JP3479999B2 - Waveform data generator - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [0001] The present invention relates to a waveform data generator.
More specifically, waveform data for generating a tone signal or the like is described in detail.
Data from the waveform memory to generate a waveform
Data generator. [0002] 2. Description of the Related Art Recently, with the development of digital technology,
Digitalization of electronic musical instruments has progressed, and
Reads the waveform data stored in the
Pitch conversion is performed by filter processing. [0003] In such pitch conversion, digital
Conventionally, as a waveform data generator that performs filter processing,
There is one as shown in FIG. FIG. 8 shows a conventional waveform data generator.
1 has four registers 2, 3, 4, 5 for waveform data.
Provided in each of the waveform data registers 2, 3, 4, and 5.
From the waveform memory (not shown) at each sampling timing.
The four consecutive waveform data read out are sequentially stored.
You. This waveform memory is controlled by an address counter (not shown).
Addressed. [0005] The waveform data generating device 1 has a coefficient memo.
The coefficient memory 6 has an address for reading a waveform.
Addressed by the decimal data of the address counter.
From the coefficient memory 6 to the above four waveform data
The corresponding four coefficient data are stored in registers 7, 8, 9, 1
Read to 0. The coefficient memory 6 has a predetermined value.
Multiple sets of filtered filter coefficient data are stored.
The numerical data is a set of four data. The data stored in the registers 2, 3, 4, 5
The waveform data and the data stored in the registers 7, 8, 9, and 10 are stored.
The obtained coefficient data is respectively supplied to the corresponding multipliers 11, 1
2, 13 and 14, and each of the multipliers 11, 12, 1
Reference numerals 3 and 14 denote input waveform data and coefficient data, respectively.
And outputs the result to the adder 15. [0007] The adder 15 is connected to each of the multipliers 11, 12, 1
The multiplication results input from 3 and 14 are added, and the output waveform and
And output. [0008] SUMMARY OF THE INVENTION However, the conventional
In a waveform data generator, the coefficient data is
Is stored in the coefficient memory dedicated to
Sufficient coefficient data can be stored in terms of cost
The overall calculation accuracy in the pitch conversion process is poor.
There was a title. That is, the pitch conversion process is performed by a digital filter.
Perform pitch conversion with high accuracy when realizing with
Then, instead of the primary filter (linear interpolation),
It is necessary to perform digital filter processing.
Digital filter processing
In order to realize the processing itself with high accuracy, fill
It is necessary to increase the word length of the
Becomes large. In this case,
At the expense of the calculation accuracy of the
Noise increases, and the accuracy of the pitch conversion process deteriorates. That is, a filter coefficient in an electronic musical instrument
In general, the word length of 16 bits or more
Have been. Therefore, a conventional waveform data as shown in FIG.
In the case of the data generator 1, the decimal part of the address counter is
If it is 15 bits, the coefficient memory 6 has
Filter coefficients with different word lengths
Unit stores (16 × 4) bits × 32768 words
Requires a large storage capacity, and the coefficient memory 6 and thus the waveform data
The size of the entire data generator 1 becomes large,
Problem. Therefore, the upper bit of the decimal part of the address counter is
Read two sets of coefficients according to the
To obtain a given set of filter coefficients by interpolation
Is also conceivable. However, it is difficult to find the filter coefficients by interpolation.
In some cases, this is not a basic solution. That is, by interpolation processing,
To maintain the filter coefficient accuracy, the stored filter
Data pairs must be somewhat large.
Address counter to address the coefficient memory.
Even if the upper part of several parts is 10 bits, the coefficient memory
The size is (16 × 4) bits × 1024 words
Requires a large storage capacity. Accordingly, the present invention provides a waveform conversion method for pitch conversion.
While maintaining good calculation accuracy of data and coefficients,
Can reduce the size of hardware and reduce costs
It is an object of the present invention to provide a waveform data generator. [0016] SUMMARY OF THE INVENTION Waveform data generation of the present invention
The device is a means for storing waveform data, and
Stores coefficient data for waveform data calculation in the storage area of
Waveform storage means;The bits that make up the integer and decimal parts
And is represented by an integer bit for each sampling timing.
An address counter for updating a count value to be provided; Sun
Integer part of this address counter for each pulling timing
The count value represented by the bits constitutingThe waveform record
For reading predetermined waveform data stored in the storage means.
Waveform addressAsMeans for generating a waveform address,
Each time the sampling counter is
The bits that make up the fractional part and the fractional part bit
The lower predetermined bits of the integer part composing the waveform address
The added one is generated by the waveform address generating means.
Corresponding to the waveform data read by the
Reading corresponding predetermined coefficient data from the waveform storage means.
As a coefficient address forGenerate coefficient address to generate
Means and a waveform output from the waveform address generating means
Address and a coefficient output from the coefficient address generating means.
Number addressesAnd are selected in turn,The waveform memory hand
Address selecting means for outputting to the stage;
From the waveform storage means based on the address from the stage
Data and coefficient dataAnd alternatelyReading means to read
Read from the waveform storage means by the reading means.
Waveform dataOf the read waveform data
The coefficient data read from the waveform storage means corresponding to the set
Multiply-accumulate operation by data setFilter that performs processing and outputs
Processing means, and
I have. [0017] [0018] [0019] [0020] According to the present invention, a large amount of waveform data is stored.
The waveform data is stored in the waveform storage means designed to
Simply storing the waveform data in the free space that normally occurs
The coefficient data for data calculation is stored. And from this waveform storage meansWaveform data
In order to read the
Generate an address. This waveform address has an integer part and
It has bits that make up the decimal part, and the sampling timing
Updates the count value represented by the integer part bit for each bug
It has an address counter, and every sampling timing
Represented by the bits that make up the integer part of this address counter.
Count value. In addition, this waveform address
Predetermined coefficient data corresponding to the waveform data read from
In order to read the set of parameters,
To generate a coefficient address. This coefficient address is
Decimal number of this address counter for each sampling timing
Waveforms above the bits that make up the part and the decimal part bits
Adds lower specified bits of the integer part of the address.
Consisting of [0022]The waveform is obtained by the address selecting means.
The waveform address and coefficient output from the address generator are
And the coefficient addresses output from the dress generation means.
Alternately selected and output to the waveform storage means.
Waveform recording based on the address from the address selection means.
The waveform data and the coefficient data are sequentially and alternately read from storage means.
Protrude. [0023]Furthermore, by the filter processing means,Reading means
Waveform data read byAnd the read
Read from the waveform storage means corresponding to the set of
Sum-of-products operation using the set of output coefficient dataPerform processing. Therefore, for storing the coefficient data,
There was no need to provide a separate storage means, which was conventionally required
Hardware can be reduced by the amount of coefficient memory.
Waveform data and coefficient data with a simple circuit configuration.
Address setting for reading the data
Calculation accuracy between waveform data and coefficient data
It is important to reduce the size of the waveform data generator while maintaining good
And cost can be reduced. [0025] EXAMPLES Hereinafter, the present invention will be specifically described based on examples.
I will tell. FIGS. 1 to 6 show a waveform data generator according to the present invention.
FIG. 3 is a diagram showing an embodiment of a storage device.
ad Only Memory)
PCM sound that outputs waveform data
Applied to the source system. FIG. 1 shows a waveform data generator 10 according to the present invention.
FIG. 3 is a circuit block diagram of FIG. In FIG. 1, a waveform data generator 10
Is a sound source LSI (Large Scale Integrated circuit) 1
1 and a waveform memory 12.
Address counter 13, address selector 14, address
Les register 15, data register 16, and pitch converter
17 and the like. The waveform memory 12 is constituted by a ROM.
Are divided into a waveform data area and coefficient data area.
ing. The waveform memory 12 stores a waveform in the waveform data area.
Shape data is stored in advance, and pitch conversion is performed in the coefficient data area.
Coefficient data is stored in advance. The waveform memory 12 stores the address of the
The packet configuration is 20 bits, and the address
Addressing unit 13 and the address selector 14
Is The address counter 13 is provided with a CPU (not shown).
(Central Processing Unit)
At each sampling timing, the waveform
Waveform address (upper bit) A for reading data
a and a coefficient address for interpolating the read waveform data.
Output address (lower bit) Ad to address selector 4
I do. The waveform address Aa and the coefficient address Ad are:
For example, it is composed of 20 bits. The address counter 13 stores the address
In the setting, for the pitch conversion by the instruction from the CPU
Four adjacent waveform data from the waveform memory 12
To read the four adjacent addresses
A pair, for example, as shown in FIG.
1, Aa0, (Aa0) +1, (Aa0) +2 or (Aa
1) -1, Aa1, (Aa1) +1, (Aa1) +2
Is the waveform address Aa and the sampling timing
It is output sequentially every time. The address counter 13 is provided with a CPU
From the four waveform data
The corresponding set of four filter coefficients is read from the waveform memory 12.
To find out, a set of four addresses, for example, FIG.
As shown in (b), Ad0_a, Ad0_b, Ad0
_C, Ad0_d, Ad1_a, Ad1_b, Ad1_
c, Ad1_d as coefficient address Ad, sampler
Output sequentially at each timing. The waveform address Aa and the coefficient address Ad
Means the count value of the address counter 13 and the value shown in FIG.
They have such a relationship. That is, the address counter unit 13
Bit data (for example, 20 bits
Data) as the waveform address Aa, and
Lower bit data of the default value (for example, 16-bit data)
Waveform address to specify a set of coefficient data
Add the lower 2 bits of Aa and add the value of the additional 2 bits
, Four coefficient addresses Ad, for example, A
dx_a, Adx_b, Adx_c, Adx_d (x
Produces an arbitrary numerical value). Note that the address counter
The lower bit data of the count value of the unit 13 is 16 bits
Therefore, even if the above 2 bits are added, 18 bits
It is. Therefore, the address counter unit 13 further
To the waveform address Aa.
Generates a 20-bit coefficient address Ad with the same number of bits as
are doing. Therefore, the coefficient address Ad is
When "00" is added to the beginning, the waveform memory 12
Starting address of waveform memory 12 by several addresses Ad
The part will be addressed. Therefore, the waveform
A coefficient data area is formed at the beginning of the address of the memory 12.
And the lower bits of the address counter 13 are 16 bits.
, The coefficient data KD is (16 + 2) bits
And the address will be specified from "00000".
“3FFFF” is an address area of the coefficient data KD.
You. The coefficient data is as shown in FIG.
The waveform address Aa, for example, (Aa0) -1, Aa
Waves read at 0, (Aa0) +1, (Aa0) +2
In the shape data, a coefficient address Ad, for example, Ad0_a,
The relations read out by Ad0_b, Ad0_c, and Ad0_d
Coefficient data array that is related to multiply numerical data
Is stored in the waveform memory 12. The waveform address generated as described above
Aa and the coefficient address Ad are sent to the address selector 14.
Output, and the address selector 14
Aa and the coefficient address Ad are converted as shown in FIG.
And alternately select them and output them to the address register 15
I do. The address register 15 has an address select.
The output of the data 14 at the sampling timing.
And outputs it to the data bus AA. Therefore, the data
As shown in FIG. 2D, the waveform address A
a and the coefficient address Ad are output alternately, and the waveform memory 1
2 is set. According to the address setting of the address bus AA,
Waveform data and coefficient data from the waveform memory 12 alternately.
The data is read and the data bus DD is connected to the tie shown in FIG.
, The waveform data WD, for example, WD00, WD0
1, WD02, WD03 and coefficient data KD, for example, K
D0a, KD0b, KD0c, and KD0d are output alternately.
It is. The data bus DD is connected to the data register 16.
The data register 16 is connected to the data bus D
Waveform data WD and coefficient data KD input through D
Are sequentially held, and at the timing shown in FIG.
Waveform data WD and coefficient data as the register output DR.
For example, KD00, KD0a,...
In this way, the output is alternately output to the pitch converter 17. The pitch conversion unit 17 receives the above-mentioned wave
The shape data WD and the coefficient data KD are fetched and digitized.
Executes the total filter operation and outputs the reproduced waveform output
I do. The pitch conversion unit 17 is provided as shown in FIG.
, A waveform register 20, a coefficient register 21, and a product-sum operation
It is constituted by a part 22. The waveform register 20 and the coefficient register 21
From the data register 16 as shown in FIG.
The data register output DR is input and the waveform register 20
And the coefficient register 21 waveform the data register output DR.
Data and coefficient data, and separate
The waveform data is stored in the waveform register 20 and the coefficient data is stored in the coefficient register.
The star 21 holds. The waveform register 20 stores the held waveform data.
As shown in FIG. 5B, the waveform register output WR,
For example, WD00, WD01, WD02, and WD03
And outputs the result to the product-sum operation unit 22.
The obtained coefficient data is stored in a coefficient register as shown in FIG.
Star output KR, for example, KD0a, KD0b, KD0
c and KD0d are output to the product-sum operation unit 22. The product-sum operation unit 22 is separated and input.
Multiply the waveform register output WR by the coefficient register output KR
After performing accumulation, that is, product-sum operation processing,
Output as shape. That is, the product-sum operation unit 22 is configured as shown in FIG.
Thus, the multiplier 30, the adder 31, the accumulation register 32,
And an output register 33. The multiplier 30 has the waveform register output W
R and the coefficient register output KR are input and the multiplier 30
Is a waveform register output WR and a coefficient register which are sequentially input.
The output KR is multiplied by the output KR and output to the adder 31. Sand
The multiplier 30 has a waveform register output WR, for example,
WD00, WD01, WD02, WD03 and coefficient register
Output KR, for example, KD0a, KD0b, KD0c,
KD0d are sequentially input, and the multiplier 30 outputs these waveforms.
A register output WR and a coefficient register output KR, for example, WD
00 and KD0a, WD01 and KD0b...
Then, the data is output to the adder 31. The adder 31 further includes an accumulation register 32
Is output, and the adder 31
The result of the multiplication is added to the output of the accumulation register 32 to perform accumulation.
Output to register 32. The accumulation register 32 further includes
Asynchronous clear signal as shown in (e) is input and accumulated
The register 32 stores the asynchronous clear signal and the next asynchronous clear.
A predetermined time is output from the adder 31 until the signal is input.
Accumulates the input signal and outputs the accumulation result to the output
Output to the register 33. That is, the accumulation register 32
As shown in FIG. 5 (e), an asynchronous clear signal is input.
Is cleared to "0", and then
For each time, the addition result input from the adder 31 is sequentially accumulated
And the accumulation result, for example, Ac0a, Ac0a + Ac
0b, Ac0a + Ac0b + Ac0c, Ac0a + Ac
0b + Ac0c + Ac0d is output to the output register 33.
You. The output register 33 is an accumulation register 33
The accumulated result input from the
As shown in FIG. 5F, the final accumulation result, for example, Ac
0a + Ac0b + Ac0c + Ac0d is used as the reproduced waveform output
And output. Next, the operation will be described. The waveform data generator 10 includes a waveform memory 1
2 stores the coefficient data together with the waveform data.
Read four consecutive waveform data from waveform memory 12
And four sets corresponding to these four waveform data
Address setting for reading coefficient data of
You. That is, the add of the waveform data generator 10
The counter 13 is operated by an instruction from a CPU (not shown).
For each sampling timing, as shown in FIG.
Waveform address for reading waveform data from memory 12
Aa, for example, (Aa0) -1, Aa0, (Aa0)
+1 and (Aa0) +2 to complement the read waveform data.
The coefficient address Ad of the coefficient data KD, for example,
For example, Ad0_a, Ad0_b, Ad0_c, Ad0_d
Is generated and output to the address selector 14. The address selector 14 has an address counter.
Address Aa and coefficient address input from the
2 are sequentially and alternately selected as shown in FIG.
And outputs it to the address register 15. The address register 15 has an address select.
The address output from the data 14 is sequentially
2d, and stored in the waveform memory as shown in FIG.
12 is output to an address bus AA connected to the address bus AA. According to the address setting of the address bus AA,
Waveform data and coefficient data from the waveform memory 12 alternately
And the data connected to the data register 16
At the timing shown in FIG.
WD, for example, WD00, WD01, WD02, WD
03 and coefficient data KD, for example, KD0a, KD0b,
KD0c and KD0d are output alternately. The data register 16 stores the input waveform data.
The data WD and the coefficient data KD are shown in FIG.
The data register output DR
The shape data WD and the coefficient data KD are, for example, WD00,
.. KD0a,...
Output. The pitch conversion unit 17, as shown in FIG.
A product-sum operation is sequentially performed on the waveform data WD and the coefficient data KD,
In synchronization with the asynchronous clear signal, the output level shown in FIG.
The register output is output as a reproduction waveform. As described above, the waveform data generation of this embodiment
The device 10 stores the waveform data in the waveform memory 12 together with the waveform data.
Numerical data is also stored, and continuous data is stored from the waveform memory 12.
Read out the four waveform data and read these four waveforms
To read four sets of coefficient data corresponding to the data
Address setting of the address counter 13 and the address
This is performed by the selector 14. As a result,
In addition, it is necessary to prepare a separate coefficient memory for storing coefficient data.
There is no need to use waveform data and coefficient data for pitch conversion.
Hardware scale while maintaining good computational accuracy
By reducing the size, the waveform data generator 10 can be downsized.
And the cost can be reduced. FIG. 7 shows another example of the waveform data generating apparatus of the present invention.
It is a figure which shows the Example of. This embodiment has the same waveform data as the above embodiment.
This is applied to the data generator, and
In the following description, the reference numerals used in the above embodiments are used as they are.
I do. The waveform data generating device 10 of the present embodiment
The waveform data similar to the above is stored in the waveform memory 12 of FIG.
And the coefficient data of two groups with different filter characteristics.
Data KD. Further, the address counter section 13
As in the embodiment, the upper bits of the count value are assigned to the waveform address.
Aa is output to the address selector 14 and
A coefficient address Ad is generated based on the lower bits, and
Output to the selector 14. The coefficient address Ad is, as shown in FIG.
Are the lower bits of the count value of the address counter unit 13.
In the same manner as above, the waveform address is assigned to Ab as a bit representing a set.
Add the lower 2 bits of Aa and specify the group.
1 bit of the waveform address Aa
Of the filter characteristic selection flag bit Sl of
It is 19 bits with 3 bits added. And last
And 20 bits, which is the number of bits of the waveform address Aa.
In order to make the coefficient address A
Output as d. Therefore, the coefficient address Ad is (16
+3) bits to specify the address.
0000H ”to“ 7FFFFH ”are the addresses of the coefficient data KD.
It becomes a dress area. Then, the filter characteristic selection flag bit S
The group of coefficient data KD selected by l
The filter characteristic selection flag bit Sl is composed of one bit.
Therefore, depending on the value of the filter characteristic selection flag bit Sl,
Thus, the grouping of two coefficient data KD having different filter characteristics is performed.
Loop can be selected, filter characteristics selection flag
Group of coefficient data KD whose bit Sl is "0"
Indicates that the coefficient address Ad is "00000H" to "3FFF
The range of “FH” becomes the data area, and the filter characteristic selection is performed.
Of the coefficient data KD whose selection flag bit Sl is “1”.
In the loop, the coefficient address Ad is "40000H" to "7
The range of “FFFFH” is the data area. The address counter section 13 detects such a wave.
Address address selector Aa and coefficient address Ad
14, and the address selector 14 outputs
Similarly, the waveform address Aa and the coefficient address Ad are sequentially
The addresses are alternately selected and output to the address register 15. The address register 15 has an address select.
The output of the data
Output to the memory AA, and specify the address of the waveform memory 12.
You. The waveform memory 12 stores the waveform as described above.
Along with data WD, two groups with different filter characteristics
The coefficient data KD of the loop is stored and the waveform address
Be addressed by Aa and coefficient address Ad
As a result, waveform data is read out and two groups
Four sets of coefficient data KD are read from one of the loops.
And are sequentially output to the data bus DD. The waveform data output to data bus DD
The data WD and the coefficient data KD are the same as in the above embodiment.
And is output to the pitch conversion unit 17.
It is. The pitch conversion unit 17 performs the waveform decoding in the same manner as in the above embodiment.
Data WD and coefficient data KD are multiplied and summed to produce a reproduced waveform
Is output as Therefore, in the present embodiment,
Similar to the embodiment, a separate coefficient memory for storing coefficient data is provided.
No need to prepare, reduce hardware scale,
To reduce the size and cost of the
Can be. Further, in this embodiment, the waveform memory
12 shows two coefficient groups for different filter characteristics.
Select the coefficient data from two groups
So that the playback output waveform can be
The pitch conversion can be performed more appropriately. [0073] According to the present invention, waveform data is stored.
In the free space of the large-capacity waveform storage means for
Waveform data and coefficient data with a simple circuit configuration
Address can be set to read the
Separate coefficient memory for storing data must be provided
There is no hardware
Data can be reduced, and the waveform
While maintaining good calculation accuracy between data and coefficient data.
Shape data generator can be downsized,
Cost can be reduced.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a circuit block diagram of an embodiment of a waveform data generating device according to the present invention. FIG. 2 is an operation timing chart of the waveform data generation device of FIG. 1; FIG. 3 is an explanatory diagram of address generation of a waveform address Aa and a coefficient address Ad by the waveform data generation device of FIG. 1; FIG. 4 is a detailed circuit block diagram of a pitch converter of FIG. 1; FIG. 5 is an operation timing chart of a pitch conversion unit. FIG. 6 is a detailed circuit block diagram of a product-sum operation unit in FIG. 4; FIG. 7 is an operation timing chart of an embodiment of the waveform data generation device of the present invention. FIG. 8 is a circuit block diagram of a conventional waveform data generation device. [Description of Signs] 10 Waveform data generation device 11 Sound source LSI 12 Waveform memory 13 Address counter unit 14 Address selector 15 Address register 16 Data register 17 Pitch conversion unit 20 Waveform register 21 Coefficient register 22 Product-sum operation unit 30 Multiplier 31 Adder 32 accumulation register 33 output register

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int. Cl. ⁷ , DB name) G10H 1/00-7/12

Claims (1)

(57) Claims 1. A means for storing waveform data, wherein a waveform storage means for storing coefficient data for calculating waveform data in a partial storage area thereof, an integer part and a decimal number Has bits that make up the
Count value represented by an integer part bit at each timing
An address counter for updating, at each sampling timing settling of the address counter
A count value represented by bits constituting the number parts, the waveform address generating means for generating a waveform address for reading predetermined waveform data stored in waveform memory means, said address counter small for every sampling timing
The bits that make up the fractional part and the fractional part bit
The lower predetermined bits of the integer part composing the waveform address
The added one is generated by the waveform address generating means.
Corresponding to the waveform data read by the
Reading corresponding predetermined coefficient data from the waveform storage means.
A coefficient address generating means for generating a coefficient address for generating a waveform address, and a waveform address output from the waveform address generating means and a coefficient address output from the coefficient address generating means are sequentially and alternately selected to store the waveform data. and an address selection means for outputting the unit, the waveform data and the coefficient data from said waveform storage means based on the address from said address selecting means, sequentially and alternately
Reading and reading means, and a set of waveform data read out from said waveform storage means by the reading means, to the waveform data read out set in
Coefficient data correspondingly read from the waveform storage means
And a filter processing means for performing a sum-of-products calculation process by a set of and outputting the result.