JP3124349B2 - Sound source device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a sound source device used for, for example, a tone generator or a sound effect generator.

[0002]

2. Description of the Related Art Conventionally, as a sound source device as described above,
For example, there is a type in which an input tone signal is sampled by a sampling signal of a predetermined frequency, these sampling values are stored in a storage unit, and the sampling values are read out from the storage unit. In such a tone generator, when a tone having a pitch different from that of an input tone signal is to be generated, a sampling value is read from a storage unit at a speed different from that at the time of sampling. But,
In this case, the formant of the read tone signal is different from the input tone signal, that is, the formant is moving, and the tone becomes unnatural.

When a formant is not moving, that is, when a fixed formant tone is to be obtained, a filter is used. However, when a desired formant is to be obtained by the filter, the configuration and scale of the filter are complicated and large. .

Japanese Patent Application Laid-Open No. Sho 62-65098 discloses an example of a fixed formant that can change the pitch. In this method, phonemes of an input voice signal are sequentially cut out, and are sequentially generated repeatedly at a pitch different from the pitch of the input voice signal.

[0005]

However, Japanese Patent Application Laid-Open No. Sho 62-62
According to the technology disclosed in Japanese Patent Application Laid-Open No. 65098, when playing back at a pitch different from that of an input audio signal, if the pitch becomes higher than the pitch of the input audio signal, the phonemes cut out from the input audio signal are truncated to shorten the extraction length. Setting was done. When such processing is performed, a characteristic portion of a phoneme of an input audio signal is cut off, and there is a problem that a change in timbre is increased.

[0006]

According to a first aspect of the present invention, there is provided a waveform storing means for storing a plurality of cycles of a waveform, and an address signal is supplied to the waveform means to generate the waveform. Reading means for reading a waveform in a desired section at a speed irrelevant to the pitch to be calculated, multiplying means for applying a window function to the read waveform, and trying to generate a waveform obtained by applying the window function. And a waveform synthesizing means for synthesizing at a period of the pitch to be performed, so that the highest level portion of the waveform in the desired section read out from the waveform storage means coincides with the highest level portion of the window function applied. The address supplied from the reading means to the waveform storage means is set.

According to a second aspect of the present invention, there is provided a waveform storing means for storing a waveform of a plurality of cycles, and a desired reading start time for the waveform means.
Reading means for supplying an address signal starting from an address and reading a waveform in a desired section at a speed independent of the pitch to be generated; multiplying means for applying a window function to the read waveform; a waveform synthesis means for the waveform synthesizing in a period of pitch to be the generation multiplied, a modulation signal generating means, comprises, the modulation signal from the modulation signal generating means, read start address of the reading means Is modulated.

[0008]

According to the first aspect of the present invention, the waveform of the desired section from the waveform storage means is independent of the pitch at which the waveform is to be generated by the reading means (for example, the same pitch as the original pitch at the time of sampling). High). The read waveform is multiplied by a window function by a multiplying means, and the waveform subjected to the window function is synthesized by a waveform synthesizing means at a period of a pitch to be generated. In order for the reading means to read the waveform from the waveform storage means, the address given to the waveform storage means has the highest level of the waveform in the desired section.
The threshold is the highest level of the window function applied
And the waveform output from the multiplying means is the characteristic of the waveform stored in the waveform means .
The level decrease is small and the original waveform has
Cement information you are less likely to be lost.

Also, in the second invention, similarly to the first invention, the waveform of the desired section from the waveform storage means is independent of the pitch at which the waveform is to be generated (for example, the original sound at the time of sampling). At the same pitch as the pitch), a window function is applied to the read waveform, and the waveform subjected to the window function is synthesized at a period of the pitch to be generated by the waveform synthesis means. You. Since the read start address given to the waveform storage means by the read means is modulated by the modulation signal from the modulation signal generation means, when read repeatedly, the monotony of the read waveform is reduced. You.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the present invention will be described in detail based on one embodiment in which the present invention is applied to a tone generator of a tone generator. As shown in FIG. 1, the tone generator has waveform memories 2 and 4, and the waveform memories 2 have different tone waveforms A, B, C,.
. Are stored for a plurality of cycles, respectively, by sampling a sampling signal of a predetermined frequency. The waveform memory 4 also stores a plurality of sampling values obtained by sampling the tone waveforms A, B, C,. Each cycle of each of the waveforms A, B, C,... Is assigned a waveform number having a different value, as shown in FIG.

An address signal is supplied from the adder 6 to the waveform memory 2, and each sampling value is read out according to the address signal. The address signal is also supplied from the adder 8 to the waveform memory 4, and each sampling value is read out in accordance with the address signal.

These address signals have a read speed coefficient R
A, phase 1 signal from two-phase counter 9, phase
e2 signal, a noise signal from the noise generator 10 which is a modulation signal generating means, a head address of a cycle corresponding to the waveform number from the waveform number tables 12 and 14, and a cycle corresponding to the waveform number from the peak address tables 16 and 18. Are generated on the basis of the address of the characteristic portion in the above. For the sake of simplicity, first, the head address from the waveform number tables 12 and 14, the phase1 signal,
A description will be given of a case where an address signal is generated based on the phase2 signal (that is, the noise generator 10, the peak address tables 16, 18, the adders 72, 74, 7).
6, 78 and the state where the multipliers 73 and 75 do not exist will be described), and then the case where an address signal is generated in consideration of a peak address, a noise signal and a read speed coefficient RA will be described.

The waveform number tables 12 and 14 generate the start address of the cycle corresponding to the input waveform number. The start address from the waveform number table 12 is supplied to the adder 6 and the waveform number table 14 is generated. Is supplied to the adder 8.

The waveform number in the waveform number table 12 is obtained by adding the count value from the waveform number counter 20 and the waveform information a by the adder 22, and the waveform number in the waveform number table 14 is Count value from counter 24 and waveform information a
Are added by the adder 26. The waveform information a represents the waveform to be read, for example, the first waveform number among the waveform numbers of A.

The waveform number counters 20 and 24 generate count values based on the pulse signals C1 and C2, the tone generation start information, the pitch coefficient PA and the tone generation start data selection signal from the two-phase counter 9, respectively. FIG. 1 shows only the pulse signals C1 and C2 and the pitch coefficient PA.

The two-phase counter 9 has an accumulator 28, as shown in FIG. 2, which accumulates frequency information F generated by, for example, pressing a key on a keyboard.
Each time a sampling signal (which has the same frequency as the sampling signal used to obtain the sampling values stored in the waveform memories 2 and 4) is accumulated, the accumulation is performed. The value of the frequency information F is selected so as to generate the carry-out signal CO as shown in FIG. 3A when the frequency information is accumulated over the corresponding period.

The carry-out signal from accumulator 28 is
The frequency divider 30 divides the frequency by ２. However, as shown in FIG. 3, the frequency divider 30 is set by the sound generation start information generated following the sound generation start data selection signal generated in response to the key depression.
Occurs, the Q output of the frequency divider 30 goes low,
The Q output becomes H level, and thereafter the carry-out signal CO
Is input, the Q and XQ outputs are inverted.

Therefore, as shown in FIGS. 3B and 3C, the Q signal and the XQ signal have a period twice as long as the key pressing frequency. The Q signal is differentiated in a differentiating circuit 32, and the differentiating circuit 32 generates a pulse signal C1 corresponding to the rising of the Q signal, as shown in FIG. Further, the XQ signal is supplied to the differentiating circuit 34 and the differentiating circuit 3
4 generates a pulse signal C2 in response to the rising edge of the XQ signal, that is, the falling edge of the Q signal, as shown in FIG. As is clear from comparison with FIGS. 3D and 3E, the pulse signals C1 and C2 have a period twice as long as the key-pressing frequency, and C2 has a phase corresponding to one period of the key-pressing frequency from C1. Is out of alignment.

The two-phase counter 9 further includes a counter 36 for increasing the count value by one each time a sampling signal is supplied. The counter 36 is connected via an OR circuit 37 to sound generation start information or a pulse signal C1. Is reset each time is supplied. Accordingly, as shown in FIG. 3 (f), the value of the phase1 signal, which is the count value of the counter 36, increases by one each time a sampling signal is generated from when the sound generation start information is generated. When this occurs, the count value becomes 0, and thereafter, the count value is increased by one each time a sampling signal is generated until the pulse signal C1 is generated again.

The counter 38 counts the sampling signal supplied through the AND circuit 39. The AND circuit 39 counts the sampling signal only when the H level signal is supplied from the RS flip-flop 40. To supply. RS flip-flop 40
It is reset when the tone generation start information is supplied, and is set by the pulse signal C2. Therefore, the counter 38
Does not perform the counting operation until the pulse signal C2 is generated.

The counter 38 includes an OR circuit 41
Is reset each time the tone generation start information or the pulse signal C2 is supplied via the. Therefore, the phase 2 signal, which is the count value of the counter 38, is reset when the sound generation start information is generated as shown in FIG.
The value is set to 0, and this state is maintained until the pulse signal C2 is supplied for the first time, and thereafter, every time the sampling signal is supplied until the pulse signal C2 is supplied and the next pulse signal C2 is supplied. Increment the value by one. FIG.
As is clear from comparison with (f) and (g), phas
The phase2 signal is shifted in phase by one period of the key press frequency from the e1 signal.

As shown in FIG. 4, the waveform number counter 20 has a multiplier 42 for multiplying a pitch coefficient PA generated according to a depressed key of the keyboard by M times. In addition,
Since the period is set to be twice as long as the pitch at which the pulse signals C1 and C2 are to be generated, M is set to 2. However, the pulse signals C1 and C2 have a period three times the pitch,
Alternatively, if the period is set to be 4 times, M is 3
Or 4.

The output of the multiplier 42 is connected to one input of an adder 44. The other input of the adder 44 is supplied with the output of a flip-flop 46 described later. The output D of the adder 44 is input to the subtractor 50 and the comparator 52. Further, the subtractor 50 and the comparator 5
2 is supplied with a value N obtained by adding 1 to the final waveform number of the waveform to be read. The subtracter 50 subtracts N (the value obtained by adding 1 to the final waveform number) from the output D of the adder 44. Further, the comparator 50 includes an adder 44.
Is compared with N (the value obtained by adding 1 to the final waveform number).

The output of the subtractor 50 is supplied to a contact 54a of a changeover switch 54. The output of the adder 44 is supplied to a contact point 54b of the changeover switch 54. The contact 54c of the changeover switch 54 is connected to a contact 56a of another changeover switch 56, and the other contact 56b of the changeover switch 56 is supplied with an initial value. The contact 56 c of the changeover switch 56 is connected to the flip-flop 46.

The contact 56c of the changeover switch 56
In the state where the sound start data selection signal is not supplied,
As shown in FIG. 4, the contact 56a is switched to the contact 56b and the contact 56c is switched to the contact 56b in a state where the tone generation start data selection signal is supplied. The flip-flop 46 holds the input signal every time the pulse signal C1 or the sound generation start information is supplied via the OR circuit 58.

Therefore, for example, when the pitch coefficient PA is 0.75
Assuming that the initial value is 0 and the final waveform number is 5, when the tone generation start data selection signal is generated, the changeover switch 56
Is switched to the contact 56b side, and the initial value 0
Is supplied to the flip-flop 46 via the changeover switch 56, and when the sound generation start information is supplied to the flip-flop 46 via the OR circuit 58, the flip-flop 46
Holds the initial value 0.

The initial value 0 of the flip-flop 46 is obtained by multiplying the pitch coefficient PA by 1.5 by the multiplier 42 and the adder 4.
4 is added. The subtracter 5 subtracts the sum D (1.5) from
The value N (6) obtained by adding 1 to the final waveform number at 0 is subtracted, and at the same time, the added value D of the adder 44 in the comparator 52 is added.
(1.5) is compared with N (6), but since D (1.5) is smaller than N (6), the contact 54c of the changeover switch 54 is switched to the contact 54b side. .

Further, the contact 56c of the changeover switch 56
Since the tone generation start data selection signal has disappeared, the contact 56c has been switched to the contact 56a side. Therefore, the sum D (1.5) of the adder 44 is
When the pulse signal C1 is supplied through the OR circuit 58, the addition value D (1.5) is held in the flip-flop 46.

Similarly, the output of the flip-flop 46 becomes 3.0, 4.5 each time the pulse signal C1 is generated.
And change. Next, when the pulse signal C1 is generated, the addition value D of the adder 44 becomes 6, the addition value D becomes equal to N, and the comparator 54 switches the contact 54c of the changeover switch 54 to the contact 54a side. , Subtractor 50
Is held in the flip-flop 46. Hereinafter, similarly, the output of the flip-flop 46 becomes 1.
5, 3.0,... The output of the flip-flop 46 is output as the count value of the waveform number counter 20. In practice, the integer parts 0, 1, 3, 4, 0... Of this output are output.

The waveform number counter 24 has the same configuration as the waveform number counter 20, as shown in FIG. 5, except that an RS flip-flop 60 which is set by tone generation start information and reset by the pulse signal C2 is added. ,
Further, a changeover switch 62 for switching the coefficient M or 1 to the multiplier 42 is added. This changeover switch 6
2 indicates that when flip-flop 60 is set,
When flip-flop 60 is reset to supply coefficient 1, switch 62 is switched to supply coefficient M. The OR circuit 58 is supplied with a pulse signal C2 instead of the pulse signal C1.

Therefore, the initial value is 0 and the pitch coefficient is 0.7
Assuming that the last waveform number is 5, when the sound generation start data selection signal is generated, the contact 56c of the changeover switch 56 is switched to the contact 56b, the initial value 0 is supplied to the flip-flop 46, and the sound generation start information is output. Is supplied to the flip-flop 46 via the OR circuit 58, the flip-flop 46 holds the initial value 0.
Also, the RS flip-flop 60 according to the sounding start information.
Is set, the changeover switch 62 supplies the coefficient 1 to the multiplier 42.

Then, according to the disappearance of the tone generation start data selection signal, the contact 56c of the changeover switch 56 is set to the contact 56a.
Side. Accordingly, the output 0 of the flip-flop 46 and the pitch coefficient PA (0.75) multiplied by 1 by the multiplier 42 are added by the adder 44, and the added value D (0.75) is calculated by the comparator 44. 1 for the final waveform number at 52
Is compared with the value N (6) obtained by adding, but since the added value D is smaller than N, the contact 54c of the changeover switch 54 is switched to the contact 54b. Accordingly, when the addition value D (0.75) of the adder 44 is supplied to the flip-flop 46 and the pulse signal C2 is supplied to the flip-flop 46, the addition value D (0.75) is supplied to the flip-flop 46.
Is held.

At the same time, the RS flip-flop 60 is reset by the pulse signal C2, and the changeover switch 62 switches to supply the coefficient M (2) to the multiplier 42. Therefore, the output 0.75 of the flip-flop 46,
The pitch coefficient PA (0.75) multiplied by M (2) in the multiplier 42 is added to 1.5 in the adder 44. The added value D (2.25) is compared with N (6) in the comparator 52. Since D is smaller than N, the added value D (2.25) is supplied to the flip-flop 46. When the pulse signal C2 is supplied to the flip-flop 46 via the OR circuit 58, the flip-flop 46 sets the sum D
(2.25) is retained. Hereinafter, similarly, the pulse signal C2
Each time occurs, the output of flip-flop 46 becomes 3.
75 and 5.25.

Next, the addition value D of the adder 44 becomes 6.75, which is larger than N (6).
4 is switched to the contact 54a side, the subtraction value of the subtractor 50 (6.75-6 = 0.75) is supplied to the flip-flop 46, and when the pulse signal C2 is supplied, the flip-flop 46 Is held. Thereafter, similarly, every time the pulse signal C2 is supplied, 2.25, 3.75... Are held in the flip-flop 46. In fact, the integer of the held output is output from the flip-flop 46. The output is 0, 0, 2, 3, 5, 5, 0, 2, 3,.

The count value from the waveform number counter 20 is supplied to an adder 22, added to the waveform information a, and supplied to the waveform number table 12. Since the waveform information a represents the first one of the waveform numbers of the waveforms to be read out of the plurality of types of waveforms A, B, C,... Stored in the waveform memory 2, for example, Assuming that a is 0 corresponding to the waveform A, the output of the adder 22, that is, the waveform number is 0, 1, 3, 4, 0 every time the two-phase counter 9 generates the pulse signal C1.
... changes. Since the count value from the waveform number counter 24 is added to the waveform information a in the adder 26, the output of the adder 26, that is, the waveform number becomes 0, 2, 3,. 5,
0 and so on. Note that the sound generation stop information in which the count operation of phase 2 is stopped and the first pulse signal C2
The output value 0 of the waveform number counter 24 output during this period is not actually used.

The pitch coefficient PA supplied to the waveform number counters 20 and 24 is such that when the waveform specified by the waveform information a is to be read at the key pressing frequency, the period to be read is the original waveform. . For example, when 0.75 is given as the pitch coefficient as described above, the period of the waveform to be read is 3/4 times the original waveform. .

The waveform number table 12 includes an adder 22
And outputs the start address of the cycle corresponding to the waveform number supplied from the adder 2.
The head address of the cycle corresponding to the waveform number supplied from 6 is output. The head address from the waveform number table 12 is supplied to the adder 6, where it is added to the phase1 signal from the two-phase counter 9. phase
In one signal, a pulse signal C1 is generated, and the pulse signal C1 is generated again.
Since the value is increased each time a sampling signal is generated until “1” is generated, the waveform specified by the waveform information “a” from the waveform memory 2 by the output of the adder 6 changes the key-pressing frequency to “2”. It is read out first over the period.

For example, as described above, the waveform information a specifies the waveform number 0, and when the sound generation start information is generated, the value of the waveform number counter 20 is 0. Is supplied to the adder 6, and the waveform number 0 is supplied as shown in FIG.
And one half cycle of waveform number 1 are read out. Next, when the pulse signal C1 is generated, the value of the waveform number counter 20 is 1, so that the waveform number counter 12 supplies the start address of the cycle of the waveform number 1 to the adder 6. Therefore, the cycle of the waveform number 1 and the half cycle of the waveform number 2 are read according to the change of the phase1 signal. Hereinafter, the waveform is read out in the same manner.

Similarly, the top address of waveform number 0 is generated when the pulse signal C2 is supplied from the waveform number table 14, and this is supplied to the adder 8, and
8 from the waveform memory 4 in accordance with the change of the ase2 signal.
As shown in (d), the cycle of the waveform number 0 and the half cycle of the waveform number 1 are read, and subsequently, the cycle of the waveform number 2 and the half cycle of the waveform number 3 are read. Hereinafter, a waveform is similarly read from the waveform memory 4.

FIGS. 8B and 8D obtained as described above.
Of the waveform A is cut out from the waveform A,
The waveform is read out at the same cycle as the sampling cycle, and the read waveform is repeated at the pitch cycle at which the waveform is to be generated. A waveform different from the pitch of a certain original waveform A can be obtained.

As described above, although the formants do not move, the waveform is discontinued by cutting out a part of the waveform, for example, as shown in FIGS. 8 (b) and 8 (d). In order to solve this, in this embodiment, the multiplier 62 applies the waveform read from the waveform memory 2 to
The multiplier 64 multiplies the window function read from the waveform memory 4 by a window function called cos2, and then multiplies the waveform read from the waveform memory 4 by a window function called cos2.

The window functions cos1 and cos2 are generated by a cosine waveform generation circuit 67. As shown in FIG. 6, the cosine waveform generation circuit 67
8 and 69, and these cosine tables 68 and 69
Stores curve signals as shown in FIGS. 7 (e) and 7 (f), which are generated when X changes from −π to π in the equation of 1/2 (1 + cosX), as window function signals.
The input address to the cosine table 68 is phase
One signal is obtained by multiplying the frequency information F and the coefficient K1 by the multiplier 70. The input address to the cosine table 69 is obtained by multiplying the phase2 signal by the frequency information F and the coefficient K1 by the multiplier 71. can get.

The reason why the phase 1 signal and the phase 2 signal are multiplied by the frequency information F and the coefficient K1 is as follows. If the phase1 signal and the phase2 signal are simply supplied to the cosine tables 64 and 66 as address signals, the final value of the address changes according to the pitch. When the pitch is changed, a phase 1 signal, phase
The number of increment operations for obtaining the two signals changes, and as a result, the peak values of the phase1 signal and the phase2 signal change in inverse proportion to the frequency.
4 and 66, the final value of the address supplied changes according to the pitch. In order to prevent such a situation and make the final value of the address the same, the phase1 signal and the phase2 signal are multiplied by the frequency information F and the coefficient K1.

Therefore, when the peak address table 16 and the noise generator 10 are ignored and considered, this apparatus reads out the waveform from the waveform memories 2 and 4 at a pitch different from the original waveform, and furthermore, the formant of the original waveform is read. , And is not affected by the discontinuity of the waveform.

Note that actually, the phase1 signal, ph
The ase2 signal is supplied to the adders 6 and 8 after being multiplied by the read speed coefficient RA by the multipliers 73 and 75. These multipliers 73 and 75 are used when changing the formants of the waveforms read from the waveform memories 2 and 4, and when the formants are not changed,
The reading speed coefficient RA is set to 1.

However, in this embodiment, the start address of a certain waveform number from the waveform number table 12 is
Using the peak address from the peak address table 16, the frequency information 1 / F, and the coefficient K, the window function cos
The most characteristic part of the cycle of the waveform number read at the point of the maximum peak of 1, cos2 is made to coincide.

That is, the peak address tables 16 and 18
As shown in FIG. 9A, the value S representing the difference between the most characteristic portion in the cycle assigned with the waveform number, for example, the portion having the highest level, and the start address of the waveform number, It is stored for each waveform number. Accordingly, when the waveform number is supplied from the adder 22, the peak address table 16 outputs the difference between the most characteristic part in the cycle of the waveform number and the start address of the cycle. When the adder 72 adds the start address of the waveform number supplied from the adder 72, the address of the most characteristic portion of the waveform can be obtained. This is the same even when the output of the waveform number table 14 and the output of the peak address table 18 are added by the adder 74.

However, in order to match this characteristic part to the part of the maximum level of the window function, the window functions cos1, cos
It is necessary to start reading from the waveform memories 2 and 4 in synchronization with the occurrence of s2. Therefore, in the adders 72 and 74, the reciprocal (1/1 /) of the frequency information F representing the key press frequency is used.
F) The period T of the key press frequency is obtained by multiplying the coefficient K by the coefficient K (which is proportional to the period of the key press frequency), and this is output from the waveform number tables 12 and 14 and the output from the peak address tables 16 and 18. Is subtracted from the sum of the two to obtain the reading start address from the waveform memories 2 and 4. This state is shown in FIGS. 9B and 9C. With this configuration, as is clear from FIGS. 9B and 9C, even if the key pressing frequency is changed and the period T is changed, the characteristic portion of the waveform is always the most of the window functions cos1 and cos2. Matches high level parts. Outputs of the adders 72 and 74 are supplied to adders 6 and 8.

In addition, noise signals from the noise generator 10 are actually supplied to the outputs of the adders 72 and 74 by the adders 76 and 78. When the noise signal is added to the outputs of the adders 72 and 74 and supplied to the adders 6 and 8, the head addresses of the waveforms read from the waveform memories 2 and 4 undergo modulation and fluctuate. Thereby,
It is possible to reduce the monotony of the generated tone.

In the above embodiment, the present invention is applied to the sound source device of the musical sound generating device, but may be applied to the sound source device of the sound effect generating device.

In the above-described embodiment, the waveform number counters 20 and 24 are configured such that the count value changes in a manner similar to the temporal change of the formants of the waveforms read from the waveform memories 2 and 4. For example, if there is no temporal change in the formant of the waveform to be read and it is not necessary to change the count value of the waveform number counters 20 and 24 in the same manner as the temporal change of the formant, the configuration of the waveform number counters 20 and 24 Can also be simplified. Although the cosine function is used as the window function in the above embodiment, another known function may be used.

Further, the count values of the waveform number counters 20 and 24 may be changed by operating an operator such as a pedal. Also, the waveform number counters 20, 2
Instead of 4, a random signal generator may be used to randomly change the order in which waveforms are read.

The waveform number counters 20, 24
May be configured to decrease after the count value increases, or may be configured to decrease the count value.

In the above embodiment, the noise generation circuit 10 is used as the modulation signal generation circuit. However, a circuit that generates a periodic function may be used instead.

[0055]

As described above, according to the present invention, the waveform storing means storing the waveforms of a plurality of cycles has no relation to the pitch at which the waveform is to be generated (for example, the original pitch at the time of sampling and the original pitch). Read the desired section of the original waveform at the same pitch)
The read waveform is multiplied by a window function by a multiplying means, and the waveform subjected to the window function is synthesized by a waveform synthesizing means at a period of a pitch to be generated. Of the desired section of the waveform
The highest level part is the highest level part of the window function.
Address to be supplied to the waveform storage means to match the minute
Set. Therefore , when reading at a pitch different from the original waveform, the original waveform always has the highest level , even if the read pitch is changed.
The higher part of the bell is multiplied by the higher part of the window function.
Therefore, the formant information of the original waveform is less likely to be lost, and a change in timbre can be reduced. Further, in the present invention, since the read start address of the reading means is modulated by the modulation signal from the modulation signal generating means, even when a repeated tone is generated, the monotone of the generated tone can be reduced. .

[Brief description of the drawings]

FIG. 1 is a block diagram of one embodiment of a sound source device according to the present invention.

FIG. 2 is a block diagram of a two-phase counter used in the embodiment.

FIG. 3 is an operation explanatory diagram of the two-phase counter of FIG. 2;

FIG. 4 is a waveform number counter 20 used in the embodiment.
It is a block diagram of.

FIG. 5 is a waveform number counter 24 used in the embodiment.
It is a block diagram of.

FIG. 6 shows a cosine waveform generating circuit 67 used in the embodiment.
It is a block diagram of.

7 is an operation explanatory diagram of the cosine waveform generation circuit 67 of FIG. 7;

FIG. 8 is an explanatory diagram of waveform reading in the embodiment.

FIG. 9 is a diagram showing a relationship between a portion having a high level of a window function and a most characteristic portion of a read waveform in the embodiment.

[Explanation of symbols]

 2 4 Waveform memory 12 14 Waveform number table 16 18 Peak address table 67 Cosine waveform generation circuit

Claims (2)

(57) [Claims] 1. A waveform storage means for storing waveforms of a plurality of cycles, a read means for supplying an address signal to the waveform means and reading a waveform in a desired section at a speed independent of a pitch to be generated, and multiplying means for multiplying the read Dasa a window function waveform, and a waveform synthesis means for the waveform obtained by multiplying the the window function synthesized in a period of pitch to be the generation, provided, read from the waveform memory means Of the desired section
As best level is high portion of the waveform is coincident with the highest level part of the window function to be applied above the reading means, wherein the this <br/> to set the address to be supplied to the waveform storage means Sound source device. 2. A waveform storage means for storing waveforms of a plurality of cycles, and an address signal starting from a desired read start address is supplied to the waveform means at a speed independent of a pitch to be generated. Reading means for reading a waveform in a desired section; multiplying means for applying a window function to the read waveform; and waveform synthesizing means for synthesizing a waveform obtained by applying the window function at a period of the pitch to be generated. And a modulation signal generating means, wherein the modulation signal from the modulation signal generating means modulates the read start address of the reading means .