JPH0252446B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to a digital graphic equalizer, and more particularly, to a graphic equalizer that outputs a digital audio signal such as a PCM audio signal by varying the level for each frequency band.

Conventional Technology Digital filters used in conventional digital graphic equalizers are generally designed with a focus on the cutoff frequency (center frequency), and are not designed with a focus on plateau characteristics in frequency characteristics. Ta.

Problems to be Solved by the Invention For the above reasons, in the conventional FIR digital filter, the band emphasis characteristic K _b (boost amount
L _b ) and band attenuation characteristic K _c (cut amount L _c ), ripples occur in the plateau characteristics above frequency f _H and below frequency f _L , and in both cases, echoes due to dispersion occur. Another problem was that the coloring, that is, the tone, changed unnaturally.

On the other hand, in order to reduce the ripple in the flat region above the frequency f _H of the low-frequency emphasis characteristic and the ripple in the flat region below the frequency f _L of the low-frequency attenuation characteristic, for example, the A technique using an IIR digital filter is known, but since it is an IIR digital filter, the phase is non-linear,
In addition, the high-frequency SN ratio deteriorates due to rounding errors,
Furthermore, there was a problem that oscillation was likely to occur.

Furthermore, conventional digital filters have had the problem that it is difficult to obtain desired band emphasis and attenuation characteristics.

Furthermore, when conventional equalizers set the equalizer characteristics variably, overflows and underflows occur during the filter multiplication process, especially in digital ones, and as a result, the input and output volume cannot be maintained constant, resulting in a loss of natural hearing. There was a problem that I could not get it. This was also the case with analog equalizers.

The present invention has less ripples in the flat area, so it is possible to obtain a natural tone, and also has a good signal-to-noise ratio, making it suitable for HiFi use, and with less coefficient data, desired band emphasis and attenuation characteristics can be obtained. Another object of the present invention is to provide a digital graphic equalizer that can maintain input and output volume constant when equalizer characteristics are varied.

Means for solving the problem In FIG. 1, the control device 9 and the DSP (digital signal processor) 4 are
This is an example of a means for changing a filter coefficient according to a designation from a filter, adding and normalizing coefficients for each band to obtain a new coefficient, and a means for changing a frequency characteristic using the new coefficient.

Function FIR digital band filters 19 ₁ to 19 _o (FIG. 4) virtualized for each of a plurality of bands (FIG. 4) Out of each of the plurality of coefficients 1 to l, the coefficients of the corresponding band filter corresponding to the specification are specified from the characteristic input section 15. Change accordingly,
Adding the coefficients of multiple bands 1 to n to multiple coefficients 1 to l ( _o 〓 ^k=1 h _ki (i=1 to l), the total sum M= _o 〓 ^k=1 h _k1 + _o 〓 ^{k =1} h _k2 +...+ _o 〓 ^k=1 h _kl is determined, M is normalized to a predetermined digital word length L, and the coefficients h ₁ to h _l are the same as the number l of each coefficient.
is obtained, and the frequency characteristics of the FIR digital filter 50 are changed based on the obtained coefficients.

Embodiment FIG. 1 shows a block diagram of an embodiment of the equalizer of the present invention. In the figure, 15 is a characteristic input section;
Reference numeral 6 denotes _a display unit _, as _shown in FIG _.
16 ₂ , . . . 16 _o , a λ display element group 16 is provided. 4 is a DSP, which has n-band equalizer characteristics set in the characteristic input section 15 as shown in FIG.
It performs digital filter calculations to realize K ₁ to K _o , and has internal coefficient memory 4 ₂ , program ROM 4 ₁ , etc.

A digital audio signal supplied to the input terminal 1 via an optical fiber (not shown) is converted into an electrical signal by the optical interface 2, and is converted into an electrical signal, for example, via a serial/
The signal is parallel-converted and supplied to the latch circuit 3, where it is latched. The output of latch circuit 3 will be described later.
The equalizer characteristics are varied in the DSP 4 according to the specifications of the characteristic input section 15, and the equalizer characteristics are supplied to the latch circuit 5, which is converted into an optical signal by parallel/serial conversion by the optical interface 6, and is taken out from the terminal 17 and sent to the optical fiber. is supplied to

Substantially, as shown in FIG. 4, the DSP 4 changes the respective coefficients of the n-band digital band filters 19 ₁ to 19 _o by rewriting the coefficient memory 4 ₂ to obtain desired equalizer characteristics K ₁ to _{K o} . For example, the digital bandpass filter shown in FIGS. 6A and 6B can be considered.

6A and 6B respectively show the hardware configuration (circuit diagram) of the digital bandpass filter and its schematic block system diagram. In FIG _. 6A, a low _- pass filter (low _- pass filter ₎ that _obtains the characteristics shown by _the solid line in FIG. filter) 22, 1 sample delay section 35 ₁ to 35 _l , multiplication section 36 ₁ to 36 _l ,
The adders 37 ₂ to 37 _l each constitute a high-pass filter (high-pass filter) 24 that obtains the characteristics shown by the solid line in FIG _{.
.about.34.sub.n} , and multiplication sections (level adjustment sections) 33 and 38 constitute a full range filter (substantially a delay circuit and an attenuator) 23.

Coefficients of multipliers 31 ₁ to 31 _l of low-pass filter 22
a ₁ to a _l are the function windows shown in FIG. 10A, and on the other hand,
Coefficient b ₁ of multiplier section 36 ₁ to 36 _l of high-pass filter 24
_.about.bl is a _{functional window} shown in FIG.

In this case, as shown in FIG. 7A, the low-pass filter (solid line) and the wide range filter (dotted chain line) are combined to obtain the characteristics indicated by the broken line, and as shown in FIG. 7B,
The high-pass filter (solid line) and the whole range filter (dotted chain line) are combined to obtain the characteristic shown by the broken line, and these are combined to obtain the band attenuation characteristic K _c shown in FIG.

On the other hand, when obtaining band emphasis characteristics, the coefficients a ₁ to a _l of the multipliers 31 ₁ to 31 _l of the low-pass filter 22 are variably set to obtain the characteristics shown in FIG. The coefficients b ₁ to b _l of the multipliers 36 ₁ to 36 _l of the filter 24 are variably set to obtain the characteristics shown in FIG. As in the case of obtaining the above-mentioned band attenuation characteristic, in this case, as shown in FIG. As shown in FIG. 3, the low-pass filter (solid line) and the wide range filter (dotted chain line) are combined to obtain the characteristic shown by the broken line, and these are combined to obtain the band emphasis characteristic K _b shown in FIG.

As described above, in the present invention, since the flat characteristic above the frequency f _H and the flat characteristic below the frequency f _L are each related to the whole range filter, there are fewer ripples in the flat area compared to the conventional one.

FIG _. 8 is a simplified version of the bandpass filter _shown in _FIG .
The coefficient b _i (b ₁ to b _l ) of 4 is added to the adder 40 (40 ₁ to 4
The coefficients _{h 1i (} h ₁₁ to _{h 1l} ) (Fig. 11) added in the multiplier 31 (31 1 to 31 l) are set as the coefficients of the multiplier 31 (31 ₁ to 31 _l ), and are substantially as shown in Fig. 6A. It is equivalent to a circuit diagram.

The band filters shown in FIGS. 6A and 8 are applied to the band filters 19 ₁ to 19 _o shown in FIG. 4, respectively, but the band filter shown in FIG. ₁ to 19 _o and adder 2
1 (indicated by 50), and in practice, the present invention uses a bandpass filter as shown in FIG. In the case shown in the fifth diagram, h ₁₁ to h _1l shown in the eighth diagram are coefficients of the first band, h ₂₁ to _{h 2l} are coefficients of the second band,
h ₃₁ ~ _{h 3l} is the coefficient of the third band, ..., h _o1 ~ h _ol is the nth band coefficient.
The coefficients of the bands are added to the adder units 41 ₁ , 41 ₂ ,..., 4
1 _l , calculate _o 〓 ^k=1 h _k1 , _o 〓 ^k=1 h _k2 ,..., _o 〓 ^k=1 h _kl , respectively, and further multiply by the coefficient λ in the multipliers 42 ₁ to 42 _l to obtain the respective results. By supplying the multipliers 31 ₁ to 31 ₂ , . . . 31 _l , the band filters 19 ₁ to 19 _o shown in FIG.
A result equivalent to calculating each of these individually can be obtained using a single band filter.

Next, switching of equalizer characteristics will be explained. DSP4 when switching equalizer characteristics
This switching operation is controlled by the CPU 10 in the control device 9.
The CPU 10 is configured to operate according to the flowchart shown in FIG. 9 based on control signals from the ROM 11 and RAM 12.

Switch position control data is input by operating the switch of a predetermined band in the characteristic input section 15 (step 101 in FIG. 9A), and a predetermined segment of the display element group corresponding to this band is displayed (step 102). ), filter coefficients (for example, h ₃₁ , h ₃₂ , . . . h _3l for the third band) corresponding to the specified equalizer characteristics are retrieved from the ROM 11 (step 103). When the filter coefficient is retrieved from the ROM 11, the normalizing unit 10 ₁
In (Fig. 1), the sum M of the filter coefficients, that is, M= _o 〓 ^k=1 h _k1 + _o 〓 ^k=1 h _k2 +...+ _o 〓 ^k=1 h _kl is determined (step 104), and then, When the digital word length is L, a coefficient of (2 ^L -1)/M=λ is obtained, and each filter coefficient set is multiplied by λ (multiplying units 41 ₁ to 42 _l in FIG. 5) to obtain λ _o 〓 ^k=1 h _k1 , λ _o 〓 ^k=1 h _k2 , . . . , λ _o 〓 ^k=1 h _kl , that is, the filter coefficients are normalized (step 105). In this case, λ _o 〓 ^k=1 h _ki (i=
It is necessary to perform the operations 1 to 1) using a number of bits that is sufficiently longer than the digital word length L to ensure sufficient accuracy.

By normalizing in this way,
Multipliers 31 ₁ to 31 _l play a central role in the system
(FIG. 5), word length L and coefficient word length (e.g., equal to word length L) can be used, and there is no need to use a particularly long bit length.
The cost of the system can be reduced, and since no overflow or underflow occurs, the input/output volume can be kept constant, and deterioration of the SN ratio due to overflow or level drop can be prevented.

This normalized filter coefficient is RAM
12, the coefficients λ are displayed on the λ display element group 16λ (step 106), and these are sequentially sent to the latch circuit 8 at interrupt timing (step 111 in FIG. 9B). At the same time, address data is supplied to the address memory 7, and addresses corresponding to each latch data are designated (step 112). At this time, the CPU 10 sequentially counts up from 1 to l (steps 113 to 115),
The coefficient λ is output (step 116). The coefficient λ is taken out from the terminal 18 via the latch circuit 13 and the optical interface 14, and is supplied to a volume control volume, etc., to return the volume to the value before normalization.

The DSP 4 is configured to operate according to the flowchart shown in FIGS. 12A and 12B, and after being initialized (step 121 in FIG. 12A), the previously performed calculation results are output at interrupt timing. is input (step 131), and the next sampling data from the latch circuit 8 is input (step 13).
2) The address specified in the address memory 7 is written into the coefficient memory ₄₂ (step 133).
When the filter coefficients are written to all addresses (step 134), the memory section ₄ of the coefficient memory 42
₂₁ , 4 ₂₂ , ..., 4 _2o is switched to the memory section of the inactive page (step 135 in FIG. 12), and thereby new coefficients h ₁ , h ₂ , ..., h _l are obtained, and a filter operation is performed for each sampling (step 136).
New equalizer characteristics can be obtained. In addition, the CPU
By switching the switch S according to the operation of 10, 2
It is also possible to switch to the memory section of the non-operating page among the two memory sections (for example, 4 ₂₁ and 4 ₂₂ ) and control the band filters 19 ₁ to 19 _o using one system.

In this way, each time the equalizer characteristic is specified in the characteristic input section 15, λ _o 〓 ^k=1 h _ki (i=1 to l) is calculated and new coefficients h ₁ , h ₂ ,..., h _l are calculated. obtained, and the equalizer characteristics are varied.

Effects of the Invention According to the digital graphic equalizer of the present invention, in particular, since the flat characteristic substantially involves a whole area filter, it is possible to reduce ripples in the flat area compared to the conventional one. of
It is possible to obtain a more natural tone than that using an FIR digital filter, and since it uses an FIR digital filter, it is possible to obtain a more natural tone than that using an FIR digital filter.
Compared to a filter that uses a filter, there is no deterioration in the SN ratio in the high range, and there are no problems such as oscillation, and because the set of coefficients is converted into a set of one system, there is less Various characteristics can be combined using coefficient data, which provides a large degree of freedom in characteristics.Also, since it can be realized with a single filter configuration, the configuration is simple.For example, it is compact and can be used with high-speed processors such as DSP , can be constructed at a low cost, and furthermore, since the sum of filter coefficients is normalized to a predetermined word length, there is no need to use a particularly long bit in the multiplication section, and the system can be made at a low cost. Since no overflow or underflow occurs, the input/output volume can be kept constant, and the SN
It is possible to prevent ratio deterioration, and by outputting the coefficient λ, it is possible to return the volume to the value before normalization in conjunction with the volume control volume.
By displaying , it is possible to easily set the desired characteristic by operating the characteristic variable switch of the characteristic input section.

[Brief explanation of the drawing]

FIG. 1 and FIG. 2 are a block diagram and a schematic diagram of a part of an embodiment of the equalizer of the present invention, respectively;
Figures 3 and 4 are equalizer characteristic diagrams and block system diagrams of bandpass filters that obtain equalizer characteristics, Figure 5 is a circuit diagram of an FIR digital filter used in the equalizer of the present invention, and Figure 6 is a diagram of the FIR digital bandpass filter. Fig. 7 is a frequency characteristic diagram of the filter, Fig. 8 is a circuit diagram of the FIR digital band filter, Fig. 9 is a flowchart for explaining the operation of the CPU, and Figs. 10 and 11 are diagrams of the filter. Figure 12, which shows numerical values, is a flowchart for explaining the operation of the DSP. 1,...Digital audio signal input terminal, 4
...DSP, 4 ₂ ... coefficient memory, 7 ... address memory,
8...Latch circuit, 9...Control device, 10...CPU,
10 ₁ ...Normalize section, 11...Program
ROM, 12... RAM for CPU work, 15... Characteristics input section, 16... Display section, 17... Audio signal output terminal, 18... Coefficient output terminal, 30 ₁ to 30 _l ... 1 Sampling delay section, 31 ₁ to 31 _l ... Multiplication section, 32 ₂
~ _32l , ₄₁₁ ~ _41l ...addition section, 50...FIR digital filter.

Claims (1)

[Scope of Claims] 1. A characteristic input section for specifying desired equalizer characteristics for each of a plurality of bands, and specification from the characteristic input section among a plurality of coefficients of each virtual FIR digital band filter for each of the plurality of bands. The coefficients of the filter of the corresponding band corresponding to the above are changed according to the specification, the coefficients of the plurality of bands are added for each of the plurality of coefficients to obtain the sum, the sum is normalized to a predetermined digital word length, and the coefficient of the filter is changed according to the specification. a coefficient changing means for obtaining the same number of coefficients as each of the plurality of coefficients; and one coefficient changing means whose frequency characteristics are changed based on the coefficients obtained by the coefficient changing means.
A digital graphic equalizer featuring an FIR digital filter. 2. A patent claim characterized in that the coefficient changing means is provided with means for outputting a coefficient such that (2 ^L −1)/M=λ, where M is the summation and L is the digital word length. The digital graphic equalizer according to item 1. 3. A patent claim characterized in that the coefficient changing means is provided with means for displaying a coefficient of (2 ^L -1)/M=λ, where M is the summation and L is the digital word length. The digital graphic equalizer according to item 1.