JPS5842480B2 - electronic musical instruments - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an electronic musical instrument that obtains tones based on digitally stored logical orthogonal functions.

Conventionally, typical electronic musical instruments using digital technology can be known from U.S. Pat. No. 3,515,792, U.S. Pat.

U.S. Pat. It is read out repeatedly to obtain a digitally expressed musical tone, to which amplitude modulated envelopes such as attack and release are added digitally, and then converted to analog and converted into sound.

This system is also configured to use the memory in a time-divisional manner to obtain musical tones for a plurality of keys in order to achieve polyphonic synthesis.

The biggest drawback of this system is that it is extremely difficult to provide audibly desirable changes such as changes in the spectral structure of the waveform over time.

In other words, the musical tones that are generated can only be generated in waveforms that are read out from a fixed memory, and the variety of changes that occur in the musical tones of general musical instruments, such as the variety of timbres that change from moment to moment, cannot be obtained. .

The system shown in US Pat. No. 3,809,786 also solves many of the musical problems encountered in the systems described above.

The basic configuration of this system is to express the frequency selected by the memorized harmonic coefficient and key as a digital number, and calculate the Fourier components for each point individually in real time while establishing the waveform period, and then add them together at a specific time. Digitally calculates the musical tone amplitude and converts it into sound.

That is, by using a method of obtaining musical tones using the Fourier algorithm, a system with a high degree of freedom that not only solves the problems of the above-mentioned system but also facilitates the acoustical manipulation of various musical tones has been obtained.

However, a sinusoidal orthogonal function system is used to perform the inverse Fourier transform.

A sinusoidal orthogonal function system is an orthogonal function system that takes analog values, and even if it is quantized and encoded, the function value must be expressed with a constant word length.

This means that a relatively large multiplier must be used when multiplying the harmonic coefficient and each harmonic.

For example, if each harmonic amplitude is expressed in n bits and the harmonic coefficient is expressed in m bits, multiplication by Xm bits is required. If this is done, the disadvantage is that the computation time increases.

An object of the present invention is to provide an electronic musical instrument that can realize digitally diverse musical tones with a simplified configuration.

In order to achieve the above object, the electronic musical instrument of the present invention includes a first storage device that encodes and stores each coefficient for a plurality of tones;
A circuit that reads required coefficients from the first storage device and calculates a synthesis coefficient for a logical orthogonal function in order to form the timbre selected by the timbre selection device, and a plurality of waveform calculation devices that temporarily store the calculated synthesis coefficients. a second storage device that temporarily stores the synthesis coefficients given from the second storage device, and multiplies the row vector components of the logical orthogonal function by the synthesis coefficients each time an operation request signal is input; A plurality of waveform calculation devices that calculate one sample value of the waveform amplitude by cumulatively adding the multiplication results, and a calculation request is made to the plurality of waveform calculation devices at a timing that is a predetermined times the musical tone period corresponding to the key pressed. means for providing a signal, a D/A converter that converts the waveform amplitude value calculated by the waveform calculation device and expressed in digital code into an analog quantity, and an envelope added in response to key presses and key releases. It is characterized by being equipped with a circuit that generates musical tones.

The principle and embodiments of the present invention will be explained in detail below.

First, the principle of the present invention will be explained.

The musical tone synthesis method of the present invention is carried out by calculating digital sample values representing musical waveform amplitudes from stored coefficients for logical orthogonal functions according to a predetermined program.

In general, logical orthogonal function systems are well known as orthogonal function systems that take binary values (+1, -1) and are suitable for digital circuits, as opposed to orthogonal functions that take analog values such as sine wave orthogonal function systems or complex exponential orthogonal function systems. It is being

In the embodiment of the present invention, a Walsh-type Hadamard matrix, which is one of the logical orthogonal function systems, is employed.

This Walsh form Hadamard matrix is obtained by performing row permutation on the natural form Hadamard matrix, and is arranged in descending order of the number of changes in the sign of the row vector components.

For example, the 8th order Walsh-shaped Hadamard matrix [H8]
is expressed as shown in equation (1) below.

In this example, the waveform amplitude is calculated by inversely transforming the coefficient values for this orthogonal function (orthogonal matrix).
In addition to the property that transformation and inverse transformation using the Walsh-form Hadamard matrix can be performed using the same matrix operation, the generalized low-frequency components can be expressed from components with little change to components with many changes as coefficients for each row vector. It has the property of having a spectral decomposition function covering high frequency components.

In this embodiment, when calculating the waveform amplitude, the coefficients for each row vector are sifted (weighted) using a predetermined program, thereby causing a spectrum change of the musical tone that is generated from time to time, which is very audible. It is configured to generate pleasant musical tones.

Furthermore, as mentioned above, a logical orthogonal function such as the Walsh-type Hadamard matrix is a function system that takes two values, +1 and -1, and operations on coefficients and row vectors can be achieved by addition and subtraction, and are essentially exclusive OR gates. achieved.

This can greatly simplify the scale and size of the device when inversely converting the waveform amplitude.

For detailed explanations regarding logical orthogonal functions, please refer to the Institute of Electronics and Communication Engineers: Nonlinear Problem Study Group materials (material number NLP70-1).
4) is described in Tsuguhiko Kato's ``Walsh function''.

On the other hand, a method of establishing a musical tone period specific to an electronic musical instrument is carried out by periodically and intermittently operating a calculation device having a function of providing a predetermined clock and calculating a waveform amplitude value at high speed with a fixed clock.

On the other hand, if the calculation device is configured to operate with a clock that is a predetermined multiple of the musical tone period, the calculation device must be designed to operate over a relatively wide range of frequencies, which burdens the design of the calculation timing circuit. It has the disadvantage of increasing the

In other words, this calculation device is programmed to digitally calculate and output at least one waveform amplitude each time a calculation request signal is received, and to suspend calculation processing until the next calculation request signal is received, or is it equipped with dedicated hardware. Consists of clothing.

Further, the generation timing of the calculation request signal is set by a keyboard switch, and is generated at a period that is a predetermined times the musical tone period corresponding to the selected key (for example, a period of 1/64).

Therefore, when the waveform calculator receives the calculation request signal, it calculates the waveform amplitude digitally using a predetermined program, but it is configured to perform relatively high-speed calculation so that the calculation can be completed until the next calculation request signal is generated. It has become.

Also, a plurality of waveform calculators are used to perform polyphonic synthesis.

FIG. 1 is an explanatory diagram showing the structure of an embodiment of the present invention according to the above-described principle.

In the figure, key press and key release signals from a keyboard switch 1 are input to a key press detection and assignment device 2, which detects these key signals and assigns them to the waveform calculator to be used. It controls the generation of a calculation request signal, and further controls an envelope generator for performing amplitude modulation such as attack and release on the calculated waveform.

The calculation request signal generator 3 receives the allocation and control signals from the key press detection and allocation device 2, and sends a calculation request signal to the waveform calculator at a period that is a predetermined times the musical tone period corresponding to the selected key. .

Furthermore, the envelope control signal is sent to the envelope generator 4.
generates an envelope waveform for amplitude modulation of the attack, release, etc. of musical tones formed in response to key presses and key releases.

In this embodiment, this envelope waveform is generated as an analog quantity, and when converting the output of the waveform calculator (to be described later), that is, the digitally displayed musical waveform, into analog, the musical tone generated by controlling the unit quantized voltage with the envelope is generated. Used to simulate amplitude modulation such as attack and release.

-The coefficient calculation section, which is the main part of the invention of strength wood, has a coefficient memory (q
c1)5>(qc2)6, and the coefficient values for each row vector of the Walsh Hadamard matrix for calculating the desired musical sound waveform are digitally expressed and stored.

The contents of the coefficient memories 5 and 6 are stored in the coefficient calculation execution control device 8.
The reading is controlled via the address decoder 7.

Coefficient memory (qc 1 )5. (qc 2) The contents of 6 are the tuplets of the timbre selection device.) A set of coefficients selected by Ta, Tb, etc. is put into an adder 9 and summed to calculate a set of coefficients, that is, a synthesis coefficient.

For example, when the tone color tablets Ta and Tb are closed, the coefficients of the coefficient memories (qel) 5 and (qc2) 6 are summed for each row vector component of the Walsh-form Hadamard matrix and stored in the register 10. .

The first step uses the adder 9 and the coefficient register 10 to calculate the combined coefficient according to the states of the tone tablets Ta and Tb.
The operations and the calculated results are stored in the next stage coefficient buffer memory 1.
The second operation of transferring to 10 and 11□ is controlled by the coefficient calculation execution control device 8, and these two operations are repeatedly executed.

The result calculated in the coefficient register 10 in this manner is obtained by the waveform calculation unit (CUl) 1! h and (Cu2) 15□ are written into coefficient buffer memories (B1) ILl and (B2) 112 provided correspondingly.

Coefficient buffer memory (B1) 111. (B2) 112
Read and write are executed in two time periods in a time-division manner.

That is, as described above, there is a timing at which the composite coefficients of the coefficient register 10 are written and a timing at which they are read based on requests from the waveform calculators 150, 15□.

Waveform calculator (CUl) 151. (Cu2')
152 are coefficient buffer memo 1J111, 112 respectively
Based on the data from , each calculation clock (φt
)14t, (φ2)142 This is a calculator that digitally calculates the musical waveform amplitude based on the calculation timings, and performs calculations intermittently based on the calculation request signal.

Here, two waveform calculators 151.15□ are provided,
Although it is possible to generate musical tones in response to two pressed keys, the number can be increased as necessary.

In order to provide intermittent operation timing to ensure the stop and execution of the waveform calculators 151 and 152, flip-flops 131 and 13□ are provided, and are set by the calculation request signal from the calculation claim signal generator 3. Q
Controlled by output.

The waveform calculators 151 and 152 are reset by a signal indicating that a predetermined calculation has been completed.

Furthermore, readout of the coefficient buffer memories 11□ and 112 is controlled via address decoders 121 and 122, respectively, according to commands from the waveform calculators 151 and 152.

The digital output of the waveform calculator 151.15□ is converted to a digital-to-analog converter (DACI) 161° (DAC2) 1
62 to convert it into an analog quantity, and add the envelope output of the envelope generator 4 as described above.

This (DACI)161. (DAC2) 16□ is preferably a multiplication type D/A converter with variable unit quantization step voltage.

The outputs of (DACl) 161 and (DAC2) 162 are output as musical sounds from the acoustic device 18 via the summing amplifier 17.

The above is an overview of the embodiments of the present invention, and the main parts thereof will be explained in more detail below.

Now, coefficients for n Walsh-type Hadamard matrices of X1 to Xn are stored in the coefficient memory ((el)5, and the same coefficients of y1 to yn are stored in (qc2)6.

Here, n=2X (x=1.2,...).

The following description assumes that the tone tablets Ta and Tb are closed.

The coefficient calculation execution control device 8 first clears the coefficient register 10.

Next, the contents of the coefficient memory (qcl) 5 are stored in the coefficient register 10 via the adder 9.

The coefficient register 10 is preferably constituted by a shift register, the output of which is input to the adder 9, and the output thereof is inputted to the adder 9.
and coefficient register 10 constitute an accumulator.

When all the contents of the coefficient memory (qcl) 5 have been stored, the coefficient calculation execution control device 8 reads out the contents of the coefficient memory (qc2) 6, and sequentially shifts the coefficient register 10 in synchronization to calculate each composite coefficient for each row vector. 21~Z
Calculate 8.

That is, x1+yl=2. , x2+y2=Z2,...
.

The first operation described above is thus completed, and the calculation result is then sent to the coefficient buffer memo 'J110, 11□.

At this time, the coefficient arithmetic execution control device 8 generates a synchronizing signal DMA that indicates the readout status of the coefficient buffer memories 111 and 112, and a synchronization signal DMA that indicates the read status of the coefficient buffer memories 111 and 112. .

1120 A signal MP that separates two times, that is, a time used for writing and a time used for reading, and a signal BW that identifies whether or not the coefficient register 10 is in the second operation, that is, transferring synthesis coefficients, are connected to line 12. A waveform calculator (CUI) 151. (Cu2) Notify 152.

Further, the synthesis coefficients are sent to the coefficient buffer memory (B1) 111 via line 12. (B2) Input to 112.

The coefficient buffer memories 111 and 112 perform different operations at two times.

FIG. 2 shows the timing relationship between the above-mentioned signal MP, synchronization signal DMA, and coefficient readout signal RMA.

That is, the signal MP periodically changes "1" and "O" at high speed to form a first timing for writing and a second timing for reading.

The synchronization signal DMA stores the threat coefficient sent from the coefficient register 10 in the coefficient buffer memory January 1. ．． 1120, the address of the coefficient buffer memory 110 is specified at a second timing via the data selector 31 in the waveform calculator 15 of FIG. .

At this time, if a coefficient is being sent out, a signal B identifies it.
W becomes "1", the write signal is set to "1" via the AND circuit 32, and a new coefficient is written to the predetermined location at the first timing.

On the other hand, reading of the coefficients for waveform calculation is selectively performed at the second timing using the read signal RMA from the arithmetic execution program control device 34.

Next, FIG. 3 shows a detailed circuit example of the waveform calculator 15 (150, 15□).

That is, the following calculations are performed in the buffer memory 11.
Waveform calculation is performed using the synthesis coefficient values read from the , but in order to lower the pitch, the calculation clock is prohibited by the flip-flop 13, and waveform calculation is performed intermittently, and the musical sound waveform amplitude is digitally expressed. is calculated.

Two calculation methods for calculating the musical waveform amplitude in a waveform calculation unit using a Walsh-type Hadamard matrix will be described below, and first the principle and embodiment of the first calculation method will be explained.

The n-th order Walsh-type Hadamard matrix is expressed by the following equation.

[Hn]-[hij:i, j-1,2,...
..., n] (2) where hij = +1 or -1 Waveform amplitudes S1, S2, ..., calculated when the threat coefficient is Zl, z2, ......, Zn・・・
..., Sn are calculated by the matrix operation shown by the following equation.

At this time, the synthesis coefficients z1, z2, ......, zn
is scaled (weighted) temporally.

In other words, the scaling value given by the time function is Z1
~ Let αn(t) correspond to Zn, then αt(t)Zt, α2(L)Z2, etc. for each coefficient
- The spectral structure of the generated musical tone is made variable by performing the multiplication αn(t)Zn.

This brings about a variety of audible tonal changes.

A formula for executing formula (3) is shown below.

The procedure for calculating a musical tone waveform by the method shown by equation (4) will be explained using the circuit diagram shown in FIG.

Now, when the calculation request signal is generated, flip-flop 1
3 is set and A is sent to the calculation execution program control device 34.
Operation clock φ1. φ2 is given.

At this time, the calculation is always executed at the second timing.

First, the arithmetic execution program control device 34 receives the read signal R.
Read coefficient 21 by MA.

In addition, the αn(t) memory 36 is sent via the address decoder 35 by the read control signal ADC of the time function αn(1).
Read out α, (t) and use multiplier 40 to calculate α1(j)
A multiplication of Zl is performed.

Further, in response to the control signal WDC, the coefficient h is sent to the Walsh-type Hadamard matrix memory 42 via the address decoder 43.
ll is read out and multiplier 41 calculates αx(t)Ztht
x is calculated.

Since this multiplier 41 only takes either +1 or -1 for hij, it can essentially be a complementer or an exclusive OR gate, and the purpose can be achieved with a very small number of gates.

αx(t)Z t h t t is stored in register 46 via adder 44 .

Next, the arithmetic execution program control device 34 performs RMA, ADC
22, α2(t) and h12 are read out by WDC and α2(t)Z 2 h 12 is calculated, and the adder 44 calculates α1z1 hll previously calculated.
The result is stored in the register 46.

In this way, the first sample point S at the generated waveform amplitude is calculated sequentially,
is found.

Then, 81 is stored in the register 47, and its output is D.
It is converted into an analog quantity by AC16.

Next, the operation execution program control device 34 clears the register 46, resets the flip-flop ★★13, temporarily stops the operation, and maintains this state until the next operation request signal is generated.

That is, one waveform sample value is calculated every time a calculation request signal is generated.

The table below shows the calculation request signal τK (K=1.2.3,...
. . .) shows the execution process of the calculation.

FIG. 4 shows a detailed circuit example of the arithmetic execution program control device 34 for executing the above-mentioned arithmetic operations.

In the figure, an input is input from an AND gate 33, an n-ary counter 51 and an n-ary counter 52 are connected in series, and a q-ary counter 54 is connected via a changeover switch (SW2) 53 that switches between the MSB of the n-ary counter 52 and an external clock. Connecting.

The outputs of each stage of the n-ary counter 51, n-ary counter 52, and q-ary counter 54 are RMA, WDC, and AD, respectively.
Take out as C.

A connection point between the n-ary counter 51 and the n-ary counter 52 is connected to the register 47 and the flip-flop 13 via the monostable multivibrator 56, and further to the register 46 via the monostable multivibrator 57.

Further, the q-ary counter 54 is reset by switching the input level of the monostable multivibrator 55 using a key switch (Kl) 58.

With this configuration, each counter 5L52, 54 counts with the operation clock from the AND gate 33, and each memory, that is, the coefficient buffer memory, the Walsh-type Hadamard memory, and the αn(t) meseri outputs each counter R.
Data zn, hij, αn(1) calculated according to the order of the equations shown above according to MA, WDC, and ADC are output.

At the same time, the storage control Hals Ad of the register 46
, a reset pulse Cp for the register 46, a storage control pulse Sp for the register 47, a reset pulse for the flip-flop 13, etc.

The changeover switch (SW2) 53 is the M of the n-ary counter 52.
This switch selects whether to input the clock from the SB to the q-ary counter 54 or to have the counter 54 count using an external clock.

The q-adic counter 54 is used to select a set of αri(t) over time, where q sets of αn(t) can be specified, and waveforms having essentially q types of spectral structures are periodically Repeated.

The part 60 surrounded by the dashed line is the key switch (K1) 58
It is used to give a change only during the first predetermined time (several milliseconds) of the musical tone produced by the monostable multivibrator 55. The four outputs are set to a low level to enable the q-ary counter 54 to operate.

Next, a second calculation method for calculating the musical waveform amplitude in a waveform calculation unit using a Walsh-type Hadamard matrix will be explained.

This calculation method uses a relatively low-order Walsh-type Hadamard matrix (H) = (hij; J 1.2, m) when calculating one waveform amplitude value, for example, S1. = Waveform amplitude value generated using n and Sl, S2,...
・, Sn and coefficients z1, z2, ......, zn
correspond as follows.

From the above equation (8), similarly to the first method, the calculation request signal τK
(K = 1.2, ......) The procedure for calculating the waveform amplitude ※※Sl, S2, Sn is as follows: When calculating one sample value, the first
For the method, the steps are reduced to 1/m.

As for the configuration of this second method, in FIG. 3, a switch (SWl) 37 is inserted on the input side of the multiplier 41, and the contact P on the multiplier 40 side is short-circuited and connected directly to the coefficient buffer memory 11. It is designed to switch to a y-contact.

In this way, as shown in equation (9), the procedure can be simplified by omitting the temporal coefficient αn(t).

FIG. 5 shows a detailed circuit example of an arithmetic execution program control device 34 corresponding to FIG. 4 for executing arithmetic operations by this second arithmetic method.

In the same figure, input is made from an AND gate 33, m-ary counters 61, 62, and 63 are connected in series, and m-ary counter 6L
WDC is taken out from 62, RMA is taken out from m-ary counter 63, and monostable multivibrators 56 and 57 are connected between m-ary counters 61 and 62 as in FIG. 4.

The operation is almost the same as that in FIG. 4, but the time coefficients are omitted, so the configuration can be simplified depending on the procedure.

As explained above, according to the present invention, the coefficients corresponding to the timbre selected by the timbre selection device are read out from the storage device stored in advance, the synthesis coefficients of a logical orthogonal function, for example, a Walsh-type Hadamard matrix, are calculated, and the Send the coefficients to multiple waveform calculation devices to calculate the amplitude value of the desired timbre waveform,
It is designed to operate at a timing that is a predetermined times the musical tone period corresponding to the pressed key.

With such a configuration, by arbitrarily setting the coefficients, an arbitrary tone waveform corresponding to the coefficient can be obtained, and as described above, the spectral structure of the waveform can be changed over time.

Furthermore, as described above, polyphonic synthesis can be performed using a waveform calculator that performs multiple intermittent operations, and a variety of tones similar to those of general musical instruments can be realized with a simple configuration.

[Brief explanation of the drawing]

FIG. 1 is an explanatory diagram showing the configuration of an embodiment of the present invention, FIG. 2 is an explanatory diagram of the timing used in the embodiment of the present invention, and FIG. 3 is a detailed explanatory diagram of the main part of the embodiment of FIG. Figures 4 and 5
The figure shows a more detailed circuit example of FIG. 3, in which 1 is a keyboard switch, 2 is a key press detection and assignment device, 3 is a calculation claim signal generator, 4 is an envelope generator,
, 6 is a coefficient memory, 7, 121.122 is an address decoder, 8 is a coefficient calculation execution control device, 9 is an adder, 10
are coefficient registers, 118 and 112 are coefficient buffer memories, 138 and 132 are flip-flops, 14□, 14□
is the calculation clock, 151, 15□ is the waveform calculator, 168
．． 16□ is a D/A converter, 17 is a summing amplifier, 18 is an audio device, 31 is a data selector, 34 is an arithmetic execution program control device, 36 is an αn(t) memory, 37 is a changeover switch, 40° and 41 are multiplication units. 42 is a Walsh-type Hadamard matrix memory, 44 is an adder, and 46 and 47 are registers.

Claims (1)

[Claims] 1. A first system that encodes and stores each coefficient for a plurality of tones.
a storage device, a circuit for reading out necessary coefficients from the first storage device to form the timbre selected by the timbre selection device and calculating a synthesis coefficient for the logical orthogonal function, a circuit for temporarily storing the calculated synthesis coefficients and a plurality of them; a second storage device that sequentially sends out the combination coefficients to the waveform calculation device; temporarily stores the synthesis coefficients given from the second storage device; and stores the row vector fraction of the logical orthogonal function and the synthesis coefficients every time the calculation request signal is input; a plurality of waveform calculation devices that calculate one sample value of the waveform amplitude by performing each multiplication and cumulatively adding the multiplication results; a D/A converter that converts the waveform amplitude value calculated by the waveform calculation device and expressed in digital code into an analog quantity;
and a circuit that generates musical tones with envelopes added in response to key presses and key releases. 2. A third storage device in which the waveform calculation device has a timing for selectively sequentially writing the calculated and sent out synthesis coefficients and a timing for reading out the coefficients as necessary; and a fourth storage device for storing the logical orthogonal function. a storage device, a multiplier that multiplies the coefficients read from the third and fourth storage devices by a logical orthogonal function, a circuit that cumulatively adds the outputs of the multiplier a predetermined number of times, and an operation request signal. 2. The electronic musical instrument according to claim 1, further comprising means for calculating within a certain period of time the amplitude of one waveform of a musical tone generated each time the musical tone is generated, and stopping the calculation until the next calculation request signal is received. . 3. The electronic musical instrument according to claim 1, wherein the circuit for calculating the synthesis coefficients for the logical orthogonal function generates the coefficients for the Walsh-type Hadamard matrix by weighting them with respect to time.