JP4983858B2 - Filter device and filter processing program - Google Patents

Description

  The present invention relates to a filter device and a filter processing program suitable for use in, for example, an electronic musical instrument.

  An electronic musical instrument uses a digital filter whose filter characteristics can be dynamically changed according to the manner in which musical sounds are generated. For this reason, in many cases, a filter coefficient having desired filter characteristics is obtained by referring to a coefficient table in which filter characteristics and filter coefficients are stored in association with each other.

  As this type of technology, for example, in Patent Document 1, a cutoff frequency value is expressed by an integer part I and a fractional part F, and a coefficient table storing filter coefficients corresponding to the integer part I of the cutoff frequency value is provided. When the fractional part F of the cutoff frequency value is zero, the filter coefficient corresponding to the integer part I of the cutoff frequency value is read from the coefficient table and output as it is, and when the fractional part F is not zero, the coefficient table is changed to the integer part I. The corresponding filter coefficient A and the filter coefficient B corresponding to the integer part I + 1 are read out, and the filter coefficient C obtained by linear interpolation A + F (AB) of the read filter coefficients A and B with the decimal part F is output. A filter device is disclosed.

Japanese Examined Patent Publication No. 6-91418

By the way, in the conventional filter device described above, when the filter frequency characteristics are continuously changed, if the filter coefficient is changed suddenly, the output becomes discontinuous and causes noise, so the filter coefficient is set in small increments. Or by interpolating filter coefficients.
However, if the filter coefficient is set in small increments, the coefficient table is enlarged. On the other hand, since the filter coefficient is interpolated, a separate coefficient interpolator is provided, resulting in an increase in circuit scale and a decrease in processing speed.
Further, in order to improve the sound quality of the output, it is only necessary to increase the number of calculation bits for the filter operation. However, if the number of calculation bits is increased, there is a problem that the circuit scale and the processing speed are further reduced.

Accordingly, a first object of the present invention is to provide a filter device and a filter processing program capable of continuously changing the frequency characteristics without enlarging the coefficient table and without causing an increase in circuit scale or a reduction in processing speed. There is.
A second object of the present invention is to provide a filter device and a filter processing program capable of improving sound quality without inviting an increase in circuit scale and a decrease in processing speed.

According to the first aspect of the present invention, the input waveform is divided into a plurality of frequency bands, the detection means for detecting the envelope level for each divided frequency band, and the envelope level detected by the detection means has a predetermined value. Instructing means for instructing filter operation of the first number of calculation bits for the frequency band exceeding, and instructing filter operation of the second number of calculation bits for the frequency band below a predetermined value, corresponding to the plurality of frequency bands Provided with a filter means for applying any one of the first and second calculation bit numbers instructed by the instruction means to the input waveform in the corresponding frequency band.

According to the second aspect of the present invention, band dividing means for dividing the input waveform into a plurality of frequency bands, and the number of calculation bits associated with the waveform type and pitch of the input waveform for each of the plurality of frequency bands in advance. Storage means for storing, assignment means for reading out the corresponding number of operation bits from the storage means and assigning them to each frequency band in accordance with a musical tone parameter representing the waveform type and pitch of the input waveform, and for each of the plurality of frequency bands And a filter unit that is provided correspondingly and applies a filter operation of the number of operation bits assigned by the assigning unit to an input waveform in a corresponding frequency band.

According to the third aspect of the invention, the input waveform is divided into a plurality of frequency bands, the detection step for detecting the envelope level for each of the divided frequency bands, and the envelope level detected in the detection step is a predetermined value. An instruction step for instructing a filter operation with a first number of calculation bits for a frequency band exceeding 1, a command for instructing a filter operation with a second number of calculation bits for a frequency band equal to or less than a predetermined value, and corresponding to the plurality of frequency bands And a filtering step in which any one of the first and second calculation bit numbers indicated in the instruction step is applied to the input waveform in the corresponding frequency band. .

In the invention according to claim 4 , a band dividing step for dividing the input waveform into a plurality of frequency bands, and a calculation bit number associated with the waveform type and the pitch of the input waveform for each of the plurality of frequency bands in advance. An allocation step of storing and assigning the corresponding number of operation bits to each frequency band in accordance with a musical sound parameter indicating the waveform type and pitch of the input waveform, and executing corresponding to the plurality of frequency bands A filtering step of applying a filter operation of the number of operation bits assigned in the assignment step to an input waveform in a corresponding frequency band is executed by a computer.

  In the present invention, the input waveform is divided into a plurality of frequency bands, the envelope level for each frequency band is detected, and a filter operation with the first number of calculation bits is performed on the frequency band where the detected envelope level exceeds a predetermined value. When a filter operation of the second number of calculation bits is instructed to a frequency band equal to or less than a predetermined value, the filter means provided corresponding to each frequency band performs a filter operation of the specified number of calculation bits on the input waveform. Therefore, high-accuracy filtering is performed in the frequency band where the waveform level is large and easily audible, and low-accuracy filtering is performed in the frequency band where the waveform level is small and audibly inconspicuous, thereby increasing the circuit scale and processing. It is possible to improve the sound quality on hearing without incurring a decrease in speed.

According to the first and third aspects of the present invention, the input waveform is divided into a plurality of frequency bands to detect the envelope level for each frequency band, and the detected envelope level exceeds the predetermined value. When a filter operation with the number of operation bits of 1 is instructed and a filter operation with the second number of operation bits is instructed in a frequency band equal to or less than a predetermined value, the filter means provided corresponding to each frequency band is instructed Since a number of filter operations are performed on the input waveform, high-accuracy filtering is performed in the frequency band where the waveform level is large and easily audible, and low-accuracy filtering is performed in the frequency band where the waveform level is small and audible. Further, it is possible to improve the sound quality on hearing without inviting an increase in circuit scale or a decrease in processing speed.
According to the second and fourth aspects of the invention, the number of calculation bits associated with the waveform type and pitch of the input waveform is stored in advance for each of a plurality of frequency bands, and the waveform type and pitch of the input waveform are stored. When the corresponding number of operation bits is read out and assigned to the filter means for each frequency band in accordance with the musical sound parameter representing, each filter means performs the filter operation for the assigned operation bit number on the input waveform of the corresponding frequency band. Therefore, it is possible to immediately specify filtering of the optimum number of calculation bits corresponding to the change in waveform type or pitch, and the configuration for detecting the envelope can be omitted, so that the audibility is not incurred without increasing the circuit scale or reducing the processing speed. The above sound quality can be improved.

It is a block diagram which shows the whole structure of the electronic musical instrument provided with the filter apparatus by 1st Example. It is a block diagram which shows the partial structure of the sound source 6 provided with the filter apparatus by 1st Example. 3 is a block diagram illustrating a configuration example of a filter unit 12. FIG. It is a figure which shows an example of the filter coefficient memorize | stored in the coefficient table. 3 is a block diagram showing a configuration of an amplitude interpolator 14. FIG. It is a figure which shows the mode of the time-dependent change (the figure (A)) of the coefficient (alpha) which the interpolation coefficient generator 14a generate | occur | produces, and the time-dependent change (the figure (B)) of the coefficient (beta). It is a block diagram which shows the structure of the filter apparatus by 2nd Example. It is a block diagram which shows a product-sum operation processing logic. It is a flowchart which shows the operation | movement of the word length allocation process which the control part 10 performs. 6 is a flowchart showing an operation of a filter operation timer interrupt process executed by the low-pass filter 22; It is a figure which shows the structure of the word length allocation table TBL in a modification. It is a flowchart which shows the operation | movement of the word length allocation process which the control part 10 in a modification is performed.

  The filter device and the filter processing program according to the present invention are applied to an electronic musical instrument, an effect adding device, and the like. Hereinafter, a filter device (or a filter processing program) mounted on an electronic musical instrument will be described as an example, and this will be described with reference to the drawings.

A. First Embodiment (1) Overall Configuration FIG. 1 is a block diagram showing the overall configuration of an electronic musical instrument provided with a filter device according to a first embodiment. In this figure, reference numeral 1 denotes a keyboard for generating performance information including key-on / key-off events, note numbers, velocities, and the like in response to pressing and releasing keys. Reference numeral 2 denotes a switch unit composed of various switches arranged on the musical instrument panel, and generates a switch event corresponding to the operated switch. Reference numeral 3 denotes a ROM for storing various control programs. A RAM 4 is used as a work area for the CPU 5 and temporarily stores various register / flag data.

  The CPU 5 executes various control programs stored in the ROM 3, for example, generates musical tone parameters corresponding to performance information generated by the keyboard 1 and sends them to the sound source 6, or operates in response to a switch event generated by the switch unit 2. Control each part of the instrument, such as changing the mode. Note that the musical sound parameters generated by the CPU 5 include parameters for modifying the waveform by changing the frequency characteristics of the generated waveform, such as timbre switching and effect application, in addition to no-on / off events for instructing sound generation or muting. Reference numeral 6 denotes a sound source that generates a waveform corresponding to a musical sound parameter supplied from the CPU 5. The sound source 6 includes a filter device according to the present invention, and a specific configuration thereof will be described later. A D / A converter 7 converts the output of the sound source 6 into an analog waveform signal. Reference numeral 8 denotes a sound system that amplifies the waveform signal supplied from the D / A converter 7 and produces sound from the speaker 9.

(2) Configuration of Sound Source 6 Next, the configuration of the sound source 6 will be described with reference to FIGS. FIG. 2 is a block diagram showing a partial configuration of the sound source 6 including the filter device according to the present invention. In FIG. 2, reference numeral 10 denotes a control unit that generates a timbre number TN, phase data PD, and cut-off frequency data CF in accordance with a musical tone parameter supplied from the CPU 5. Further, the control unit 10 generates the interpolation start trigger TR at the timing of switching the cutoff frequency data CF generated corresponding to the musical tone parameter. Reference numeral 11 denotes a waveform generator configured by a known waveform memory reading system. The waveform generator 11 reads out and outputs the waveform data of the timbre specified by the timbre number TN supplied from the control unit 10 according to the phase data PD.

A filter unit 12 filters the waveform data output from the waveform generator 11 and supplies it to the amplitude interpolator 14 at the next stage. For example, as illustrated in FIG. 3, the filter unit 12 includes delay elements 12 a to 12 d that delay the input by one sample and output, and a coefficient multiplier 12 e to which filter coefficients a0 to a2 and b1 to b2 described later are respectively supplied. To 12i and an adder 12j for adding the outputs of the coefficient multipliers 12e to 12i. The filter unit 12 having such a configuration performs a filter operation of frequency characteristics according to filter coefficients a0 to a2 and b1 to b2 output from a coefficient table 13 described later.
Specifically, input x [n], output x [n-1] of delay element 12a, output x [n-2] of delay element 12b, output y [n-1] of delay element 12c, and delay element 12d Output y [n-2], the output y [n] is generated by the filter operation expressed by the following equation (1).
y [n] = a0 * x [n] + a1 * x [n-1] + a2 * x [n-2] -b1 * y [n-1] -b2 * y [n-2] (1)

  When the filter unit 12 is operated as a low-pass filter, the coefficient table 13 stores filter coefficients a0 to a2 and b1 to b2 corresponding to various cutoff frequency data CF in the ROM as illustrated in FIG. The filter coefficients a0 to a2 and b1 to b2 corresponding to the cutoff frequency data CF supplied from the control unit 10 are read out and supplied to the coefficient multipliers 12e to 12i of the filter unit 12.

  As shown in FIG. 5, the amplitude interpolator 14 includes an interpolation coefficient generator 14a, a ring buffer 14b, multipliers 14c to 14d, and an adder 14e. The interpolation coefficient generator 14a generates coefficients α and β that change with time at a constant rate from the time when the interpolation start trigger TR is input from the control unit 10, that is, the time corresponding to the N cycles of the waveform from the filter coefficient switching time. When α becomes “0” (coefficient β becomes “1”), a signal “write” is generated to instruct the ring buffer 14b to start buffer clear and write. The amplitude interpolator 14 returns the coefficient α to the initial setting of “1” and the coefficient β to “0” when the ring buffer 14 b has finished writing the waveform data for N cycles (N is an integer).

The ring buffer 14b sequentially reads and outputs the accumulated waveform data while accumulating the input waveform data for N cycles of the waveform (N is an integer). In other words, the ring buffer 14b delays the input waveform data by N cycles of the waveform and outputs it. When the signal write is supplied from the interpolation coefficient generator 14a, the accumulated waveform data is once cleared. Start writing the newly input waveform data.
The multiplier 14c multiplies the waveform data output from the ring buffer 14b by the coefficient α supplied from the interpolation coefficient generator 14a and outputs the result. The multiplier 14d multiplies the waveform data output from the previous filter unit 12 by the coefficient β supplied from the interpolation coefficient generator 14a, and outputs the result. The adder 14e adds the multiplication results of the multipliers 14c and 14d and outputs the result.

(3) Operation of Amplitude Interpolator 14 Next, the operation of the amplitude interpolator 14 according to the gist of the present invention will be described with reference to FIG. FIG. 6 illustrates the change over time of the coefficient α (FIG. 6 (a)) and the change over time of the coefficient β (FIG. 6 (b)) generated by the interpolation coefficient generator 14a. In this figure, (A) represents a steady state before the interpolation start trigger TR is supplied from the control unit 10. In the steady state (A), since the interpolation coefficient generator 14a is under the initial setting with the coefficient α being “1” and the coefficient β being “0”, the amplitude interpolator 14 receives only the waveform data read from the ring buffer 14b. Output.

  Then, it is assumed that the filter coefficient switching time t1 when the interpolation start trigger TR is input from the control unit 10 to the interpolation coefficient generator 14a is reached. At this time, since the interpolation coefficient generator 14a sets the coefficient α to “1” and the coefficient β to “0”, the amplitude interpolator 14 stores the waveform data read from the ring buffer 14b (accumulated before switching the filter coefficient). Only waveform data). Therefore, even if a discontinuity occurs in the output of the filter unit 12 when the filter coefficient is switched, the output of the filter unit 12 is muted by the multiplier 14d, so that noise generation can be suppressed.

  Next, in the transition state (B) after the filter coefficient switching time t1, the amplitude interpolator 14 multiplies the coefficient α that changes with time at a constant rate from “1” to “0” within the time corresponding to the N cycles of the waveform. Waveform data before switching the filter coefficient and waveform data after switching the filter coefficient obtained by multiplying “0” to “1” by a coefficient β (= 1−α) that changes with time at a constant rate within the time corresponding to N cycles of the waveform. Perform linear interpolation to add As a result, the amplitude level is continuously controlled in such a manner that the waveform data before the filter coefficient switching is cross-fade to the waveform data after the filter coefficient switching.

  When the coefficient α becomes “0” and the waveform data before switching the filter coefficient is muted, the coefficient β reaches “1” and only the waveform data after switching the filter coefficient is output. In the unit 14, in accordance with the signal write supplied from the interpolation coefficient generator 14a, the ring buffer 14b once clears the accumulated waveform data, and then starts writing the newly input waveform data. Subsequently, when reaching the time point t3 when the writing of waveform data for N waveforms (N is an integer) is completed in the ring buffer 14b, the interpolation coefficient generator 14a initializes the coefficient α to “1” and the coefficient β to “0”. Return to. As a result, the amplitude interpolator 14 returns to the steady state (A) in which only the waveform data read from the ring buffer 14b is output.

  Thus, in the first embodiment, the waveform data before the filter coefficient switching output from the filter unit 12 is accumulated in the ring buffer 14b until the filter coefficient is switched, while the waveform N period ( Waveform data delayed by N) is read and output. After the filter coefficient is switched, the waveform data before switching the filter coefficient read from the ring buffer 14b is subjected to amplitude interpolation so that the waveform data output from the filter unit 12 after the filter coefficient switching is crossfaded and output.

  The waveform data after switching the filter coefficient output from the filter unit 12 is updated while the accumulated content of the ring buffer 14b is updated from the waveform data before switching the filter coefficient to the waveform data after switching the filter coefficient from the time when the crossfading is completed. After the update is completed, the waveform data after the filter coefficient switching output from the filter unit 12 is accumulated in the ring buffer 14b, and the waveform data delayed from the ring buffer 14b by the waveform N period (N is an integer). Is read and output. As a result, discontinuity of the filter output is avoided, so that it is possible to continuously change the frequency characteristics without increasing the coefficient table, increasing the circuit scale, or reducing the processing speed.

In the first embodiment described above, an example in which the amplitude interpolator 14 performs linear interpolation is disclosed in order to simplify the description. However, the gist of the present invention is not limited to this, and includes a mode of nonlinear interpolation.
For example, the optimum interpolation calculation form is selected according to the level difference of the waveform data output by the filter unit 12 before and after the filter coefficient switching, and after the filter coefficient switching from the waveform data before the filter coefficient switching in the selected interpolation calculation form. It is also possible to perform non-linear interpolation on the waveform data. In this way, since the waveform data before and after the filter coefficient switching is smoothly connected in a more natural form, it is possible to contribute to improvement in sound quality.

  Further, in the first embodiment, the hardware image is expressed in order to clarify the functional configuration, but the gist of the present invention is not limited thereto, and the disclosed functional elements can of course be implemented by software processing. .

B. Second Embodiment Next, a second embodiment will be described with reference to FIGS.
(1) Configuration FIG. 7 is a block diagram showing the configuration of the filter device according to the second embodiment. Elements common to the first embodiment described above are given the same reference numerals. In FIG. 7, the filter unit 12 includes filter units 12-1 to 12-3 that classify an input waveform supplied from a waveform generator 11 (not shown) into first to third frequency bands. Each of the filter units 12-1 to 12-3 includes a bandpass filter 20 that classifies an input waveform into first to third frequency bands, an envelope detector 21 that detects an output envelope of the bandpass filter 20, and a later-described filter unit. A low-pass filter 22 that performs a filter operation with the number of operation bits specified by the signal WL supplied from the control unit 10.

  As in the first embodiment, the control unit 10 generates the timbre number TN and the phase data PD according to the musical tone parameters supplied from the CPU 5 and supplies them to the waveform generator 11 (not shown), while detecting the envelope. A signal WL for designating the number of calculation bits at the time of filter calculation is generated according to the detection result of the detector 21. When the signal WL is “1”, a 16-bit filter operation is instructed, and when the signal WL is “0”, an 8-bit filter operation is instructed.

  As in the first embodiment, the low-pass filter 22 is composed of the second-order IIR digital filter shown in FIG. 3, and the above-described equation (1), that is, the input x [n] and the output x [n] of the delay element 12a. n-1], output x [n-2] of the delay element 12b, output y [n-1] of the delay element 12c, and output y [n-2] of the delay element 12d, y [n] = a0 X [n] + a1 * x [n-1] + a2 * x [n-2] -b1 * y [n-1] -b2 * y [n-2] product-sum operation (filter operation) Is performed with 16 or 8 arithmetic bits. The low-pass filter 22 is preset with filter coefficients a0 to a2 and b1 to b2 adapted to the corresponding frequency band.

  By the way, when the product-sum operation e = a × c + b is performed by 16 bits or 8 bits, each word length of a, b, and c is 16 bits, and the upper / lower 8 bits of a are the higher order of ah, al, and c, respectively. / If the lower 8 bits are ch and cl, respectively, the 16-bit multiply-add operation is e = ah × ch × 10000h + ah × cl × 100h + al × ch × 100h + al × cl + b, which can be expressed by a combination of 8-bit operations. Here, 10000h represents a 16-bit upper shift, and 100h represents an 8-bit upper shift. On the other hand, the 8-bit multiply-accumulate operation has a form in which the lower 8 bits in the 16-bit multiply-accumulate operation are omitted, and is represented by e = ah × ch × 10000h + b.

  Therefore, in order to perform the product-sum operation e = a × c + b with 16 bits or 8 bits, the processing logic shown in FIG. 8 is executed. In FIG. 8, 40 is a selector for selecting either the upper 8 bits ah / lower 8 bits al of a, and 41 is a selector for selecting either the upper 8 bits ch / lower 8 bits cl of c. A multiplier 42 multiplies the output of the selector 40 and the output of the selector 41. A shifter 43 shifts the output of the multiplier 42 by 16 bits or 8 bits. An adder 44 accumulates the multiplication results. 45 is a latch that holds the previous multiplication result, 46 is a latch that holds the accumulated value output from the adder 44, and 47 is an adder that adds b to the value held by the latch 46.

When performing 16-bit multiply-accumulate operations in the above processing logic, ah × ch × 10000h, ah × cl × 100h, al × ch × 100h, and al × cl while appropriately switching the selectors 40 and 41 and the shifter 43. Multiply and accumulate the result, then add b. On the other hand, the 8-bit multiply-accumulate operation adds b after shifting the multiplication result of ah × ch by 16 bits.
Based on such processing logic, the low-pass filter 22 determines that y [n] = a0 · x [n] + a1 · x [n−1] + a2 · x [n−2] −b1 · y [n−1] −. A filter operation of b2 · y [n−2] is performed. That is, according to the processing logic of 16 bits / 8 bits specified by the signal WL, a0 · x [n], a1 · x [n−1], a2 · x [n-2], −b1 · y [n −1] and −b2 · y [n−2] terms are summed up.

(2) Operation Next, operations of the word length assignment process executed by the control unit 10 and the filter operation timer interrupt process executed by the low-pass filter 22 in the second embodiment having the above-described configuration will be described.

(A) Operation of Word Length Allocation Processing The control unit 10 executes the word length allocation processing shown in FIG. 9 every time the musical tone parameter supplied from the CPU 5 is updated. When this process is executed, the control unit 10 advances the process to step SA1 in FIG. 9, and stores the initial value “1” in the pointer i. The pointer i designates the detection result ENV [i] of the envelope detector 21 included in the filter unit 12-i. Next, in step SA2, it is determined whether or not the detection result ENV [i] corresponding to the pointer i is equal to or less than a predetermined reference value. If the detection result ENV [i] is less than or equal to the reference value, the determination result is “YES”, the process proceeds to step SA3, “0” is set in the register WL [i], and the low pass is performed so as to perform the 8-bit filter operation. The filter 22 is instructed.

  On the other hand, if the detection result ENV [i] exceeds the reference value, the determination result is “NO”, the process proceeds to step SA4, “1” is set in the register WL [i], and a 16-bit filter operation is performed. The low-pass filter 22 is instructed to do so. In step SA5, the pointer i is incremented and incremented, and in the subsequent step SA6, whether or not the incremented pointer i exceeds “3”, that is, each of the filter units 12-1 to 12-3. It is determined whether or not the low-pass filter 22 has been instructed about the number of calculation bits for the filter calculation. If the instruction has not been completed, the determination result is “NO”, and the above steps SA2 and after are repeated. When the instruction on the number of calculation bits is completed, the determination result is “YES”, and this processing is completed.

  Thus, every time the musical tone parameter supplied from the CPU 5 is updated, the control unit 10 determines whether or not the envelope level of each frequency band exceeds a predetermined reference value, and the envelope level is below the reference value. If there is, the low-pass filter 22 that performs low-pass filtering in the frequency band is instructed to perform 8-bit arithmetic. On the other hand, if the envelope level exceeds the reference value, the low-pass filter 22 that performs low-pass filtering in the frequency band is instructed to perform 16-bit arithmetic. It is supposed to be.

(B) Filter calculation timer interrupt process The low-pass filter 22 interrupts and executes this process at regular intervals. When the execution timing comes, the processing logic shown in FIG. Based on the filter operation, that is, steps SB1 to SB16 are executed.
First, in step SB1, “1” is set to the filter number m for identifying the filter units 12-1 to 12-3, and the addition value b is reset to zero. Next, in step SB2, “1” is set to the item number s. The term number s designates the term of the filter operation expression represented by the above-described equation (1). Specifically, the term number s = 1 is a0 · x [n], the term number s = 2 is a1 · x [n−1], and the term number s = 3 is a2 · x [n-2]. The term number s = 4 designates -b1 · y [n-1], and the term number s = 5 designates -b2 · y [n-2].

  In step SB3, the signal value of the term number s is stored in the register a and the filter coefficient of the term number s is stored in the register c. In the subsequent step SB4, “1” is set in the product-sum operation term k and the register DL1 Reset to zero. The product-sum operation term k is ah × ch × 10000h (k = 1), ah × cl × 100h (k = 2), al × ch × 100h (k = 3) and al × cl in the processing logic of FIG. Specify (k = 4). The register DL1 holds the value of the latch 45 in the processing logic illustrated in FIG.

Next, in step SB5, processing according to the value of the product-sum operation term k is executed. That is, when the product-sum operation term k is “1”, that is, in order to execute ah × ch × 10000h in the processing logic of FIG. 8, the process proceeds to step SB6 and the upper 8 bits of the register a (signal value) The ah is stored in the register DS1, the upper 8 bits ch of the register c (filter coefficient) are stored in the register DS2, and “16” is set in the register LS that holds the bit shift value.
When the product-sum operation term k is “2”, that is, to execute ah × cl × 100h in the processing logic of FIG. 8, the process proceeds to step SB7, and the upper 8 bits ah of the register a (signal value) are set. The lower 8 bits cl of the register c (filter coefficient) are stored in the register DS1, respectively, and “8” is set in the register LS holding the bit shift value.

When the product-sum operation term k is “3”, that is, to execute al × ch × 100h in the processing logic of FIG. 8, the process proceeds to step SB8 and the lower 8 bits al of the register a (signal value) are set. The upper 8 bits ch of the register c (filter coefficient) are stored in the register DS1 in the register DS2, and “8” is set in the register LS that holds the bit shift value.
When the product-sum operation term k is “4”, that is, to execute al × cl in the processing logic of FIG. 8, the process proceeds to step SB8, and the lower 8 bits al of the register a (signal value) are stored in the register DS1. In addition, the lower 8 bits cl of the register c (filter coefficient) are stored in the register DS2, respectively, and the register LS holding the bit shift value is reset to zero.

When the data corresponding to the value of the product-sum operation term k is thus set in the registers DS1, DS2, and LS, the process proceeds to step SB10. In step SB10, the result of multiplying the value of the register DS1 and the value of the register DS2 is shifted by the upper bits by the value of the register LS, and the value is added to the registers DL1 and DL2. In step SB11, the product-sum operation term k is incremented and stepped.
Next, in step SB12, it is determined whether the value of the register WL [m] corresponding to the filter number m is “1” and the product-sum operation term k is smaller than “5”. That is, a 16-bit operation instruction is received from the control unit 10 and it is determined whether or not the product-sum operation is in progress.

If it is in the middle of the product-sum operation according to the 16-bit operation instruction, the determination result is “YES”, and the process returns to step SB5. Thereafter, steps SB5 to SB12 are repeated until the product-sum operation of the product-sum operation term k = 4 is completed. When all the product-sum operations are completed and the product-sum operation term k becomes “5”, the determination result in step SB12 is “NO”, and the process proceeds to step SB13. On the other hand, if an 8-bit calculation instruction is received, the determination result in step SB12 is “NO”, and the flow advances to step SB13.
In step SB13, the result of adding the addition value b to the register DL2 is set in the register e, and then the value of the register e is reset to the addition value b. In step SB13, the item number s is incremented.

  Subsequently, in step SB14, it is determined whether or not the incremented item number s exceeds “5”, that is, whether or not the filter operation has been completed. If the filter operation has not been completed, the determination result is “NO”, and the process returns to step SB3 described above. Thereafter, steps SB3 to SB13 are repeated until the filter operation is completed. When the filter calculation is performed for all terms from a0 · x [n] to −b2 · y [n−2], the determination result in step SB14 is “YES”, and the flow proceeds to step SB15. In step SB15, the filter number m is incremented, and in the subsequent step SB16, whether or not the incremented filter number m exceeds “3”, that is, whether all the filter units 12-1 to 12-3 have performed the filter operation. Judge whether. If it is halfway, the determination result is “NO”, and the process returns to step SB2. On the other hand, when the filter operation is finished in all the filter units 12-1 to 12-3, the determination result is “YES”, and this processing is completed.

As described above, the filter device according to the second embodiment classifies the input waveform supplied from the waveform generator 11 (not shown) into the first to third frequency bands, and the input waveform for each frequency band. An envelope level is detected, and if the detected envelope level exceeds a preset reference value, a 16-bit filter operation is performed, and if it is less than the reference value, an 8-bit filter operation is performed.
In other words, high-accuracy filtering is performed in the frequency band where the waveform level is large and easily audible, and low-accuracy filtering is performed in the frequency band where the waveform level is small and not audibly conspicuous, thereby increasing the circuit scale. As a result, it is possible to improve the sound quality on hearing without incurring a decrease in processing speed.

C. In the second embodiment described above, the number of calculation bits for the filter operation is specified according to the envelope level of the input waveform. Instead, for example, as shown in FIG. The control unit 10 is provided with a word length allocation table TBL for reading out the number of operation bits WL [n, p, 1] to [n, p, 3] to be given to the filter units 12-1 to 12-3 according to the height p. In accordance with the musical tone parameter supplied from the CPU 5, the control unit 10 reads out the corresponding operation bit number WL [n, p, 1] to [n, p, 3] from the word length allocation table TBL, and the filter unit 12- It is good also as an aspect supplied to 1-12-3.

Specifically, the control unit 10 interrupts and executes the word length allocation process shown in FIG. 12 at regular intervals. When the execution timing is reached, the process proceeds to step SC1, and it is determined whether or not the musical tone parameter supplied from the CPU 5 indicates a change in waveform or pitch. If the change of the waveform or pitch is not instructed, the determination result is “NO”, and this process is completed without doing anything.
On the other hand, if a change in waveform or pitch is instructed, the determination result is “YES”, the process proceeds to step SC2, and the number of calculation bits WL [n, n corresponding to the changed waveform type n or pitch p is reached. p, 1] to [n, p, 3] are read from the word length assignment table TBL, and each is sent to the filter units 12-1 to 12-3 to complete the process.
According to such a modified example, it is possible to immediately instruct the filtering of the optimum number of calculation bits corresponding to the change in the waveform or the pitch, and the configuration for detecting the envelope can be omitted, so that the circuit scale increases and the processing speed decreases. It is possible to improve the audible sound quality without inviting sound.

1 Keyboard 2 Switch 3 ROM
4 RAM
5 CPU
6 Sound source 7 D / A converter 8 Sound system 9 Speaker 10 Control unit 11 Waveform generator 12 Filter unit 13 Coefficient table 14 Amplitude interpolator 14a Interpolation coefficient generator 14b Ring buffer 14c, 14d Multiplier 14e Adder

Claims (4)

Detecting means for dividing an input waveform into a plurality of frequency bands and detecting an envelope level for each divided frequency band;
The filter operation of the first calculation bit number is instructed for the frequency band in which the envelope level detected by the detection means exceeds the predetermined value, and the filter operation of the second operation bit number is instructed for the frequency band below the predetermined value. Indicating means;
Filter means provided corresponding to the plurality of frequency bands, and applying a filter operation of any of the first and second calculation bit numbers indicated by the instruction means to an input waveform of the corresponding frequency band. A filter device characterized by: A band dividing means for dividing the input waveform into a plurality of frequency bands;
Storage means for storing in advance the number of calculation bits associated with the waveform type and pitch of the input waveform for each of the plurality of frequency bands;
In accordance with a musical tone parameter indicating the waveform type and pitch of the input waveform, an assigning unit that reads out the corresponding number of calculation bits from the storage unit and assigns it to each frequency band;
And a filter unit that is provided corresponding to the plurality of frequency bands and that performs a filter operation of the number of operation bits assigned by the assigning unit on an input waveform in the corresponding frequency band. A detection step of dividing an input waveform into a plurality of frequency bands and detecting an envelope level for each of the divided frequency bands;
The filter operation of the first calculation bit number is instructed for the frequency band in which the envelope level detected in the detection step exceeds the predetermined value, and the filter operation of the second calculation bit number is instructed for the frequency band below the predetermined value. An instruction step to
A filtering step that is executed corresponding to the plurality of frequency bands, and that performs a filter operation of any one of the first and second calculation bit numbers indicated in the indicating step on an input waveform in the corresponding frequency band; A filter processing program that is executed by the program. A band dividing step for dividing the input waveform into a plurality of frequency bands;
For each of the plurality of frequency bands, the number of calculation bits associated with the waveform type and pitch of the input waveform is stored in advance, and the number of calculation bits corresponding to the musical sound parameter representing the waveform type and pitch of the input waveform is stored. An assignment step of reading out the number of operation bits and assigning it to each frequency band;
A filtering step that is executed in correspondence with the plurality of frequency bands and that performs a filtering operation of the number of calculation bits assigned in the assignment step on an input waveform in the corresponding frequency band, Filter processing program.

Images