter coefficient data stored in the memory in advance
And keep this filter coefficient data as a key.
To interpolate according to code KC and touch data TD
The filter coefficients may be generated more densely. Ma
Further, a plurality of musical tone signal generating devices according to the present invention are provided,
The tone signal generated in each series is converted to a key code KC or touch
Interpolation synthesis according to the data TD. the above
In the embodiment, the sampling frequency is set to the pitch of the tone signal.
Musical sound by pitch asynchronous method which is always constant regardless of
Although waveform sample data is being generated,
The pitch synchronization method in which the wave number is synchronized with the pitch of the tone signal
To generate musical sound waveform sample data
The present invention can be applied. In addition, musical sound memo
Without reading all the waveform sample data stored in the
In the high range, for example, once every two or once every four
Control the address signal so that it is decimated and read
You may do so. [0044] As described above, according to the present invention,
Integer of the address signal that changes according to the pitch of the power tone
Generates sound waveform sample data according to the
Corresponding to the m-th order filter coefficient according to the decimal part of the
Of the coefficient data (where n <m)
Corresponding to the n coefficient data and the integer part of the address signal
Using the generated tone waveform data for n sample points
Then, the m-th order filter operation is performed for the tone waveform data for n sample points.
Data for the sound wave data
Of sound waveform sample data prepared by data generator
The function is relatively coarse, corresponding to the integer part of the address signal.
To simplify the circuit configuration
And in practice less than m
Not limited to n sample points of tone waveform data
That the filter operation only needs to be performed on the
Also simplifies the configuration of digital filter circuits
You can enjoy both,
Effective filter operation is the decomposition of the fractional part of the address signal.
M for accurate sound waveform sample data with high performance
It is equivalent to performing the following precise filter operation.
This makes it possible to deal with high-resolution musical sound waveform sample data.
Various effects brought by the precise filter operation
As a result, for example, unnecessary noise components can be reliably
The advantage of being able to obtain the tone signal of
It has an excellent effect that it can be In addition, according to the present invention, the filter coefficient
Since the supply means has a filter coefficient interpolation means,
With filter coefficient generation meansRememberMore than predetermined order
M-th order filter coefficient, The filter coefficient generating means
Without increasing the storage capacity of filter coefficientsEasy departure
It has the effect of being able to live. Also Phil
The filter coefficient supply means includes a plurality of filters having different characteristics.
Corresponds to control data for controlling the tone from the characteristics
Select the filter characteristics to
Continuous to achieve the corresponding mth order filter characteristic
From the m filter coefficients corresponding to each order,
N (where n <m) coefficients according to the decimal part of the address signal
As we have selected and supplied data,
Simple filter configuration for m-order precise filter operation
In addition to being able to obtain the noise removal function by
Realizes tone control function according to arbitrary tone control data
It has an excellent effect that it can be performed.In addition, billing
According to the invention described in Item 2, the filter coefficient interpolation means
Is stored in the filter coefficient generating means.
Filter coefficients of a higher order than the number of filter coefficients
Storage capacity of filter coefficients in the filter coefficient generation means
It can be easily generated without increasing the amount
It works.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram showing a basic configuration of an embodiment of a tone signal generating apparatus according to the present invention. FIG. 2 is a waveform sample at each stage based on a general sampling frequency conversion theory. FIG. 3 is a waveform diagram showing an example of data. FIG. 3 is a diagram showing an example of a spectrum envelope of each waveform sample data in FIG. 2. FIG. 4 is a diagram for explaining in principle a digital filter operation in the present invention. FIG. 6 is a graph showing an example of an amplitude-frequency characteristic of an FIR low-pass filter realized according to the present invention. FIG. 6 is an actual measurement diagram showing a spectrum of a sine wave signal passing through a low-pass filter having the characteristics of FIG. FIG. 8 is a block diagram showing an example of a filter coefficient supply unit and a digital filter operation unit in the embodiment. FIG. 9 is a timing chart showing a timing relationship of arithmetic and other operations in the embodiment of FIG. 8; FIG. 10 is a diagram for explaining a digital filter arithmetic operation in another embodiment of the present invention; FIG. 12 is a block diagram showing another embodiment of the present invention related to FIG. 10; FIG. 12 is a timing chart showing timing relations of arithmetic and other operations in the embodiment of FIG. 11; Tone 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.

Continuation of front page    (56) References JP-A-56-19099 (JP, A)                 JP-A-55-161296 (JP, A)                 JP-A-57-2114 (JP, A)                 JP-A-57-2115 (JP, A)                 JP-A-57-2116 (JP, A)                 JP-A-62-115194 (JP, A)                 JP-A-61-239713 (JP, A)                 JP-A-54-65020 (JP, A)                 JP-A-51-9348 (JP, A)                 JP-A-51-14016 (JP, A)                 JP-A-58-31395 (JP, A)                 Tokiko Hei 6-31989 (JP, B2)

Claims (1)

(57) [Claims] 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 musical tone for generating musical tone waveform sample data in accordance with the integer part of the address signal Waveform data generating means, control data generating means for generating control data for controlling a musical tone, and a predetermined order for a plurality of filter characteristics exhibiting different characteristics.
The filter coefficients consisting of
Select a filter characteristic corresponding to the control data from the filter characteristics of several, the ad from the filter coefficient comprising a predetermined order corresponding to the filter characteristic selected
Multiple signals according to the predetermined upper bits of the decimal part of the address signal.
Filter coefficient generating means for reading a filter coefficient and a plurality of filter coefficients read out from the filter coefficient generating means interpolating according to a predetermined lower-order bit of a decimal part of the address signal to thereby obtain a filter order higher than the predetermined order. A filter coefficient interpolation means capable of generating a large number of m-th order filter coefficients, and controlling the reading of the filter coefficients by the filter coefficient generation means and the interpolation by the filter coefficient interpolation means according to the decimal part of the address signal; Filter coefficient supply means for selecting and supplying n (where n <m) coefficient data from m filter coefficients corresponding to each successive order for realizing an mth-order filter characteristic, Using the n pieces of coefficient data and tone waveform data for n sample points generated by the tone waveform data generating means, an m-th order filter operation is performed. Tone signal generation device with a digital filtering unit performed for n sample points worth of tone waveform data. 2. 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 musical tone for generating musical tone waveform sample data in accordance with the integer part of the address signal Waveform data generation means, filter coefficient generation means for storing filter coefficients of a predetermined order, and reading out a plurality of filter coefficients of different orders according to predetermined upper bits of a decimal part of the address signal; Filter coefficient interpolating means for interpolating a plurality of filter coefficients of different orders read from the filter according to predetermined lower-order bits of a decimal part of the address signal to generate a filter coefficient of an order larger than the predetermined order; and The tone waveform data generated by the tone waveform data generating means in accordance with the filter coefficient generated by the coefficient interpolation means. A tone signal generator comprising a digital filter operation means for performing a filter operation on data.