JPH06202666A - Waveform generating device and waveform storage device - Google Patents

Abstract

PURPOSE:To skip and read data in a waveform memory when a high-pitch music waveform signal is reproduced by providing compressed waveform ampli tude value data between uncompressed waveform amplitude value data. CONSTITUTION:Data are read out of the waveform memory 410 according to four successive addresses outputted from an adder 408. The 1st data is a linear sample, the 2nd data corresponds two compressed samples, and the 3rd data is a linear sample. The four read data are inputted to a decoder 411 in sequence. An envelope generator 413 generates an envelope signal corresponding to timbre data. A multiplier 414 multiplies the amplitude value data by the envelope signal after interpolation. The amplitude value data given the envelope are accumulated by a channel accumulator 415 and outputted as final waveform data. Thus, waveform data can be obtained from any location in the waveform memory and the data in the waveform memory can be skipped and read.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a waveform generator having a waveform memory storing waveform sample data and a waveform storage device (waveform memory) used in the waveform generator.

[0002]

2. Description of the Related Art Conventionally, as a method of generating a musical tone waveform used in electronic musical instruments and the like, waveform amplitude values at sequential sampling points of a predetermined waveform are digitally stored in a waveform memory and the stored contents of the waveform memory are repeated. A waveform generation method is known in which the waveform is read out and a musical tone waveform is generated.

To read from the waveform memory, specifically,
It is performed as follows. First, frequency information (so-called F number) F corresponding to the pitch of a musical tone to be generated is sequentially accumulated by an accumulator, and accumulated values qF (q = 1,
2, 3, ...) is used as an address to read the waveform memory in a predetermined cycle. Further, since the waveform data read by using only the integer part I of the accumulated value qF as an address contains quantization noise, the interpolation method is used to improve this. That is, the waveform data of a plurality of adjacent sample points is read out, interpolation is performed based on the decimal part d of the accumulated value qF, the amplitude value at an arbitrary position is obtained, and this is output as waveform data.

In this way, a musical tone waveform having a desired pitch is output. The waveform data in the waveform memory sequentially stores the waveform amplitude values at the sampling points as they are (so-called linear). Since the F number value increases as the pitch increases, the integer part I of the accumulated value qF also increases, and it may be necessary to skip the data in the waveform memory. Since the waveform memory stores the waveform data linearly, the correct waveform amplitude value at each sample point can always be read even if the address is skipped.

On the other hand, the waveform data is compressed for the purpose of suppressing the capacity of the waveform memory. As a compression method, for example, DPCM (differential PC
M), LPC (Linear Predictive Coding), and ADPCM. All of these are methods of obtaining the current waveform data (reproduction value) by using the past waveform data (reproduction value) and the newly read data.

[0006]

However, in the case of performing the data compression as described above, since the waveform data at that time is always calculated using the data before a certain time, it is stored in the waveform memory. Data must be read continuously. If even one data is not read, the data cannot be reproduced from the next time. Therefore, since the musical tone waveform signal having a high pitch is reproduced, the waveform memory cannot be skipped and read, and the pitch width is limited.

In view of the problems in the above-mentioned conventional example, the present invention can generate a waveform by skipping the waveform memory without limiting the pitch width and by compressing the data in the waveform memory. The purpose is to provide a device.

[0008]

In order to achieve this object, the present invention provides uncompressed waveform amplitude value data for each predetermined plurality of sample points and exists between the sample points of the uncompressed waveform amplitude value data. For sample points,
Waveform storage means provided with compressed waveform amplitude value data compressed so that it can be converted to uncompressed waveform amplitude value data using the immediately preceding uncompressed waveform amplitude value data, and a frequency corresponding to the pitch of the musical tone waveform to be generated. Frequency information generating means for generating information, accumulating means for repeatedly accumulating the frequency information at a predetermined speed and outputting an accumulated value, and two or more from the waveform storing means based on the accumulated value. Reading means for reading a plurality of uncompressed waveform amplitude value data and compressed waveform amplitude value data in a continuous range including the uncompressed waveform amplitude value data, and the compressed waveform amplitude value data read by the reading means. Conversion means for converting into amplitude value data, uncompressed waveform amplitude value data read by the reading means based on the accumulated value, and uncompressed waveform amplitude value obtained by the conversion means From over data, selecting means for selecting a non-compressed waveform amplitude value data to be used for interpolation, using a non-compressed waveform amplitude value data selected by said selecting means,
Interpolation means for performing interpolation processing based on the accumulated value and outputting the interpolation result as waveform data is provided.

The read means may determine the read address of the waveform storage means on the basis of the higher value of the accumulated value. In the embodiment described later, the read address of the waveform memory is determined by the integer part I of the accumulated value qF. Further, it is preferable that the selecting means selects data to be used for interpolation based on a middle value of the accumulated value. In the embodiment described later, data to be used for interpolation is selected based on the integer part i of BJ as a result of multiplying the decimal part J of the accumulated value qF by the constant B. further,
The interpolation processing in the interpolation means may be performed based on the lower value of the accumulated value. In the embodiment described later, the interpolation process is performed based on the decimal part j of BJ obtained by multiplying the decimal part J of the accumulated value qF by the constant B.

Further, the uncompressed waveform amplitude value data is provided for each of a plurality of predetermined sample points, and the immediately preceding uncompressed waveform amplitude value data is set for the sample points between the sample points of the uncompressed waveform amplitude value data. Waveform storing means provided with compressed waveform amplitude value data compressed so as to be converted into uncompressed waveform amplitude value data, and frequency information generating means for generating frequency information according to the pitch of a musical tone waveform to be generated. And repeatedly accumulating the frequency information at a predetermined speed,
An accumulating unit that outputs an accumulated value, and based on the accumulated value,
Reading means for reading a plurality of uncompressed waveform amplitude value data and compressed waveform amplitude value data in a continuous range including two or more uncompressed waveform amplitude value data from the waveform storage means; Selecting means for selecting uncompressed waveform amplitude value data and compressed waveform amplitude value data to be used for interpolation from the plurality of uncompressed waveform amplitude value data and compressed waveform amplitude value data read by the reading means; Using the conversion means for converting the compressed waveform amplitude value data thus converted into the uncompressed waveform amplitude value data, the uncompressed waveform amplitude value data selected by the selecting means, and the uncompressed waveform amplitude value data obtained by the converting means. And an interpolation means for performing interpolation processing based on the accumulated value and outputting the interpolation result as waveform data.

Further, the waveform storage device according to the present invention is
The uncompressed waveform amplitude value data is provided for each of a plurality of predetermined sample points, and the sample points between the sample points of the uncompressed waveform amplitude value data are uncompressed using the immediately preceding uncompressed waveform amplitude value data. It is characterized in that compressed waveform amplitude value data compressed so as to be converted into waveform amplitude value data is provided.

[0012]

The waveform storage means stores the amplitude value data of a plurality of sequential sampling points. The amplitude value data is either uncompressed waveform amplitude value data or compressed waveform amplitude value data. The uncompressed waveform amplitude value data is provided for each of a plurality of predetermined sample points. The compressed waveform amplitude value data is provided between the uncompressed waveform amplitude value data. The compressed waveform amplitude value data can be converted to the uncompressed waveform amplitude value data by using the uncompressed waveform amplitude value data immediately before it.

When a musical tone waveform is generated, first, frequency information corresponding to the pitch of the musical tone waveform to be generated is generated. This frequency information is repeatedly accumulated at a predetermined speed, and the accumulated value is output. Based on the accumulated value, a plurality of uncompressed waveform amplitude value data and compressed waveform amplitude value data in a continuous range including two or more uncompressed waveform amplitude value data are read from the waveform storage means. The read compressed waveform amplitude value data is converted into non-compressed waveform amplitude value data.

On the other hand, from the uncompressed waveform amplitude value data read from the waveform storage means and the uncompressed waveform amplitude value data obtained by conversion, the uncompressed waveform amplitude value to be used for interpolation based on the accumulated value. The data is selected. The selected uncompressed waveform amplitude value data is used to perform interpolation processing based on the accumulated value, and the interpolation result is output as waveform data.

The data used for interpolation may be selected from the data read from the waveform storage means, and the compressed waveform amplitude value data of the selected data may be converted into the non-compressed waveform amplitude value data.

[0016]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 shows a block configuration of an electronic musical instrument to which a waveform generating apparatus according to an embodiment of the present invention is applied. This electronic musical instrument includes a tone color switch 1, a detector 2, a keyboard 3, a detector 4, a tone generator 5, and a digital-analog converter (DAC).
6 and a sound system 7.

The tone color switch 1 is a switch for designating a tone color of a generated musical tone. The detector 2 detects the operation of the tone color switch 1 and detects the tone color number T of the designated tone color.
C is output toward the sound source unit 5. The keyboard 3 is a keyboard provided with a plurality of keys for the performer to play. Detector 4
Detects the performance of the keyboard 3 and outputs the performance data KD to the tone generator section 5.
Output to. The performance data KD includes a key on / off signal indicating a key on / off and a node code indicating a pitch.

The tone generator 5 produces and outputs a tone signal (digital waveform amplitude value data) based on the tone color number TC and the performance data KD from the detectors 2, 4. The sound source unit 5 is
It has a waveform memory inside. The DAC 6 is the sound source unit 5
The digital-to-analog conversion is performed on the musical tone signal from and the analog musical tone signal is output. Sound system 7
Actually emits a musical tone based on the analog musical tone signal output from the DAC 6.

FIG. 2 shows the waveform data format of the waveform memory provided inside the tone generator section 5 of FIG. The waveform memory has a plurality of storage banks, and the waveform data for each tone color is stored in each bank. FIG. 2 shows waveform data of one timbre.

The waveform data is composed of sequential sampling point amplitude value (waveform amplitude value) data of the tone waveform of the tone color. However, in this embodiment, uncompressed data with the amplitude value as it is (hereinafter referred to as “linear sample”) and difference data from the amplitude value of the immediately preceding sample point (hereinafter, “compressed sample”).
Called) is mixed. Specifically, 16-bit linear samples, 8-bit compressed samples, and 8-bit compressed samples are repeated blocks arranged in this order.

In FIG. 2, D0 is a 16-bit linear sample stored at an address (offset address from the beginning) 0 which is a read start position, and D1 is an 8-bit compressed sample stored in the first half of the next address 1. , D2
Indicates an 8-bit compressed sample stored in the latter half of address 1. Similarly, D3, D6, D9, ..., Dm
Indicate 16-bit linear samples stored at addresses 2, 4, 6, ..., (2/3) m, respectively. D4,
D7, D10, ..., Dm + 1 are addresses 3, 5 and 5, respectively.
7 ... Shows 8-bit compressed samples stored in the first half of (2/3) m + 1. D5, D8, D11, ..., Dm + 2
Are addresses 3, 5, 7, ..., (2/3) m +, respectively.
8 shows an 8-bit compressed sample stored in the latter half of 1.

A set of linear samples followed by two compressed samples is called a block. For example, the data D0, D
1, D2 is one block, data D3, D4, D5 is one block, ....

FIG. 3 is a diagram showing the relationship between the waveform data stored in the waveform memory as shown in FIG. 2 and the original tone waveform.
Reference numeral 300 denotes a tone waveform to be sampled, and reference numerals 301 to 306 denote sampling points. Sampling points 301 and 304 indicated by black circles take linear samples, and sampling points 302, 303, 305 and 30 indicated by white circles are taken.
At 6, a compressed sample is taken.

For example, assuming that the linear sample sampled at the sampling point 301 is D3 in FIG. 2, the data D4 at the next sampling point 302 is the sampling point 3
A difference 311 between the amplitude value at 02 and the amplitude value at the sampling point 301 is obtained. Similarly, sampling point 3
The data D5 of 03 is the difference 3 between the amplitude value at the sampling point 303 and the amplitude value at the sampling point 302.
Twelve.

In the electronic musical instrument of this embodiment, interpolation is performed from waveform data corresponding to four adjacent sample points. If the waveform data is configured as shown in FIG. 2 described above, there are the following three combinations in order to obtain the waveform data of four adjacent points.

(1) Four points are taken in the order of linear sample, compressed sample, compressed sample, and linear sample.
For example, there are four points of data D0, D1, D2, D3 in FIG. In order to obtain the linear amplitude value of the sample point corresponding to the compressed sample D1, the linear sample D0 is used and D0 + D1 = D1 'is calculated. Further, in order to obtain the linear amplitude value of the sample point corresponding to the compressed sample D2, the calculated linear sample D1 'is used to obtain D1
It is calculated by '+ D2 = D2'. For the sake of simplicity, a linear amplitude value obtained from a compressed sample (for example, the above D
1'and D2 ') are called calculated linear samples. Interpolation is performed on the basis of the obtained data D0, D1 ', D2', D3, and the sampling points of the data D0 and the data D
Obtain the waveform amplitude value of the target sample point between the 1'sample points.

(1-) compressed sample, compressed sample,
Four points are taken in the order of linear sample and compressed sample. For example, there are four points of data D1, D2, D3 and D4 in FIG. In order to obtain the calculated linear samples D1 'and D2' corresponding to the compressed samples D1 and D2, the linear sample D
0 is required. The calculated linear sample D4 'is calculated using the linear sample D3. Obtained data D1
Interpolation based on ′, D2 ′, D3, D4 ′,
The waveform amplitude value of the target sample point between the sample points of the data D1 'and the sample points of the data D2' is obtained.

(1) Four points are taken in the order of compressed sample, linear sample, compressed sample, and compressed sample. For example, there are four points of data D2, D3, D4, D5 in FIG. Calculated linear sample D corresponding to compressed sample D2
To obtain 2 ', linear sample D0 and compressed sample D1 are needed. Calculated linear samples D4 ', D5
′ Is calculated using the linear sample D3. Interpolation is performed based on the obtained data D2 ', D3, D4', D5 'to obtain the waveform amplitude value of the target sample point between the sample point of the data D2' and the sample point of the data D3 '.

From the above, if the addressing is performed four times starting from the linear sample (for example, in FIG.
If the 16-bit × 4 data of the addresses 0, 1, 2, and 3 are read out), the amplitude values of four adjacent points can be obtained without fail. In other words, no matter where in the waveform memory the reading is started, the amplitude values of four adjacent points can be obtained without fail by reading only two blocks.

Next, a detailed description will be given of the sound source section of the present embodiment, which has the waveform memory as described above and is capable of skip reading.

FIG. 4 shows a block configuration of the sound source unit 5 of FIG. The sound source unit 5 includes a sound generation channel assignment unit 401, a tone color data supply unit 402, an F number generation unit 403, a counter 404, multipliers 405 and 406, and adders 407 and 40.
8, an auxiliary counter 409, a waveform memory 410, a decoder 411, an interpolator 412, an envelope generator 413, a multiplier 414, and a channel accumulator 415.

The tone color number TC is the tone color data supply section 402.
To enter. The performance data KD is input to the sound generation channel allocation unit 401. Pronunciation channel allocation unit 401
Assigns channels on a time-division basis based on the input performance data KD, and sends a key on / off signal to the tone color data supply unit 402 in a time slot of each channel.
The key code indicating the pitch is output to the F number generation unit 403, respectively.

The tone color data supply section 402 outputs the tone color data corresponding to the tone color number TC in the time slot of each channel to the F number generation section 403 and the envelope generation section 4.
Supply to 13. Further, the tone color data supply unit 402 adds the start address (start address for storing the tone color data of the waveform memory 410) SA corresponding to the tone color to the adder 407.
Output to.

The F number generating section 403 generates an F number corresponding to the tone color data of the musical tone to be generated and the key code in each time slot. The counter 404 sequentially accumulates the F numbers and accumulates the accumulated value qF (q = 1, 2, 3,
...) is output. The integer part I of the accumulated value qF (value higher than a predetermined bit position (decimal point position)) is multiplied by a predetermined constant A in a multiplier 405. The multiplication result AI becomes an offset (address 0, 1, 2, ... In FIG. 2) from the start address of the waveform data. The multiplication result AI is added to the start address SA by the adder 407, and the addition result SA + AI is input to the adder 408.

Here, the constant A of the multiplier 405 is selected so that the linear sample of the waveform memory is addressed by the addition result SA + AI. As can be seen from FIG. 2, addresses 0, 2, 4, ...
Since the linear sample is stored in, the value of the constant A is “2”. As a result, the addition result of the adder 407 is always the address for addressing the linear sample.

The auxiliary counter 409 repeatedly outputs a count value according to the number of samples used for interpolation and the number thereof. In this embodiment, as described above, the amplitude values of four adjacent points can be obtained without fail by reading the data (16 bits × 4) at four consecutive addresses, so the auxiliary counter 409 is set to “0”. "1", "2", and "3" are sequentially output in this order.

The adder 408 adds the output from the auxiliary counter 409 to the addition result (address of the linear sample) of the adder 407, and outputs sequentially. As a result, four consecutive addresses are sequentially output from the adder 408.

Data is read from the waveform memory 410 by four consecutive addresses sequentially output from the adder 408. 1st data is linear sample, 2nd
The second is for two compressed samples, the third is for linear samples, and the fourth is for two compressed samples. The read four data are sequentially input to the decoder 411.

On the other hand, the decimal part J of the counter 404 (value lower than the predetermined bit position (decimal point position)) is the multiplier 40.
It is multiplied by 6 with a predetermined constant B. In this embodiment, the value of the constant B is "3". As can be seen from FIGS. 2 and 3, this value has three sections from the linear sample to the next linear sample, and is selected so that it can be distinguished which section contains the target sample point to be interpolated. . Integer part i of multiplication result BJ (“0” “1” “2” or “3”)
Is input to the decoder 411, and the decoder 411 sequentially outputs linear amplitude value data of four adjacent sample points based on the value of the integer part i.

Hereinafter, the meaning of the integer part i and the decoder 41
The outline of how 1 selects four points based on the value of the integer part i will be described.

For example, it is assumed that the four pieces of data of addresses 4, 5, 6, and 7 in FIG. 2 are input from the waveform memory 410 to the decoder 411. Speaking of samples, linear sample D6,
This means that D9 and the compressed samples D7, D8, D10 and D11 are input to the decoder 411.

In FIG. 3, the sample point 301 corresponds to D6, the sample point 302 corresponds to D7, the sample point 303 corresponds to D8, the sample point 304 corresponds to D9, the sample point 305 corresponds to D10, and the sample point 306 corresponds to D11. Then, the following cases need to be classified according to where the target sample point to be obtained by interpolation is now.

(2-) When the target sample point is between the sample points 301 and 302 At this time, the fractional part J output from the counter 404 in FIG.
Is 0 ≦ J <(1/3), the result BJ obtained by multiplying the constant B (= 3) is 0 ≦ BJ <1 and the multiplication result B
The integer part i of J is “0”. When the integer part i = 0 is input to the decoder 411, the input sample data D6 is input.
~ D11 based on sample points 301, 302, 30
Each amplitude value of 3,304 (all linear values) is obtained and sequentially output. That is, the linear sample D6, the calculated linear sample D7 '(= D6 + D7), the calculated linear sample D8' (= D7 '+ D8), and the linear sample D
The four amplitude values of 9 are sequentially output.

(2-) When the target sample point is between the sample points 302 and 303 At this time, the fractional part J output from the counter 404 in FIG.
Is (1/3) ≦ J <(2/3), the constant B
As a result of multiplying (= 3), BJ is 1 ≦ BJ <2,
The integer part i of the multiplication result BJ becomes “1”. Decoder 41
When the integer part i = 1 is input, 1 is the sampling point 302, 30 based on the input sample data D6 to D11.
The amplitude values of 3, 304 and 305 (all linear values) are obtained and sequentially output. That is, the calculated linear sample D
7 '(= D6 + D7), calculated linear sample D8' (=
D7 '+ D8), linear sample D9, and calculated linear sample D10' (= D9 + D10).

(2-) When the target sample point is between the sample points 303 and 304 At this time, the fractional part J output from the counter 404 in FIG.
Is (2/3) ≦ J <1, the result BJ obtained by multiplying the constant B (= 3) is 2 ≦ BJ <3, and the multiplication result B
The integer part i of J is “2”. When the integer part i = 2 is input to the decoder 411, the input sample data D6 is input.
~ D11 based on sample points 303, 304, 30
Each amplitude value of 5,306 (all linear values) is obtained and sequentially output. That is, the calculated linear sample D8 '(=
D7 '+ D8), linear sample D9, calculated linear sample D10' (= D9 + D10), and calculated linear sample D11 '(= D10' + D11).

Next, referring to FIGS. 5 and 6, the decoder 41
1 will be described in detail. In FIG. 5, the decoder 411 includes a latch 501, a selector 502, a data decompression unit 503, an adder 504, a selector 505, a delay unit 506, a load pulse generation unit 507, and an output register unit 508.

The output from the waveform memory 410 is the latch 5
01 and the selector 502. Latch 501
Is for latching particularly linear samples of the input data. Latch 501 output (16 bits)
Is input to the input terminal A of the selector 505.

The selector 502 outputs the output data from the waveform memory 410 for two compressed samples (8 bits × 2 = 1).
6 bits), it is for separating two compressed samples. The upper 8 bits of the 16-bit input from the waveform memory 410 are input to the input terminal A of the selector 502. The lower 8 bits of the 16-bit input from the waveform memory 410 are input to the input terminal B of the selector 502.

The data decompression unit 503 decompresses the compressed sample (8 bits) output from the selector 502 into 16 bits (adds 8-bit "0" to the higher order). The output from the data decompression unit 503 is input to the adder 504. The output data from the delay unit 506 is input to the adder 504. The adder 504 adds these and outputs the addition result (16 bits) to the input terminal B of the selector 505.

The adder 504 is for adding a compressed sample and a linear sample (which may be a calculated linear sample) to obtain a calculated linear sample corresponding to the compressed sample.

That is, when a linear sample is first input from the waveform memory 410, the delay section 506 holds the linear sample via the latch 501 and the selector 505. On the other hand, data for two compressed samples is input to the decoder following the linear samples.

The first compressed sample is separated by the selector 502, further expanded by the data expansion unit 503 to 16 bits and output. At this time, the immediately preceding linear sample is output from the delay unit 506, and the adder 504 adds the linear sample and the compressed sample. Thereby, the adder 504 outputs the calculated linear sample of the sample point corresponding to the compressed sample, and is output via the selector 505 and the delay unit 506. The next compressed sample is similarly processed, and the calculated linear sample corresponding to the compressed sample is output.

As a result, the delay unit 506 sequentially outputs the linear amplitude value data of the six sample points included in the input data for four addresses.

The linear amplitude value data output from the delay unit 506 is loaded into the output register 508.
The loading is performed based on the load pulse output from the load pulse generator 507.

The load pulse generating section 507, based on the integer part i output from the multiplier 406 of FIG. 4, outputs four data to be loaded out of the six linear amplitude value data sequentially output from the delay section 506. Output a load pulse to load only. The output register 508 is
This is a FIFO (First-In First-Out) register, and the loaded amplitude value data is read based on a pulse φSP of a predetermined cycle. The amplitude value data read from the output register is input to the interpolation unit 412.

Next, the operation of the decoder 411 will be described in more detail using a specific example.

FIG. 6 shows a time chart for explaining the operation of the decoder 411. PA-PE of FIG. 6 shows the time chart in the position shown by the same PA-PE of FIG. Here, a case where the data at the addresses 4, 5, 6, and 7 in FIG. 2 (that is, the sample data D6 to D11) are sequentially input will be described as an example.

As shown in PA of FIG. 6, the waveform memory 41
0 to linear sample data D6 at address 4 in FIG.
, Compressed sample data D7 and D8 at address 5, linear sample data D9 at address 6, and address 7
The compressed sample data D10 and D11 of
It is assumed that the data is input to 11. PB indicates the output of the latch 501. Latch 501 has linear samples D6 and D9.
Is latched and output.

PC shows a timing chart of the output of the selector 502. The selector 502 is the waveform memory 41.
The compressed sample data D7 is selected and output in the first half of the time slot in which the compressed sample data D7 and D8 are output from 0, and the compressed sample data D is output in the latter half of the time slot.
8 is selected and output.

PD shows a timing chart of the output of the delay unit 506. The delay unit 506 first holds and outputs the linear sample D6 input via the latch 501 and the selector 505.

While the linear sample D6 is being output, the compressed sample D7 is input to the adder 504 via the data expansion section 503 as shown in the PC. Adder 504
Then, the linear sample D6 and the compressed sample D7 are added, and the calculated linear sample D7 '= D6 + which is the linear amplitude value of the sample point corresponding to the compressed sample D7.
D7 is output. This calculated linear sample D7 'is
It is held by the delay unit 506 via the selector 505,
It is output from the delay unit 506 subsequent to the linear sample D6 like PD.

When the calculated linear sample D7 'is output from the delay unit 506, the next compressed sample D8 is output from the selector 502 as shown in the PC.
The compressed sample D8 is input to the adder 504 via the data expansion unit 503. The adder 504 adds the calculated linear sample D7 'and the compressed sample D8, and outputs the calculated linear sample D8' = D7 '+ D8 which is the linear amplitude value of the sample point corresponding to the compressed sample D8. This calculated linear sample D8 'is the selector 50
5 and the delay unit 506, the signal is continuously output like a PD.

As a result, as shown in PD of FIG. 6, linear sample (originally linear amplitude value) D6, calculated linear sample (linear amplitude value of sample point corresponding to compressed sample D7) D7 '= D6 + D7, and calculated The linear sample (the linear amplitude value of the sampling point corresponding to the compressed sample D8) D8 '= D7' + D8 is output to the output register 5
It is sequentially output toward 08. Next sample D9 ~ D11
Are processed in the same manner, and the linear sample D9, the calculated linear sample D10 ', and the calculated linear sample D1 are processed.
1'is sequentially output to the output register 508.

Six linear amplitude value data which are sequentially output like PD are loaded into the output register 508, and the load pulse indicating the loading timing is generated at the timing shown in FIG.

First, when the value of the integer part i output from the multiplier 406 of FIG. 4 is "0", which is the case of the above (2-), the amplitude value data D6, D7 ', D8'. , D
A load pulse is generated so that 9 is loaded into output register 508. When the value of the integer part i is "1",
Since it is the case of (2-) above, the amplitude value data D7
A load pulse is generated so that ', D8', D9, D10 'are loaded into output register 508. When the value of the integer part i is "2", it is the case of (2-) described above, so that the amplitude value data D8 ', D9, D10', D11 'are loaded in the output register 508 so that A pulse is generated.

PE in FIG. 6 indicates a data string read from the output register 508. The four amplitude value data loaded as described above are sequentially read out as S0, S1, S2, S3 based on the pulse φSP of a predetermined cycle, and output toward the interpolating unit 412.

As described above, the decoder 411 sequentially outputs the linear amplitude value data of four adjacent sample points based on the value of the integer part i ("0""1" or "2"). A timing generator (not shown) that sends a timing signal is provided to each unit in order to perform the above-described operation.

Referring again to FIG. 4, the fractional part j of the multiplication result BJ is input to the interpolation section 412. The interpolation section 412 performs interpolation processing based on the amplitude value data of the four points from the decoder 411 and the value of the decimal part j, and outputs post-interpolation waveform data.

FIG. 7 shows a block configuration of the interpolation section 412. The interpolation unit 412 includes an auxiliary counter 701, a coefficient generation unit 702, a multiplier 703, an adder 704, a gate 705,
The delay unit 706 and the latch 707 are provided.
In this embodiment, a cubic interpolation method is used in which interpolation is performed assuming a cubic curve passing through four adjacent sample points. The cubic interpolation method is a conventionally known technique, and will be briefly described below.

The auxiliary counter 701 has four amplitude value data (S0, S1, S) sequentially input from the decoder 411.
2, S3), the count values c of "0", "1", "2" and "3" are sequentially output. Coefficient generator 70
2 outputs a coefficient Ac (j) obtained by substituting the decimal part j from the multiplier 406 into the function Ac corresponding to the count value c from the auxiliary counter 701.

According to the conventionally known cubic interpolation method, A0 (j) .S0 + A1 (j) .S1 + A2 (j) .S
The interpolated amplitude value can be obtained by 2 + A3 (j) · S3.

Therefore, in the multiplier 703, c = 0, 1,
In each of the cases 2 and 3, Ac (j) is multiplied by Sc, and the multiplication result is input to the adder 704. Adder 704
After being held in the delay unit 706, the addition result of is input to the adder 704 by opening the gate 705 at a predetermined timing. In this loop circuit, Ac (j)
-Sc is accumulated to obtain the result of the above equation, that is, the interpolated amplitude value. The interpolated amplitude value data is stored in the latch 70.
7 and is output to the next envelope applying multiplier 414.

Referring again to FIG. 4, the envelope generator 413 generates an envelope signal according to the tone color data. The multiplier 414 multiplies the amplitude value data after interpolation by the envelope signal. The amplitude value data to which the envelope has been added is accumulated by the channel accumulator 415 and output as final waveform data.

With the above arrangement, the waveform data can be obtained from any position in the waveform memory, and the data in the waveform memory can be skipped and read. There is also the effect of data compression.

In the above embodiment, the linear sample is 16 bits and the compressed sample is 8 bits. However, the number of bits of the linear sample is reduced and the range of the compressed sample may be widened by using the surplus bits. . FIG. 8 shows a modification of the first embodiment, in which linear samples are represented by 14 bits and the remaining 2 bits (upper) are used for the shift-up amount of the compressed samples. These 2 bits are indicated by EXP.

When EXP = 0, the next two 8-bit compressed samples are not shifted up, and the compressed samples range from -128 to +127. EXP =
When it is 1, the next two 8-bit compressed samples are shifted up by 1 bit, and the compressed sample is -2.
It ranges from 56 to +255. However, the resolution is 2. When EXP = 2, 2-bit shift up is performed for the next two 8-bit compressed samples,
Compressed samples range from -512 to +511. However, the resolution is 4. When EXP = 3,
The next two 8-bit compressed samples are shifted up by 3 bits, and the compressed samples are changed from -1024 to +.
It takes up to 1023. However, the resolution is 8.

In order to do this, in FIG. 5 of the above embodiment, the upper 2 bits EXP of the output from the latch 501 are added to the data decompression unit 503, and the other 14 bits are added to the upper "0". Then, the data is input to the selector 505, and the data decompression unit 503 shifts up the compressed sample as described above according to the value of EXP.

In the above embodiment, the linear sample is provided for every 3 sample points, and the interpolation is performed using 4 sample points, but these can be appropriately changed. For example, a linear sample may be provided for every 5 sample points, and interpolation may be performed using 2 sample points.

Hereinafter, such a second embodiment will be described.

FIG. 9 shows the number of 5 used in the second embodiment.
The format of the waveform data that has a linear sample for each sample point is shown. D0 is a 16-bit linear sample, and D1, D2, D3, and D4 are 8-bit compressed samples. Similarly, the following D5 is a 16-bit linear sample, D6, D7, D8 and D9 are 8-bit compressed samples, ....

FIG. 10 is a diagram showing the relationship between the waveform data of FIG. 9 and the original tone waveform. Reference numeral 1000 indicates a musical tone waveform to be sampled, and reference numerals 1001 to 1006 indicate sampling points. Sampling points 1001,1 indicated by black circles
A linear sample is taken at 006, and a compressed sample is taken at sampling points 1002 to 1005 shown by white circles.

For example, assuming that the linear sample sampled at the sampling point 1001 is D5 in FIG. 9, the data D6 at the next sampling point 1002 is the difference 1011 between the amplitude value at the sampling point 1002 and the amplitude value at the sampling point 1001. . Similarly, the data D7 of the sampling point 1003 is the sampling point 10
A difference 1012 between the amplitude value at 03 and the amplitude value at the sampling point 1002 is obtained.

In the second embodiment, interpolation is performed from two adjacent sample points. If the waveform data is configured as shown in FIG. 9, in order to obtain the waveform data of two adjacent points,
There are five combinations:

(3-) Two points (sample points 1001 and 1002 in FIG. 10) such as data D5 and D6 in FIG. 9 are used. (3-) Two points (sample points 1002 and 1003) such as data D6 and D7 are used. (3-) Two points (sample points 1003 and 1004) such as data D7 and D8 are used. (3-) Two points (sample points 1004 and 1005) such as data D8 and D9 are used. (3-) Two points (sample points 1005 and 1006) such as data D9 and D10 are used.

Therefore, starting from the linear sample, 4
In any of the above cases, if addressing is performed once,
This means that the amplitude values of two adjacent points can be obtained without fail.

As described above, the second embodiment in which the linear sample is provided for every 5 sample points and the interpolation is performed by using the 2 sample points has the same configuration as that of FIGS. Good. However, in FIG. 4, the constant A multiplied by the multiplier 405 is “3”, the constant B multiplied by the multiplier 406 is “5”, and the auxiliary counter 409 outputs “0” and “1”. .

The constant A of the multiplier 405 is a multiplier for allowing the linear sample to be addressed by the waveform memory read address output from the adder 407. Here, since linear samples are stored at addresses 0, 3, 6, ... As shown in FIG. 9, the value of the constant A must be "3".

Further, the constant B of the multiplier 406 is multiplied by the decimal part J from the counter 404, and the target sample is obtained from the above (3-) to (3) by the integer part i of the multiplication result BJ.
Since it is for distinguishing which section of −) is included, the constant B must be set to “5” so that the integer part i takes 0 to 4.

The auxiliary counter 409 repeatedly outputs a count value in accordance with the number of samples used for interpolation, and the number thereof. However, since it is two-point interpolation here, "0" and "1" are repeatedly output. Need to do so. Similarly, in the interpolator in FIG. 7, the auxiliary counter 701 has the value “0”.
It is necessary to repeatedly output "1" and "1".

Further, the decoder of FIG. 5 may have the same structure, but in the second embodiment, since linear samples are provided for every 5 samples and two-point interpolation is performed, the timing of the latch 501 and the load pulse generator 507 are used. The load pulse and the like have different timings. This will be described below.

FIG. 11 shows a time chart for explaining the operation of the decoder 411 in the second embodiment.
PA to PE in FIG. 11 are time charts at the positions indicated by the same PA to PE in FIG. Here, a case will be described as an example where the data at the addresses 3, 4, 5, and 6 in FIG. 9 (that is, the sample data D5 to D10) are sequentially input.

As shown in PA of FIG. 11, the waveform memory 4
10 to linear sample data D at address 3 in FIG.
5, compressed sample data D6 and D7 at address 4, compressed sample data D8 and D9 at address 5, and linear sample data D10 at address 6 are sequentially decoded by the decoder 4
It is assumed that the data is input to 11. PB indicates the output of the latch 501. Latch 501 has linear samples D5 and D10.
Is latched and output.

PC shows a timing chart of the output of the selector 502. The selector 502 is the waveform memory 41.
The compressed sample data D6 is selected and output in the first half of the time slot in which the compressed sample data D6 and D7 are output from 0, and the compressed sample data D6 is output in the latter half of the time slot.
Select 7 to output. Similarly, the compressed sample data D8 is selectively output in the first half of the time slot in which the compressed sample data D8 and D9 are output, and the compressed sample data D9 is selectively output in the latter half of the time slot.

PD shows a timing chart of the output of the delay unit 506. The delay unit 506 first holds and outputs the linear sample D5 input via the latch 501 and the selector 505.

While the linear sample D5 is being output, the compressed sample D6 is input to the adder 504 via the data decompression unit 503 as shown on the PC. Adder 504
Then, the linear sample D5 and the compressed sample D6 are added, and the calculated linear sample D6 '= D5 + which is the linear amplitude value of the sample point corresponding to the compressed sample D6.
D6 is output. This calculated linear sample D6 'is
It is held by the delay unit 506 via the selector 505,
It is output from the delay unit 506 following the linear sample D5 like PD.

When the calculated linear sample D6 'is output from the delay unit 506, the next compressed sample D7 is output from the selector 502 as shown in the PC.
The compressed sample D7 is input to the adder 504 via the data expansion unit 503. The adder 504 adds the calculated linear sample D6 'and the compressed sample D7, and outputs the calculated linear sample D7' = D6 '+ D7 which is the linear amplitude value of the sample point corresponding to the compressed sample D7. This calculated linear sample D7 'is the selector 50
5 and the delay unit 506, the signal is continuously output like a PD.

Thereafter, similarly calculated linear sample D8
', D9' are output. Further, the next linear sample D10 is subsequently output. As a result, the PD in FIG.
, Linear sample or calculated linear sample D
5, D6 ', D7', D8 ', D9' and D10 are sequentially output to the output register 508.

Six linear amplitude value data which are sequentially output like PD are loaded into the output register 508, and a load pulse indicating the loading timing is generated as shown in FIG.

First, when the value of the integer part i output from the multiplier 406 of FIG. 4 is "0", which is the case of (3-) above, the amplitude value data D5 and D6 'are output registers. A load pulse is generated, as loaded at 508. When the value of the integer part i is “1”, the above (3-
Therefore, a load pulse is generated so that the amplitude value data D6 ', D7' is loaded into the output register 508. When the value of the integer part i is "2", this is the case of the above (3-). Therefore, the amplitude value data D7 ',
A load pulse is generated so that D8 'is loaded into output register 508. When the value of the integer part i is "3", it is the case of (3-) described above, so that the load pulse is generated so that the amplitude value data D8 'and D9' are loaded into the output register 508. . When the value of the integer part i is “4”, this is the case of the above (3-).
A load pulse is generated so that the amplitude value data D9 'and D10 are loaded into the output register 508.

PE in FIG. 11 indicates a data string read from the output register 508. The two amplitude value data loaded are sequentially S0, based on the pulse φSP of a predetermined cycle.
It is read out as in S1 and output toward the interpolation unit 412.

As described above, the decoder 411 determines, based on the value of the integer part i (“0” “1” “2” “3” or “4”), the linear amplitude value of two adjacent sample points. Output data sequentially.

In this second embodiment, since the linear samples are provided every 5 samples, the data compression rate is
Is higher than the example.

Also in the second embodiment, as described in the first embodiment, the number of bits of linear samples may be reduced and the range of compressed samples may be widened by using the surplus bits. The configuration of FIG. 8 may be applied as it is.

[0105]

As described above, according to the present invention,
Linear (non-compressed) amplitude value data is provided for each of a plurality of predetermined sampling points, and compressed data is provided between them. For reading, a continuous range including two or more linear amplitude value data is read and the read data is read. Since the amplitude value of the target sampling point is obtained, the waveform memory can be skipped and read without the pitch width being limited (pitch up limitation). Further, the waveform memory is data-compressed, so that the capacity is small.

[Brief description of drawings]

FIG. 1 is a block configuration diagram of an electronic musical instrument to which a waveform generation device according to an embodiment of the present invention is applied.

[Figure 2] Waveform data format diagram

FIG. 3 is a diagram showing a relationship between waveform data and an original musical tone waveform.

FIG. 4 is a block configuration diagram of a sound source section of the electronic musical instrument of the embodiment.

FIG. 5 is a block configuration diagram of a decoder of a sound source section.

FIG. 6 is a time chart for explaining the operation of the decoder.

FIG. 7 is a block configuration diagram of an interpolation unit of a sound source unit.

FIG. 8 is a waveform data format diagram in a modified example of the embodiment.

FIG. 9 is a waveform data format diagram of the second embodiment.

FIG. 10 is a diagram showing the relationship between waveform data and the original tone waveform in the second embodiment.

FIG. 11 is a time chart for explaining the operation of the decoder in the second embodiment.

[Explanation of symbols]

1 ... Tone switch, 2 ... Detector, 3 ... Keyboard, 4 ... Detector, 5 ... Sound source section, 6 ... Digital-analog converter (DA
C), 7 ... Sound system, 403 ... F number generator, 404 ... Counter, 405, 406 ... Multiplier, 41
0 ... Waveform memory, 411 ... Decoder, 412 ... Interpolation unit,
413 ... Envelope generator.

Claims (3)

[Claims] 1. Uncompressed waveform amplitude value data is provided for each of a plurality of predetermined sample points, and immediately before the uncompressed waveform amplitude value data for sample points located between the sample points of the uncompressed waveform amplitude value data. Waveform storing means provided with compressed waveform amplitude value data compressed so that it can be converted into non-compressed waveform amplitude value data, and frequency information generating means for generating frequency information according to the pitch of a musical tone waveform to be generated. And accumulating means for repeatedly accumulating the frequency information at a predetermined speed and outputting an accumulated value, and based on the accumulated value, two or more non-compressed waveform amplitude value data from the waveform storing means. Read means for reading a plurality of uncompressed waveform amplitude value data and compressed waveform amplitude value data in a continuous range including the compressed waveform amplitude value data read by the reading means. Conversion means for converting into a data, and based on the accumulated value, the uncompressed waveform amplitude value data read by the reading means and the uncompressed waveform amplitude value data obtained by the converting means. Selecting means for selecting the waveform amplitude value data, and using the uncompressed waveform amplitude value data selected by the selecting means, an interpolation means for performing interpolation processing based on the accumulated value and outputting the interpolation result as waveform data. A waveform generation device comprising: 2. Uncompressed waveform amplitude value data is provided for each of a plurality of predetermined sample points, and immediately before the uncompressed waveform amplitude value data for sample points between the sample points of the uncompressed waveform amplitude value data. Waveform storing means provided with compressed waveform amplitude value data compressed so that it can be converted into non-compressed waveform amplitude value data, and frequency information generating means for generating frequency information according to the pitch of a musical tone waveform to be generated. And accumulating means for repeatedly accumulating the frequency information at a predetermined speed and outputting an accumulated value, and based on the accumulated value, two or more non-compressed waveform amplitude value data from the waveform storing means. Reading means for reading a plurality of uncompressed waveform amplitude value data and compressed waveform amplitude value data in a continuous range, and a plurality of uncompressed waveform amplitude values read by the reading means based on the accumulated value Selecting means for selecting the non-compressed waveform amplitude value data and the compressed waveform amplitude value data to be used for interpolation from the data and the compressed waveform amplitude value data; and the compressed waveform amplitude value data selected by the selecting means. Converting means for converting into data, using the uncompressed waveform amplitude value data selected by the selecting means and the uncompressed waveform amplitude value data obtained by the converting means, performing interpolation processing based on the accumulated value,
A waveform generation device, comprising: an interpolation unit that outputs an interpolation result as waveform data. 3. An uncompressed waveform amplitude value data is provided for each of a plurality of predetermined sample points, and immediately before the uncompressed waveform amplitude value data for the sample points between the sample points of the uncompressed waveform amplitude value data. The waveform storage device is characterized in that compressed waveform amplitude value data compressed so as to be converted into non-compressed waveform amplitude value data is provided.