JPS61141497A - Waveform generator for electronic musical instrument - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of Industrial Application The present invention relates to a waveform generator for an electronic musical instrument, and in particular uses an inverse fast Fourier transform (IFFT) to generate more complex tones. The present invention relates to a waveform generator for an electronic musical instrument that generates musical sound waveforms through pipeline processing.

Prior Art and Problems Conventional electronic musical instrument generators include large-scale integrated circuits (L).
There are some sound source circuits having a digital circuit configuration suitable for SI (for example, Japanese Patent Publication No. 52-31731, Japanese Patent Application Laid-Open No. 58-65493, etc.).

However, since these conventional devices use a slow-operating read-only memory (ROM) as a frequency storage device that sends frequency information according to the key address code, the waveform to be generated is fixed and the waveform If it is difficult to make changes or to save data, you can generate approximate waveforms and synthesize them, similar to using a waveform memory with a large address with a small waveform memory. Generates waveforms with high frequency variable accuracy
1, there were problems such as difficulty in producing a natural tone.

Therefore, the present invention provides an electronic musical instrument that solves the above problems by connecting a plurality of column units in series to perform IFFT calculations, and thereby selectively outputting data obtained by converting input frequency information into time information. The purpose of the present invention is to provide a waveform generator.

Means for Solving the Problems The waveform generation device for an electronic musical instrument according to the present invention includes an inverse fast Fourier transform section and a DA conversion and selection output means. The inverse fast Fourier transform section is supplied with frequency information data of the musical tone signal waveform to be generated, and a plurality of column units for performing butterfly calculations at each stage of the inverse fast Fourier transform are connected in series, corresponding to the number of stages of the inverse fast Fourier transform. Ori,
Convert input frequency information data to time information data. Further, the DA conversion and selection output means selects one data set from the output time information data sets of the inverse fast Fourier transform section, converts it into an analog signal waveform, and outputs it as a musical tone signal waveform.

The inverse fast Fourier transform (I
FFT>, there are many documents (e.g., E, O, BR
"Fast Fourier Transform" by IGHAM, jointly translated by Miyaji and Miyaji,
Since it is well known, detailed explanation will be omitted, but its principle is the discrete Fourier transform (OFT). and its inverse transformation (
10FT: Inverse Discrete Fourier Transform This is an algorithm that reduces the number of multiplications and shortens the calculation time by dividing it into multiple times.In equations (1) and 2, xp is the sample value series of the time function, and Xk is the sample value sequence of the frequency spectrum. value series, and p and k are respectively 0 to N-1
Indicates a positive integer in the range up to . Also, in the formula ('l)~■,
When pk@:S is set, Wpk is expressed as wplc = ws = cos ((2πs)/N) −jsin((2πS)/N) (4). This Wl is a variable called a phase rotation factor.

When calculating the width expression directly, W is a complex number and x
Since p may also be a complex number, it requires Nz multiplications of complex numbers and N(N-1> additions of complex numbers), whereas according to FFT, the number of data N in the input data string is N/2, which is half, and 1, which is the number of stages (number of columns in the calculation example)
It requires the number of multiplications represented by the product of 0(12N) and the addition of twice as many complex numbers. Therefore, assuming that the calculation time is proportional to the number of multiplications, FFT requires significantly more calculations than the above direct calculation method. It can save time.

Similarly, IFFT calculation can also significantly reduce calculation time compared to IDFT.

Since the one column unit in the inverse fast Fourier transform section is configured to be able to constantly perform the butterfly calculation of one predetermined stage, the inverse fast Fourier transform unit in which this column unit is connected in series for the number of stages The section allows continuous pipeline processing of sequentially arriving input frequency information data. Hereinafter, each embodiment of the present invention will be described with reference to the drawings.

Embodiment FIG. 1 shows a block system diagram of a first embodiment of the apparatus of the present invention. In the figure, frequency information data of a musical tone signal waveform to be generated (this is output from, for example, a central processing unit (CPU) within an electronic musical instrument) is input to an input terminal 1. This frequency information data indicates that the musical tone signal to be generated is shown in Fig. 2 (A).
If it has a frequency spectrum as shown by f1 to r6, then the real part data and imaginary part data are generated based on the frequency (fI is the fundamental wave + 12 to f6 are harmonics), amplitude, and phase. It consists of data.

These real loss part data and imaginary part data are each 16 bits, for example, and are supplied to the 32-bit parallel data input terminal 1, and the address signal and access signal are input from the input terminal (not shown) to the inverse fast Fourier transform unit. 2 and controller 3, respectively.

The inverse fast Fourier transform unit 2 has one column unit 2-1 to 2-m connected in series, and a predetermined phase circuit to the column units 2-1 to 2-m separately based on the output signal of the controller 3. and a coefficient memory 2-0 supplying the factor W. The above-mentioned value 1 is selected to be the number of stages 1002N, where N is the number of input data. Further, each of the column units 2-1 to 2-1 has the same circuit configuration, for example, as shown in FIG. 3.

In FIG. 3, input data consisting of real part data and imaginary part data inputted to an input terminal 10 is supplied to two sets of memory units 11 and 12, respectively. The memory unit 11 consists of N/2 stage shift registers 11a and 11b, where the number of data is N, and the memory unit 12
Similarly, it is made up of N/2 stage shift registers 12a and 12b. Note that the output terminals of the shift registers 11b and 12b are also connected to the input terminals of the shift registers 11a and 12a. Memory units 11 and 12 are used for data sorting and buffering for synchronization with the next stage.
Each has two functions. The selector 13 is supplied with the Il]llll signal from the controller 3 to an input terminal 17 and supplies both output data from the shift registers 11a and 11b to the butterfly operation circuit 14, or outputs both output data from the shift registers 12a and 12b to the butterfly operation circuit 14. The signal is supplied to the arithmetic circuit 14. Two sets of memory units are provided as shown in 11.12 in the butterfly calculation circuit 1.
This is to keep 4 in constant operation.

The butterfly arithmetic circuit 14 is supplied with the phase rotation factor of a predetermined stage to this column unit from the input terminal 15, and is also supplied with real part data and imaginary part data from the selector 13, respectively, and performs butterfly computation of inverse fast Fourier transform. and send the obtained data to output terminal 1.
Output to 6.

As shown by 2-1 to 2-1 in FIG. 1, the number of column units having such a configuration is equal to the number of stages, and since they are connected in series, sequentially arriving data is processed in a pipeline. be able to. Such configurations are known per se (for example interplates).

Takahashi: “Modular one-dimensional array FFT processor”
, Proceedings of the 29th National Conference of the Information Processing Society of Japan.

P, 131-132).

Note that the circuit section consisting of the memory units 11 and 12 and the selector 13 in FIG.
It is also possible to perform the IFFT calculation using a so-called in-brace type algorithm.

The time information data extracted from the final stage column unit 2-ra in FIG. . Each of the section memories 4-1 to 4-n has a storage capacity capable of writing time-open information data for a period of data output length T, and only one of them can be read by the output read clock pulse of the controller 3. one or more section memories perform a write operation by the corrupted clock pulse. Here, section memory 4-1 to 4-
The reason why nWA (usually n = 2) is provided is that the time information data set of 1 after the IFFT calculation has been completed in the inverse fast Fourier transform unit 2 is being read from the section memory of 1. This is because the next time information data set may be retrieved after the completion of the IFF* calculation, so it is necessary to store this data. In addition, in this embodiment, when one time information data set of the superposition method is written to one interval memory, at that time, one time information data set immediately before it is read from another interval memory. Even during the process, the next time information data set written in the first section memory starts to be read out immediately. 1. For example, the time information data set shown in FIG. 2(B) is stored at time t from the section memory 4-1.
In the case where data is read out for a period of data output length T from a to j+, assuming that the time information data set shown in FIG. The period up to is section memory 4.
The time information data is read from time t2, and the latest time information data starts being read from the section memory 4-2 from time t2. Therefore, the time information data output from the memory section 4 to the DA converter 5 becomes as shown in FIG. 2(D).

For convenience, the signal waveforms shown in FIGS. 2(B) to 2(D) are
Although shown as an analog signal waveform, the section memory 4-1
It goes without saying that the output signal waveforms of 4-n are actually binary digital signal waveforms.

In this way, if the next IFFT operation result is obtained in the middle of the data output f%T obtained from one IFFT operation, the latest IFF is immediately updated as soon as it is written to the interval memory.
To output the T calculation result, Zunawara time information data,
The output can be switched immediately regardless of the data output length T. Therefore, the timbre can be changed moment by moment by specifying another tone even as a sub-tone after the specified tone begins to be output until the specified tone disappears. In addition, the IFFT calculation speed can currently be approximately 1ms to 1018 times, so the real-time performance required for switching sounds during a performer's performance (the time it takes from specifying a sound until the sound begins to be produced) is +ms
degree) will be ensured sufficiently.

The time information data selectively output from the memory 4 as described above is digital-to-analog converted by the DA converter 5 into an analog signal, and then converted into an analog signal by the low-pass filter 6.
The IIF wave is further outputted to the output terminal 8 through the amplifier 7 as a musical tone signal waveform.

The output musical tone signal waveform of the output terminal 8 is output from the speaker (shown in the figure).

Pronounced with higher quality and natural tone than without).

In the first embodiment described above, only one piece of time information data is output from the memory unit 4, so although it is possible to generate a complex waveform, it is limited to only a single sound, and multiple sounds (multiple voices) can be generated. It is not possible. In contrast, the second and third embodiments described below are waveform generators capable of generating and outputting multiple sounds (multiple voices).

FIG. 4 is a block system diagram of the second embodiment of the device of the present invention. In the figure, the same components as in FIG. 1 are denoted by the same reference numerals, and their explanations will be omitted. The n series of time information data selectively read out from the section memories 4-1 to 4-n are supplied to the corresponding DA converters 18-1 to 18-0. The DA converters 18-1 to 18-n constitute the [)A converter 18, which performs digital-to-analog conversion on input time information data, and sends the iqed analog signal to the corresponding fader 19-1-19. −Supply to 〇. Fader 19-1
・-19-0 are electronic volume controllers each having a known configuration, and can be turned into either the on state or the off state independently of each other by a control signal from the controller 3.
When in the on state, the output analog signal level is attenuated over time and finally becomes zero, giving the input analog signal a time attenuation characteristic, and when in the off state, the level of the input analog signal is not attenuated.
It is configured to pass through as is. Therefore, for example, time 1. During the period ■ from tl to tl, the DA converter 18
-1, when an analog signal is output as shown in FIG. 5(A), the output of the analog signal is started from the DA converter 18-2 as shown in FIG. 5(B) at time t2 on the way. , the fader 19-1 is set at time t2.
At time t4 corresponding to (usually jz = j4), switching from the off state to the on state is controlled. Therefore, the envelope of the output analog signal of the fader 19, -1 attenuates from time t4 as shown by broken lines 1a and Ib in FIG. 5(A), and becomes zero at time ts.

The output analog signals of the faders 19-1 to 19-n are supplied to an adder 20, where they are added and synthesized and then output to the low-pass filter 6. Therefore, in this embodiment, the above time t
During the period from 4 to t5, the output analog attenuation signal of the fader 19-1 is superimposed on the output analog signal of the fader 19-2.

In this way, for example, at time t2, the fader 19-1
The next sound is input at time t4, and the previous sound is input at time t4.
When the sound is turned off, fading is performed to give the previous sound a time decay characteristic. The faders 19-2 to 19-n also perform similar fading.

This fading can also be created by manipulating the input data in the inverse fast Fourier transform section 2, but it is better to provide independent faders for each system like 19-1 to 19-n for better real-time performance. This is desirable because complexity for input data manipulation is avoided.

Note that the plurality of systems n may or may not match the number of voices, and need not be fixed to match the number of voices. It is sufficient to use vacant systems one after another.

Next, a third embodiment of the device of the present invention will be described with reference to a block system diagram shown in FIG. In FIG. 6, the same components as those in FIG. 4 are designated by the same reference numerals, and their explanations will be omitted. In this embodiment, the faders 21-1 to 21-n have a digital electronic volume configuration, for example, a shift register, and the shift position is controlled every moment by control signals from the controller 3, and the input digital After selectively imparting a level attenuation characteristic to the digital level information (time information data), the digital level information is supplied to the DA converters 22-1 to 22-n constituting the DA converter 22. That is, in this embodiment, compared to the second embodiment, the connection order of the DA converter and the fader is changed, and the fader 19-
This embodiment has the simple feature that 1 to 19-n are configured as circuits for performing digital signal processing as shown by 21-1 to 21-n, and has the same effects as the second embodiment.

Next, a fourth embodiment of the device of the present invention will be described with reference to the block system diagram shown in FIG. In FIG. 7, the same components as those in FIG. 4 are designated by the same reference numerals, and their explanations will be omitted. In FIG. 7, sampling variable setters 23-1 to 23-n are provided corresponding to DA converters 18-1 to 18-0, respectively. Sampling variable setter 23-1 to 23-n
variably controls the sampling frequency of the corresponding DA converters 18-1 to 18-n based on the output control signal of the controller 3. Here, the sampling frequency is the DA conversion speed of the input sample values of the DA converters 18-1 to 18-n, and the T changes according to the sampling frequency. Therefore, for example, depending on the sound ^, the sampling variable setting device 23
-1 to 23-n, the sampling frequency of the corresponding one DA converter 18-1 is varied by a predetermined sampling variable setter 18-1. Data regarding this pitch is included in the input data 3 of the input terminal 1 and is identified by the controller 3.

According to this embodiment, a plurality of musical scales can be obtained by varying the sampling frequencies of the DA converters 18-1 to 18-n using one input data. That is, for example, for one input data, it is not necessary to share a scale a semitone higher, a scale a quarter tone higher, and a scale lower a quarter tone, and provide various input data corresponding to all the scales, so the standard pattern of input data ( (prepared in advance on the input data transmitting side) can be reduced.

In addition, in FIG. 7, if time information data that is a calculation result is output from any one of the section memories 4-1 to 4-n, the time information data is not supplied. Only the DA converter 18-1 is operated by the output of the sampling variable setting unit 23-1, and during its operation, the DA converter starts reading out the latest time information data from the other section memory 4-j at the same time. Vessel 18-1
The operation of the DA converter 18-j is stopped and an analog signal is switched and outputted at the designated sampling frequency only from the DA converter 18-j. Therefore, the adder 20 includes the DA converters 18-1 to 18-1.
An analog signal is supplied from only one of the DA converters 18-0.

However, by providing faders 19-1 to 19-n between the DA converters 18-1 to 18-n and the adder 20, a plurality of voices can be obtained.

In each of the above embodiments, the inverse fast Fourier transform unit 2. Controller 3. Memory section 4.0A conversion section 18.22. Faders 21-1 to 21-n, sampling variable setting device 2
3-1 to 23-n, etc. can be constructed with digital circuit elements without using analog circuit elements such as capacitors and resistors, and are therefore suitable for large-scale integration (LSI).

Effects of the Invention As described above, according to the present invention, the IFFT operation from input frequency information data to time information data can be pipeline-processed using a single system of inverse fast Fourier transform units. New input data one after another
FFT calculation can be started, and a musical tone signal waveform with a natural tone, that is, very close to the intended original sound, can be generated, and the tone can be easily controlled by slight changes in input data. Since time information data, which is the latest IFFT calculation result, is selectively output, it is possible to specify another sound and change the timbre from moment to moment even from the time the specified sound starts to the time it disappears. It is possible to ensure real-time switching of the tone,
In addition, the switching time is kept constant, making it easy to play, and since it uses a digital signal processing method, it is suitable for large-scale integrated circuits.Furthermore, since it is configured with a fader, it is possible to obtain time decay characteristics with sufficient real-time performance. Furthermore, since the sampling variable setting device is provided corresponding to each of the plurality of systems of DA converters, it has many features such as being able to share input data for a plurality of scales.

[Brief explanation of drawings]

1, 4, 6, and 7 are block system diagrams showing each embodiment of the device of the present invention, and FIG. 2 shows a frequency spectrum and signal waveform for explaining the operation of the block system shown in FIG. 1. FIG. 3 shows one example of the essential parts of the device of the present invention.
FIG. 5 is a signal waveform diagram for explaining the operation of the block system shown in FIG. 4. 1... Input terminal, 2... Inverse fast Fourier transform unit, 2
-0...Coefficient memory, 2-1 to 2-1...Column unit, 3...Controller, 4...Memory section, 4-1
~4-n... section memory, 5.18-1 ~ 18-n,
22-1 to 22-n...DA converter, 8...Music signal waveform output terminal, 11.12...Memory unit, 1
3...Selector, 14...Butterfly calculation circuit, 1
5... Phase rotation factor input terminal, 17... Control signal input terminal, 18.22 DA conversion section, 19-1 to 19-n
. 21-1 to 21-n...Fader, 20...Adder, 23-1 to 23-n...Sampling variable setting device.

Claims (4)

[Claims] (1) Frequency information data of the musical tone signal waveform to be generated is supplied, and a plurality of column units that perform butterfly calculations at each stage of the inverse fast Fourier transform are connected in series for the number of stages of the inverse fast Fourier transform, and the input frequency An inverse fast Fourier transform unit that converts information data into time information data, and an analog signal of a predetermined data set selected from among the time information data sets extracted from the inverse fast Fourier transform unit. 1. A waveform generator for an electronic musical instrument, comprising: DA conversion and selection output means for outputting a musical tone signal waveform as a musical tone signal waveform. (2) The DA conversion and selection output means is provided in parallel to the output end of the inverse fast Fourier transform section, and converts the time information data extracted from the inverse fast Fourier transform section into data output lengths. A patent claim characterized by comprising a plurality of section memories that are temporarily stored in sequence, and a DA converter that sequentially converts the time information data sequentially taken out from the plurality of section memories into digital-to-analog and outputs the resultant data. A waveform generator for an electronic musical instrument according to item 1. (3) The DA conversion and selection output means is provided in parallel to the output end of the inverse fast Fourier transform section, and converts the time information data extracted from the inverse fast Fourier transform section into data output lengths. A plurality of systems are provided corresponding to a plurality of section memories that are stored once in sequence and the output sides of the plurality of section memories, respectively, and convert the output time information data of the section memories from digital to analog, and D that separately outputs analog signals to which attenuation characteristics are selectively applied.
Claim 1, characterized in that it consists of an A converter, a fader, and an adder that adds and synthesizes analog signals taken out from a plurality of systems of the DA converters and faders and outputs the resultant signal as a musical tone signal waveform. The waveform generator for the electronic musical instrument described above. (4) The DA conversion and selection output means is provided in parallel to the output end of the inverse fast Fourier transform section, and converts the time information data extracted from the inverse fast Fourier transform section into data output lengths. A plurality of systems are provided corresponding to a plurality of section memories that are stored once in sequence and the output sides of the plurality of section memories, respectively, and the output time information data of the section memories is at least separately digital-to-analog converted. a DA converter that outputs the plurality of analog signals;
It is characterized by comprising a sampling variable setting device that is provided corresponding to each of the plurality of DA converters in the A conversion section and changes the sampling frequency of the corresponding DA converter according to the input frequency information data. A waveform generator for an electronic musical instrument according to claim 3.