JPS6219689B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an analog data measurement system, and more particularly to a method for correcting dynamic errors caused by a disturbance prevention filter.

FIG. 1 shows a general configuration example of an analog measuring device currently in use. Data is captured from the process through detector 1, and noise filter 2
After passing through, the signal is amplified to an appropriate level by a preamplifier 3, digitally converted by an AD converter 4, and finally captured by a data reader 5. However, 6 is a decoder that provides a timing signal.

The accuracy evaluation of the signal that has passed through this entire series of blocks is the error rate _Es , which is generally given by the following equation.

E _s =E ₁ +E ₂ +E ₃ +E ₄ +E ₅ where Ei (i=1 to 5) indicates the error rate occurring in each device alone. These errors are called static errors, and can be calculated in advance from the input/output characteristics of each block, regardless of temporal changes in the input waveform.

For example, possible error generating factors inherent in the preamplifier 3 include input/output nonlinearity, input/output gain characteristics, initial drift, power supply voltage fluctuation, etc., and in the AD converter 4, quantization errors, random noise, etc. are considered. The noise filter 2 is composed of resistors R ₁ and R ₂ and a capacitor C as shown in FIG. The main purpose is to prevent external disturbances, and it has a noise reduction of about 60 dB. The static characteristic of this noise filter is generally evaluated as E ₂ =0 because it shows linearity when a constant value is given to the input. The current level of technology makes it possible to suppress this total static error rate E _s to within 1%.

On the other hand, the error that occurs when actual data is used as an input value is defined as a dynamic error. The actual analog waveform does not have a constant value as in the accuracy evaluation above, but changes very rapidly over time. Third
The analog values shown by 11 in Figures a and b are typical examples of spot measurement waveforms of analog data, which are often seen in temperature measurement, flow rate measurement, etc., where a is a sharp rising waveform and b is a sudden falling waveform. This is what is shown. The causes of dynamic error occurrence will be explained with reference to waveform 11 in FIG. What is dynamic error?
It is determined by the quality of the tracking characteristics to the changing waveform. 1st
In the block diagram shown in the figure, the noise filter 2 and preamplifier 3 are thought to mainly occur.
It is. The dynamic error of the preamplifier 3 is determined by the semiconductor elements and capacitive elements that make up the preamplifier 3, but for the former, many types of semiconductor elements have been developed that can obtain sufficient high frequency characteristics for the rate of change (frequency) of analog data. For the latter, both can be ignored as dynamic error factors by adopting a direct current amplification type (DC amplifier) that does not include a capacitive element in the signal line.

On the other hand, since the original purpose of the noise filter 2 is to prevent disturbances, it inevitably has a large time constant when trying to obtain noise reduction above a certain level. This time constant is the main cause of dynamic errors. The dynamic error factors will be explained by analyzing the response when the waveform No. 11 in FIG. 3 is passed through this noise filter.

Noise reduction in the filter shown in Figure 2
In order to have 60dB, select the following value.

R ₁ = R ₂ = 1.5 (kΩ) C = 100 (μF) From this, the time constant T of this circuit is as follows.

T = C × (R ₁ + R ₂ ) = 300 (ms) On the other hand, if the waveform in Figure 3 14 is expressed as a function of time, v _OUT (0-t ₂ ); value between t = 0 and t ₂ (Fig. 3 1
4) a: slope of t ₀ → t ₁ T: time constant u(t-t _i ): becomes a unit function. From this, the dynamic error δ becomes the following equation.

From this formula, the error is: slope a, time constant T, time t
It can be seen that it is expressed as a function. The larger the slope a (that is, it changes rapidly) and the larger the time constant T (that is, the larger the response delay), the larger the dynamic error becomes.

Let us now calculate the error rate E _d at time t ₁ when a=1 T=0.3 t ₀ =0 t ₁ =1.

E _d = δ/v _io t ₁ ≒ 30 _{( %} ) v _io t ₁ ; true value at the time of t ₁ (Fig. 3, 13). That is, it can be seen that a very large error occurs when trying to obtain a. E _d (FIG. 3, 15) is the error at t ₁ .

In recent advanced systems, there is a strong demand for obtaining analog data in real time and with high precision. However, conventional measurement methods not only do not take measures against this dynamic error, but also
It was often overlooked.

An object of the present invention is to provide an apparatus that can capture highly accurate analog values by sequentially correcting dynamic errors that occur during analog data measurement as described above.

Correction is performed for the portion where the dynamic error is large. The main points of the amendment concept are as follows.

Analog waveforms are unspecified curves. As is well known, all curves can be expressed as a set of ramp functions. Also, a characteristic of analog data is that it approximates a linear function when there is a sudden change in rise or fall. (Depending on the characteristics of the detector, etc.) From these conditions, it is possible to learn and store the waveform of data after digital conversion, and to perform corrections sequentially from when the waveform is stable.

The present invention mainly focuses on abrupt change points, and the measurement method is based on a spot measurement method.

FIG. 4 shows an embodiment of the correction device 7 of the present invention, which is placed between the AD converter 4 and the data reading device 5 as shown in FIG. The apparatus is comprised of a tilt detection timing storage device 21, an error calculation device 22, and a correction device 23 in FIG. As shown in equation (1) above, the dynamic error is determined by the slope and time constant, so a constant is used as a constant for recognizing the correction points.
Define C1 using the following formula.

C1=a×T where a: slope T: time constant In the slope detection device 21, the slope is determined from the difference (Δv _i ) between each sampling point of the digital values input at a constant pitch.

a=Δv _i /t _p where t _p ;sampling pitch If a×t _i >C1 at this time, the sampling point (t _i ) is stored as t _i ]0 (correction factor learning start). In this way, we search for a stable point while calculating the slope each time (t ₆ in Figure 5). From the stage when t ₆ is obtained, the error calculation device 22 starts operating and calculates the error using the following equation.

However, t ₀ : Correction factor learning start timing Subsequently, the correction device 23 operates and adds an error value to the input value to obtain an output v _p .

v _p = v _io + δ _i However, v _io ; The end timing of analog input value correction after digital conversion is C2: A constant for end timing, so that the influence of dynamic errors can be ignored.
This device can be realized relatively easily in both hardware and software configurations.

By inserting the above-mentioned device into the measurement system, the error in the data after t ₆ in the waveform of FIG. 4 is only due to the broken line approximation, and the accuracy is greatly improved.

The most effective method of utilizing the present invention is spot measurement, which covers a very wide range of applications, including high temperature measurement using thermocouples, flow rate measurement, and gas concentration measurement. Most of these measured waveforms show curves like those shown in Figure 5, and the maximum number of waveforms in the measurement system not equipped with this device is
A total error of 10% will occur.

FIG. 5 shows waveforms in the case of the present invention.
1 represents the original input, 32 represents the waveform approximated by a broken line, 33 represents the waveform after passing through the filter, and 34 represents the waveform after correction.

The present invention is not only applicable to analog measurements in general, but also describes a correction method for a system including response lag, and has a wide variety of applications.

Specific application examples include molten steel temperature measurement in the steelmaking process and exhaust gas measurement in oxygen flow rate controlled tunnel ventilation in the same process.

As a result of the present invention, a more accurate analog measurement system with excellent dynamic characteristics could be realized.

[Brief explanation of the drawing]

Fig. 1 is a block diagram of a conventional analog data measurement, Fig. 2 is an example of a noise filter circuit, Fig. 3 is a dynamic error caused by passing through a noise filter, Fig. 4 is an example of an analog data dynamic error correction device, Fig. 5 is an enlarged waveform of Fig. 3 and is a diagram explaining the principle of dynamic error correction, and Fig. 6 is a block diagram of a measuring device with analog data dynamic error correction. 1...detector, 2...noise filter, 3...
...Preamplifier, 4...A-D converter, 5...Data reader, 6...Timing decoder, 7...
Dynamic error correction device.

Claims (1)

[Claims] 1 In a method for correcting an output signal of an analog filter, the output signal corresponding to the input signal of the filter is sampled at a predetermined sampling period, the slope a of the output signal is calculated for each sampling period, and the slope and the filter are The correction of the output signal is started from the time when the product of the time constant T and the time constant T becomes larger than a predetermined value _C1 , and the straight line representing the slope of the output signal at that time and the steady state of the input signal of the filter are A method for correcting dynamic errors in analog signals, characterized in that the difference between a straight line representing a value and a value approximated by two straight lines is calculated at each sampling time point, and the output signal is corrected.