JP3095018B2 - Music generator - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a musical tone generator for generating a musical tone waveform using waveform data stored in a waveform memory. This tone generator can be used, for example, as a sound source of an electronic musical instrument.

[0002]

2. Description of the Related Art Conventionally, waveform data stored in a waveform memory is read out at a speed corresponding to a pitch of a generated musical tone, and a musical tone for forming a musical tone waveform by controlling an envelope or the like of the read waveform data. Generators are known. In a tone generator using such a waveform memory, the method of controlling the timbre during waveform reproduction is limited. A plurality of waveform data is prepared in a waveform memory, and waveform data of a timbre corresponding to performance data is selected from the waveform data, and read out to form a musical tone. But,
When a waveform played by a specific performance expression and having a characteristic corresponding to the performance expression is stored in the waveform memory, even if only one performance expression is picked up, a short time slur, a long time slur, etc. The shapes of generated musical sound waveforms vary variously, and it is not practical to store all of these musical sound waveforms. Therefore, when it is desired to control the timbre in accordance with the performance data, a general method is to process the waveform data read from the waveform memory with a digital filter having a frequency characteristic corresponding to the performance data.

In any case, the reading of the waveform data is controlled only in accordance with the pitch of the musical tone to be generated, and the time axis of the read waveform data cannot be freely controlled regardless of the pitch. There was a problem. For example, if the reading rate is increased, the pitch is increased, but the entire length of the waveform is shortened. Conversely, if the reading rate is decreased, the pitch is decreased, but the entire length of the waveform is increased. Even when individual time lengths such as a rising portion, a center portion, and a falling portion in one waveform are controlled, the pitch is determined in the same manner.

If the time axis of the waveform data to be read can be freely controlled, the types of timbres that can be generated from one type of waveform data during waveform reproduction can be increased. For example, different timbres can be generated by changing the attack length without changing the pitch. Also, the performance scene can be greatly expanded. For example, when reading a slur waveform, if the waveform is contracted in the time axis direction without changing the pitch, a slur that has a shorter time than the slur at the time of recording and a slur that has a longer time can be generated by extending the slur. When reading the vibrato waveform, if the waveform is extended in the time axis direction without changing the pitch, the vibrato becomes slower, and if the waveform is reduced, the vibrato becomes faster. In any case, it is necessary to expand or contract the waveform in the time axis direction regardless of the pitch.

On the other hand, in the field of recording / reproducing technology, a technology for extending a voice waveform in the time axis direction without changing the pitch in order to make it easy to hear the words of the fast voice, a technology for returning the pitch of the reproduced voice to the original pitch at the time of double speed reproduction. Etc. are known. It is conceivable that such a technique is applied to the above-described tone generator. However, the pitch of the musical tone waveform data dynamically fluctuates as the waveform data progresses. The conventional time axis compression / expansion technology described above is a technology for voice that does not require pitch control, and is not easily applied to a case where pitch control in units of cents is required, such as a sound source of an electronic musical instrument. Also, while the musical tone waveform must be controlled in a different manner for each tone depending on the performance data,
The conventional time-axis compression / expansion technology is for uniformly processing input waveform data, and the read rate of musical tone waveform data cannot be freely controlled in accordance with the pitch of musical tones to be generated. was there.

For this reason, when waveform data having characteristics corresponding to the performance expression is stored in a waveform memory, and the read-out tone waveform is partially skipped or repeated at the time of reading, the shape of the tone waveform is changed. If one looks closely at the original waveform data, each cycle is not constant. Therefore, when attempting to partially extract or repeat the read waveform data as it is when reading from the waveform memory, the connection at the boundary is poor and the waveform processing operation for joining is difficult. there were.

[0007]

SUMMARY OF THE INVENTION The present invention has been made in order to solve the above-mentioned problems, and it is possible to freely control the pitch of a musical tone to be generated and the time axis of waveform data read out from a waveform memory. It is an object of the present invention to provide a musical sound generating device and a musical sound generating method capable of controlling and improving connection of a plurality of partial waveform data.

[0008]

According to the first aspect of the present invention, in a tone generator, a plurality of unit waveform data obtained by dividing a plurality of continuous tone waveforms by one or a plurality of cycles are stored. Read-out means for generating a read address that increases at a speed corresponding to a designated musical tone pitch, reading the plurality of unit waveform data from the waveform memory by the read address, and a virtual address that changes with time. Virtual address output means for outputting a read address different from the read address in accordance with a difference between the read address and the virtual address, and the read means replaces the current read address with the replacement read address. And control means for controlling the plurality of unit waveform data to be read as the read address. Therefore, the compression / expansion of the time axis can be controlled by the virtual address, and the pitch of the generated musical tone and the time axis of the waveform data read from the waveform memory can be freely controlled. Also,
Even during the reading of the waveform data, the compression ratio in the time axis direction can be precisely controlled. Since the unit waveform data is divided in units of one or a plurality of cycles, the connection of the unit waveform data is improved.

According to a second aspect of the present invention, in the musical sound generating device, a waveform memory storing a plurality of unit waveform data in which a plurality of continuous musical tone waveforms are divided in units of one or a plurality of periods; Reading means for generating a read address that increases at a speed corresponding to the selected tone pitch, reading the plurality of unit waveform data from the waveform memory based on the read address, and a virtual address for outputting a virtual address that decreases and changes with time Output means; and control means for performing control to change the read address generated by the read means so as to read a cycle at a past position on a time axis according to a difference between the read address and the virtual address. And

According to a third aspect of the present invention, in the musical tone generator, a waveform memory storing a plurality of unit waveform data in which a plurality of continuous musical tone waveforms divided in units of one or a plurality of periods is provided. Reading means for generating a read address which increases at a speed corresponding to the selected tone pitch, reading the plurality of unit waveform data from the waveform memory in accordance with the read address, and a virtual means temporally varying in accordance with parameters relating to tone control. Virtual address output means for outputting an address, and control for changing the read address generated by the read means so that the musical tone waveform is expanded or contracted in time in accordance with a difference between the read address and the virtual address. And control means for performing the operation.
According to the fourth aspect of the present invention, there is provided a musical sound generator.
And a plurality of continuous musical tone waveforms in one or more cycles
Stores a plurality of unit waveform data separated by
At the speed corresponding to the specified musical tone pitch
Generating an increasing read address, and
Reading the plurality of unit waveform data from the waveform memory
Read-out means and a virtual address which decreases and changes with time.
Output means for outputting a virtual address;
What is the read address according to the difference with the virtual address?
Generate a different alternate read address and the read means
The replacement read address in place of the read address of the
Reading the plurality of unit waveform data as an output address
Control in such a way that the past position on the time axis
Control means for reading out the cycle of
is there. According to the fifth aspect of the present invention, a musical sound generator is provided.
, One or more musical tone waveforms of continuous plural cycles
Multiple unit waveform data separated by the cycle of
According to the stored waveform memory and the specified tone pitch
Generating a read address that increases with speed,
The plurality of unit waveform data from the waveform memory
Reading means for reading the tone and parameters related to the tone control.
Virtual address that outputs a virtual address that changes over time
Address output means, the read address and the virtual address.
Read address different from the read address according to the difference between
An address is generated, and said read means outputs a current read address.
And the replacement read address is used as the read address.
Control to read out the plurality of unit waveform data
By and this, be such a tone waveform is telescopic time axis
Control means.

A musical tone generating apparatus according to the invention described above has a waveform storing a plurality of unit waveform data in which a plurality of continuous musical tone waveforms are divided in units of one or a plurality of periods and whose period length is standardized. A memory for generating a read address that increases at a speed corresponding to a designated tone pitch, reading means for reading the plurality of unit waveform data from the waveform memory based on the read address, and outputting a time-varying virtual address; A difference between the read address and the virtual address by comparing the read address with the virtual address by using a cycle number of the unit waveform data as a unit. An alternate read address that differs by an integer multiple of the standardized cycle length from the read address in accordance with a difference between the address and the read address. A scan occurs, the reading means those having a control means for controlling to read said plurality of unit waveform data of the replacement read address as the read address instead of the current read address.

Further, a waveform memory storing a plurality of unit waveform data in which a plurality of continuous musical tone waveforms are divided in units of one or a plurality of periods, and a read address which increases at a speed corresponding to a designated musical tone pitch. Reading means for reading the plurality of unit waveform data from the waveform memory in accordance with the read address, and number output means for outputting a number of the unit waveform data read by the reading means, which changes with time. Virtual address output means for outputting a virtual address, and detecting that the difference between the number and the virtual address is greater than or equal to a predetermined value, and changing the read address generated by the read means so as to reduce the difference. Some have control means for performing control.

[0013] Each of the above-mentioned musical tone generating devices according to the first to third aspects of the present invention, and each means constituting the above-mentioned musical tone generating device related to these musical tone generating devices can be realized by hardware. it can. Further, it is possible to create a program for causing a computer to function as each unit described above. This program can be recorded on a recording medium, and the program can be read from the recording medium by a computer.

[0014]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS First, a waveform preparation stage will be described with reference to FIGS. FIG. 1 is a principle explanatory view of a processing method of musical tone waveform data stored in a waveform memory used in a musical tone generating device of the present invention. In the drawing, 1 is a sample value of an original waveform having a long cycle (cycle length L ₁ ), 2 is a sample value of an original waveform having a short cycle (cycle length L ₂ ), and 3 is a standardized cycle length CL (Cycle Length). Is a sample value of the stored waveform of FIG. When storing a musical sound waveform, an original waveform having a long cycle is compressed at a compression rate α ₁ = CL / L ₁ ,
The original waveform having a short cycle is expanded at a compression rate α ₂ = CL / L ₂ to form a storage waveform whose cycle is standardized to CL and stored in the waveform memory.

The compression and decompression are realized by changing the sampling frequency of the waveform data by using a so-called sampling rate conversion technique.
The number of samples per cycle changes. For example, if the sampling frequency of waveform data of 100 samples per cycle is multiplied by 1.5, waveform data of 150 samples is obtained per cycle, the original waveform is expanded at a compression rate α ₂ = 150/100, and the cycle length is A storage waveform standardized to 150 samples is formed and stored in the waveform memory. in this way,
In the waveform preparation processing, a stored waveform in which the length of the original waveform is multiplied by α by compression and expansion is generated.

More specifically, a point in time at which the phase becomes the same is detected from the musical sound waveform data, and this point is designated as a breakpoint of the original waveform. A process of interpolating the sample values so as to have the determined number of samples is performed to obtain the sample values 3 of the stored waveform having the standardized cycle length CL. Since the sampling frequency is constant, the cycle length can be represented by the number of samples per cycle. In the above-described waveform processing, the cycle length is compressed or expanded by increasing or decreasing the number of samples per cycle. Note that α ₂ = CL /
L ₂ is takes a value greater than 1, also stretched in order to use the same calculation formula as that of the compression is expressed at a compression rate alpha.

As a point of a cycle break, when a musical tone waveform has been read out to one point, the waveform is relatively smooth even if it is directly connected to a waveform starting from another arbitrary point at that time. And a position where noise is relatively unlikely to occur is selected.
In the present invention, it is called an in-phase point. For example, a zero cross point having an amplitude value of zero is preferable.
The range delimited by the same phase point is each one cycle of the waveform data of a plurality of cycles, and the length is standardized to a predetermined length.
The unit waveform data after the normalization processing is stored in the waveform memory.

As described above, in the present invention, each cycle length of the waveform data is standardized, and is processed into unit waveform data in which the cycle length CL is set to a constant value, and then stored in the waveform memory. The compression rate α of the normalized unit waveform data for the waveform is also stored in the cycle data memory,
At the time of reproduction, the original waveform can be reproduced as it is using this compression ratio α, or the read-out waveform data can be compressed and decompressed in the time axis to generate a tone waveform having a tone different from the original waveform. Since the length of each cycle of the waveform data of a plurality of cycles to be stored in the waveform memory is standardized in advance, it is possible to easily switch from reading one cycle of the waveform data to reading another cycle when reading the waveform data. become.

FIG. 2 is a schematic block diagram of a first device for creating a waveform memory used in the tone generator of the present invention. In the figure, 11 is a recording / reproducing unit, 12 is a cycle length standardizing unit, 13 is each cycle length designating unit, 14 is a waveform data writing unit, and 15 is a cycle data writing unit.

The waveform input is digitally recorded in the recording / playback unit 11. At the time of this reproduction, the recorded waveform data of a plurality of cycles is automatically detected by the cycle length designation unit 13 or the length of each cycle of the waveform data is designated by the operation of the user's operation element. The length of the cycle, L ₁ and L ₂ in the example of FIG. 1, is determined, and the compression ratio α is determined. The compression ratio α is a value obtained by dividing the normalized cycle length by the cycle length of the original waveform. The cycle length normalizing unit 12 performs compression and decompression based on the determined compression ratio α to create waveform data in which each cycle length CL of a plurality of cycles of waveform data is normalized to a predetermined length. The waveform data writing unit 14 creates a waveform memory with the standardized waveform data, and the periodic data writing unit 15 converts the compression rate α or the compression rate α ′ in cent units into create.

In the embodiment of the tone generator of the present invention, which will be described later, the waveform data prepared by the above-mentioned method is used. However, the waveform data can be created by the following device. FIG. 3 is a schematic configuration diagram of a second device for creating a waveform memory used in the tone generator of the present invention. In the figure, the same parts as those in FIG. Reference numeral 21 denotes a separating unit, and reference numeral 22 denotes an aperiodic waveform writing unit.
Compared to the first device shown in FIG. 2, the separating unit 21 separates the recorded waveform data into a periodic component and an aperiodic component. And the periodic component is stored in the waveform memory with the periodic length set to a predetermined length, as in the first device shown in FIG. The aperiodic waveform and the periodic waveform are stored so that they can be read out synchronously.

As the separating section, for example, a filter for separating a frequency band having a large number of periodic components, a gate circuit for separating a period having a large number of periodic components in one waveform, or the like can be used. In such tone generation using non-periodic waveform data and periodic waveform data, two channels of a sound source are used, and one of the channels reads out non-periodic waveform data in a normal manner, and the other channel reads periodic waveform data. While performing the time axis compression / expansion.

FIG. 4 shows a tone generator according to the present invention.
FIG. 5 is a schematic explanatory diagram of a storage format of waveform data stored in a waveform memory. FIG. 4A shows an example in which a series of all waveforms from rising to falling is stored, and FIG.
This is an example in which a waveform of an attack portion and a waveform of a loop portion are cut out and stored in a waveform memory.

As shown in FIG. 4A, the performance sound of a natural musical instrument having a plurality of cycles has the waveform data of one cycle equal to a predetermined standardized length, in other words, one cycle of the waveform data. A predetermined number of sample values are stored at each address of the waveform memory in order from rising to falling. In the figure, A0 to An-1 are the starting addresses of the periods 1 to n-1 of the normalized waveform data. On the original waveform of a plurality of cycles, each in-phase point for each cycle is determined, and normalization is performed so that the interval between the in-phase points becomes the cycle length CL. As a result, the start addresses Ai of each cycle i after the standardization are at regular intervals equal to the cycle length CL.

In FIG. 4B, when a performance sound of a natural musical instrument is reproduced, a waveform of an attack portion and a waveform of a loop portion are extracted. As in FIG. 4A, A0 to A
n is the start address of the period 1 to period n of the normalized waveform data, and the start address is at regular intervals equal to the period length CL. The waveform data of the attack part is represented by the start address AS =
The period of the waveform data from A0 is 0 to the period m-1, and the waveform data of the loop portion is the period m to n of the waveform data from the start address LS = Am. One period of the modulation frequency of a modulated waveform having a periodicity such as a steady waveform, vibrato, tremolo, and torill is used as the waveform data of the loop portion. The modulation frequency is, for example, several Hz to vibrato or the like.
The frequency is about several tens Hz. Also, a plurality of cycles of the modulation frequency may be used as the loop unit.

In order to smooth the portion returning from the end of the loop to the end, the waveform obtained by cross-fading the end of the waveform cut out for the loop and the waveform at the end is used as the end of the loop. May be stored. Alternatively, the middle of the waveform of the attack portion may be extracted and cross-faded on both sides of the waveform, and the length of the attack portion may be shortened and stored. In FIG. 4B as well, the compression ratio α for each cycle is stored as cycle data. However, since the compression ratio does not change so much in portions other than the attack portion, it may be stored only once in a plurality of cycles.

In the above description, the normalized waveform data has the same period length regardless of the pitch, in other words, the same number of samples per period. However, the cycle length CL to be normalized can be changed according to the pitch and the like. FIG. 5 is an explanatory diagram showing first and second examples of normalizing the cycle length. FIG. 5A shows a first example, in which the lengths of the periods in one waveform data are made uniform regardless of the timbre or pitch. For example, one cycle length is 1k (1024) samples. In the following example, 1k samples will be described as a reference value, but this value is a numerical example for description.

FIG. 5B shows a second example.
The length of the cycle to be aligned is different for each octave range. Allocate memory banks for each range,
In the range of F # 1 to 1024, the number of samples in one cycle is 1024, but the number of samples is halved every one octave up, and the range of G7 to F # 8 is 8 samples. Within one waveform data, they are set to the same cycle length (number of samples).

A plurality of periods may be defined as one unit waveform data. For example, in the range of G1 to F # 2, two periods are set to one and the unit waveform data is 1024 samples. When reading the waveform memory, the number of digits of the cycle number for specifying the cycle of the plurality of unit waveforms is reduced, and the number of bits of the counter can be reduced. This is particularly effective in the high octave range because the number of samples is small.

FIG. 6 is an explanatory diagram showing a third example of the standardization of the cycle length. The length of the period to be aligned in each of the sound ranges divided into four semitones is different. A memory bank is allocated to each range, and the range of G0 to A # 0 has 1536 samples in one cycle, but the number of samples is set to 5/6 to 3/4 every time the range is increased by one. As for the number of shift-down operations, the operation of the shift register added to the 1024-bit counter will be described later. In one waveform data, the number of cycles (the number of samples) is equalized.

In the above description, each cycle length in one rising to falling waveform is set to a different length depending on the range of the waveform. However, the cycle length of the attack is increased, and the cycle length of the sustain is increased. For example, the sound may be switched in accordance with the sounding period during sounding. Further, the above-described example is an example of waveform data belonging to one tone color. When waveform data of a plurality of tone colors is prepared, the cycle length may be determined independently for each tone color.

In the above-described modified example, the cycle length is varied depending on the tone range, the tone generation period, the timbre, etc. However, even if the cycle length is not standardized at all, the detection of the end of the unit waveform is one sample period. The number can be easily detected by a counter that counts the number, and the process can move to the start address of the next unit waveform. Also, the connection between the unit waveforms is improved. When preparing a plurality of variations of waveform data (a waveform with strong / weak touch, a waveform with / without modulation, etc.) in one tone color, the cycle of each waveform data becomes in phase with each other. So that By doing so, it is possible to switch to reading an arbitrary cycle of waveform data of a different variation in the middle of reading a waveform data of one variation while suppressing noise.

FIG. 7 is an overall configuration diagram for explaining an embodiment of a musical sound generating apparatus according to the present invention. In the figure, 31 is a performance input section, 32 is a setting input section, 33 is a control section, 34 is a sound source section, 35 is a control register section, 36 is a waveform generation section,
7 is a volume control unit, 38 is a CH accumulation unit, 39 is a DAC,
0 is a sound system.

The performance input unit 31 includes a MIDI keyboard,
It is a performance operator such as a MIDI guitar, a wheel operator, a pedal operator, a joystick, or a combination of arbitrary operators among them, and an automatic performance device, and supplies performance information such as a MIDI event. The setting input unit 32 is a display, a panel switch, a slider, a jog dial, and the like.
The user inputs the setting information and displays the setting information and the like. The control unit 33 includes a CPU, a ROM, a RAM, and other peripheral circuits, and performs various settings of the musical tone waveform generator according to the setting information, and controls sound generation for the sound source unit 34 according to the performance input.

The control register section 35 in the tone generator section 36 receives and holds data such as tone color designation data, pitch data, envelope data, and sound generation start / end data from the control section 33. The waveform generating unit 34 includes the control register unit 3
5 to generate waveforms for a plurality of channels by time division. The volume control unit 37 adds a volume change characteristic from the rise to the end of the musical tone to the generated waveform of each channel.
While generating a decay, sustain, release (ADSR) type envelope, the waveform generated by the waveform generator 36 is multiplied by the envelope to control the volume.

The above-described waveform generator 36 and volume controller 37
Is performed independently for each sounding channel. The CH accumulating unit 38 combines the waveforms of the plurality of channels to which the envelope characteristic is added, and supplies the synthesized waveform to the DAC (D / A converter) 39. The DAC 39 outputs the waveform returned to the analog signal to the sound system 40.

FIG. 8 is an internal configuration diagram of the waveform generator shown in FIG. In the figure, 51 and 52 are adders, 53 is an F number generator, 54 is a counter, 55 is a waveform selector, 56
Is an LPF, 57 is a start address and cycle length storage section, 58 is a cycle data storage section, 59 is a first cycle number register, 60 is a second cycle number register, 61 is a first processing section, and 62 is a second processing section. , 63 is a waveform memory, 64
Denotes a first interpolation unit, 65 denotes a second interpolation unit, 66 denotes a cross-fade synthesizing unit, and 67 denotes a virtual address counter unit.

In this embodiment, two series of readings are performed from the waveform memory 63 by shifting the read address for each tone generation channel. The virtual address counter 67 indicates the trajectory of the position from which the waveform data is to be read out from the waveform memory 63 with the passage of time, and the waveform selector 55 switches the two series with the target indicated by the virtual address VA as the target. Makes the address to be read follow. At the same time, the cross-fade synthesizing unit 66 is controlled to select one of the two streams or to combine the two streams. 1
When reading consecutive cycles of two waveform data sequentially,
Reading is performed using only one of the two sequences. When the read cycle jumps, one cycle reads the cycle following the cycle read immediately before, while the other stream reads a new cycle at the jump destination, and cross-fade synthesizes the two waveforms. Part 66
To improve the connection of the waveforms.

In this embodiment, an F-number (frequency number) calculation method is employed as phase data indicating an address of a sample point in a unit waveform stored in the waveform memory 63. The numerical value proportional to the pitch frequency of each key is accumulated by the counter unit 54, and the sample value is read out in real time using the integer part as the address of the sample point in the unit waveform.

The note number (in cents) that is proportional to the pitch frequency of each key is added to a pitch shift input such as pitch bend (in cents) in an adder 51, and an LPF (low-pass filter) is added in an adder 52. The number is added to the output of 56 and input to the F-number generator 53. Other input data of the pitch shift include detune data for specifying a shift from the reference pitch, low-frequency waveform data generated by an LFO (low-frequency oscillator), pitch envelope data generated by a pitch envelope generator, and the like. The data is supplied to the F-number generator 53 alone or as a mixture of a plurality of these data.

The LPF 56 is a digital low-pass filter. The LPF 56 receives the compression ratio α ′ for each unit waveform from the periodic data storage unit 58, performs filtering so as to make a smooth change, and outputs the compression ratio α ′ to the addition unit 52. Output. The cycle data storage unit 58 stores a compression rate α ′ obtained by converting the compression rate α of each cycle used when processing the selected waveform data into cent units. Since the input data of the note number and the pitch shift described above are both in cents, the portion to be multiplied can be added.

By adding the compression ratio α ′, the frequency is multiplied by the compression ratio, and the cycle length of the normalized unit waveform data is changed to the fluctuating cycle length before the standardization. Will be returned. If it is not necessary to return to the fluctuating recording waveform data using the standardized cycle length as it is,
There is no need to add the compression ratio α '.

As described with reference to FIG. 5B, in the second example of the period length standardization, the period length (the number of samples) of the unit waveform used for generating the musical tone differs depending on the musical range. . The F-number generator 53 outputs frequency information (F-number) corresponding to the tone pitch in consideration of this. Further, as described with reference to FIG. 6, in the third example of the standardization of the cycle length, the banks 1 to 3 and the banks 4 to 4
6, the banks 7 to 9 have the same F number because the count range (the number of bits to be masked) of the counter unit 54 is the same as described below.

The counter section 54 is a counter for accumulating the F number. The integer part needs to be able to count up to 1024 at the maximum, and the decimal part uses about 15 bits in order to generate a musical tone of an accurate pitch asynchronously with the pitch. Therefore, the total is about 25 bits. The counter section 54 is reset by note-on, accumulates the F number of each channel in each sampling period, and reads a pointer p for reading out the unit waveform of each series.
1 and p2 are output.

The waveform selecting section 55 includes the control section 3 shown in FIG.
3, the start address A0 and the cycle length CL of the waveform data to be read designated by the CPU in
It is received from the start address and cycle length storage unit 57. The start address and cycle length storage unit 57 stores a start address of the rising edge of the selected waveform data,
Period length of unit waveform of the waveform (value common to each unit waveform)
I remember.

The first and second cycle number registers 59 and 60 in the waveform selector 55 hold cycle numbers CN (cycle numbers) 1 and 2 which are read in each series, respectively. The cycle numbers CN (cycle numbers) 1 and 2 are values corresponding to “each standardized cycle i” in the description with reference to FIG. 4, in other words, the cycle numbers CN (cycle numbers) 1 and 2 2 specifies “each period i after standardization”. The start address of each unit waveform in one waveform is calculated by the following equation for each series. ADS1 = A0 + CL × CN1 ADS2 = A0 + CL × CN2 As described above, the head address of each unit waveform can be easily obtained without previously storing it, and it is easy to arbitrarily connect a plurality of unit waveform data. is there.

The waveform selector 55 has a head address ADS
1 and ADS2 are sent to the processing units 61 and 62, and an instruction is sent to the first and second processing units 61 and 62 to change the designated range of the pointers p1 and p2 according to each cycle length CL.
Further, the output from the counter unit 54 is monitored, and the timing at which the pointers p1 and p2 pass through the range of the standardized cycle length CL, that is, the timing at which reading of one designated cycle is completed, is detected. The connection with the next waveform is controlled at that timing. However, the end timing differs depending on the bank.

As for the unit waveform data, one cycle of the unit waveform data is arbitrary n
It is standardized by the cycle length (the number of samples) obtained by multiplying the number represented by bits by 2 ^m . For example, FIG.
In the case of the standardization example of (b), m = 3, n for bank 1
= 7, m = 3, n = 6 for bank 2, m = 3, n = 5 for bank 3, and m = 3, n = 5 for bank 8.
3, n = 0. The counter unit 54 has an integer part of m + n.
= 10 bits, and specifies the read address of the sample point within one cycle of the unit waveform data by the integer part of the pointers p1 and p2. In the case of bank 1, the end of the unit waveform data can be determined by detecting that the 10th bit of the most significant bit has been inverted from 1 to 0.
, The ninth bit in the case of bank 3, the eighth bit in the case of bank 3, and the third bit in the case of bank 8, the end of the unit waveform data can be determined by detecting the inversion from 1 to 0, respectively. .

As described above, if the normalization shown in FIG. 5B is performed, when a plurality of unit waveform data are sequentially read out from the waveform memory, even if the waveform data has different cycle lengths, one unit data is obtained. One address counter is shared, and the determination of the end of the unit waveform data can be made only by the determination of the upper bit of the counter unit 54. Although not shown, when the waveform data is read from the waveform memory in the first and second streams by simultaneous processing by time division, the first waveform data to be read in the first stream and the second waveform are read in the second stream. Even if the number of samples per cycle with the second waveform data is different, one address counter is shared, and the determination of the end of the unit waveform data for each series is made only by the determination of the upper bit of the counter unit 54 Can be.

FIG. 9 is a diagram showing the internal structure of the first and second processing sections shown in FIG. FIG. 9A shows a second example of the standardization of the cycle length described with reference to FIG.
Is used in the third example of the standardization of the cycle length described with reference to FIG. In the figure, 71 is an upper bit mask unit, 72 is an addition unit, 73 is a shift unit, and 74 is an addition unit.

In FIG. 9A, the upper bit mask unit 71 receives the cycle length CL and outputs all bits as it is in the case of the bank 1 shown in FIG. In the case of bank 3, the upper 2 bits are masked to 9 bits, and in the case of bank 3, the upper 2 bits are masked to 8 bits. 63. The same applies to the banks 4 and thereafter. as a result,
Addresses AD1, AD2 output from the processing units 61, 62
In bank 1, starting from the top address ADS1, ADS2, the range of 1024 samples changes at a speed corresponding to the F number, and in bank 2, 51
Addresses that change in a range corresponding to each bank are output, such as a range corresponding to two samples, a range corresponding to 256 samples in the bank 3, and the like.

As described above, if the normalization shown in FIG. 5B is performed, when a plurality of unit waveform data are sequentially read from the waveform memory, even if the waveform data has different cycle lengths, one unit data is obtained. One address counter is shared, and a read address that changes in the same phase can be created only by changing the bit mask. Further, as described above, when reading out the waveform data from the waveform memory in the first and second streams by simultaneous processing by time division, the first waveform data to be read out in the first stream and the second waveform to be read out in the second stream. Even when the cycle lengths of the two waveform data are different, it is possible to create a read address of each series that changes in the same phase by sharing one address counter and changing the bit mask for each series.

In FIG. 9B, the upper bit mask section 71 has 10 bits without masking in the case of the banks 1 to 3 shown in FIG. 6, and has 9 bits in the case of the banks 4 to 6, In the case of the banks 7 to 9, the bit is reduced to 8 bits, and the banks 10 to 12 and thereafter are similarly masked to reduce the bit. In addition, in the case of banks 1, 4, and 6, pointers p1 and p2 after the bit mask are
The addition unit 74 adds the output shifted down by 1 bit by 3 (to に よ り times), and outputs a pointer 3/2 times as a result. In the case of the banks 2, 5, 7,..., The output is added to the output (down to 1/4) shifted down by 2 bits by the shift unit 73, and as a result, a 5/4 pointer is output. , 6, 9..., The pointer after the bit mask is output as it is.

Therefore, the addresses AD1 and AD2 output from the processing sections 61 and 62 start from the top addresses ADS1 and ADS2 in the bank 1 shown in FIG.
The range of 36 samples changes at a speed corresponding to the F number. In bank 2, a range of 1280 samples from the start address, in bank 3, a range of 1024 samples, in a bank 4, a range of 768 samples, and so on. An address that changes within a range corresponding to the bank is output.

Next, returning to FIG. 8, reading of a waveform from the waveform memory 63 will be described. Address AD1 = (ADS1 + p1), AD2 =
The waveform data stored in (ADS2 + p2) is read. Here, the read pointers p1 and p2 start from a value 0 (almost 0) at the start of reading in each cycle,
For each sampling period, a rate corresponding to the musical tone pitch (a value determined by the F number and the shift amount of the shift unit 73)
To increase. Here, the start addresses ADS1, ADS2
Has only an integer part, and the F number and the pointers p1 and p2 have an integer part and a decimal part.
Is also an address consisting of an integer part and a decimal part. From the waveform memory 63, the read address AD
The two sample values of the address indicated by the integer part of 1, AD2 and the next address are read out, and the first and second values are read out.
Are output to the interpolation units 64 and 65.

The first and second interpolators 64 and 65 interpolate between two sample values read for each stream in accordance with the decimal part of the read addresses AD1 and AD2, respectively. Specifically, two series of interpolated sample values corresponding to the integer part and the decimal part of the read addresses AD1 and AD2 are output. When a crossfade is instructed, the crossfade synthesizing unit 16 gradually reduces the level of the pre-switching sequence among the input interpolated sample values for the two sequences from the maximum value (fade-out) and sets the post-switching sequence. Level gradually increases from zero (fade in)
Then, the sample values of the waveform data to be output are obtained by adding the two series of interpolated sample values after the level control. If there is no instruction for cross-fade, the level of the series that has just faded in is held at the maximum value, the level of the other series is held at zero, and waveform data obtained by combining the two series is output.

FIG. 10 is a flow chart for explaining a situation in which, in the tone generator according to the present invention, generation of a tone according to the instruction is started in the channel assigned to the tone generation of the tone generator in response to the tone generation start instruction. is there. FIG. 10 (a)
Is a main flowchart, and FIG. 10B is a flowchart of a key-on event. In FIG. 10A, in S81, the apparatus is initialized. S8
In step S2, processing related to key switch input is performed, in waveform generation S83, processing related to performance operator input is performed, and in step S84, processing related to setting operator input is performed.

FIG. 10B illustrates a process when a key-on event (an event for instructing sound generation) occurs in the key switch process of S82.
In, the key pitch is set in a parameter NN register, and the key pressing intensity (speed) is set in a parameter VEL register. In S86, the sound channel (ch) by the key-on event is assigned as the sound channel AS. In S87, the waveform selection information and the envelope information of the currently selected timbre TC are set in the control register of the tone generation channel of ASch. More specifically, information indicating a storage position of a waveform, an attack length m, a loop length n, a pitch corresponding to a pitch NN, each level of an envelope, each rate, and the like are written. In S88, a note-on is instructed to the sound channel of the ASch, and waveform reading, control of the volume envelope, and the like are started.

In the setting of the control register at the time of the tone generation instruction in S87 described above, if the compression / expansion is set not to be performed in the channel, the operation of the waveform generator after the note-on is as follows. First, as step 1,
Using only the first sequence, the waveform data stored in the waveform memory is read out from the address A0 (refer to FIG. 4A) of the waveform data corresponding to the selected timbre TC at the cycle 0. To start. Here, CN1 of the first cycle number register 59 is 0. Next, as step 2, reading is continued while updating the pointer p1 at a speed corresponding to the pitch. As step 3, the pointer p
When 1 reaches the cycle length CL (a constant length in the example of FIG. 4A), CN1 is incremented by one. Here, the position indicated by the pointer p1 has returned to the leading 0 according to the operation of the upper bit mask unit 71 in the processing unit 61 shown in FIG. Thereafter, steps 2 and 3 are repeated.

The virtual address counter 67 has a 10-bit integer part and a 5-bit decimal part on the scale of the cycle number.
This is a counter having 5 bits, and the time axis compression / expansion at the time of waveform reading is controlled according to the output value. The number of bits of the integer part is selected so as to be the maximum value of the number of periods of one waveform or more. The decimal part is for finely controlling the traveling speed VF (virtual F number) of the virtual address. The virtual address counter 67 inputs a loop address for specifying a range for repeatedly reading the same unit waveform data, a progress rate for reading one waveform, and the like, and outputs a virtual address to the waveform selector 55. The count value of the virtual address counter 67 advances as the time progresses or the cycle progresses. This count value is hereinafter referred to as a virtual address VA.

As a first example, for the virtual address VA, the above-mentioned traveling speed VF is set at every predetermined time (for example, every 10 milliseconds, every 2 milliseconds, or every 100 sampling cycles). Accumulate. If this value has a decimal part, the amount to be advanced in one accumulation can be specified as 1.2 or 0.8. The traveling speed VF in this case is the traveling rate of the virtual address VA based on the absolute time. Therefore, even if the pitch of the waveform is changed by changing the F number (the progress rate of the real address), the reproduction time of the waveform data is not changed.

As a second example, every time one unit waveform is read, the traveling speed VF is accumulated for the virtual address VA. In this case, the traveling speed VF is a relative rate based on the traveling rate (corresponding to the F number) of the real address when the waveform data is read without time axis compression / expansion.
For example, if the traveling speed VF is 2, the virtual address advances at twice the speed of the real address, and the reproduction time of the waveform is halved. In the following description, only the case where the virtual address VA progresses as time progresses as in the first example will be described.

The virtual address starts from 0 and indicates the locus of the position from which unit waveform data is to be read from the waveform memory 63 as time elapses. The waveform selector 55 causes the cycle numbers CN1 and CN2 stored in the first and second cycle number registers 59 and 60 to follow the cycle number indicated by the virtual address VA. That is, two series of read addresses are made to follow. This state is shown in FIGS. 11 to 17 as described later. At the same time, the cross-fade synthesizing unit 66 is controlled to select one of the two streams or to combine the two streams. More specifically, the cycle number VA indicated by the virtual address counter 67 and the series currently fading in the cross-fade section 66 or the series whose level is held at the maximum value, that is, the cycle number CN of the current series Then, it is determined whether to continue reading the cycle in the series or to start reading the cycle of a different number from the other series and switch to that series.

In the above description, the compression rate α ′ is read from the cycle data storage unit 58 using the current cycle number CN1 or CN2, and the pitch is controlled according to the read current waveform unit waveform data. It was assumed that Instead of this, as shown by a dotted arrow from the virtual address counter 67 to the periodic data storage 58 in FIG. By reading out the compression ratio α ′, the waveform data of the sequence currently fading in the cross-fade unit 66 or the sequence whose level is held at the maximum value corresponds to the progress of reading when the entire waveform is viewed. Reading can be performed with the changed pitch.

In the setting of the control register at the time of sounding instruction in S87 shown in FIG. 10B, if compression / expansion is set in the channel concerned, the waveform generator shown in FIG. A sound source operation for compressing / expanding the time axis of the waveform read from the waveform memory 63 while performing cross-fading by the cross-fade synthesizing unit 66 as necessary using both the first and second streams is started.

First, waveform data is output using one of the series of read addresses (tentatively AD1). If the trajectory of the cycle number CN1 of this series is sufficiently close to the trajectory of the virtual address VA, the reading of this series is continued. on the other hand,
When the trajectory of the cycle number CN1 of this sequence is far from the trajectory of the virtual address VA, the switching of the reading is instructed. Readout using the cycle number CN2 of
Crossfade is performed from this stream to the other stream. After the end of the crossfade, the waveform data is read at the read address (AD2) in the other series, the difference between the locus of the cycle number CN2 of this series and the virtual address VA is determined, and the same operation is performed.

The operation of the waveform generating section shown in FIG. 8 for the note-on in S87 shown in FIG. 10B will be described. The counter 54 accumulates the F number at each sampling cycle with zero as an initial value, and outputs the accumulated value to the processing units 61 and 62. On the other hand, the virtual address counter 67 outputs a virtual address VA that changes with time according to the traveling speed VF with zero as an initial value. The waveform selecting unit 55 sets the cycle number CN1 = 0 as an initial value, and firstly obtains waveform data corresponding to the tone color TC selected in the first series (see FIG. 4A for the storage format of the waveform data). Of the cycle 0 is performed. At this time, the interpolated sample value of the first interpolator 64 of the first system is output from the crossfade unit 66.

At the end of the reading of the cycle number CN1 = 0, the cycle number C subsequently read in the first stream
N1 + 1 = 1 is set to the new cycle number CN1, and the virtual address VA at that time is set to the new cycle number CN1.
= 1, and it is determined whether the difference is equal to or more than 周期 cycle. If the difference between the virtual address VA and CN1 = 1 is less than 周期 cycle, switching is not performed, and reading of the cycle corresponding to CN1 = 2, 3,. At the end of the cycle, the same operation as described above is repeated.

On the other hand, the virtual address VA is 1 /
Switching is performed when the signal is advanced by two cycles or more, or when the signal is delayed by １ ／ cycle or more. The waveform selection unit 55
A cycle number corresponding to the virtual address VA is set in CN2,
Subsequently, in the first stream, the cycle corresponding to CN1 is read out, and in the second stream, the above-described C is read.
A new cycle corresponding to N2 is read out, and the first interpolator 64 of the first stream is read by the crossfade unit 66.
Is controlled so as to perform cross-fading from the interpolation sample value of the second series to the interpolation sample value of the second interpolation unit 65 of the second series. This crossfade corresponds to the period C in the second sequence.
It is set to end before the reading of N2 ends.

After that, when the reading of the period CN2 is completed in the second period of the fade-in, the value of CN2 is incremented by one and the virtual address VA at that time is incremented.
And it is determined whether or not the difference is equal to or more than 周期 cycle.
The subsequent operation is the same as the operation after the determination is made in the first series described above. That is, when switching is necessary, crossfading is performed again toward the first stream, and when switching is not required, reading is continuously performed in the second stream. The criterion for determining whether or not to perform switching is not limited to the 周期 cycle as described above.
It can be arbitrarily determined, such as 5/4 cycle or 3 cycle.
In particular, the virtual address VA is also determined by the read address cycle CN.
As in the case of the first and CN2, if the cycle number of the unit waveform data is designated and compared with the read address cycles CN1 and CN2, the cycle numbers are compared. Can be easily performed with a small number of bits.

Generally, when switching is not performed,
The current sequence of the two sequences (the sequence that has just been faded in by the cross-fade unit 66 or the sequence that has been held at the maximum level) continues to be read in the cycle that follows the cycle that was just read out. Is output at the maximum level from the crossfade section. When the switching is performed, reading of the cycle following the cycle in which the current series was read immediately before is performed, and reading of the cycle to be switched corresponding to the virtual address VA is performed in the other series. A crossfade is performed from the current stream to the other stream.

FIG. 11 shows the progress of the first and second series of read addresses according to the virtual addresses described above.
17 to FIG. FIG. 11 shows a first example in which only the reproduction time is compressed at a constant pitch. In the figure, reference numeral 91 denotes a first series read address output from the first processing unit when no compression / expansion is performed, 92 denotes an output from the virtual address counter unit, 93 and 95 denote the first series read addresses, and 94 and 9.
6 is a second series of read addresses. The horizontal axis is time, and the vertical axis is the address of the waveform memory 63. When compressing the reproduction time, the output 9 of the virtual address counter unit is used.
The slope of 2 is steeper than the slope of the first series of read addresses 91 when no compression / expansion is performed. First, the first series read from the waveform memory 63 is output from the cross-fade synthesizing unit 66 using the first series of read addresses 93 as addresses.

Then, the first series of read addresses 9
The position 3 is delayed from the output 92 of the virtual address counter unit, and when the delay reaches a predetermined number of cycles, the read address 94 of the second stream is used as an address to advance the address by the predetermined number of cycles. , The second sequence is output from the crossfade synthesizing unit 66. At this time, instead of switching instantaneously, the cross-fade synthesizing unit 66 gradually increases the ratio of the second stream by using the output of two streams during a predetermined time before and after the switching.

Thereafter, the second series of read addresses 94
Also lags behind the output 92 of the virtual address counter.
Then, when this delay reaches the above-mentioned predetermined number of cycles, the read address 95 of the first series is set as an address,
The waveform memory 63 is read from the position where the address has been advanced by the predetermined number of cycles, and the first series is output from the cross-fade synthesizing unit 66. Also in this case, in a predetermined time before and after the switching, the ratio of the first stream is gradually increased by using the output of the two streams, and the output is output from the cross-fade synthesizing unit 66. Similarly, the read address 95 of the first stream is switched to the read address 96 of the second stream. As a result, two systems are alternately switched using the output 92 of the virtual address counter unit as the target value, and
One entire waveform is read while locally skipping the read address.

FIG. 12 shows a second example in which only the reproduction time is compressed at a constant pitch. In the figure, the same parts as those in FIG. In the example shown in FIG. 11, the output 92 of the virtual address counter section is set linearly so that the compression is uniformly performed during the entire waveform period. In the second example, the compression is gradually performed. It is a curve that increases the rate. Also in this case, compression can be realized by alternately switching the two systems in the same manner as in the example shown in FIG. At predetermined times before and after the switching, the two series of outputs are cross-fade and output from the cross-fade synthesizing unit 66. Note that, instead of fixing the predetermined number of cycles, which is the difference in delay, to a fixed value, it is also possible to adaptively control by setting a larger value as the compression ratio increases.

FIG. 13 shows a first example in which only the reproduction time is extended at a constant pitch. In the figure, the same parts as those in FIG. When the reproduction time is extended, the slope of the output 92 of the virtual address counter unit is set to the first series of read addresses 9 when the compression / expansion is not performed.
Slower than 1 slope. First, the first series read from the waveform memory 63 is output from the cross-fade synthesizing unit 66 using the first series of read addresses 93 as addresses.

When the advance of the first series of read addresses 93 from the output 92 of the virtual address counter reaches a predetermined number of cycles, the second series of read addresses 94 are used as addresses and the above-described predetermined number of cycles are used. The waveform memory 63 is read from the position delayed by only the address, and the second series is output from the cross-fade synthesizing unit 66. At predetermined times before and after the switching, the two series of outputs are cross-fade and output from the cross-fade synthesizing unit 66. Thereafter, when the advance of the second series of read addresses 94 from the output 92 of the virtual address counter unit reaches the above-mentioned predetermined number of cycles, the first series of read addresses 95 are again used as addresses, and The waveform memory 63 is read from the position where the address is delayed by the number of cycles, and thereafter, similarly, two systems are alternately switched. Eventually, two systems are alternately switched using the output 92 of the virtual address counter unit as the target value, and the address of one entire waveform is read while the read address is locally repeated.

FIG. 14 shows a second example in which only the reproduction time is extended at a constant pitch. In the figure, the same parts as those in FIG. In this example, the output 92 of the virtual address counter section is set linearly so that the example shown in FIG. 13 is uniformly expanded during the entire waveform period. In the second example, The curve is such that the compression ratio gradually decreases (the expansion ratio increases). Also in this case, decompression can be realized in the same manner as in the example shown in FIG. Note that, instead of fixing the predetermined cycle number, which is the difference in advance, to a fixed value, it is also possible to adaptively control by setting a smaller value as the compression ratio becomes lower.

The F number corresponds to the pitch of the musical tone to be generated, so that the counter 54 increases at a speed corresponding to this pitch. On the other hand, the traveling speed VF of the virtual address VA can be arbitrarily set independently of other tone characteristics such as the pitch of the tone. The traveling speed VF may take not only a positive value but also a negative value, or may be largely changed during the generation of a musical tone. When the traveling speed VF takes a negative value, in reading in each cycle, the real address advances in the positive direction and is sequentially cross-fade to a cycle at a past position on the time axis. The read position appears to move in the direction. When the traveling speed VF is changed during the generation of a musical tone, the waveform data can be partially compressed and expanded differently. For example,
Assuming that the traveling speed VF is high in the address range of the attack portion of the waveform data and the traveling speed VF is low thereafter.
The attack part is compressed, and the expanded waveform data after that is output.

FIG. 15 shows an example in which only the pitch is raised while the reproduction time is constant. In the figure, the same parts as those in FIG. Reference numeral 97 denotes a first series of read addresses output from the first processing unit. To increase the pitch, the F number output from the F number generation unit 53 is increased, and the slope of the first series of read addresses 91 when no compression / expansion is performed is determined by the output 9 of the virtual address counter unit.
Steeper than the slope of 2. First, the first series read from the waveform memory 63 is output from the cross-fade synthesizing unit 66 using the first series of read addresses 93 as addresses.

Then, the first series of read addresses 9
Reference numeral 3 designates a position which advances from the output 92 of the virtual address counter unit and, when the advance reaches a predetermined number of cycles, the read address 94 of the second stream is used as an address and the address is delayed by the above-described predetermined number of cycles. From the waveform memory 63, and the second sequence is
Output from. At this time, instead of switching instantaneously, the cross-fade synthesizing unit 66 gradually increases the ratio of the second stream by using the output of two streams during a predetermined time before and after the switching.

Thereafter, the second series of read addresses 94
Also proceeds from the output 92 of the virtual address counter. Therefore, when the advance reaches the above-described predetermined number of cycles, the waveform memory 63 is read from a position where the address is delayed by the above-described predetermined number of cycles, using the read address 95 of the first series as an address. The sequence is output from the crossfade synthesizing unit 66. Also in this case, in a predetermined time before and after the switching, the ratio of the first stream is gradually increased by using the output of the two streams, and the output is output from the cross-fade synthesizing unit 66. Similarly, the switching is performed from the read address 95 of the first stream to the read address 96 of the second stream, and further to the read address 97 of the first stream. Eventually, in this case as well, the output 92 of the virtual address counter section is used as a target value to alternately switch two systems, and the address of one entire waveform is read while the address is locally repeated.

FIG. 16 shows an example in which only the pitch is lowered while the reproduction time is constant. In the figure, the same parts as those in FIG. To lower the pitch, F
The F number output from the number generation unit 53 is reduced to make the slope of the first series of read addresses 91 gentler than the slope of the output 92 of the virtual address counter unit when no compression / expansion is performed. First, the first series read from the waveform memory 63 is output from the cross-fade synthesizing unit 66 using the first series of read addresses 93 as addresses.

The first series of read addresses 93 lags behind the output 92 of the virtual address counter unit. When the delay reaches a predetermined number of cycles, the second series of read addresses 94 are used as addresses. The waveform memory 63 is read from a position where the address is delayed by the predetermined number of cycles, and the second series is output from the cross-fade synthesizing unit 66. At that time, for a predetermined time before and after the switching, the cross-fade is performed using the outputs of the two streams and output from the cross-fade synthesizing unit 66. Thereafter, similarly, two systems are alternately switched using the output 92 of the virtual address counter section as a target value, and the address of one entire waveform is read while the address is locally skipped.

As described with reference to FIGS. 15 and 16, even if the pitch of the musical tone to be generated is changed, the same time as the original one can be obtained without changing the shape of the time change of the virtual address VA. Waveform data in which only the pitch is changed on the axis can be obtained. Since the compression / expansion of the time axis is controlled by the virtual address, the pitch can be changed with high accuracy, unlike a normal pitch changer.

In the example described with reference to FIGS. 11 to 16, the virtual address VA is made to proceed continuously, but it may be made to jump to another value after having made progress to a predetermined value. FIG. 17 shows an example in which compression and decompression are performed while looping virtual addresses. In the figure, the same parts as those in FIG. 98,100 is the second
Is a second series of read addresses output by the processing unit, and 99 is a first series of read addresses output by the first processing unit.

For example, the waveform data as shown in FIG. 4B is prepared, and the virtual address VA is set to the loop end LE.
Is reached, the loop progresses to return to the loop start address LS. Here, the loop start address LS and end address LE can be designated only by their cycle numbers. In the example of FIG. 4B, “m” may be designated as the LS address and “n” may be designated as the LE address. Alternatively, “n” may be designated as the LE and “nm” may be designated as the loop size. In the loop progress operation, when the integer part of the virtual address VA that sequentially proceeds reaches LE = n, (VA−n + m) is calculated, and the calculated value is set in the virtual address VA as the return address of the loop.

Even when the loop progress is set to the virtual address, if only the virtual address counter 67 is changed to have the above-described loop progress function, the waveform generator 3
The operation of the other blocks in 6 does not need to be changed. The waveform selection unit 55 determines the virtual address V
A is input and compared with the current series period CN, and the read address of the waveform data is made to follow the virtual address in the same procedure as described above.

This situation is illustrated in FIG. 17 and FIG.
1, the first and second series of read addresses 93,
After the replacement of the virtual addresses 94, 95, and 96, the virtual address VA becomes the loop end LE = n, and when returning to the loop start LS = m, the virtual address VA is replaced with the read address 97 of the first system, so that the virtual address VA becomes a predetermined value. The waveform memory 63 is read from a position where the address is advanced by the number of cycles, and thereafter, similarly, the read addresses 9 of the first and second series are read.
3, 94, 95 and 96 are replaced. At a predetermined time before and after the switching, cross-fading is performed using two series of outputs. In the example shown, taking into account crossfade,
The cycle number of the loop end of the virtual address VA is set slightly before the last cycle number stored in the waveform memory.

In this embodiment, the loop can be read out of the waveform while controlling the time axis compression / expansion only by making the virtual address loop proceed, and the configuration is simple. In comparison,
When trying to control the loop with the read address of the waveform, in addition to the process of detecting that the read address has reached the loop end address and returning the read address, the process of returning the virtual address to the loop start must be performed. .

In the description with reference to FIGS. 15 to 17, the output 92 of the virtual address counter is changed linearly, but it may be changed in a curve. FIG.
In the example of FIG. 16, the F number output from the F number generating unit 53 is changed in a curved line with respect to time, so that the inclination of the first and second series of read addresses 91 is changed in a curved line. You may.

In the above description, as shown in FIGS. 5 and 6, waveform data having different cycle length (number of samples) of one unit waveform data for each bank is handled by one counter unit 54 shown in FIG. In order to make it possible, a configuration in which the upper bit mask is performed in the first and second processing units shown in FIGS. 8 and 9 is adopted. Instead, the operation of the counter unit 54 may be controlled so that when the predetermined final address corresponding to the bank is determined, the address becomes the head address.

In the above description, crossfading is always performed at the time of switching between two systems, but crossfading is not always necessary. This is because the unit waveforms have the same phase with each other, so that even if they are connected as they are, no large noise will occur. Further, a compression period and an expansion period can be provided in one entire waveform.

In addition to the connection of discontinuous unit waveform data in one waveform data, the unit waveform data may be connected between two different waveform data. At the time of connection to the new waveform, the start address of the new waveform and the number of the cycle to be connected may be indicated to the sound source unit. Even if two different waveforms are read out so as to have the same frequency and the waveforms are connected at the same phase point, it is possible to prevent a large noise at the time of connection.

Next, a specific example in which the shape of the waveform data stored in the waveform memory is controlled and reproduced will be described with reference to FIGS. FIG. 18 is an explanatory diagram of a first specific example in which the shape of waveform data is controlled and reproduced by the tone generator of the present invention. This is an example of extending or contracting one entire waveform. In the figure, 111 is an original waveform, 112 is a pitch-up waveform, 113 is a pitch-down waveform, 114 is a compressed waveform, 115 is an equal-length waveform, and 116 is an expanded waveform. The horizontal axis represents time, and the vertical axis represents amplitude. In each case, one waveform from the rise to the attenuation of one musical sound waveform is schematically divided into four sections from the left, an attack period (A), a decay period (D), a sustain period (S), and a release period (R). It is shown in FIG.

The pitch-up waveform 112 is a waveform reproduced by increasing the F number at the time of reading from the waveform memory, increasing the reading speed, and increasing the pitch. The generation period of one waveform is reduced in accordance with the pitch. ing. The pitch-down waveform 113 is a waveform reproduced by setting the F number to a small value, lowering the reading speed and lowering the pitch, and the generation period of one waveform extends in accordance with the pitch.

The compression waveform 114 is the pitch-up waveform 11
2, a waveform obtained by processing any one of the pitch-down waveforms 113 and shortening the generation period of one waveform from the original waveform 111 without changing the pitch of the waveform before processing. Isometric waveform 1
15 is pitch up waveform 112, pitch down waveform 1
13 is a waveform obtained by processing any one of the waveforms 13 and reproducing the waveform generation period with the same length as the original waveform 111 without changing the pitch of the waveform before processing. Of course, it can be reproduced even if read at the same pitch as the original waveform 111. The expanded waveform 116 is a waveform obtained by processing any one of the pitch-up waveform 112 and the pitch-down waveform 113 and extending the generation period of one waveform longer than the original waveform 111 without changing the pitch of the waveform before processing.

FIG. 19 is an explanatory diagram of a second specific example in which the shape of waveform data is controlled and reproduced by the tone generator of the present invention. This is an example in which one waveform is partially expanded or contracted. In the figure, 111 is the original waveform shown in FIG.
Is a waveform with a rising portion compressed, 122 is a waveform with a stationary portion compressed, 123 is a waveform with a falling portion compressed, 124
Represents a waveform in which a rising portion is extended, 125 is a waveform in which a steady portion is extended, and 126 is a waveform in which a falling portion is extended.
One waveform is schematically shown as in FIG. The rising part is
The attack period and the decay period are described above. The steady portion is a sustain period, and the falling portion is a release period.

A waveform 121 with a rising portion compressed, a waveform 122 with a stationary portion compressed, and a waveform 1 with a falling portion compressed
Reference numeral 23 designates a part of the period of one waveform as the original waveform 11
This is a waveform that is reproduced with being reduced from the corresponding part of No. 1. Waveform 124 with rising portion extended, waveform 1 with steady portion extended
25, the waveform 126 whose falling part is expanded is respectively
This is a waveform reproduced by extending the period of a part of one waveform beyond the corresponding part of the original waveform 111. In the example shown in FIG. 19, the pitch of the waveform data is the same as the pitch of the original waveform, but the pitch can be changed.

As described above, according to the tone generating apparatus of the present invention, not only can the pitch of the original waveform 111 be changed, but also the entire one waveform or a partial waveform of one waveform regardless of the pitch. It can be compressed and expanded on the time axis. In FIGS. 18 and 19, if the time change shape of the virtual address VA is not changed even if the pitch of the waveform data is changed, the overall waveform shape (anywhere on the time axis, what shape Is the waveform output?)
Is retained. The above-mentioned “overall waveform shape” does not mean the volume envelope, but can be seen even in the case of waveform data that has been subjected to waveform processing for making the volume envelope constant at the stage of waveform preparation. it can.

In the above description, the virtual address is generated to control the compression / expansion on the time axis. However, the virtual address does not necessarily have to be generated as long as the difference between the virtual address and the read address can be obtained. Obtaining the difference is substantially equivalent to generating a virtual address, and such is also included in the technical scope of the present invention. This is because the difference is obtained by subtracting the read address from the virtual address described above, and the virtual address can be obtained by adding the difference to the read address.

In the above description, the start address ADS
1, ADS2 and the read pointers p1 and p2 read the waveform data from the waveform memory 63. As the start address, the first start address A0 of the series of waveforms shown in FIG. By increasing the number of bits of the pointers p1 and p2, the pointers p1 and p2
In such a case, a plurality of unit waveform data may be sequentially read, or read addresses may be switched.

In the above description, the cycle length of the unit waveform is standardized. However, even if the cycle length is not standardized, for example, the unit waveform data is stored in the unit waveform data in which the head address is stored in advance. It is only necessary to attach a number for specifying data. By detecting the difference between the virtual address and the virtual address based on the number of the unit waveform data read by the reading means, the process of determining the difference between the read address and the virtual address is simplified, and the control is facilitated. At this time, the compression ratio α ′ is stored in the memory in advance corresponding to the unit waveform data specified by such a number. The cycle length of the unit waveform can be obtained, for example, from the difference between the start address of the unit waveform that is adjacent later and the start address of this unit waveform. The cycle length of a unit waveform can be stored in advance for each unit waveform.

The expansion and contraction of the entire waveform described with reference to FIG.
By appropriately performing the partial expansion and contraction of the waveform described in the above and the modes illustrated in FIG. 11 to FIG. 17 in the setting process of step S87 shown in FIG. The expansion and contraction can be controlled. For example, control can be performed according to a selected tone color, or control can be performed based on a pitch or intensity specified by performance data. In addition, control may be performed by volume, tempo, rhythm type, effect type, various envelope waveforms, performance timing, chord type, key, and the like.

In the above description, the musical tone of a natural musical instrument is used as the original waveform. However, it is sufficient that the musical tone includes a periodic component such as a musical tone from an electronic musical instrument, a human voice, and the like. It is collectively referred to as musical sounds. Further, the compression ratio α ′ does not necessarily need to be set faithfully from the measured values, but can be set by arbitrarily processing. A waveform having a time-varying characteristic approximating the compression rate α '
It may be generated by a pitch envelope generator. Specifically, there is a method of creating an envelope by approximating it as a polygonal line envelope, or by reading a memory storing envelope samples at different sampling intervals in time. In this case, if the waveform component of the above-mentioned pitch shift is taken into the parameters of the envelope generator, the configuration becomes simple.

The tone generator according to the present invention may be executed by a software tone generator which generates a tone by executing a program by a CPU instead of a hardware tone generator. Specifically, instead of realizing the waveform generation unit 36 shown in FIG. 7 by hardware using a memory and a logic circuit or the like, using a memory and a CPU to realize a function equivalent to the CPU by a program Can be.
Further, the functions of the entire block of the sound source unit 34 shown in FIG. 7 can be realized by the CPU using a program.
Further, the functions of the control unit 33 and the sound source unit 34 in FIG. 7 can be implemented by a program that is realized by one common CPU. In this case, the control unit 33 is directly connected to the DAC 39.

In a personal computer, a dedicated CPU, DSP, ROM,
If the function of the sound source section 34 is executed using a RAM or the like,
A software sound source program is executed as a system program such as an IOS and an application program operating under the operating system, and the functions of the control unit 33 and the sound source unit 34 are controlled by a CP of a personal computer.
U. The above-described software sound source program is stored in a ROM or a magnetic disk, an optical disk, or the like, and is loaded into a RAM and executed at the time of execution. This program can be supplied using an optical disk or a communication network.

[0108]

As is apparent from the above description, the pitch of the musical tone and the compression / expansion of the waveform data on the time axis can be freely controlled, and the compression ratio in the time axis direction can be controlled even during the reading of the waveform data. Can be precisely controlled. There is an effect that connection between a plurality of partial waveform data is improved.

[Brief description of the drawings]

FIG. 1 is a principle explanatory diagram of a method of processing musical tone waveform data stored in a waveform memory used in a musical tone generating device of the present invention.

FIG. 2 is a schematic configuration diagram of a first device for creating a waveform memory used in the musical tone generating device of the present invention.

FIG. 3 is a schematic configuration diagram of a second device for creating a waveform memory used in the musical tone generating device of the present invention.

FIG. 4 is a schematic explanatory diagram of a storage format of waveform data stored in a waveform memory in the tone generator of the present invention.

FIG. 5 is an explanatory diagram showing first and second examples of normalization of a cycle length.

FIG. 6 is an explanatory diagram showing a third example of normalization of the cycle length.

FIG. 7 is an overall configuration diagram for describing an embodiment of a musical sound generating device according to the present invention.

8 is an internal configuration diagram of the waveform generator shown in FIG.

FIG. 9 is an internal configuration diagram of first and second processing units shown in FIG. 8;

FIG. 10 is a flowchart for explaining a situation in which generation of a musical tone is started in response to a sound generation start instruction in the musical sound generating device of the present invention.

FIG. 11 shows a first method of compressing only the reproduction time at a constant pitch.
It is explanatory drawing of the example of.

FIG. 12 shows a second method of compressing only the reproduction time at a constant pitch.
It is explanatory drawing of the example of.

FIG. 13 shows a first method of extending only the reproduction time at a constant pitch.
It is explanatory drawing of the example of.

FIG. 14 shows a second method of extending only the reproduction time at a constant pitch.
It is explanatory drawing of the example of.

FIG. 15 is an explanatory diagram of an example in which only the pitch is increased while the reproduction time is constant.

FIG. 16 is an explanatory diagram of an example in which only the pitch is lowered while the reproduction time is constant.

FIG. 17 is an explanatory diagram of an example of compressing and expanding virtual addresses while looping them.

FIG. 18 is an explanatory diagram of a first specific example in which the shape of the waveform data is controlled and reproduced by the tone generator of the present invention.

FIG. 19 is an explanatory diagram of a second specific example in which the tone generation device of the present invention controls and reproduces the shape of the waveform data.

[Explanation of symbols]

1 sample value of the original waveform having a long cycle, 2 sample values of the original waveform having a short cycle, 3 sample values of the stored waveform having a standardized cycle length, 51, 52 adding sections, 53 F number generating section, 54 counter section, 55 waveform selector, 56 LP
F, 57 Start address and cycle length storage unit, 58
Cycle data storage unit, 59 first cycle number register,
60 second cycle number register, 61 first processing unit,
62 second processing section, 63 waveform memory, 64 first interpolation section, 65 second interpolation section, 66 cross-fade synthesis section, 67 virtual address counter section, 91 readout address of the first series when no compression / expansion is performed , 92 Output of virtual address counter section, 93, 95, 97, 99 1st
Read addresses of the series, 94, 96, 98, 100 read addresses of the second series

Claims (5)

(57) [Claims] 1. A waveform memory storing a plurality of unit waveform data obtained by dividing a plurality of continuous tone waveforms in units of one or a plurality of cycles, and a read address increasing at a speed corresponding to a designated tone pitch. Reading means for reading the plurality of unit waveform data from the waveform memory according to the read address; virtual address output means for outputting a time-varying virtual address; and a difference between the read address and the virtual address. Generating an alternate read address different from the read address in accordance with
Control means for controlling the read means to read the plurality of unit waveform data using the replacement read address as the read address in place of a current read address. 2. A waveform memory storing a plurality of unit waveform data obtained by dividing a plurality of continuous tone waveforms in units of one or a plurality of cycles, and a read address which increases at a speed corresponding to a designated tone pitch. It generates a reading means for reading the plurality of unit waveform data from the waveform memory by the read address, a virtual address output means for outputting a virtual address which changes temporally reduced, and the read address and the virtual address According to the difference ,
Control means for controlling the change of the read address generated by the read means so as to read a cycle at a past position on a time axis. 3. A waveform memory storing a plurality of unit waveform data obtained by dividing a plurality of continuous tone waveforms by one or a plurality of cycles, and a read address which increases at a speed corresponding to a designated tone pitch. Reading means for reading the plurality of unit waveform data from the waveform memory based on the read address; virtual address output means for outputting a virtual address that changes with time according to a parameter relating to musical tone control; wherein in accordance with the difference between the virtual address and,
Control means for controlling the read address generated by the read means so that the musical tone waveform is expanded and contracted on a time axis. 4. A musical tone waveform having a plurality of continuous periods of one or more
Multiple unit waveform data separated by multiple cycles
Waveform memory that stores data Read-out address that increases at the speed corresponding to the specified tone pitch
Address and generates the waveform memory according to the read address.
Reading means for reading the plurality of unit waveform data from A virtual address that outputs a virtual address that decreases with time
Output means, According to the difference between the read address and the virtual address,
Generate an alternate read address different from the
The read means replaces the current read address with the replacement read address;
An output address as the read address and the plurality of unit waves.
By controlling the shape data to be read, the time axis
A control method to read the cycle at the past position above
Steps and A musical sound generator comprising: 5. A musical tone waveform having a plurality of continuous cycles of one or more
Multiple unit waveform data separated by multiple cycles
Waveform memory that stores data Read-out address that increases at the speed corresponding to the specified tone pitch
Address and generates the waveform memory according to the read address.
Reading means for reading the plurality of unit waveform data from Changes over time according to parameters related to musical tone control
Virtual address output means for outputting a virtual address; According to the difference between the read address and the virtual address,
Generate an alternate read address different from the
The read means replaces the current read address with the replacement read address;
An output address as the read address and the plurality of unit waves.
By controlling to read the shape data,
Control means for causing the sound waveform to expand and contract on the time axis, A musical sound generator comprising: