JP2790066B2 - Tone signal generator and waveform memory read-out interpolator - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a tone signal generator for outputting a digital tone signal by reading time series sample data of a tone waveform from a memory, and a waveform for performing an interpolation operation of waveform data in the tone signal generator. The present invention relates to a memory read interpolation device.

[0002]

2. Description of the Related Art As a sound source of an electronic musical instrument, time series sample data (hereinafter referred to as "waveform data") of a musical tone waveform is stored in a memory in advance, and the waveform data is sequentially read out in response to a musical tone generation instruction. 2. Description of the Related Art A sound source of a waveform readout type for outputting a signal is known. The sound source of the waveform readout method is roughly classified into a pitch synchronous method sound source and a pitch asynchronous method sound source.

Here, the pitch synchronous type sound source outputs a tone signal by reading waveform data from a waveform memory in accordance with a read frequency proportional to the pitch of a tone to be generated. Also, pitch asynchronous type sound source
The waveform data is always read from the waveform memory in accordance with a constant read frequency regardless of the pitch of a musical tone to be generated, and a tone signal is generated based on the waveform data.

When a waveform data sequence is created by sampling a musical sound waveform, the waveform data sequence includes, in addition to the spectrum corresponding to the original waveform, the spectrum as an axis of symmetry which is an integral multiple of the sampling frequency. , Image spectra of spectral distributions arranged symmetrically are included. For example, as shown in FIG. 13, when a musical tone waveform before sampling is composed of a spectrum of 24 kHz or less, if this musical tone waveform is sampled with a sampling clock having a sampling frequency fs = 48 kHz, the resulting waveform data string has 48 ± 24 kHz, 96 ± 2 as shown in FIG.
The image spectrum in the frequency range of 4 kHz,..., M · fs ± 24 kHz,. If the image spectrum thus generated overlaps the spectrum of the original musical sound waveform on the frequency axis, it will be called aliasing noise. This aliasing noise is extremely difficult to remove once generated, and if waveform data containing aliasing noise is reproduced as a musical tone as it is, it will be heard as harsh noise. Therefore, in order to reproduce high-quality musical sound, it is necessary to take care that no aliasing noise occurs during the entire process from the sampling of waveform data to the reproduction of the waveform data as a musical sound signal.

[0005] As described above, the pitch synchronous type sound source
The tone waveform is reproduced by reading the waveform data in accordance with the read frequency proportional to the pitch of the tone to be generated. Therefore, in the case of this sound source, as long as the waveform data sequence is obtained from a musical sound waveform that does not include a spectrum having a frequency equal to or more than half the sampling frequency (ie
If the waveform data sequence itself does not include aliasing noise), a tone waveform that does not include aliasing noise is reproduced, regardless of the frequency at which the waveform data is read. This is because the upper limit frequency of the spectrum of the read waveform data increases and decreases in proportion to the increase and decrease of the read frequency. FIGS. 15 and 16 show this example. FIG. 15 shows the spectrum distribution of a waveform obtained by reading out the waveform data showing the spectrum distribution in FIG. 14 at a readout frequency of 50 kHz. 3 shows a spectrum distribution of a waveform obtained by reading at a read frequency of FIG.

However, when a tone signal is generated by a pitch-synchronized tone generator, the output frequency of the waveform data constituting the tone signal changes depending on the pitch of the tone signal, which complicates the processing of the subsequent device. For example, when two or more types of tone signals having different pitches are mixed and generated, there is a problem that the two cannot be directly added because the respective waveform data are output at different timings.

On the other hand, a pitch asynchronous type sound source
When a musical tone generation instruction is given, pitch information called an F number is generated in proportion to the designated pitch (tone frequency), and this F number is accumulated for each sampling period to obtain a pitch. An address signal that changes at a corresponding rate is generated, and waveform data corresponding to the address signal is reproduced from a memory. In a pitch asynchronous type sound source, these series of processes are executed in synchronization with a fixed sampling frequency, and waveform data of a fixed sampling frequency is output irrespective of the pitch. Therefore, in the pitch asynchronous type sound source, there is no problem as in the case of the pitch synchronous type sound source.

By the way, in an electronic musical instrument, it is necessary to generate many kinds of musical tone signals having slightly different pitches from one kind of musical tone waveform stored in a memory. Therefore, it is necessary to use a real number including an integer part and a decimal part as the F number. When the F number including the decimal part is used, the address signal generated in each sampling period generally also includes the decimal part, but the waveform data corresponding to the address signal, for example, the address “3.5” Is not stored in the memory and cannot be read.

If the waveform data corresponding to the integer part of the address signal is read out from the memory in each sampling period, a tone waveform completely different from the original tone waveform is reproduced. FIG. 17 shows that, when the waveform data (sampling frequency 48 kHz) having the spectrum distribution shown in FIG. 14 is stored in the memory, the F number = 0.8 is accumulated in synchronization with the clock of the sampling frequency 50 kHz. This is an example of the spectrum distribution of the read waveform when the waveform data corresponding to the integer part of the address signal as the accumulation result is read from the memory. In the same figure, the mark of “x” indicates the aliasing noise corresponding to the spectrum indicated by the broken line in FIG. 14 included in the original waveform data in the memory. As shown in FIG. 17, when the waveform data is reproduced based on only the integer part of the address signal in the pitch asynchronous reproduction, a large number of aliasing noise components are generated.

In order to prevent the above problems, interpolation processing is generally performed on a pitch asynchronous type sound source. That is, in each sampling period, a plurality of continuous waveform data including the waveform data corresponding to the integer part of the address signal is read from the memory, and these waveform data are determined corresponding to the decimal part of the address signal. By multiplying and adding the interpolation coefficient, the interpolation of the waveform data corresponding to the address signal is performed.

[0011]

According to the pitch asynchronous type sound source using such interpolation processing together, a tone signal having an arbitrary pitch can be generated. However, this pitch-asynchronous sound source has a problem that aliasing noise is likely to occur when the pitch of a musical tone to be generated is high. Hereinafter, this problem will be described.

First, it is assumed that waveform data (sampling frequency 48 kHz) having a spectrum distribution shown in FIG. 14 is stored in a memory. It is assumed that the F number is 1, the F number is accumulated in synchronization with a clock having a sampling frequency of 50 kHz, an address signal is generated, and waveform data corresponding to the address signal is reproduced by interpolation using waveform data in the memory. Suppose you did. In this case, since the sampling frequency at which the waveform data is reproduced is 50/48 at the time of acquiring the waveform data, the upper limit frequency of the spectrum of the reproduced waveform (excluding the image spectrum) is 24.
kHz × (50/48) = 25 kHz.

Next, when the F number is smaller than 1, for example, when the waveform data is reproduced under the condition of F number = 0.8, the reproduced waveform at F number = 1 is displayed along the time axis. (1 / 0.8) times, and expand the expanded waveform by 5
The same waveform as that re-sampled by the sampling clock of 0 kHz is reproduced.

Accordingly, the spectrum distribution of the reproduced waveform in this case is as shown in FIG. That is, by setting F number = 0.8, the spectrum SA of the reproduced waveform is obtained.
Is obtained by compressing the spectrum of the reproduced waveform at the time of F number = 1 by 0.8 times along the frequency axis. Although the image spectrum indicated by the broken line is removed by the interpolation processing, the image spectrum S obtained by sampling at the sampling frequency of 50 kHz is removed.
B, SB,... Are newly added as the spectrum of the reproduced waveform. Since the frequency range of this image spectrum is narrowed to 0.8 times that of the case where the F number is 1, the spectrum S of the originally required waveform corresponding to the F number of 0.8 is used.
A (upper limit frequency 20 kHz) does not overlap and does not cause aliasing noise.

On the other hand, when the F number exceeds 1,
For example, when the waveform data is reproduced under the condition of F number = 1.2 in synchronization with a clock having a sampling frequency of 50 kHz, the reproduced waveform at the time of F number = 1 is represented by (1/1) along the time axis. .2), and the same waveform as that obtained by re-sampling the compressed waveform with a sampling clock of 50 kHz is reproduced.

Therefore, the spectrum distribution of the reproduced waveform in this case is as follows. First, by setting F number = 1.2, as shown in FIG. 19, the spectrum SA of the reproduced waveform becomes 1.2 times the spectrum of the reproduced waveform when F number = 1 along the frequency axis. It will be stretched. Further, although the image spectrum indicated by the broken line is removed by the interpolation processing, as shown in FIG.
The image spectrum SB, SB,... Obtained by sampling at the sampling frequency of Hz is newly added as a spectrum of the reproduced waveform. Since the frequency range of the image spectrum SB is 1.2 times as wide as that of the case where the F number is 1, the spectrum spectrum SB overlaps with the spectrum SA of the originally required waveform corresponding to the F number of 1.2, and the overlapping portion is folded back. It becomes noise.

The present invention has been made in view of the above-mentioned circumstances, and overcomes the disadvantages of the above-mentioned pitch asynchronous type sound source. It is an object to provide a generator and a waveform memory readout interpolator.

[0018]

The invention according to claim 1 is
In a tone signal generating apparatus which responds to a tone generation instruction and outputs a digital tone signal of a predetermined sampling frequency representing a tone waveform of a pitch designated by the tone generation instruction, the tone waveform is sampled at regular time intervals. A waveform memory storing the obtained waveform data, an address signal generating means for outputting an address signal that changes at a rate corresponding to the instructed pitch in accordance with an operation cycle frequency higher than the sampling frequency, and the waveform memory. Executing an interpolation operation according to the operation cycle frequency using a predetermined number of pieces of waveform data corresponding to the integer part of the address signal in the stored waveform data and an interpolation operation coefficient corresponding to a decimal part of the address signal. By interpolating means for outputting waveform data having the designated pitch, Downsampling means for performing processing for removing a spectrum having a frequency equal to or higher than a predetermined high cut-off frequency to the waveform data output by the interpolation means, and outputting the result in accordance with the sampling frequency. The gist is a signal generator.

According to a second aspect of the present invention, n (n is an integer)
A time-division control using the time-division channels of the waveform memory read-out interpolation device which reads out waveform data from the waveform memory for each time-division channel and generates interpolation waveform data based on the read waveform data; Address generating means for sequentially generating an address signal corresponding to each of the time-division channels, and reading out waveform data from the waveform memory based on an integer part of each address signal, and reading each of the waveform data in accordance with a decimal part of each address signal. Interpolating means for respectively generating interpolated waveform data corresponding to the time-division channels and outputting the interpolated waveform data in a time-division manner, wherein the interpolation means comprises k groups (k is a divisor of n) K arithmetic means for performing an interpolation operation corresponding to each group divided into a plurality of groups, and a waveform data corresponding to an integer part of the address signal of each time-division channel. Data from the waveform memory to one of the k arithmetic means for executing an interpolation operation corresponding to the time-division channel, and converts the decimal part of the address signal of each time-division channel into the k arithmetic means. Supply means for delivering to one of the calculation means that performs an interpolation calculation corresponding to the time-division channel, and time-division multiplexing and output of n pieces of interpolation waveform data which are interpolation calculation results of the k pieces of calculation means. The present invention provides a waveform memory reading and interpolating device, which comprises an output unit.

According to a third aspect of the present invention, there are provided n (n is an integer)
Time-division control using the time-division channel of each of the time-division channels, in response to a musical tone generation instruction, a digital musical tone signal of a predetermined sampling frequency representing a musical tone waveform of the pitch indicated by the musical tone generation instruction. In the output tone signal generating apparatus, a waveform memory storing waveform data obtained by sampling a tone waveform at regular time intervals, and, for each of the time-division channels, a rate corresponding to the designated pitch. Address signal generating means for outputting a changing address signal in accordance with an operation cycle frequency higher than the sampling frequency; and for each of the time-division channels, corresponding to an integer part of the address signal in the waveform data stored in the waveform memory. And a predetermined number of waveform data and interpolation calculation coefficients corresponding to the decimal part of the address signal are used. By performing the inter-operation in accordance with the operation cycle frequency, interpolation means for outputting waveform data having the indicated pitch, and for each of the time-division channels, for the waveform data output by the interpolation means, Downsampling means for performing processing for removing a spectrum equal to or higher than a predetermined high-frequency cutoff frequency and outputting the result in accordance with the sampling frequency, wherein the interpolation means sets the n time-division channels to k (k (K is a divisor of n), k operation means for executing an interpolation operation corresponding to each group, and waveform data corresponding to an integer part of an address signal of each of the time-division channels.
From the waveform memory, the k arithmetic units are passed to one of the k arithmetic units that performs an interpolation operation corresponding to the time-division channel, and the decimal part of the address signal of each time-division channel is converted to the k arithmetic units. Wherein the tone generator comprises: a supply unit for delivering the interpolation operation corresponding to the time-division channel; and an output unit for time-division multiplexing and outputting the interpolation operation results of the k operation units. The gist is a signal generator.

[0021]

According to the first aspect of the present invention, processing is performed in synchronization with the operation cycle frequency higher than the final sampling frequency until waveform data corresponding to the designated pitch is obtained. Then, when waveform data corresponding to the designated pitch is obtained, high-frequency removal processing is performed on the waveform data in advance, and the waveform data subjected to this processing is converted into a final sampling frequency. .

The invention according to claim 2 solves the above-mentioned problem by speeding up the interpolation operation as described below. First, in order to obtain waveform data that does not include aliasing noise in an arithmetic operation such as an interpolation operation, it is necessary to sufficiently increase the operation cycle frequency when repeating the operation.
For that purpose, it is necessary to execute each operation in a very short time. However, in a so-called double tone type tone signal generating device, that is, a tone signal generating device of a method of independently generating a plurality of tone signals using a plurality of time division channels, an interpolation calculation to be executed within a calculation cycle period is performed. Since the number of times increases in proportion to the number of time-division channels, the demand for shortening the individual processing time becomes extremely severe. In particular, since the interpolation calculation involves processes such as reading of waveform data from a waveform memory and a sum-of-products calculation, it is difficult to increase the speed, and this has been an obstacle in taking measures to cope with the above demand.

According to a second aspect of the present invention, the above-mentioned request is satisfied without imposing severe requirements on the arithmetic circuit. That is, according to the present invention, each of the interpolation operations corresponding to the n time-division channels is shared by the k operation means and executed in parallel. The output is divided and multiplexed. As described above, according to the waveform reading and interpolating apparatus of the present invention, the number of interpolation operations to be performed by the individual operation units is reduced. Can be output. Therefore, according to the waveform reading and interpolating apparatus of the present invention, it is possible to prevent the generation of aliasing noise in the double tone type tone signal generating apparatus.

According to a third aspect of the present invention, by combining the technical ideas of the first and second aspects of the present invention, a double tone tone signal generation in which the generation of aliasing noise is prevented. An apparatus is provided. That is, when the invention according to claim 1 is to be applied to a generating device for a double tone type tone signal, a strict demand for speeding up the interpolation operation is generated. The present invention addresses this need using the technical idea of the second aspect of the present invention. Since the specific contents have already been described as the operation of the invention according to claim 2, the duplicate description will be omitted here.

[0025]

EXAMPLES In order to make the present invention easier to understand,
An example will be described. These examples illustrate one embodiment of the present invention and do not limit the present invention. It can be arbitrarily changed within the scope of the present invention.

A. 1. Configuration of Embodiment (1) Overall Configuration FIG. 1 is a block diagram showing the overall configuration of an embodiment of an electronic musical instrument provided with a tone generator according to the present invention. In FIG. 1, reference numeral 1 denotes a keyboard, which has a large number of keys, a key switch whose on / off state is switched by pressing and releasing the keys, and a speed detecting mechanism for detecting a key pressing speed. ing. Reference numeral 2 denotes a key press detection circuit which scans the state of a key switch of each key of the keyboard 1 and generates a key-on signal KON and a key code KC representing the key pressed when a key press operation is performed. . Reference numeral 3 denotes a touch detection circuit which detects a key pressing speed via the speed detecting mechanism and generates touch information IT indicating an initial touch.

Reference numeral 4 denotes an assigner circuit for assigning sound channels. This electronic musical instrument is configured to use 16 sound channels and form musical tones by time division control. The assigner circuit 4 outputs the key-on signal KO
When N occurs, a tone generation channel CH for performing a tone generation process corresponding to the key code KC received from the key press detection circuit 2 is selected from tone generation channels not used in the tone generation process at that time. When the number of musical tones to be simultaneously generated by pressing a key exceeds 16 tones, the assigner circuit 4 is configured to execute a truncation process in accordance with a well-known late arrival priority rule.

Reference numeral 5 denotes a waveform generator based on a waveform readout method which embodies the tone signal generator according to the present invention. The tone generator uses the tone generation channel CH selected by the assigner circuit 4 and controls the tone corresponding to the key code KC by time division control. A process for forming high waveform data is performed. The waveform generator 5 finally outputs waveform data having a sampling frequency of 50 kHz. However, the waveform generator 5 does not generate aliasing noise until waveform data of a musical sound waveform corresponding to the key code KC is obtained. For example, processing such as reproduction of waveform data is performed at a higher operation cycle frequency (200 kHz in this embodiment). After the waveform data of the musical tone waveform corresponding to the key code KC is once generated in synchronization with the clock having a frequency of 200 kHz by this processing, a spectrum having a frequency equal to or more than 1/2 of the sampling frequency of 50 kHz is removed from the waveform data. The waveform data (sampling frequency: 200 kHz) subjected to the high-frequency removal processing is converted into waveform data having a sampling frequency of 50 kHz and output. The specific configuration of the waveform generator 5 will be described later.

Reference numeral 6 denotes a digital filter, which performs a filtering process on the waveform data output from the waveform generator 5 for imparting an effect corresponding to the touch information IT, adjusting a tone color, and the like. Reference numeral 7 denotes an envelope generator which generates and outputs an envelope signal ENV when a key-on signal KON is supplied. Reference numeral 8 denotes a multiplier, which is an envelope signal E output from the envelope generator 7.
NV is multiplied by the output signal of the digital filter 6, and the multiplication result is output as a tone signal. Reference numeral 9 denotes an accumulator, which accumulates and outputs a musical tone signal over all tone generation channels. Reference numeral 10 denotes a D / A converter for converting the output signal of the accumulator 8 into an analog signal. Reference numeral 11 denotes a sound system, which amplifies an analog signal supplied from the D / A converter 10 and produces a tone from a speaker SP.

(2) Configuration of Waveform Generator 5 Here, the configuration of the waveform generator 5 which is the most characteristic part in this embodiment will be described. This waveform generator 5
As shown in FIG. 2, a waveform memory 12, an address generator 13, an interpolator 14, a down-sampling FIR
And a filter 15. Hereinafter, each part of the waveform generator 5 will be described in detail.

Waveform Memory 12 The waveform memory 12 stores a plurality of types of musical sound waveform data. Which musical tone waveform data to read from the waveform memory 12 is selected based on the timbre, tone range, and the like of the tone signal to be generated.
Each waveform data is time-series digital data obtained by performing A / D conversion on a musical tone waveform at a constant sampling frequency. When generating waveform data, as sampling pre-processing of the A / D conversion, A process of removing a spectrum having a frequency equal to or more than 1/2 is performed on the musical tone waveform. Therefore, when reading out the waveform data of any musical tone waveform, if all the waveform data are read out without thinning out, the musical tone waveform containing no aliasing noise is reproduced.

Address Generator 13 FIG. 3 shows the structure of the address generator 13. As shown in the figure, an address generator 13 includes a pitch information generator 13a for generating pitch information corresponding to a key code KC, and a rate corresponding to the key code KC by accumulating the pitch information. And a pitch information accumulating section 13b for generating an address signal which increases in the range.

In this embodiment, in order to calculate the waveform data corresponding to the key code KC at each calculation cycle period of 1/200 kHz, an address signal designating the phase of the waveform data to be calculated is synchronized with a clock of 200 kHz. Must be output. The address signal is calculated by accumulating the pitch information every time the operation cycle period is switched. The F number (that is, from the waveform reading to the tone signal output) assuming a sampling frequency of 50 kHz as the pitch information is calculated. If the F number used in a conventional pitch asynchronous type sound source that performs signal processing in synchronization with a fixed sampling frequency is used as it is, the address signal as a result of accumulation is four times the rate corresponding to the key code KC. Will increase. Therefore, in this embodiment, in order to change the address signal at a rate corresponding to the key code KC, an F ^* number having a value of 1/4 of the F number is generated by the pitch information generation unit 13a.

As described above, this electronic musical instrument has one
A tone signal is formed for each tone generation channel using six tone generation channels.
The generation of the F ^* number by a and the pitch information accumulating unit 13b
Is also generated for each sounding channel by time division control. That is, in the present embodiment, the calculation of the F ^* number and the address signal corresponding to each sounding channel is performed using each time slot obtained by further dividing the calculation cycle period of 1/200 kHz into 16.

The integer part IAD of the address signal generated by the pitch information accumulating unit 13b is supplied to the waveform memory 12 as a read address. Further, the decimal part FAD of the address signal is supplied to the interpolation unit 14. Further, the pitch information accumulating section 13b outputs a carry signal INC when the accumulation of the F ^* number causes a carry from a decimal part to an integer part of the address signal.

Interpolator 14 The interpolator 14 is determined by a predetermined number of waveform data WD including those corresponding to the integer part IAD of the address signal among the waveform data stored in the waveform memory 12 and the decimal part FAD of the address signal. An interpolation operation using the interpolation coefficient sequence is executed in synchronization with a clock having a frequency of 200 kHz, and waveform data WP obtained as a result of the interpolation operation is output.

FIG. 4 is a block diagram showing the configuration of the interpolation unit 14. Since the present embodiment has 16 sound channels as described above, it is necessary to configure so that interpolation processing corresponding to each of these channels can be performed. Here, if the interpolation processing for 16 channels is performed by a completely time-division processing within one operation cycle period (1/200 kHz), the demand for the processing speed of the interpolation processing corresponding to each sounding channel is extremely severe. turn into. Therefore, in the present embodiment, as shown in FIG. 4, as a hardware for performing the interpolation processing, a system (A) including the waveform data storage block 16a and the interpolation circuit 17a is used.
Block), a system (B block) composed of the waveform data storage block 16b and the interpolation circuit 17b, a system (C block) composed of the waveform data storage block 16c and the interpolation circuit 17c, and a system (C block) composed of the waveform data storage block 16d and the interpolation circuit 17d. D block) are provided in parallel, and interpolation processing for four channels is assigned to each of the four systems, thereby alleviating the demand for the processing speed.

The interpolation unit 14 is supplied with 800 kHz four-phase clocks φ _A , φ _B , φ _C , and φ _D. A block to D
Block operation timing of each of which is controlled by a clock φ _A ~φ _D. These four-phase clocks φ _A , φ _B , φ _C , φ _D
Is synchronized with the sounding channel switching timing. Therefore, in the following, the four-phase clock φ _A ,
Each clock of φ _B , φ _C and φ _D is called a channel clock.

The channel clock φ _A is 1, 5, 9, 1
3 in each of the sounding channels. A block is synchronized to the channel clock phi _A, it executes the interpolation processing corresponding to each of these sound channel. Further, the channel clock φ _B is generated in each of the tone generation channels 2, 6, 10, and 14, and the B block executes an interpolation process corresponding to each tone generation channel in synchronization with the channel clock φ _B. The same applies to the other blocks, C block 3 in synchronism with the channel clock phi _C,
An interpolation processing corresponding to the respective sound channel of 7, 11 and 15, D block, performs interpolation processing corresponding to each sound channel of 4, 8, 12, 16 in synchronism with the channel clock phi _D.

Next, the internal configuration of each of the blocks A to D will be described. First, in the block A, the waveform data storage block 16a has shift registers for four channels corresponding to each of the tone generation channels 1, 5, 9, and 13. Each shift register is used for interpolation processing of the channel. 8 waveform data to be stored. The interpolation circuit 17a receives the waveform data of four channels consisting of eight waveform data for each channel from the waveform data storage block 16a during one sampling period (1/200 kHz), and performs interpolation processing of four channels. This interpolation processing is performed by convolving the interpolation coefficient corresponding to the decimal part FAD of the address signal of each channel into eight waveform data for each channel.

The writing of the waveform data to the waveform data storage block 16a is performed only when the necessity arises. That is, each of the eight channels corresponding to each tone generation channel stored in the waveform data storage block 16a.
Each of the pieces of waveform data can be continuously used for the interpolation process even if the sampling cycle is switched, as long as the integer part IAD of the address signal of the channel does not change. Therefore, when the integer part IAD of the address signal of a certain sounding channel increases and the carry signal INC is output, the waveform data corresponding to the increased integer part IAD is read out from the waveform memory 12 and a new channel of the channel is read. The waveform data is stored in the waveform data storage block 16a as waveform data for interpolation processing. By thus interposing the waveform data storage block 16a,
Compared with the method of directly reading out the waveform data used for the interpolation processing from the waveform memory 12, the demand for the reading speed of the waveform memory 12 is relaxed. The other blocks B to D have exactly the same configuration as the block A described above.

Down-sampling FIR filter 15 The down-sampling FIR filter 15
Is converted into waveform data having a sampling frequency of 50 kHz and output. Here, the waveform data before the conversion of the sampling frequency is 25 k.
If a spectrum exceeding Hz is included, aliasing noise occurs in the waveform data after the conversion of the sampling frequency. Therefore, in the present embodiment, a down-sampling FIR filter 15 having a filter characteristic of removing a high-frequency spectrum of 25 kHz or more is used, and after removing a spectrum of 25 kHz or more from waveform data of a sampling frequency of 200 kHz, the frequency is reduced to 50 kHz.
Is converted to the sampling frequency.

FIG. 5 shows the configuration of the down-sampling FIR filter 15. As shown in the figure, the down-sampling FIR filter 15 includes a ring buffer 15a, a channel synchronization counter 15b, a selector 15c, and a control unit 1.
5d and a filter operation processing unit 15e.

The ring buffer 15a is composed of a RAM or the like, and under the control of the control unit 15d, sequentially stores the waveform data WP output from the interpolation unit 14 and corresponding to each channel, and should be used for filter processing. The waveform data is sequentially delivered to the filter operation processing unit 15e.

The channel synchronization counter 15c has a frequency of 200 kHz × 16 synchronized with the switching timing of the sound channel.
= Counts clock phi _sync of 3.2 MHz, and outputs the count result as address data WA.

The selector 15c has a first input terminal supplied with the address data WA output from the channel synchronization counter 15c, a zeroth input terminal supplied with the address RA output from the control unit 15d, and a selector terminal connected to the selector 12c. given the selection signal φ _{SE L} synchronized with the 8MHz clock. When the selection signal φ _SEL is “1”, the address data WA is selected by the selector 15c and supplied to the ring buffer 15a as an address. When the selection signal φ _SEL is “1”
, The ring buffer 15a is in the write mode, and the waveform data WP output from the interpolation unit 14 at this time is
Is written in the area corresponding to the address data WA in the ring buffer 15a. As described above, since the address data WA is incremented each time the sounding channel is switched, the waveform data WP, WP,... Sequentially output from the interpolation unit 14 sequentially writes successive addresses in the ring buffer 15a. It will be rare. on the other hand,
When the selection signal φ _SEL is “0”, the address data RA output from the control unit 15d is selected by the selector 15c and supplied as an address to the ring buffer 15a. Further, while the selection signal φ _SEL is “0”, the ring buffer 15a is in the read mode, and the waveform data WP is read from an area of the ring buffer 15a corresponding to the address data RA.

The selection signal φ _SEL is switched in accordance with the output timing of the waveform data from the interpolation section 14. That is, the above-described interpolation unit 14 performs one operation cycle (1/200 kHz).
During z), waveform data of 16 channels is output, and the waveform data of each channel is supplied to the down-sampling FIR filter 15 in synchronization with a 3.2 MHz clock. For this reason, in this embodiment, the selection signal φ _{SEL is set} to “1” once every four 12.8 MHz clocks generated, and the ring buffer 1 of the waveform data is generated.
Write to 5a. In other periods, the selection signal φ _{SEL is set} to “0”, and the waveform data is read from the ring buffer 15a.

Next, the filter operation section 15e will be described. In the filter operation unit 15e, reference numeral 15f denotes a latch circuit for delaying and outputting the waveform data WP read from the ring buffer 15a, and the delay amount of this delay processing is set by the control unit 15d. An adder 15g adds the interpolation waveform data WP read from the ring buffer 15a and the output of the latch circuit 15f. A multiplier 15h multiplies the output of the adder 15g by a filter coefficient supplied from the controller 15d and outputs the result. An accumulator 15i accumulates the output of the multiplier 15h.

The filter operation unit 15e uses the past eleven pieces of waveform data WP stored in the ring buffer 15a, and performs high-frequency removal processing with a cutoff frequency of 25 kHz on the waveform data. This high band removal processing is a well-known FIR filter operation shown in FIG. FIG.
, D, D,... Respectively represent the waveform data generated by the interpolation unit 14 for one sampling period (1/200 kHz).
z) This is a delay processing for considerably delaying. WP _i (i
= 1 to 11) are the past 11 waveform data obtained by the delay processing, and WP ₁ is the oldest waveform data, WP 1
₁₁ is the newest waveform data. a _-5 ~a ₅ is a filter coefficient to be multiplied to each waveform data _{WP i (i = 1~11),} ACC is an accumulator for accumulating the multiplication results of the filter coefficients. As shown in FIG. 7, the sampling function sin θ / θ is a constant interval (in this example, π /
Each instantaneous value when sampling at (4 intervals) is used.

Now, focusing on the filter coefficients a _{-5 to} a ₅ , there is a relationship of a _-i = a _i (i = 1 to 5). Therefore, the filter processing shown in FIG. 6 can be equivalently modified to that shown in FIG. In the filter operation shown in FIG. 8, the waveform data corresponding to the filter coefficients a _-i and a _i are added in advance, and then multiplied by the filter coefficient a _i , thereby reducing the number of times of multiplication to half that shown in FIG. .

The filter operation unit 15e in the present embodiment
Performs a filter operation on the waveform data WP according to the procedure shown in FIG. 8 under the control of the control unit 15d. That is, the filter operation unit 15 e
The g and a multiplier 15h, the process of multiplying the filter coefficients a ₄ by adding processing for multiplying the filter coefficients a ₅ by adding the waveform data WP ₁₁ and WP _1, the waveform data WP ₁₀ and WP _2, waveform data W ₉ and WP ₃ to be multiplied by a filter coefficient a ₃ ; waveform data WP ₈ and WP ₄ to be multiplied by a filter coefficient a ₂ ; waveform data WP ₇ and WP ₅ to be added to a filter coefficient a ₁ and a process of multiplying the waveform data WP ₆ by a filter coefficient a ₀ are sequentially executed, and the results are accumulated in an accumulator 15i.
To accumulate. The accumulation result of accumulator 15i is frequency 5
The data is read out in synchronization with the sampling clock of 0 kHz and output to the digital filter 6 (see FIG. 1).

The control unit 15d sequentially outputs the read address RA of the waveform data WPi necessary for each process constituting the filter operation, and sequentially outputs the filter coefficients required for each process to the filter operation unit 15e. Also,
The filter operation is performed using a time slot in which writing to the ring buffer 15a is not performed.

B. Operation of Embodiment FIG. 9 shows a filtering process performed inside the waveform generator 5. FIG. 10 shows the interpolation unit 1 in the waveform generator 5.
6 is a time chart illustrating the operation of FIG. FIG. 11 shows a down-sampling FIR filter 1 in the waveform generator 5.
6 is a time chart illustrating the operation of Example No. 5; Hereinafter, the operation of this embodiment will be described with reference to these figures in addition to the figures already referred to.

When any key on the keyboard 1 is depressed, its key word KC and key-on signal KON are output by the key press detection circuit 2, and touch information IT representing the key press speed is output by the touch detection circuit 3. You. Then, the assigner circuit 4 selects a tone generation channel CH on which a tone generation process corresponding to the key depression operation is to be performed.

Then, in the waveform generator 5, a process of forming waveform data corresponding to the key code KC is performed as follows by using the tone generation channel selected by the assigner circuit 4.

First, the pitch code generator 13a in the address generator 13 sets the key code K in the time slot corresponding to the sound channel CH among the time slots obtained by dividing the operation cycle period of 1/200 kHz into 16 parts.
An F ^* number corresponding to C is generated. In the pitch information accumulating section 13b, the address signal corresponding to the sounding channel CH is initialized to "0" by the rise of the key-on signal KON. Thereafter, every time the operation cycle period is switched and the time slot corresponding to the sounding channel CH is reached, the pitch information accumulating section 13b accumulates the F ^* number corresponding to the key code KC, and accumulates the F ^* number. The address signal resulting from the calculation increases at a rate corresponding to the key code KC.

The integer part IAD of the address signal is supplied to the waveform memory 12, and the decimal part FAD is supplied to the interpolation unit 14. When the integer part IAD of the address signal increases, the carry signal INC is supplied to the interpolation unit 14.

In the interpolation section 14, as shown in FIG. 10, interpolation processing corresponding to each sounding channel is performed by time division processing. Each block from block A to block D is 800
In synchronization with the channel clocks φ _{A to} φ _D of kHz, interpolation processing for four channels shared by the block is executed by time division processing.

Hereinafter, the operation of the interpolation unit 14 will be described by taking as an example a case where the tone generation channel CH selected in response to the key pressing operation is the fifth channel.

First, each of the waveform data storage blocks 16a to 16a
Each shift register in 16d stores waveform data necessary for interpolation processing of each tone generation channel. With respect to the fifth channel in this example, the waveform data including the waveform data WD corresponding to the integer part IAD of the address signal of the fifth channel is included.
The pieces of waveform data are stored in the shift register for the fifth channel in the waveform storage block 16a. Here, when the integer part IAD of the address signal corresponding to the fifth sounding channel increases, it is necessary to update the waveform data used for the interpolation processing corresponding to the sounding channel. This updating is performed as follows. Done.

That is, as the integer part IAD of the address signal of the fifth channel increases, the waveform data corresponding to the increased integer part IAD is read from the waveform memory 12. The carry signal INC is generated at a timing synchronized with the fifth channel. As a result, the waveform data read from the waveform memory 12 is written to the shift register for the fifth channel in the waveform data storage block 16a. By writing the new waveform data, the existing waveform data in the shift register is sequentially shifted to the subsequent stage, and the last waveform data, that is, the waveform data input to the shift register most recently is discarded.

The fractional part FAD of the address signal corresponding to the fifth channel is delivered to and latched by the interpolation circuit 17a in the time slot corresponding to the fifth channel among the time slots obtained by dividing the operation cycle period into sixteen. You. Four time slots after the time slot corresponding to the fifth channel are required, and eight pieces of waveform data are sequentially read from the fifth channel shift register in the waveform data storage block 16a, and the interpolation circuit is used. 17
a. During this time, in the interpolation circuit 17a,
These waveform data are sequentially multiplied by eight interpolation coefficients corresponding to the decimal part FAD, and the result of the multiplication is accumulated. As a result of this accumulation, an address consisting of an integer part IAD and a decimal part FAD is obtained. Waveform data WP accurately corresponding to the signal is obtained.

The above-described interpolation processing is executed for each operation cycle of 1/200 kHz, and the interpolation processing corresponding to the key code KC is performed.
Waveform data WP having a sampling frequency of 00 kHz is output to down-sampling FIR filter 15. When a tone generation channel other than the fifth channel is assigned to the tone generation processing, the same processing as described above is performed by a component corresponding to the tone generation channel. The waveform data WP for a maximum of 16 channels is 1/200 kHz.
Is output to the down-sampling FIR filter 15 during the calculation cycle.

In the down-sampling FIR filter 15, filter processing corresponding to each tone generation channel is performed by 1/80.
It is performed every operation cycle period of 0 kHz. Each step of the filter operation is performed using a time slot obtained by dividing the operation cycle period into 16 as shown in FIG. As described above, since the waveform data WP for a maximum of 16 channels is output from the interpolation unit 14 during the calculation cycle of 1/200 kHz, the waveform data WP is sent to the down-sampling FIR filter 15 at a cycle of 1 / 3.2 MHz. Supplied. These waveform data WP
Is the first of the time slots obtained by dividing the operation cycle period of 1/800 kHz into 16 as shown in FIG.
The data is sequentially written to the ring buffer 15a using the fifth, ninth, and thirteenth time slots.

Then, the filter processing shown in FIG. 8 is performed using another time slot. That is, as shown in FIG. 11, in the second time slot and the third time slot, the oldest waveform data WP ₁ and the newest waveform data W
P ₁₁ is the filter coefficient a ₅ to be applied to these waveform data and WP ₁ WP ₁₁ with sent sequentially read from the ring buffer 15a to the filter calculation unit 15e (= a
_-5 ) is sent to the filter operation unit 15e.

The following processing is performed in the filter operation section 15e. First, waveform data WP ₁ read at the second time slot is held by the latch 15f. And when it comes to the third time slot,
A waveform data WP ₁₁ read from the ring buffer 15a at that time, the waveform data WP ₁ held in the latch 15f is added by the adder 15 g. Then, the multiplication of the filter coefficients a ₅ for this sum performed by a multiplier 15h, the multiplication result accumulator 1
5i.

[0067] Similarly, the process in the fourth and sixth time slot is multiplied by the filter coefficients a ₄ by adding the waveform data WP ₂ and WP ₁₀ are waveform data WP ₃ and in the seventh and eighth time slots The process of adding WP ₉ and multiplying by the filter coefficient a ₃ is the process of adding the waveform data WP ₄ and WP ₈ and multiplying by the filter coefficient a ₂ in the tenth and eleventh time slots.
Processing for multiplying the filter coefficients a ₄ by adding the waveform data WP ₅ and WP ₇ is executed in the twelfth and fourteenth time slots. Then, in the fifteenth time slot, the waveform data WP ₆ has a filter coefficient a ₀ = “1”.
Is performed. The results of the above multiplications are accumulated by the accumulator 15i, and the accumulated results are output.

The above-described filter processing for one channel is performed with an operation cycle period of 1/800 kHz, and waveform data from which a spectrum of 25 kHz or more has been removed is obtained. Subsequently, the filtering process is performed for the other 15 channels. Therefore, the waveform data for one channel is 800 kHz / 16 = 50 k
It is output in synchronization with the frequency of Hz.

In this manner, waveform data having a sampling frequency of 50 kHz corresponding to the key code KC is obtained.
An envelope is provided by the multiplier 8 to form a tone signal. The tone signal is accumulated by the channel accumulating section 9 for all tone generation channels, and the D / A converter 1
0, output as a musical tone through the sound system 11.

A series of procedures such as readout of waveform data from the waveform memory 12, interpolation processing, and filter processing have been described in a time-series manner. Here, the present invention is focused on a change in the spectrum of the waveform data to be processed. The operation and effect of the embodiment will be described.

The interpolation process performed by the interpolation unit 14 is a process for compensating for missing waveform data. However, focusing on a change in the spectrum of the waveform data before and after the process, it can be said that the interpolation process is a high-frequency removal process. FIG. 9A shows a pass band of the high-frequency removal processing performed on the waveform data WD by the interpolation processing. As shown in this figure, the higher cutoff frequency of the higher band removal processing increases as the F ^* number increases.

Here, according to the above-described conventional technique, the frequency of the F data depends on the spectrum of the waveform data in the waveform memory.
Even if the number was larger than 1, aliasing noise occurred. However, in this embodiment, the interpolation processing is performed at 200 k
Since the calculation is performed according to the operation cycle frequency of Hz, if the upper limit frequency of the spectrum (excluding the image spectrum) of the waveform data WP obtained by interpolation is 100 kHz or less, no aliasing noise occurs. The details will be described below.

First, it is assumed that waveform data (sampling frequency 48 kHz) having the spectrum distribution shown in FIG. 14 referred to in the description of the prior art is stored in the waveform memory 12. Suppose F ^* number = 0.25 (F number =
1), an address signal is generated by accumulating the F ^* number, and using the waveform data in the waveform memory 12,
If the waveform data corresponding to this address signal is reproduced by interpolation, the upper limit frequency of the spectrum of the reproduced waveform (excluding the image spectrum) in this case is 25 kHz.

Then, F ^* number = 0.3 (F number =
1.2), the reproduced waveform when F ^* number = 0.25 is compressed to (1 / 1.2) times along the time axis, The same waveform as that obtained by re-sampling this compressed waveform with a sampling clock of 200 kHz is reproduced.

Accordingly, the spectrum distribution of the reproduced waveform in this case is as follows. First, by setting F ^* number = 0.3, the spectrum SA of the reproduced waveform becomes
As shown in the figure, the spectrum of the reproduced waveform when the F number = 0.25 is 1.2 times (= 0.3 / 0.3) along the frequency axis.
(0.25 times). Although the image spectrum indicated by the broken line is removed by the interpolation processing, the image spectrum SB obtained by sampling at the sampling frequency of 200 kHz is newly added as the spectrum of the reproduced waveform. This image spectrum SB has a frequency range of F ^* number = 0.25.
However, since it is sufficiently separated on the frequency axis from the spectrum SA of the originally required waveform corresponding to F ^* number = 0.3, it does not overlap with each other, and aliasing noise is generated. do not do.

In this embodiment, when the upper limit frequency of the spectrum (excluding the image spectrum) of the waveform data WD in the waveform memory 12 is 25 kHz, the pitch within four times the pitch of the waveform data WD. Waveform data W
P can be generated without causing aliasing noise.

Downsampling FIR filter 15
Is a high frequency cutoff frequency of 25 kHz as shown in FIG.
9C, the waveform data W corresponding to any key code KC as shown in FIG.
The spectrum of P is also limited to a band of 25 kHz or less.
Therefore, when this waveform data WP is output at a sampling frequency of 50 kHz, no aliasing noise is generated.

In the above embodiment, the operation cycle frequency until the waveform data WP corresponding to the key code KC is obtained is four times the sampling frequency 50 kHz of the waveform data finally output from the waveform generator 5. , Other frequencies may be used. In short, what is necessary is just to configure so that the arithmetic processing until obtaining the waveform data WP corresponding to the key code KC is performed at a frequency higher than the final sampling frequency. Similar effects can be obtained. Further, the operation cycle frequency and the final sampling frequency need not be an integer ratio.

[0079]

As described above, claim 1 or claim 2
According to the invention according to the above, processing is performed in synchronization with the operation cycle frequency higher than the final sampling frequency until waveform data corresponding to the designated pitch is obtained, and a high After converting to the final sampling frequency after performing the band elimination processing and outputting it as a musical tone signal, a musical tone signal is generated without generating aliasing noise even when the designated pitch is high. There is an effect that can be. Further, according to the second aspect of the present invention, it is possible to execute an interpolation operation corresponding to a plurality of time-division channels at a high speed, so that a process for generating interpolation waveform data in a tone signal generator is performed. There is an effect that the processing can be performed without generating aliasing noise.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an overall configuration of an embodiment according to the present invention.

FIG. 2 is a block diagram showing a configuration of a waveform generator 5.

FIG. 3 is a block diagram illustrating a configuration of an address generation unit 12;

FIG. 4 is a block diagram illustrating a configuration of an interpolation unit 14.

FIG. 5 is a block diagram showing a configuration of a downsampling FIR filter 15;

FIG. 6 is a diagram illustrating the contents of a filtering process performed by a downsampling FIR filter 15;

FIG. 7 is a diagram illustrating the contents of a filtering process performed by a downsampling FIR filter 15;

FIG. 8 is a diagram illustrating the contents of a filtering process performed by a downsampling FIR filter 15;

FIG. 9 is a diagram for explaining filter characteristics of the waveform generator 5;

FIG. 10 is a time chart showing the operation of the interpolation unit 14;

11 is a time chart showing the operation of the downsampling FIR filter 15. FIG.

FIG. 12 shows waveform data W at F ^* number = 0.3.
It is a figure showing the spectrum distribution of P.

FIG. 13 is a diagram for explaining a conventional example.

FIG. 14 is a diagram for explaining a conventional example.

FIG. 15 is a diagram for explaining a conventional example.

FIG. 16 is a diagram for explaining a conventional example.

FIG. 17 is a diagram for explaining a conventional example.

FIG. 18 is a diagram for explaining a conventional example.

FIG. 19 is a diagram for explaining a conventional example.

FIG. 20 is a diagram for explaining a conventional example.

[Explanation of symbols]

12: address generation unit, 13: waveform memory, 14: interpolation unit, 15: downsampling FIR filter.

Claims (3)

(57) [Claims] 1. A tone signal generator for outputting a digital tone signal of a predetermined sampling frequency representing a tone waveform of a pitch designated by the tone generation instruction in response to a tone generation instruction. A waveform memory that stores waveform data obtained by sampling at the same time; and an address signal generating unit that outputs an address signal that changes at a rate corresponding to the instructed pitch according to an operation cycle frequency higher than the sampling frequency. Performing an interpolation operation using a predetermined number of pieces of waveform data corresponding to the integer part of the address signal of the waveform data stored in the waveform memory and an interpolation calculation coefficient corresponding to a decimal part of the address signal in the operation cycle. By executing according to the frequency, the waveform data having the designated pitch is output. Interpolating means, and a down-sampling means for performing processing for removing a spectrum higher than a predetermined high cutoff frequency from the waveform data output by the interpolating means, and outputting the result in accordance with the sampling frequency. A tone signal generating device characterized by the following. 2. Time-division control using n (n is an integer) time-division channels reads waveform data from a waveform memory for each time-division channel, and interpolates waveform data based on the read waveform data. An interpolating device for generating a waveform memory, comprising: address generating means for sequentially generating address signals corresponding to each of the n time-division channels; and reading waveform data from the waveform memory based on an integer part of each of the address signals. Interpolating means for respectively generating interpolated waveform data corresponding to each of the time-division channels in accordance with the decimal part of each address signal and outputting the interpolated waveform data in a time-division manner. k arithmetic means for executing an interpolation operation corresponding to each of k groups (k is a divisor of n); Transfer the waveform data corresponding to the integer part of the source signal from the waveform memory to one of the k arithmetic means for executing the interpolation operation corresponding to the time-division channel and the address signal of each time-division channel. Supply means for delivering a decimal part of the k arithmetic means to execute the interpolation operation corresponding to the time-division channel, and n interpolation waveform data which is an interpolation operation result of the k arithmetic means. Output means for time-division multiplexing and outputting the same. 3. A time division control using n (n is an integer) time division channels, responds to a tone generation instruction for each time division channel, and generates a tone having a pitch indicated by the tone generation instruction. A tone signal generator for outputting a digital tone signal having a predetermined sampling frequency representing a waveform, comprising: a waveform memory storing waveform data obtained by sampling a tone waveform at regular time intervals; and Address signal generating means for outputting an address signal changing at a rate corresponding to the instructed pitch in accordance with an operation cycle frequency higher than the sampling frequency; and for each of the time-division channels, stored in the waveform memory. In the waveform data, a predetermined number of waveform data corresponding to the integer part of the address signal and a decimal part of the address signal are associated. Interpolating means for outputting waveform data having the designated pitch by executing an interpolation operation using a corresponding interpolation operation coefficient in accordance with the operation cycle frequency; and for each of the time division channels, Down-sampling means for performing processing to remove a spectrum higher than a predetermined high cut-off frequency from the waveform data output by the means, and outputting the result in accordance with the sampling frequency; K operation means for performing an interpolation operation corresponding to each group obtained by dividing the time-division channels into k (k is a divisor of n); and an integer part of the address signal of each time-division channel. The corresponding waveform data is drawn from the waveform memory to one of the k calculation means for executing the interpolation calculation corresponding to the time division channel. Supply means for passing the fractional part of the address signal of each time-division channel to one of the k operation means for performing an interpolation operation corresponding to the time-division channel; and Output means for time-division multiplexing and outputting an interpolation calculation result.