JP2006086681A - Data corrector and correction method of a-d converter - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide data corrector and a correction method of an A-D converter in which a correction value is updated at all times without requiring a voltage generator or mode switching in an external apparatus, delay units are not required by the number of impulse responses, an output responds quickly to an input signal, and a quantization error can be corrected stably with high precision. <P>SOLUTION: In the A-D converter 107 having an adaptive signal processing section 38 composed of an input value delay section 31, a prediction value calculating section 32, and a prediction filter coefficient calculating section 34, the prediction filter coefficient calculating section 34 comprises a step ST70 for calculating and updating a new prediction filter coefficient K used at the prediction value calculating section 32 during the next operation period using a Volterra series filter and employing a correction value ERRt calculated by the difference between a digital signal value Xt and a prediction value Yt as a coefficient. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a data correction apparatus and a data correction method for an A / D conversion apparatus that converts an analog signal incorporated in an industrial machine, an OA device, or the like into a digital signal and outputs the digital signal.

FIG. 5 is a block diagram relating to data correction of a conventional A / D converter (see, for example, Patent Document 1).
In FIG. 5, 1 is an analog input circuit, 2 is a multiplexer circuit, 3 is an A / D conversion circuit, 4 is a CPU, 30 is a plant, 100 is a plant controller, 102 is a CPU of the plant controller, and 103 is an A / D conversion device. S1, S2 are analog signals, S3 is a channel select signal, S4 is a digital signal, and D1 is data.
Hereinafter, the configuration of the portion related to data correction of the conventional A / D converter will be described with reference to FIG.
A plurality of analog signals S1 are input to the plant controller 100 including the A / D converter 103 and the CPU 102 from the plant 30 and the like, and are received by the analog input circuit 1. One of the analog signals S1 having a plurality of channels (analog signal S2) is selected by the channel select signal S3. The selected analog signal S2 is converted into a digital signal S4 by the A / D conversion circuit 3. The digital signal S4 is processed by the CPU 4, and data D1 is transmitted to the CPU 102 of the plant controller.

FIG. 6 is a flowchart showing a processing procedure of a data correction method of a conventional A / D converter. FIG. 7 is an explanatory view showing a data correction method of a conventional A / D converter.
Hereinafter, the processing procedure in the data correction method of the conventional A / D converter will be described with reference to FIGS.
Note that ST numbers in FIG. 6 indicate steps in the flowchart.
(1) Step ST301 As modes of the A / D conversion device 103, two modes of “correction value calculation mode” and “A / D conversion mode” are prepared.
(2) Step ST302 In the “correction value calculation mode”, first, the CPU 4 outputs a channel select signal S3 in order to determine which channel of the plurality of channels is to be calculated. At this time, since it is necessary to input an accurate value to the analog signal S1, a voltage generator (not shown) is connected to the analog input circuit 1.
(3) Step ST303 An accurate analog value corresponding to the point b1 in FIG. 7 is input from the voltage generator (not shown).
(4) Step ST304 Read the digital value after A / D conversion. This digital value is point a1 in FIG. The same method is performed for other points to determine the values of points a2 to a5 in FIG. What is obtained by the points a1 to a5 is the characteristic curve 17 in FIG.
(5) Step ST305 Next, the straight line 16 in FIG. 7 is obtained from these values using the least square method, and the point a11 having the minimum value and the point a51 having the maximum value are obtained.
(6) Step ST306 If these two points a11 and a51 and the minimum value point b1 and the maximum value point b5 on the ideal characteristic line 18 are known, the correction values A and B can be obtained from Equation 2 and Equation 3 of Patent Document 1. Can do.
(7) Step ST307 The obtained correction values A and B are stored in a memory IC or the like.
(8) Step ST301 If the correction value can be calculated, the operation mode of the A / D conversion device 103 is switched to the “A / D conversion mode”.
(9) Step ST308 The digital value after A / D conversion is subjected to the correction calculation of the following formula 1.
Y = AX + B (1)
Here, Y is a digital value after the correction calculation, X is a digital value before the correction calculation, and A and B are the above-described correction values.
(10) Step ST309 Next, the digital value after the correction calculation is sent to the plant controller CPU102.
Thereafter, step ST308 and step ST309 are repeated.
Thus, in the conventional data correction method of the A / D converter, the operation mode of the A / D converter 103 is divided into the “correction value calculation mode” and the “A / D conversion mode”, and the “correction value calculation mode” , A correction value is obtained in advance by a difference between a digital signal obtained by calculating the digital signal 7 after A / D conversion using the least square method and an ideal digital value with respect to an input analog value obtained by connecting a voltage generator. Then, the procedure of operating the A / D conversion device 103 in the “A / D conversion mode” and performing the correction process using the correction value has been taken.

In the prior art disclosed in Patent Document 1, the A / D converter 103 is provided with a “correction value calculation mode” which is a function other than A / D conversion, and a voltage generator generates a known accurate analog signal. Further, a high-accuracy correction value based on the difference between the input analog value and the ideal digital value must be obtained and stored in the A / D conversion device 103. When operating in the “/ D conversion mode”, there is a problem in that it cannot be updated to a new correction value. Furthermore, when continuous analog data is taken into the A / D conversion circuit 3 and output as digital data, a quantization error always occurs between the analog data before A / D conversion and the digital data after A / D conversion. The data correction method of the A / D converter has a problem that the quantization error cannot be corrected.
8A and 8B are explanatory diagrams for explaining the quantization error. FIG. 8A shows the spectrum of the input signal, and FIG. 8B shows the spectrum after the conventional A / D conversion.
For example, when a sine wave having a period of 10 KHz is input to the A / D conversion circuit 3 as continuous analog data, assuming that the spectrum of the analog signal before A / D conversion is shown in FIG. As shown in FIG. 8B, a new spectrum is generated on both sides of the spectrum of the digital signal after A / D conversion, centering on 10 KHz.

Therefore, a data correction method for an A / D converter that does not require a voltage generator in an external device, does not require mode switching, is constantly updated with a correction value, and can stably and accurately correct a quantization error. The same applicant as the present invention proposes “A / D converter data correction method” (application number P2003-186077).
FIG. 8C shows a spectrum after A / D conversion in which the quantization error is corrected.
If the quantization error is predicted and subtracted from the data after A / D conversion, as shown in FIG. 8C, the new spectrum generated on both sides centering on 10 KHz is reduced, and after A / D conversion The accuracy of data is improved. As a means for predicting the quantization error and subtracting it from the data after A / D conversion, “Data correction method of A / D conversion apparatus” (application number P2003-1886077) uses data that has been A / D converted using a prediction filter. The A / D converter can be made highly accurate by strengthening the spectrum of the analog data and bringing it closer to the spectrum of analog data before A / D conversion.
That is, the digital signal obtained by converting the input analog signal by the A / D conversion circuit is obtained as nn delayed digital signals obtained by delaying p number of samples by the sampling period of the A / D conversion circuit. Calculate the sum of products of the prediction filter coefficients previously calculated and calculate the predicted value of the digital signal, calculate the correction value from the difference between the digital signal and the predicted value, and use the correction value as the coefficient for linear prediction The LMS (Least Mean Square Error) method, which is a method, is introduced, and new prediction filter coefficients used in the next calculation cycle are calculated and updated.
JP 10-145231 A (page 4-6, FIG. 1-3, Formula 1-3)

However, in the data correction method of the A / D conversion device proposed by “Data correction method of A / D conversion device” (application number P2003-186077), the predicted value calculation unit is caused by the characteristics of LMS which is a linear prediction method. In the prediction filter coefficient calculation unit, the delay unit needs the impulse response of the digital signal. As a result, the update response speed for the digital signal in the prediction value calculation unit and the prediction filter coefficient calculation unit depends on the number of delay units. Then there was a problem of being late.
Therefore, the present invention has been made in view of this problem, and does not require a voltage generator in an external device, does not require mode switching, constantly updates a correction value, and further does not require a delay unit for an impulse response. An object of the present invention is to provide a data correction apparatus and a data correction method for an A / D converter that can correct a quantization error with a fast and stable output response to an input signal with high accuracy.

In order to solve the above problem, the invention of claim 1 relates to a data correction device for an A / D converter, and in the A / D converter for converting an analog signal into a digital signal and outputting it, the A / D conversion is performed. The apparatus includes an initial value setting means for setting an initial value necessary for data correction, an A / D conversion circuit that converts an input analog signal into a digital signal and outputs a digital signal value, and the digital signal value Sum-of-product calculation of an input value delay unit for acquiring p delayed digital signals delayed by the sampling period of the A / D conversion circuit, and the predicted filter coefficient calculated and obtained one calculation period before Calculating a predicted value of the digital signal value, outputting the predicted value as corrected digital data, and correcting the difference between the digital signal value and the predicted value. A prediction value calculation unit to calculate, a prediction filter coefficient calculation unit to calculate and update a new prediction filter coefficient to be used in the prediction value calculation unit in the next calculation cycle using a Volterra series filter with the correction value as a coefficient, and It is characterized by having.
The invention of claim 2 relates to a data correction method for an A / D converter, and in an A / D converter for converting an analog signal into a digital signal and outputting the digital signal, the data correction of the A / D converter is performed. A step of setting a necessary initial value, a step of converting an input analog signal into a digital signal by an A / D conversion circuit and outputting a digital signal value, and a sampling of the digital signal value by the A / D conversion circuit A step of obtaining a delayed digital signal delayed by p cycles, and calculating a sum of products of the delayed digital signal and a prediction filter coefficient calculated and calculated one calculation cycle before, thereby obtaining a predicted value of the digital signal value A step of calculating, a step of outputting the predicted value as corrected digital data, and a correction value based on a difference between the digital signal value and the predicted value. A step of leaving, using Volterra series filtering the correction value as a coefficient, and a step of updating by calculating a new prediction filter coefficients used in the next calculation cycle, comprising the.

According to the data correction device and the data correction method of the A / D conversion device according to the present invention, the voltage generator is not required for the external device, the mode switching is not required, the correction value is constantly updated, and the sine wave or cosine wave is updated. Since the line spectrum of the sine wave or cosine wave of the digital signal after A / D conversion of the analog signal can be enhanced and the spectrum can be smoothed, the quantization error can be corrected stably and accurately.
In addition, white noise can be reduced because the frequency dispersion is strong in the process of the adaptive signal processing unit, and as a result, the accuracy of the corrected digital data output from the A / D converter can be improved.
Furthermore, the number of delay units in the prediction value calculation unit and the prediction filter coefficient calculation unit can be reduced by the Volterra series filter, the nonlinear characteristic of the input signal that is the characteristic of the Volterra series filter can be followed, and the prediction value calculation unit There is an effect that the predicted value accuracy for the input signal is increased.

  Hereinafter, specific examples of the present invention will be described with reference to the drawings.

FIG. 2 is a block diagram showing the configuration of the A / D converter according to the present invention.
In FIG. 2, 31 is an input value delay unit, 32 is a prediction value calculation unit, 34 is a prediction filter coefficient calculation unit, 38 is an adaptive signal processing unit, 107 is an A / D converter, D2 is corrected digital data, Xt Is a digital signal value at time t, Yt is a predicted value at time t, and ERRt is a difference between Xt and Yt. Here, the same reference numerals as those in FIG. 5 indicate the same components and signals as those in FIG.
The features of the A / D converter of the present invention are as follows.
In other words, the A / D conversion device 107 converts an input analog signal S2 into a digital signal S4 by outputting an initial value setting means (not shown) for setting an initial value necessary for data correction, and outputs a digital signal value Xt. The A / D conversion circuit 3 that performs the input, the input value delay unit 31 that acquires the delayed digital signal Xi obtained by delaying the digital signal value Xt by the sampling period of the A / D conversion circuit 3, and the delayed digital signal Xi and one calculation period. The prediction value Yt of the digital signal value Xt is calculated by multiplying the prediction filter coefficient K calculated and obtained in advance, and the prediction value Yt is output as the corrected digital data D2, and the digital signal value A predicted value calculation unit 32 that calculates a correction value ERRt based on a difference between Xt and the predicted value Yt, and a Volterra series filter using the correction value ERRt as a coefficient, A prediction filter coefficient calculation unit 34 be updated to calculate a new prediction filter coefficients used in prediction value calculation unit 32, is that it includes a. That is, the A / D conversion apparatus 107 of the present invention has an adaptive signal processing unit 38 including an input value delay unit 31, a prediction value calculation unit 32, and a prediction filter coefficient calculation unit 34. The prediction value calculation unit 32 and the prediction filter The coefficient calculation unit 34 performs data correction as a nonlinear prediction method by using a Volterra series filter.

FIG. 3 is a detailed block diagram of the adaptive signal processing unit 38 according to the data correction method of the A / D conversion apparatus of the present invention, and FIG. 4 is a predicted value of the adaptive signal processing unit 38 of the A / D conversion apparatus of the present invention. 3 is a detailed block diagram of a calculation unit 32. FIG.
3 and 4, 33 is an adder, 35 is a delay unit, 36 is a multiplication coefficient storage unit, 37 is a multiplier, and 50 is a two-input multiplier. 2 denote the same components and signals as those in FIG.
Hereinafter, the data correction operation of the A / D converter 107 of the present invention will be described in detail with reference to FIGS.
The digital signal value Xt at time t is output as Xi [0] by the input value delay unit 31 by delaying p digital signal values by the sampling period of the A / D conversion circuit 3. However, the number of delay devices 35 is determined so as to meet the purpose of the present invention. However, considering real-time processing, the number of delay devices 35 is empirically 3-10.
In the predicted value calculation unit 32, (nn−1) digital signal values Xi [1]... Xi [nn−1] obtained by sequentially delaying the Xi [0] output from the input value delay unit 31 for each sample. ] Is created.
Further, (nn + 1) * (nn + 4) / 2 prediction filter coefficients K [0]... K [(nn + 1) * (nn + 4) / 2− created by the prediction filter coefficient calculation unit 34 before one calculation cycle. 1] and the digital signal value Xi [0]... Xi [(nn + 1) * (nn + 4) / 2-1] are respectively multiplied by the following formula (4), and the results are added. A predicted value Yt at time t is obtained.
Yt = ΣK [i] * Xi [i] (4)
0 ≦ i ≦ ((nn + 1) * (nn + 4) / 2-1)
The predicted value Yt is output as corrected digital data.
Next, the difference ERRt between the digital signal value Xt at time t and the predicted value Yt at time t is calculated using the following mathematical formula (5).
ERRt = Xt−Yt (5)
Further, a Volterra series filter is introduced using the difference ERRt, and a prediction filter coefficient K1 [i] at time (t + 1) is calculated according to the following equation (6).
K1 [i] = K [i] + u * ERRt * Xi [i] (6)
0 ≦ i ≦ ((nn + 1) * (nn + 4) / 2-1)
Here, u is a multiplication coefficient that determines the convergence speed up to the final predicted value, which is determined in advance by experiments so as to meet the purpose of the present invention, and is stored in the multiplication coefficient storage unit 36.
The predicted value Yt at time t created by repeating the above is the sine wave of the digital signal S4 after A / D conversion of the analog signal of the sine wave or cosine wave according to the equations (5) and (6). Alternatively, the line spectrum of the cosine wave can be emphasized and smoothed, and white noise can also be reduced, so that the quantization error generated during the conversion in the A / D conversion circuit 3 is corrected data.

FIG. 1 is a flowchart showing the processing procedure of the data correction method of the A / D converter according to the present invention, which can be handled by a program.
Hereinafter, the processing procedure of the data correction method of the A / D converter according to the present invention will be described with reference to FIG.
In addition, ST number in a figure shows each step of a flowchart.
(1) Step ST10 Sampling cycle, delay number p, prediction value calculation unit 32, prediction filter coefficient calculation unit 34, multiplication coefficient u, K [0]... K [(nn + 1) * ( nn + 4) / 2] and the like, the initial values necessary for the data correction method of the A / D converter of the present invention are set by a known initial value setting means (not shown) and method. Note that the initial value of K [0]... K [(nn + 1) * (nn + 4) / 2-1] is 0.
(2) Step ST20 The A / D conversion circuit 3 converts the analog signal S2 into a digital signal S4, that is, a digital signal value Xt.
(3) Step ST30 The input value delay section 31 delays the digital signal value Xt by the sampling period of the A / D conversion circuit 3 and outputs it as Xi [0].
(4) Step ST40 In the predicted value calculator 32, Xi [0] output from the input value delay unit 31 is sequentially delayed for each sample ((nn + 1) * (nn + 4) / 2-1) digital signals. A value Xi [1]... Xi [(nn + 1) * (nn + 4) / 2-1] is generated, and (nn + 1) * (nn + 4) / 2 prediction filter coefficients generated by the prediction filter coefficient calculation unit 34 Multiply each by K [0]... K [(nn + 1) * (nn + 4) / 2-1] and add the results to calculate the predicted value Yt (Formula 4).
(5) Step ST50 The predicted value Yt at time t is output as corrected digital data D2.
(6) Step ST60 The correction value ERRt is calculated from the difference between the digital signal value Xt at time t and the predicted value Yt at time t (Formula 5).
(7) Step ST70 In the prediction filter coefficient calculation unit 34, a new prediction filter coefficient K [0]... K [(nn + 1) * used in the next calculation cycle using a Volterra series filter with the correction value ERRt as a coefficient. (Nn + 4) / 2-1] is calculated and updated (Formula 6).
Thereafter, step ST20 to step ST70 are repeated.

  As described above, the A / D converter 107 and the data correction method thereof according to the present embodiment set and input the initial values necessary for data correction of the A / D converter 107 by known means and methods. The analog signal S2 is converted into a digital signal S4 by the A / D conversion circuit 3 to output a digital signal value Xt, and the digital signal value Xt is delayed by the sampling period of the A / D conversion circuit 3 by the delayed digital signal Xi To calculate the predicted value Yt of the digital signal value Xt by calculating the sum of products of the delayed digital signal Xi and the predicted filter coefficient K calculated and calculated one calculation cycle before, and correct the predicted value Yt. Is output as the digital data D2, and a correction value ERRt is calculated from the difference between the digital signal value Xt and the predicted value Yt. Is used to calculate and update a new prediction filter coefficient to be used in the next calculation cycle, so a voltage generator is not required for an external device, mode switching is not required, and the correction value is constantly updated. A quantization error can be corrected stably and accurately. Furthermore, white noise can be reduced because the frequency dispersion is strong in the process of the adaptive signal processing unit 38, and the accuracy of the corrected digital data D2 output from the A / D converter 107 can be further improved. Further, the number of delay units 35 in the prediction value calculation unit 32 and the prediction filter coefficient calculation unit 34 can be reduced by the Volterra series filter, the nonlinear characteristic of the input signal which is the characteristic of the Volterra series filter can be followed, and the prediction value calculation The accuracy of the predicted value for the input signal to the unit 32 increases.

  By correcting the quantization error of the digital signal output from the A / D conversion circuit 3 using a Volterra series filter, it is possible to follow the nonlinear characteristics of the input signal, which is the characteristic of the Volterra series filter, and A / D conversion Since the output of the apparatus can be stabilized and the quantization error can be corrected with high accuracy, it can be applied to the entire signal processing field using various A / D conversions.

The flowchart which shows the process sequence of the data correction method of the A / D converter of this invention. The block diagram which shows the structure of the A / D converter which concerns on this invention Detailed block diagram of the adaptive signal processing unit 38 according to the data correction method of the A / D conversion device of the present invention Detailed block diagram of the predicted value calculation unit 32 of the adaptive signal processing unit 38 of the A / D conversion device of the present invention The block diagram which concerns on the data correction of the A / D converter of a prior art The flowchart which shows the process sequence of the data correction method of the A / D converter of a prior art Explanatory drawing which shows the data correction method of the A / D converter of a prior art (A) Spectrum of input signal (b) Spectrum after conventional A / D conversion (c) Spectrum after A / D conversion corrected for quantization error

Explanation of symbols

1 Analog Input Circuit 2 Multiplexer Circuit 3 A / D Conversion Circuit 4 CPU
DESCRIPTION OF SYMBOLS 30 Plant 31 Input value delay part 32 Predicted value calculation part 33 Adder 34 Prediction filter coefficient calculation part 35 Delay element 36 Multiplication coefficient memory | storage part 37 Multiplier 38 Adaptive signal processing part 50 2-input multiplier 100 Plant controller 102 CPU of plant controller
103, 107 A / D converter S1, S2 Analog signal S3 Channel select signal S4 Digital signal D1 Data D2 Corrected digital data Xt Digital signal value at time t Yt Predicted value at time t ERRt Difference between Xt and Yt K Prediction filter coefficient

Claims (2)

In the A / D converter (107) that converts the analog signal (S2) into a digital signal (S4) and outputs it,
The A / D converter (107) includes initial value setting means for setting an initial value necessary for data correction;
An A / D conversion circuit (3) for converting the input analog signal (S2) into a digital signal (S4) and outputting a digital signal value (Xt);
An input value delay unit (31) for acquiring a delayed digital signal (Xi) obtained by delaying the digital signal value (Xt) by p sampling cycles of the A / D conversion circuit (3);
The predicted value (Yt) of the digital signal value (Xt) is calculated by calculating the sum of products of the delayed digital signal (Xi) and the prediction filter coefficient (K) calculated and obtained one calculation cycle before. A predicted value calculation unit that outputs the predicted value (Yt) as corrected digital data (D2) and calculates a corrected value (ERRt) based on a difference between the digital signal value (Xt) and the predicted value (Yt) (32),
Using a Volterra series filter with the correction value (ERRt) as a coefficient, a prediction filter coefficient calculation unit (34) that calculates and updates a new prediction filter coefficient used in the prediction value calculation unit (32) in the next calculation cycle. And a data correction device for an A / D conversion device. In the A / D converter (107) that converts the analog signal (S2) into a digital signal (S4) and outputs it,
A step (ST10) of setting an initial value necessary for data correction of the A / D converter (107);
A step (ST20) of converting the input analog signal (S2) into a digital signal (S4) by an A / D conversion circuit (3) and outputting a digital signal value (Xt);
Obtaining a delayed digital signal (Xi) obtained by delaying the digital signal value (Xt) by p number of sampling periods of the A / D conversion circuit (3) (ST30);
A step of calculating a predicted value (Yt) of the digital signal value (Xt) by performing a product-sum calculation on the delayed digital signal (Xi) and the prediction filter coefficient (K) calculated and obtained one calculation cycle before. (ST40)
Outputting the predicted value (Yt) as corrected digital data (D2) (ST50);
Calculating a correction value (ERRt) from the difference between the digital signal value (Xt) and the predicted value (Yt) (ST60);
A step of calculating and updating a new prediction filter coefficient to be used in the next calculation cycle using a Volterra series filter with the correction value (ERRt) as a coefficient (ST70). Data correction method for conversion device.

Images