JPH0691417B2 - Digital filter device - Google Patents

Description

Description: TECHNICAL FIELD OF THE INVENTION The present invention relates to a digital filter device for digitally filtering.

[Prior Art and its Problems] When a digital filter is applied to an electronic musical instrument or the like, it is desired that the characteristics of the filter can be dynamically changed. The cutoff frequency is a parameter indicating the boundary between the pass band and the cutoff region of the filter, and the tone quality and timbre can be changed by changing the cutoff frequency. Although it is possible to directly calculate each coefficient (filter coefficient) of the transfer function of the digital filter from the cutoff frequency, it is desirable to calculate a trigonometric function, etc. It is too heavy for the control unit (CPU) that has to perform other processing. Therefore, there is a method of preparing a memory (filter coefficient conversion table) for storing the filter coefficient for each value of the cutoff frequency, and accessing this conversion memory with the cutoff frequency instruction value to obtain the corresponding filter coefficient. Adopted. However, there is a problem that the storage capacity of the conversion memory becomes significantly large in order to store the filter coefficients for all the cutoff frequency instruction values.

Further, when the digital filter is applied to the tone processing, it is desirable that not only the cutoff frequency in the filter characteristic but also the resonance index can be variably controlled. However, if a conversion memory is used to obtain the filter coefficient from the variable cutoff frequency indication value and the resonance index indication value, its storage capacity is (total number of possible cutoff frequency indication values) x (possible resonance frequency). The size becomes proportional to the total number of exponents), and the storage capacity becomes unacceptable. In order to solve this problem, a method has been proposed in which not all required filter coefficients are stored in advance, but interpolation calculation is performed using the filter coefficients stored corresponding to the designated cutoff frequency.

However, performing such an interpolation operation lengthens the filtering processing time accordingly, and there is a risk that the change in the filter characteristics will not be followed even if the cutoff frequency is continuously and quickly changed.

OBJECT OF THE INVENTION Therefore, an object of the present invention is to save storage capacity while
It is an object of the present invention to provide a digital filter device capable of calculating a corresponding filter coefficient at high speed from an instruction value of a variable cutoff frequency and an instruction value of a variable resonance index.

[Object and Action of the Invention] In order to achieve the above object, a cutoff frequency instructing means for variably instructing a cutoff frequency, a resonance index instructing means for variably instructing a resonance index, and the cutoff frequency instructing means. A conversion means for converting the value of the cut-off frequency specified in 1 to the corresponding multi-bit digital data, and a filter characteristic having the maximum resonance index for each of a limited number of discrete cut-off frequencies. And a second filter coefficient storing means for storing a filter coefficient giving a filter characteristic having a minimum resonance index for each of a limited number of discrete cutoff frequencies. The storage means and the value of a predetermined number of bits from the upper side of the plurality of bits output from the conversion means are used as address data Then, reading means for respectively reading the filter coefficient having the maximum resonance index and the filter coefficient having the minimum resonance index from the first and second filter coefficient storage means, and a plurality of bits of digital data output from the converting means are read. Detecting means for detecting whether or not the value of the remaining bits on the lower side is zero, and when zero is detected by this detecting means, the filter coefficient having the maximum resonance index read by the reading means is outputted. If zero is not detected, a value obtained by adding 1 to the value of a predetermined number of bits from the upper side of the plurality of bits output from the conversion means is used as address data, and the filter coefficient read from the first filter coefficient storage means, Based on the filter coefficient read by the reading means and the value of the remaining bits on the lower side First interpolating means for interpolating to create a filter coefficient having the maximum resonance index, and when zero is detected by this detecting means, the filter coefficient having the minimum resonance index read by the reading means is output. At the same time, if zero is not detected, the filter coefficient read from the second filter coefficient storage means as the address data is a value obtained by adding 1 to the value of a predetermined number of bits from the upper side of the plurality of bits output from the conversion means. Second interpolation means for interpolating on the basis of the filter coefficient read by the reading means and the value of the remaining bits on the lower side to create a filter coefficient having a minimum resonance index; and the first interpolation means. From the second interpolation means and the output corresponding to the filter coefficient having the maximum resonance index from A third interpolation means for interpolating a filter coefficient corresponding to the resonance index designated by the resonance index designating means with reference to the output corresponding to the filter coefficient, and an input signal according to the filter coefficient from the third interpolation means. And a filtering means for filtering.

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

FIG. 1 is a block diagram showing the function of the embodiment. The first coefficient generating circuit 1 and the second coefficient generating circuit 2 in FIG.
In, a filter coefficient for the indicated cutoff frequency is generated for the minimum resonance index and the maximum resonance index. The configurations of both circuits are the same as shown in FIG. 2, but the contents of the coefficient ROM 24 to be referred to are different. That is, in the coefficient ROM referred to by the first coefficient generation circuit 1, the filter coefficient for the minimum value of the resonance index designated by the resonance volume 3 is stored for each discrete cutoff frequency, and the second coefficient is stored. The coefficient ROM referred to by the generation circuit 1 stores the filter coefficient for the maximum value of the resonance index designated by the resonance volume 3 for each discrete cutoff frequency. Each of the coefficient generating circuits 1 and 2 has an interpolation function of interpolating the filter coefficient corresponding to the cutoff frequency given from the volume 18 through the A / D converter 19 by referring to each coefficient ROM. Therefore, the upper bit (address integer part) of the output (cutoff frequency indicating value) of the A / D converter 19 is extracted by the upper bit extraction circuit 20, and the lower bit (address decimal part) is extracted by the lower bit extraction circuit 21. It is extracted with. The output of the high-order bit extraction circuit 20 is incremented by 1 by the plus 1 addition circuit 22. Output of high-order bit extraction circuit 20 and plus 1 addition circuit 22
Is output to the switching circuit 23. The switching circuit 23 multiplexes both inputs to address the coefficient ROM 24,
The output of the coefficient ROM 24 is stored in the first register 25 and the second register 26. In detail, the switching circuit 23 is the upper bit extracting circuit 20.
Output (the cut-off frequency target value integer part) coefficient ROM2
When 4 is addressed, the filter coefficient data corresponding to the cutoff frequency target value integer part output from the coefficient ROM 24 is loaded into the first register 25, and when the coefficient ROM is addressed by the output of the plus 1 adder circuit 22, The filter coefficient data that gives the cutoff frequency corresponding to the cutoff frequency target value integer part plus 1 output from the ROM 24 is loaded into the second register 26. The output of the first register 25 and the output of the second register 26 are input to the cutoff interpolation circuit 27. Further, the fractional part data of the cutoff frequency target value from the lower bit extraction circuit 21 is also input to the cutoff interpolation circuit 27, where the filter coefficient for the cutoff frequency target value is linearly interpolated using these inputs. It The output of the cutoff interpolation circuit 27 is selected by the switching circuit 29 when the zero detection circuit 28 detects that the cutoff frequency target value includes a fractional part. 1 register 25
Is selected by the switching circuit 29. The arithmetic circuit 30 calculates the remaining filter coefficients (indicated by K) from the outputs of the switching circuit 29 (indicated by the filter coefficients b1 and b2).

In this way, the first coefficient generating circuit 1 outputs the filter coefficients (Fb1, Fb2, Fb1, Fb2,
FK) is output, and the filter coefficient (Qb1, Qb in FIG. 1) relating to the maximum resonance index and the cutoff frequency specified by the volume 18 is output from the second coefficient generating circuit 2.
b2, indicated by QK) is output and input to the interpolation circuit 5 for resonance correction. Furthermore, the resonance index indicating signal from the resonance volume 3 which indicates the resonance index is also A / D.
It is input to the interpolation circuit 5 through the converter 4. Interpolation circuit 5
Then, using these inputs, the volume 3 for resonance
The filter coefficient that gives the resonance index indicated by is calculated by linear interpolation. The output of the interpolation circuit 5 and the first,
The outputs of the second coefficient generation circuits 1 and 2 are input to the switching circuit 8. The switching circuit 8 is for selecting the filter coefficients b1, b2, K to be supplied to the digital filter 9, and
By the maximum value detection circuit 6 coupled to the output of the D converter 4,
When it is detected that the indicated value of the resonance index is the maximum value, the outputs Qb1, Qb2, QK of the second coefficient generation circuit 2 are selected as the filter coefficients b1, b2, K to be sent to the digital filter 9, and A When the minimum value detection circuit 7 coupled to the output of the / D converter 4 detects that the indicated value of the resonance index is the minimum value, the output Fb of the first coefficient generation circuit 2
1, Fb2, FK are selected, and in other cases, the interpolation output 5 from the interpolation circuit 5 is selected as the filter coefficients b1, b2, K to be transferred to the digital filter 9. The digital filter 9 filters the signal to be processed based on the sent filter coefficients b1, b2, K.

As described above, in the present embodiment, the interpolation of the cutoff frequency and the interpolation of the resonance index are combined to effectively use the data of the coefficient ROM, and the storage capacity thereof is significantly reduced. Note that, in FIG. 1, the elements 6, 7, and 8 are not always necessary, and when the resonance index is maximum (or minimum) in the interpolation circuit 5, the second coefficient circuit output (or the first coefficient) is output. The coefficient b1, b2, K to be constantly supplied to the digital filter 9 can be obtained by the interpolation output by performing the processing in which the ratio of the circuit output) to the interpolation output is 100%.

Specific examples will be described below. FIG. 3 is an overall hardware block diagram of an embodiment applied to an electronic musical instrument. The CPU 101 controls the entire apparatus. The program executed by the CPU 101 is stored in the program ROM 102. Coefficients ROM 103 and 104 relate to different resonance indices. Filter coefficients for a plurality of discrete cutoff frequencies are stored respectively. The RAM 105 is used as a work memory for the CPU 101. The keyboard 106 is provided with musical instrument performance inputs. The volume 107 provides an input for variably indicating the cutoff frequency of the digital filter, and the volume 108 provides an input for variably indicating the resonance index of the digital filter. A tone signal is formed in the sound source 109. The formed musical tone signal is a digital filter (DF)
The sound is filtered by 110 and emitted through the audio device 111 and the speaker 112.

FIG. 4 shows logical information when a second-order IIR filter is used as an example of the digital filter 110. 201-205 in the figure
Is a coefficient multiplier, 206 to 209 are delay elements of 1 sample, and 210 is an adder. Among the coefficient multipliers, the coefficient K of the coefficient multiplier 201 that multiplies the input signal input, the coefficient b1 of the first-order feedback multiplier 202 and the coefficient b2 of the second-order feedback multiplier 203 depend on the cutoff frequency and the resonance index. Has a value that In particular, as shown in Figure 4, coefficients b1, bd, K is ^{b1 = 2 (A 2 -1)} / (A 2 + 2aA + 1) b2 = (A 2 -2aA + 1) / (A 2 + 2aA + 1) K = (1 + b1 + b2) / 4, and a = cosRs A = tan (fcπ / fs) fc = cutoff frequency fs = sampling frequency Rs = resonance index.

Further, the transfer function H (z) showing the relationship between the input signal input and the output signal output of the filter is Given in.

For reference, some different resonance indices are shown in Fig. 5.
4 shows frequency characteristics of a digital filter having Rs. this
It takes a considerable amount of time to directly calculate the coefficients b1, b2, K from Rs, Fs, fc, which is difficult to realize in an application requiring real-time processing such as an electronic musical instrument. To avoid this,
A coefficient memory for storing the filter coefficient for each of the possible cutoff frequency value and resonance index value may be used, but this requires a huge storage capacity.
Therefore, in the present embodiment, only the cutoff frequencies designated by the volumes 107 and 108, the discrete resonance index spaced from the resolution width of the resonance index, and the filter coefficients for the cutoff frequencies are stored in the coefficient ROMs 103 and 104. However, filter coefficients other than discrete points (sample points) are obtained by interpolation. Further, the coefficient K is not prepared as stored data because it can be easily calculated from the coefficients b1 and b2 according to the above equation.

FIG. 6 shows an example of a memory map in the case of storing the filter coefficient for each cutoff frequency with respect to one specific resonance index. Here, there are nine memory sampling points of the cutoff frequency fc. (A-b1)
Indicates the start address of the coefficient b1 memory, and (A-b2) indicates the start address of the coefficient b2 memory. The number of bits of the A / D converter (not shown in FIG. 3) for A / D converting the output of the volume 107 is set to 8 bits, and the vertical number of the cutoff frequency that can be indicated is set.
If the cutoff frequency instruction value is 128 and the cutoff frequency instruction value is 0 to 127, filter coefficients are stored in adjacent addresses of the coefficient ROM 103 at intervals of 16 times the resolution of the indicated value (in the case of 9-point sample). In Fig. 6, the number attached to each coefficient indicates the cutoff frequency sample position in hexadecimal,
For example, b10H represents a coefficient for the cutoff frequency instruction value of 10H. Therefore, volume 107 to A
The upper 4 bits of the 8-bit cutoff frequency instruction value given through the / D converter indicate the relative address of each coefficient memory as an address integer part, and the lower 4 bits indicate the address decimal part.

The filter coefficient for a variable resonance index cannot be interpolated with only the coefficient ROM having only one kind of resonance index. For this purpose, at least two types of coefficient ROMs for different resonance indices are required. An example of a memory map for this purpose is shown in FIG. In FIG. 7, Fb1 represents a memory that stores the filter coefficient b1 for the first (minimum) resonance index for each cutoff frequency, and Fb2 represents a memory that stores the filter coefficient b2 for the minimum resonance index for each cutoff frequency. Further, Qb1 and Qb2 represent memories for respectively storing the filter coefficients b1 and b2 for the second (maximum) resonance index. For example, π / 4 can be used as the value of the first resonance index and 37π / 90 can be used as the value of the second resonance index. The value of the first resonance index represents the maximum value of resonance that can be indicated by the volume 108 for resonance, and the value of the second resonance index represents the maximum value of resonance that can be indicated by the volume 108 for resonance. Now, the output of the volume 108 for resonance index (output after digital) is set to 5 effective bits, and the value is 0.
Assuming that the range is -16 (decimal), 17 levels of interpolation can be performed by the resonance interpolation process described later. FIG. 8 is a table showing the values of the resonance index (expressed in dB here) which means the 5-bit output of the volume 108. For example, output 00000 (binary) of volume 108 is -3dB
Resonance gain of, volume output 10000 is +5.0
Resonance gain in dB. In this case, the resolution of the volume resonance instruction is 0.5 dB. As shown in Fig. 5, when Rs = π / 4, the relative gain (resonance gain) with respect to the filter pass band at the cutoff frequency fc is about -3 dB, and when Rs = 37π / 90, the resonance gain is about +5 dB. Become.

9 to 11 show loci with respect to changes in the cutoff frequency of the poles of the filters having different resonance indices. Fig. 9 shows register Rs = 45 degrees, Fig. 10 shows resonance index Rs =
74 degrees and FIG. 11 are figures for the resonance index Rs = 1 degree. In these figures, the curves A1, A2, A3 are ideal values obtained when the coefficients b1, b2, K are directly obtained by the above equations. It is the trajectory of the pole. On the other hand, the curves B1, B2, and B3 are the pole interpolation loci obtained by the filter coefficients obtained by interpolating the cutoff frequency between the sample points with the numerical values shown on the left of each figure as sample points of the cutoff frequency ratio fc / fs. . Further, also FIG. 12, resonance index Rs = 45 degrees, the ideal pole with the trajectory A of the time, showing the interpolation trajectory B when a sample point of a ₈ through i _8. FIG. 13 shows a filter frequency characteristic corresponding to the interpolation locus B.

Cutoff frequency indicated by volume 107 and volume 1
FIG. 14 shows a flow of processing executed by the CPU 101 to obtain the filter coefficient corresponding to the resonance index designated by 08.

First, in 14-1, the filter coefficient that gives the cutoff frequency designated by the cutoff volume 108 among the filter coefficients related to the minimum resonance index is stored in the memory Fb of FIG.
1 and Fb2 are interpolated, and at 14-2, the filter coefficient giving the cutoff frequency indicated by the cutoff volume 108 among the filter coefficients relating to the maximum resonance index is stored in the memories Qb1 and Qb2 of FIG. Refer to and interpolate. 14-
The details of the processes 1 and 14-2 are shown in FIG. Here, the cutoff frequency instruction value of the volume 107 (resonance Rsg0
(Stored in) is assumed to take a value of 0 to 127 (decimal) in 8 bits, and the number of sample points of the cutoff frequency of the coefficient ROM is 9, and the cutoff frequency between sample points is It is assumed that the interval is 16 times the resolution of the cutoff frequency indication value. Therefore, the upper 4 bits of the cutoff frequency instruction value determine the relative value (integer part) of the address to be accessed to the coefficient ROM, and the lower 4 bits determine the fractional part of the cutoff frequency instruction value. .

First, in 10-1, the cutoff frequency instruction value (8 bits) in the register Reg0 is copied to the register Reg1. Then 10
At -2, the contents of register Reg1 are shifted to the right by 4 bits. As a result, for example, the cutoff frequency indication value is 48
When H, 04H is obtained. The numerical value of this shift result represents the integer part of the cutoff frequency instruction value, and indicates the relative address of the coefficient ROM 103 with respect to each coefficient memory. So b1
Start address of coefficient memory A-b1 (A14-A when A-1
In the case of Fb1 and 14-2, the coefficient b1 (Reg1) at the address indicated by A-Qb1) + REG1 and the start address A-b2 (14-1 of the b2 coefficient memory, A-Fb2, 14-2 , The coefficient b2 at the address indicated by A-Qb2) + REG1
Read (Reg1) (10-3). Where the coefficient b1 (Reg
1) The coefficient (REG1) has the values of coefficients b1 and b2 that give the cutoff frequency indicated by the integer part of the cutoff frequency instruction value. Next, the logical product of Reg0 and 0FH is calculated to check whether or not there is a fractional part of the cutoff frequency instruction value (10-4). When the fractional part is zero, the cutoff frequency instruction value is an integer value, so the coefficient values b1 (Reg1) and b2 (Reg2) read in 10-3 are set to the b1 and b2 registers (coefficient b1 for the cutoff frequency instruction value, respectively). , B2) to store). When there is a fractional part, each coefficient b1 (Reg is at the next address of the coefficient memory address indicated by the integer part of the cutoff frequency instruction value.
+1) and b2 (Reg + 1) are read (10-6). Then, Δb1 = b1 (Reg1 + 1) −b1 (Reg1), Δ2b = b2 (Re
g1 + 1) -b2 (Reg1) finds the differences Δb1 and Δb2 between adjacent filter coefficients referenced from each coefficient memory, and the differences Δb1 and Δb2 and the fractional part of the cutoff frequency instruction value (Reg0 ∧
0FH), b1 = b1 (Reg1) + (△ b1 × (Reg0 ∧ 0F
H) / 16) b2 = b2 (Reg1) + (△ b2 × (Reg0∧0FH) / 16)
Thus, the coefficients b1 and b2 that give the cutoff frequency indicated by the cutoff frequency instruction value Reg0 are linearly interpolated. Finally, using the coefficients b1 and b2, the remaining coefficient K is calculated by K = (1 + b1 + b2) / 4 (10-8). Even if the process is performed from 10-3 to 10-6 → 10-7 → 10-8 without checking 10-4, the result is the same.

In process 14-1, the coefficients b1, b2, K thus obtained are stored in the Fb1, Fb2, FK registers in the final process (not shown), and in process 14-2, the coefficient b1 is stored in the final process. , B2, K
Are stored in the Qb1, Qb2, and QK registers. With the above, interpolation processing regarding the cutoff frequency is completed, and Fb1, Fb2,
The FK register contains the filter coefficient corresponding to the cutoff frequency instruction value for the minimum resonance index (−3 dB), and the Qb1, Qb2, and QK registers contain the cutoff frequency instruction value for the maximum resonance index (+5 dB). Contains the corresponding filter coefficients.

Therefore, the interpolation processing regarding the resonance index is performed by the following processing. First, volume 108 in the Q register
By taking the logical product of the resonance index indicated value from and the 10H, the indicated value Q is the maximum value of resonance +5 dB (8th
(14-3). If so, the filter coefficients for the resonance index are stored in the Qb1, Qb2, and QK registers, and a copy of the contents is stored in the b1, b2, and K registers (14-4). If not established, proceed to 14-5 and set the resonance index instruction value Q.
And 0FH are ANDed to check whether the indicated value Q shows the minimum resonance index (−3 dB). If it is satisfied, the necessary filter coefficients are stored in the Fb1, Fb2, and FK registers. , Copy the contents to b1, b2, K registers (14-6). If the indicated resonance index is neither maximum nor minimum, proceed to 14-7, where Δb1 =
From Qb-1-Fb1, △ b2 = Qb2-Fb2, △ K = QK-FK,
Differences Δb1, Δb2, and ΔK between the filter coefficient related to the maximum resonance index and the filter coefficient related to the minimum resonance index after the cutoff frequency interpolation are calculated. And 14-8
Then, using the resonance index value Q and the filter coefficients Fb1, Fb2, and FK designated as the differences Δb1, Δb2, and ΔK, b1 =
Fb1 + (△ b1 × Q / 16), b2 = Fb2 + (△ b2 × Q / 16), K =
Interpolate the filter coefficient values according to the indicated resonance index according to FK + (ΔK × Q / 16). The filter coefficients b1, b2, K obtained in this way have filter coefficient values that give the cutoff frequency specified by the volume 107 and the resonance index specified by the volume 108.
Therefore, while saving the storage capacity of the coefficient ROMs 102 and 103, it is possible to generate filter coefficients corresponding to variable cutoff frequency instruction input and resonance index instruction input from the volumes 107 and 108 on a real-time basis.

Although the description of the embodiment has been completed, various modifications and changes can be made within the scope of the present invention. For example, three or more types of filter coefficients related to discrete resonance indices may be stored in the coefficient ROM. In this case, cutoff frequency interpolation processing is performed according to the upper bits of the resonance index instruction value.
Select the memory to be referred to by 14-1 and 14-2 (decide which memory to store the filter coefficient related to the resonance index is determined by the high-order bit of the resonance index instruction value)
By doing so, more efficient processing is possible. If the data stored in the filter coefficient ROM is not the coefficient value itself but the difference between the coefficient values, the storage capacity can be further saved. Further, the digital filter 110 is not limited to the second-order IIR filter, and any suitable order and type of digital filter can be used. Further, instead of manual operators such as the volumes 107 and 108, a function whose value changes with time and a function (waveform) generator which automatically generates can be used.

[Effects of the Invention] As described in detail above, the filter coefficient storage means stores the filter coefficients for each of the discrete cutoff frequency and the discrete resonance index, and then the cutoff frequency instruction given is given. In order to obtain a filter coefficient for the value and the resonance index instruction value, the filter coefficient interpolated for the cutoff frequency instruction value is referred to the above filter coefficient storage means, and the maximum and minimum resonance index 2 is obtained.
Since the filter coefficients are obtained for each set and the filter coefficients for the designated resonance index are interpolated using these two sets of filter coefficients, the storage capacity of the device can be significantly reduced. Moreover, the present invention converts the value of the specified cutoff frequency into a plurality of bits of digital data, and reads out the filter coefficient based on the value of a predetermined number of bits on the upper side, and the remaining bits on the lower side are read. When the value of is zero, the read filter coefficient is output as it is. If the value of the lower bit is not zero, the read filter coefficient, the filter coefficient read based on the value obtained by adding 1 to the upper predetermined bits, and the value of the remaining lower bit Since the filter coefficient is interpolated and created based on, it is not necessary to perform the interpolation calculation when the filter coefficient used is stored in the memory. Since it is only necessary to detect the value of the lower bit of the corresponding digital data, the filtering process time is shortened accordingly, and high-speed filter control that follows variable cutoff frequency and resonance index instruction input is possible. Become.

[Brief description of drawings]

FIG. 1 is a block diagram showing the function of an embodiment of the present invention, FIG. 2 is a functional block diagram of the first (second) coefficient generating circuit of FIG. 1, and FIG. 3 is an embodiment applied to an electronic musical instrument. Fig. 4 is a hardware block diagram of the
Next block diagram of IIR filter, FIG. 5 is a diagram showing frequency characteristics of the filter with respect to different resonance indices, FIG. 6 is a diagram showing an example of a memory map of a coefficient ROM, and FIG. 7 is a coefficient ROM giving two types of resonance indices. Fig. 8 shows an example of the memory map of Fig. 8, Fig. 8 shows the correspondence between resonance index input data and resonance index, and Fig. 9, Fig. 10, Fig. 11, Fig. 12 show the ideal pole loci of the filter. Fig. 13 shows the pole locus by cutoff frequency interpolation, Fig. 13 shows the frequency characteristics of the filter corresponding to the interpolated pole locus of Fig. 12, and Figs. 14 and 15 show the CPU of Fig. 1. 5 is a flowchart of a filter coefficient interpolation process executed by 1 ... First coefficient generating circuit, 2 ... Second coefficient generating circuit, 3 ... Resonance volume, 5 ... Resonance index interpolation circuit, 9 ... Digital filter, 18 ... Volume, 24 ... Coefficient ROM, 27 ... Cutoff interpolation circuit, 101 ...
… CPU, 102 …… Program ROM, 103, 104 …… Coefficient ROM,
107 …… Cutoff volume, 108 …… Resonance volume.

Claims (1)

[Claims] 1. A cutoff frequency instructing means for variably instructing a cutoff frequency, a resonance index instructing means for variably instructing a resonance index, and a value of the cutoff frequency instructed by the cutoff frequency instructing means. Conversion means for converting to corresponding plural bits of digital data, and a first filter coefficient for storing a filter coefficient giving a filter characteristic having a maximum resonance index for each of a limited number of discrete cutoff frequencies. A storage means, a second filter coefficient storage means for storing a filter coefficient that gives a filter characteristic having a minimum resonance index for each of a limited number of discrete cutoff frequencies, and output from the conversion means. The first and second filter coefficient memory stores the value of a predetermined number of bits from the upper side of the plurality of bits as address data. Read means for reading out the filter coefficient having the maximum resonance index and the filter coefficient having the minimum resonance index from the stage, respectively, and the value of the remaining bits on the lower side of the digital data of a plurality of bits output from the conversion means is zero. Detecting means for detecting whether or not, when zero is detected by this detecting means, the filter coefficient having the maximum resonance index read by the reading means is outputted, and when zero is not detected, the converting means is outputted from the converting means. A filter coefficient read from the first filter coefficient storage means as a value obtained by adding 1 to the value of a predetermined number of bits from the upper side of the plurality of output bits, and a filter coefficient read by the read means. And the value having the maximum resonance index by interpolation based on the value of the remaining bits on the lower side. A first interpolating means for creating a filter coefficient, and a filter coefficient having the minimum resonance index read by the reading means when the zero is detected by the detecting means, and the zero when the zero is not detected. A filter coefficient read from the second filter coefficient storage means as a value obtained by adding 1 to the value of a predetermined number of bits from the upper side of the plurality of bits output from the conversion means, and read by the read means. Second interpolation means for creating a filter coefficient having a minimum resonance index by performing interpolation based on the filter coefficient and the value of the remaining bits on the lower side, and a filter having a maximum resonance index from the first interpolation means. See the output corresponding to the coefficient and the output corresponding to the filter coefficient having the minimum resonance index from the second interpolation means. And a third interpolation means for interpolating a filter coefficient corresponding to the resonance index designated by the resonance index designating means, and a filtering means for filtering the input signal in accordance with the filter coefficient from the third interpolation means. A digital filter device.