JPS6014358B2 - electronic musical instruments - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an electronic musical instrument, and more particularly to a digital type electronic musical instrument that forms a musical tone signal through regression type calculations.

Recently, it has become possible to obtain not only steady musical sounds whose pitch, volume, and timbre are constant over time, but also unsteady musical sounds such as those of natural instruments whose pitch, volume, and timbre change in a complex manner over time. Electronic musical instruments are being developed.

As one of the electronic musical instruments that achieves this purpose, one equipped with a regression type calculation device has been proposed (for example, Japanese Patent Application No. 50-63218). By the way, musical tones are generally considered to consist of partial tones that are elements of various frequency components. Therefore, according to the above-mentioned type of electronic musical instrument, the musical tone Ft
In order to obtain sample values for each time interval T, first, sample values fm0IT) (n=0, 1, 2,... This is done by independently performing a regression operation on .) and then calculating the sum of these M sample values. Next, a description will be given of a regression calculation for forming partial functions that constitute the above-mentioned partials.

In other words, the regression calculation is fm(nT)=pm・fm{
(n-1)T} 10qmfm{(n-2)T}.・．． ．．・
・{1} or in general, K fm(nT)=k≧・akfm{(n−k)T}.・・・
．． . . . It calculates the regression equation represented by {2'.

In other words, regarding equation {1}, the m-th partial sample value fm(nT) at time t=nT is expressed as the previous two times t=(n-1)T and t=(n-2)T. It is formed as the sum of the sample values multiplied by the coefficient values pm and qm, respectively, and formula (2) is a generalization of this. It is known that the partial sound waveform consisting of a series of sample values formed by such a regression operation takes various functional forms, and in particular, the solution to the second-order difference equation shown in equation (1) is as follows. It is mathematically known that there are three ways.

If pm2 14qm>0, fm(nT)=Aa, nT+Ba2nT...
...(3}Here a・=pm+No.m2 14qma2〒p
m - year + 4qm * * A, B: constant pm2 determined by the initial value of fm (nT) If m = 0, then fm (nT) NiAgeT+B style nT, . ．． 、．
．． t4} where b=signs A, B: constant pm2 determined by the initial value of fm(nT); cosn's T1B is QTsi
T of nn=Ceno? sin (T+8 of n).・・・．・
・{5} Here.

=亥n(-qm)中COS-12-Blue A, B, C; fm
Constant Determined by the Initial Value of (nT) Among the above-mentioned function forms, the one most commonly used as a partial is the form of the equation {5}.

In equation (5), enQT is for attenuation, and equation (2) is basically a sine wave.Here, it will be proven that the regression operation of equation {1) holds true for a sine wave.

sin (Q+6) ten sin (o18) = public inocos
8, so if we set Q=(n-1)wt B=wt, we get t of Sin, t of Sin<n-2), t of tCOS of 2 in(n-1).・．． t of sin = t of tcos of this in (n - 1) - t of sin (n - 2) Furthermore, even if there is a phase shift in the sine wave (which is not a problem in terms of intelligence), the same regression occurs. The operation is valid.

That is, sin {(Q+3) 18} 10 sin {Q-8
) 18} = sin Q 16) cos 8 10 cos (Q + 3
) sina ten sin (Q-8) cos eight ten cos (Q-
8) sin0 approximately = cos8
{sin (Q+B) ten sin (Q-8)} ten sin
{cos(.

10B)+cos(Q-8)}=×osaslnQcog
Since 30 inocosQcos8, q=(n-1
) If we set t of 8=t, then Sin (t of n + 8) + Si
n {t+8 of (n+2)}=2cosasin(n-1
)'s tcos's t + book inocos (n-1)'s os
t2COS of is jn{t+a of (n1・)}.・．．
sin(n of t+8): Proved by os of tsin{n-1) of t1}-sin{(n-2)wt+8}.

Note that a periodic trigonometric function can be formed as a special case of Equation '3 or Equation (4), and the same regression operation holds true in this case as well.

Next, the past of forming a partial sound waveform is shown.

That is, as a partial function constituting the m-th partial, for example, mt of fmt=AmeQmtcos is expressed by the formula {1}
In order to form the two initial values by the operation of fm
(0)=Am, fm(T)=mT of Ame2QmTcos, and the coefficient value is pm=2×m of eQmtcos.
If we perform regression calculation with t, qm=-e2Qmt,
At each time t=2T, 3T,..., fm (rainbow)=Am
mT of e2Qmrcos2 ** fm (rainbow) = mT of Ame3omTcos3: :fm(NT
) AmeNQmTCOSN mT result is obtained.

Such calculations are performed for each of the M partials (actual calculation devices operate in a time-division multiplexed manner), and the sample value of the musical tone F(t) is calculated as the sum of these M function values. .

That is, M F(0)=m≧, fmo=m≧, Am M M F(0)=m≧, fmT=m≧, m of AmeQmTC0S
TM MF (public) = m≧, fm (phantom) = m≧, Am
e2omTcos20mT: : :M
MF(NT)nim≧・ fm(NT)nim≧lAmeNQ
A sample value time series mT of mrCOSN is obtained.

As can be seen from the above explanation, the electronic musical instrument that forms musical tones in the above manner includes a calculation device that performs regression calculations, and M coefficient values that are independently set for each of the M partials during the calculation. It is comprised of a memory that stores pm and qm, and 2M initial value memories that are independently set for each of M partials when starting calculations.

Therefore, in an electronic musical instrument having such a configuration, the amplitude of each partial can be set independently and arbitrarily depending on how the initial value is determined.
In addition, by selecting the coefficient value p (qm) as necessary, the musical sound waveform obtained can be a steady waveform that remains constant over time or that fluctuates periodically, or it can be a steady waveform that changes periodically over time, or it can be a steady waveform that changes periodically over time. Alternatively, it can be a non-stationary waveform that diverges.

Therefore, it is expected that electronic musical instruments of this type can produce musical tones that are extremely similar to the musical tones of natural musical instruments. However, when the number of partials is large, each partial generally has a different amplitude, so the number of 2M initial value memories becomes enormous.

Further, when synthesizing the musical tones of certain natural musical instruments, such as the piano, the above-mentioned method poses difficulties.

That is, on a piano, there is a difference in the amplitude of musical tones between the bass and treble ranges, and the amplitude varies greatly depending on the intensity of keystrokes. Furthermore, even when considering a specific range of sounds, there is a large difference in the amplitude of partials between low-order partials and high-order partials. For this reason, in order to be able to deal with musical tones of any amplitude, that is, to be able to store the maximum data regarding the initial values, and to ensure the necessary and sufficient accuracy to compose musical tones even for the minimum data. The initial value memory and the registers and memories constituting the arithmetic unit must have a very large number of digits, which inevitably increases the scale of the arithmetic unit and, by extension, the electronic musical instrument as a whole. The present invention has been made in view of the above-mentioned circumstances, and it is an object of the present invention to provide an electronic musical instrument that can produce steady and unsteady sounds close to the musical sounds of natural musical instruments with a small-scale and relatively simple configuration. With the goal.

In order to achieve this object, the electronic musical instrument according to the present invention provides a first initial sample value "1" that is commonly set for each of the M partials fm constituting a musical tone. an initial sample value generating means, a second initial sample value generating means for generating a second initial sample value respectively set for each partial tone fm, and a parameter pm set respectively for each partial tone fm. (m=1, 2,...,M) and qm(
m = 1, 2, ..., M) by weighting the past two sample values for each partial fm by the parameters pm and qm and then adding them together. A calculation means for calculating a new sample value of each partial tone fm, which temporarily stores the first and second initial sample values for each partial tone fm at the start of musical sound generation, and stores the first and second initial sample values for each partial tone fm in the calculation means as the past sample values. At the same time, each partial tone f outputted from the calculation means
a musical tone forming section including each element of a regression control means for temporarily storing a new sample value every m and regressing it to the calculation means as the past sample value; and a coefficient for setting the relative amplitude of each partial tone fm. am (m=1, 2, . . . , M); amplitude control means for controlling the amplitude of the sample value for each partial tone fm output from the calculation means by the coefficient am; Each element of an adding means adds the sample values of each amplitude-controlled partial sound fm output from the amplitude control means, and the sample value of each partial sound formed by the partial sound forming section is added every period T. and a musical tone forming section that adds the sample value of the musical tone to the sample value of the musical tone and converts this sample value into a musical tone. Embodiments of the invention will now be described with reference to the accompanying drawings.

FIG. 1 shows an embodiment of the present invention, in which a regression calculation device 101 that performs regression calculations calculates standardized partial functions. That is, fl is set so that the sample value (initial value) of the partial function at time t=0 is "1" for all M partials.

(t) NieQ・tCOS lt et al.

(t) = 2t of eQ2tCOS: : fM.

(t) In other words, calculate the partial function Mt of MtCOS.

Therefore, the initial value at time t=0 is common to all partials (f,. (0) = f number (0) = ... = f side (
0)=1), and the first initial value memory 1 that provides the initial value at time t=0 only needs to provide one constant value (it does not need to be a so-called memory). Time t
= Another initial value at T differs depending on the frequency of each partial as shown in FIG. 2, so the second initial value memory 12 has M independent memories for each of M partials. have The regression calculation device 1 calculates sample values of equation (3) for each time series based on the contents of the initial value memories 11 and 12.

(The configuration of the regression calculation device 10, the regression calculation operation, etc. will be described in detail later.) Next, the output corresponding to each partial of the regression calculation device 10 and the amplitude values a, ~aM corresponding to each partial
A desired partial function,
That is, f, (t) = a, eQ, tcosw, t et al. (t): 2t of a2eQ2tCOS ・….・
(7) A sample value at each time of fM(t)=aMeQMtcos■Mt can be obtained.

By accumulating these sample values in the accumulator 15 and outputting them at a predetermined timing, F(t) = fm
(t) musical tone can be formed. According to this embodiment, the partial function is standardized and calculated as described above,
By later multiplying the amplitude value, the number of register bits of the arithmetic unit 10 can be significantly reduced.

Next, an application example that embodies the above-mentioned example will be described with reference to FIG. 3.

In this embodiment, a key address device 621 and a memory address control device 622 are controlled in response to key operations on a keyboard (not shown), thereby controlling the regression calculation device 600.
and drives the musical tone sample value calculation device 610 to form a musical tone.

The musical tone sample value obtained from the musical tone sample value calculation device 610 is D
Speaker 6 via A converter 623 and audio device 624
25 to be sounded as a musical tone signal. Hereinafter, the configuration and operation of the present invention will be explained while explaining the main components.

The key address device 621 sends out a key number signal KN representing the number of the key operated on a keyboard (not shown) from a memory configured with a ROM, etc.
Address 22.

The memory address control device 622 drives the regression calculation device 600 and the tone sample value calculation device 610 in response to the key number signal KN to read out parameter values and initial values necessary for subsequent calculations. A regression calculation device 60 includes parameter memories 601p and 601q, a partial tone sample value storage control device 602, and a calculation circuit 603, and executes the above-described regression calculation. Here, the parameter memory 60
1p, 601q are the parameters p, qm (
m=1, 2, . The partial tone sample value storage control device 602 stores the two initial values fm described in the embodiment.

(0), f skin (T) (m = 1, 2, ..., M), and a calculation means such as a register that is not necessary for regression calculation. The value is transferred to the arithmetic circuit 603, and the output of the arithmetic circuit 603 is temporarily stored and transferred to the musical tone sample value arithmetic unit 6101 in order from the oldest partial tone sample value for a predetermined period of time. As can be seen from this description, this device 602 includes the initial value memories 11 and 12 described in the embodiment. The operation of this partial tone sample value storage control device 602 will be further explained.

For example, time t=0, 7, 27,..., (M-1) d, (
However, in 7: poison), the device 602 takes the standardized sample value (initial value) f, of each partial. (0), f2. (0),
..., fMo(0) are sequentially outputted by the tone sample value calculation device 610'd. At the same time, the initial value f skin (0) of each partial
, fm. (T) (m=1, 2,...,M) in the arithmetic circuit 60
3, and the calculation output fm(2r
) is memorized once. Next cycle (t=M7, (M+1
)7,...,2M7) includes a musical tone sample value calculation device 61.
0' Sample value f side (t) of each partial (m=1, 2,...
. , M), and also outputs the sample value fm to the arithmetic circuit 603.
(0), fm (bottom) (m=1, 2, . . . M) are output, while the calculation result fm (rainbow) of the calculation circuit 603 is input. Thereafter, partial tone sample values fm are similarly standardized by time-division calculations at time intervals t=0, T, momme, . . . .
(0), f side (T), fm. (2T), . . . are output sequentially. Such partial tone sample value f side (0), f skin (T
), f skin (illusion), etc. are started by the output signal of the memory address control device 622 like the parameter memories 601p and 601q, and are also started by the output signal of the memory address control device 622 (for example, by the key press signal). The contents of the partial tone sample value storage control device 602 are cleared by the clear signal CL (detected and output).

As can be seen from the above description, the arithmetic circuit 603 calculates fm(nT)=pmf based on the outputs of the parameter memories 601p and 601q and the partial tone sample value storage control device 602.
It calculates m {(n-1)T} 10qmfm{(n-2)T}, and is configured with a multiplier, an adder, and the like.

The musical tone sample value calculation device 610 includes a partial tone relative amplitude value storage device 611, a multiplier 612, and an accumulator 613, and calculates musical tone sample values based on the partial tone sample values sequentially transferred from the regression calculation device 600. Form.

Here, the partial tone relative amplitude value storage device 611 stores amplitude values am (m=1, 2, . . . , M
) is stored, and a predetermined amplitude value am is read out according to the output of the memory address control device 622.

This readout is performed as an operation of the partial tone sample value storage control device 602, and a predetermined sample value f,
(t), ra(t),..., fM(t), (t=0, T, 怀,...) and the amplitude values a, ~aM are multiplied. After that, M partial tone sample values f, (t),
et(t),...,fM(t),(t;0,T,2T"...
) is accumulated by the accumulator 613 and the time t=nT
Tone sample value F(
Output as nT)=Zm=Ifm(nT).

The process of obtaining a musical tone from this musical tone sample value F(nT) has already been described. The arithmetic operation described above is performed using the clock pulses CK and CK2 (=CK.) whose duty is seasonal as shown in FIGS. 4B and 4C, and the period T as shown in FIG. It is executed periodically by the timing pulse TP.

The period of the clock pulse CK, CK2 is the M partial sample values f, (nT), f2(nT), ..., fM mentioned above.
(nT), the pulse CK reads data from the memories 601p, 601q and the storage control device 602, while the pulse CK2 writes data from the arithmetic circuit 603 to the storage control device 602. . Moreover, the timing pulse TP is at time tF0, T
, public, . 7, f young↑,・
..., fMo(0) is formed, and in the period up to the next t=T, f,.

↑, f2. d,..., fMo(0) are formed, and in the same way, periods f, (nT), f2(nT), · until t=nT are formed.
..., fM(nT) are formed, each timing t=0,
T,...,nT,..., musical tone standard M standard value F(nT) = Z
fm(nT) is output.

Incidentally, FIG. 4A shows signals corresponding to key operations.

As described above, this invention standardizes partial tone sample values and executes regression calculations, thereby minimizing the number of register bits of various calculation devices and reducing the number of memories. Accordingly, it is possible to provide an electronic musical instrument that can reduce the scale of the device.

That is, according to the electronic musical instrument according to the present invention, steady sounds and unsteady sounds close to the musical sounds of natural musical instruments can be obtained with a small-scale and relatively simple configuration.

[Brief explanation of the drawing]

Fig. 1 is a system diagram of an embodiment of this invention, Fig. 2 is an explanatory diagram of the operation of the above embodiment, Fig. 3 is a detailed system diagram of an applied embodiment of this invention, and Fig. 4 is an embodiment of the embodiment shown in Fig. 3. 3 is a timing chart for explaining the operation of FIG. 10,600...Regression calculation device, 11,12...Initial value memory, 13...Amplitude value memory, 14,612...
... Multiplier, 15...613... Accumulator,
601p, 601q...Parameter memory, 602...Partial tone sample value storage control device, 603... Arithmetic circuit, 610
... musical tone sample value calculation device, 611 ... partial tone relative amplitude value storage device, 621 ... key address device, 622
...Memory address control device, 623...DA converter, 624...Audio device, 625...Speaker. Figure 1 Figure 2 Figure 3 Figure 4

Claims (1)

[Claims] 1 M partials f_m (m=1, 2,
..., M) with the scheduled period T
The sample value of each partial tone formed by the partial tone forming section and the partial tone forming section formed by the partial tone forming section is added every period T to form a sample value of a musical tone. , and a musical tone forming section that converts this sample value into a musical tone, and the partial tone forming section has the following (a) (b) (c) (d) (e).
An electronic musical instrument, wherein the musical tone forming section includes the following (f), (g), and (g). (a) A first initial sample value generating means for generating a first initial sample value "1" set commonly for each of the partial sounds f_m. (
b) The second set for each partial tone f_m above.
second initial sample value generation means for generating an initial sample value of . (c) Parameters p_m (m=1, 2,..., M) and q_m(
parameter generating means for generating m=1, 2, . . . , M);
(d) By weighting the past two sample values for each partial f_m by the parameters p_m and q_m and then adding them together, each partial f_m is
Arithmetic means for calculating a new sample value of _m. (E) At the start of musical sound generation, the first and second initial sample values for each partial tone f_m can be temporarily stored and added to the calculation means as the past sample values, and then output from the calculation means. regression control means for temporarily storing a new sample value for each partial sound f_m and causing the calculation means to regress it as the past sample value; (f) An amplitude coefficient generating means for generating a coefficient a_m (m=1, 2, . . . , M) for setting the relative amplitude of each partial tone f_m. (G) Amplitude control means for controlling the amplitude of the sample value for each partial tone f_m output from the calculation means using the coefficient a_m. (H) Adding means for adding sample values of each amplitude-controlled partial tone f_m output from the amplitude controlling means.