JPH0695628B2 - Digital filter - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION <Industrial field of application> The present invention relates to a digital filter that performs filtering and thinning processing, and in particular, FFT (fast fourier transform).
The present invention relates to a digital filter suitable for use in an analyzer zooming device.

<Prior Art> The FFT analyzer obtains the spectrum of this time-series data by subjecting the time-series data to a Fourier calculation process, and a zooming device for expanding the frequency axis is required to obtain a more detailed spectrum. Such a zooming device requires a high zooming ratio and variability of the zooming ratio, and a digital filter for thinning and filtering time series data is used. FIG. 6 shows the structure of such a digital filter. In FIG. 6, 1 is
It is an IIR (infinite impluse filter) type filter in which time series data is input and a filter operation is performed. this
Part of the output of the IIR filter 1 is thinned out and the
It is fed back to the shape filter 1 and subjected to filter calculation. When the calculation is performed the required number of times, it is output to the FFT analyzer 2 and the Fourier calculation is performed. That is, the IIR filter is used in multiple stages. In such a digital filter, if the thinning rate is m and the number of calculations is n.
A zooming rate of m ⁿ is obtained. The zooming rate can be changed by changing m and n.

Next, the IIR filter will be described. The IIR filter has time series input data x (n) and time series output data y.
(N) It is represented by. That is, it is a recursive filter in which the output returns to the input. FIG. 7 shows the configuration of a device that executes such calculations. In FIG. 7, time series data x (n)
Is input to the adder 3. The output of the adder 3 is the register 4
Is input and shifted. The output of the register 4 is input to the multiplier 5, and the coefficients a _{1 to} a _{N-1 of the} equation (1) are used.
And are input to the adder 3, and their sum is calculated. The output of the register 4 is also output by the multiplier 6 to the coefficients b _{1 to} b.
The product of _N-1 and _N-1 is input to the adder 7 and added. The output of the adder 3 is multiplied by the coefficient b _{0 in the} multiplier 8 and input to the adder 7. The output of the adder 7 becomes the filter calculation result. In this way, the calculation of the equation (1) is executed. As a digital filter,
In addition to the IIR type, there is the FIR (finite impluse filter) type.
However, it is disadvantageous in terms of the number of multiplications and the storage of data, and is rarely used.

<Problems to be Solved by the Invention> However, such a digital filter has the following problems. As can be seen from the equation (1), since the output of the IIR filter is recursive to the input, the filter operation is executed for all the time-series data even though the output is partly discarded. Must.

Further, since one filter is used in multiple stages, doubling the sampling frequency of time series data also requires doubling the operating speed of the filter. Therefore, the hardware becomes large.

Further, as can be seen from FIG. 7, the IIR type filter has a drawback that the structure is complicated.

<Object of the Invention> An object of the present invention is to provide a digital filter which has a simple structure and can be speeded up.

<Means for Solving Problems> In order to solve the above problems, in the present invention, the time series data is input in a digital filter that performs digital arithmetic processing on the time series data to reduce the sampling rate of the time series data. A first FIR type filter, a second FIR type filter to which the output of the first multiplexer is input, a third FIR type filter to which the output of the second multiplexer is input, the first and second The first multiplexer to which the output of the second FIR filter is input; and the second multiplexer to which the time series data and the outputs of the first and second FIR filters are input. It is characterized by having done.

The second FIR type filter and the third FIR type filter are commonly used.

<Embodiment> FIG. 1 shows an embodiment of a digital filter according to the present invention. In FIG. 1, 10 is the first FIR (finite impluse
time-series data is input. The output of the first FIR filter 10 is input to multiplexers 11 and 12. Reference numeral 13 is a second FIR type filter, the output of the multiplexer 11 is input, and the output is input to the multiplexers 11 and 12. 14 is the third
This is a FIR type filter, the output of the multiplexer 12 is input, and the output is output to the outside. Multiplexer 1
1, 12 are controlled by a controller (not shown).

Next, the FIR filter will be described. The FIR type filter has x (n) as its time series input and y (n) as its time series output. It is represented by. That is, it is a non-recursive filter in which the output does not recurs to the input. Figure 2 shows the basic structure of a FIR filter. In FIG. 2, 16 is a register, which is vertically connected. The time series data x (n) is input to the register at the end of the register 16 and sequentially shifted. Reference numeral 17 denotes a multiplier, to which the time series data input to each of the registers 16 is input, and the impulse response functions h (0) to h (h) are input to this data.
Multiply (N-1) and output. Reference numeral 18 is an adder, and the outputs of these multipliers 17 are input and added. This adder
The output of 18 becomes the time series output y (n). In this way, the calculation of the equation (2) is executed. Register 16, multiplier
Basically, 17 must be the same number as the filter order N.

Next, the operation of the embodiment shown in FIG. 1 will be described. First to third
All FIR type filters 10, 13 and 14 in FIG. 3 are used, and the decimation rate of each filter is referred to as M. The M time series data are first input to the first FIR type filter 10,
The filter operation is executed once based on the equation (2), and one output is output. That is, _assuming that the sampling frequency of the input time series data is f _s, the frequency of the output of this first FIR filter is f _s / M. This is the first
We show that the FIR filter should operate at a frequency of f _s / M. This output data is sent to the second via the multiplexer 11.
Is input to the FIR filter 13 of. The filter 13 obtains M pieces of data from the first FIR type filter 10 and performs filter calculation and thinning processing. The filter 13 is used in multiple stages. That is, the output of the second FIR type filter 13 is input to the filter 13 again by the multiplexer 11 and processed. This process (stage) is repeated as many times as necessary. Each stage is time-shared. In general, the filter operation is performed using time series data with a larger number than the thinning rate M, so that the data of each stage is held inside this filter and the data held by the stage is switched. Since the sampling frequency f _s of the time series data is set to 1 / M by the first FIR type filter 10, the operating frequency of the second FIR type filter may be f _s / M.
If the number of stages of the second FIR filter is n,
The sampling frequency of the output of this filter is f _s / (M) ^{n + 1}
Becomes The output of the second FIR type filter 13 is input to the third FIR type filter 14 via the multiplexer 12, and filter calculation and thinning processing are performed. The output of this filter is the final output. This gives a zooming rate of ^{Mn + 2} . When the zooming rate is M ² , the input time series data is passed through the first FIR type filter 10, and the output is filtered by the third FIR type filter 14 and thinned out before being output. When the zooming ratio is M, only the third FIR type filter 14 is used. These selections are made by switching the data input to each filter by multiplexers 11 and 12. Furthermore the first
Since the FIR type filter 10 and the second FIR type filter 13 of FIG. 3 do not form the final stage, the characteristics of these filters may be gentle. That is, since it is not necessary to increase the order, the structure can be simplified.

FIG. 3 shows the operation timing of each stage when the thinning rate per stage is M = 2 and the number of stages of the entire filter is 5 (the number of stages of the second FIR filter 13 = 3). In this figure, (1) is the FIR type filter 1 of FIG.
4, (2) to (4) show the third to first stages of the second FIR type filter 13, and (5) shows the operation timing of the first FIR type filter 10. In the figure → indicates the operating time.
(6) is the sampling period (1 / f _s ) of the input time series data
Represents Each filter is operated when two pieces of input data are obtained. As can be seen from this figure, the operation of the first FIR type filter is twice the sampling period of the input time series data (generally M times), and the operation of the second and third FI
The operation timings of the R-type filters 13 and 14 can be prevented from overlapping.

FIG. 4 shows another embodiment of the present invention. The same elements as those in FIG. 1 are designated by the same reference numerals and the description thereof will be omitted. In the figure, reference numeral 19 is a memory to which the output of the first FIR type filter 10 and the output of each stage of the second FIR type filter 13 are input and stored. Each stage has a storage area of M words and its address is predetermined. 20
Is a multiplexer, which receives the input time-series data and the data stored in the memory 19 and selects the third time
Output to FIR filter 14. A controller 21 controls the second FIR filter 13, the memory 19, and the multiplexer 20. The operation is the same as that of the embodiment shown in FIG. In this embodiment, the configuration is simplified by storing the output of each filter in the corresponding address of one memory.

FIG. 5 shows still another embodiment. The same elements as those in FIG. 4 are designated by the same reference numerals and the description thereof will be omitted. As can be seen from FIG. 3, the operation timings of the second FIR type filter 13 and the third FIR type filter 14 can be made not to overlap. In this embodiment, the second FIR type filter 13 and the third FIR type filter 14 are shared to simplify the structure. Fifth
In the figure, 22 is an FIR type filter, and the second and third F
It has the same structure as the IR filters 13 and 14. Reference numerals 23 and 24 denote second and third FIR filter coefficient storage units, respectively, which store impulse response coefficients of the respective filters. 25 is a multiplexer, and these are the second and third FI
The coefficients stored in the R-type filter coefficient storage units 23 and 24 are washed and output to the FIR filter 22. A controller 26 controls the memory 19, the multiplexers 20 and 25, and the FIR type filter 22. That is, the FIR filter is used as the second FIR
When used as a third filter, the second FIR type filter coefficient storage unit 23 is selected, and when used as a third FIR type filter, the third filter coefficient storage unit 24 is selected. The other operations are the same as those in the embodiment shown in FIG.

In these embodiments, the first to third FIR type filters 1
Although the thinning rates of 0, 13, and 14 are the same, they may be different.

<Effects of the Invention> As specifically described above based on the embodiments, in the present invention, the FIR type filter is used as the filter, and the first FIR type filter in the first stage and the second FIR type used in the multistage are used. The filter is divided into the third stage FIR filter at the final stage. Therefore, the first and second FIR
Since the shape filter does not become the final stage, it is not necessary to increase the order, the configuration can be simplified, and the number of multiplications can be reduced.

In addition, the first FIR type filter was installed separately to make it the first stage only. Therefore, the sampling frequency of the second and third FIR filters is 1 / M compared to the input frequency of the first stage.
A filter with a slow operation speed can be used. Also, since the first filter is also of FIR type, the actual operating speed is 1 / M of the input frequency.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of a digital filter according to the present invention, FIG. 2 is a block diagram showing the configuration of a FIR filter, FIG. 3 is a diagram showing operation timing, FIG.
FIG. 5 is a block diagram showing another embodiment, FIG. 6 is a block diagram showing the configuration of a conventional digital filter, and FIG. 7 is II.
It is a block diagram showing a configuration of an R-type filter. 10 …… First FIR type filter, 11,12,20,25 …… Multiplexer, 13 …… Second FIR type filter, 14 …… Third FIR
Shape filter, 19 ... Memory, 21,26 ... Controller. 2
2 ... FIR type filter, 23 ... Second FIR type filter coefficient storing section, 24 ... Third FIR type filter coefficient storing section.

Claims (2)

[Claims] 1. A digital filter for subjecting time-series data to a digital operation process to reduce the sampling rate of the time-series data, wherein a first FIR filter to which the time-series data is input and an output of a first multiplexer are provided. A second FIR filter to be input, a third FIR filter to which the output of the second multiplexer is input, and a first FIR filter to which outputs of the first and second FIR filters are input A digital filter comprising: a multiplexer; and the second multiplexer to which the time series data and the outputs of the first and second FIR filters are input. 2. The second FIR type filter and the third FIR type filter.
The digital filter according to claim 1, wherein the digital filter shares a common type filter.