JPH1065540A - Electronic balance for making one time of measurement into plural times in short time - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide exact A/D conversion by reducing errors generated at the time of A/D conversion by matching the end of reference integration time in a double integration system with the start of input integration time. SOLUTION: Every input integration time Tx' becomes 1/m of Tx. Since the input integration time is 1/m when the measured value of an a-th reference integration term is defined as n(a), the reference integration time becomes 1/m as well. A result (n') successively adding measured values n(0), n(1)...n(m-1) from a=0 to m-1 at the time of processing becomes n(0)+n(1)+...+n(m-1). Since the condition of n(a)=n/m is established even in the case of short input integration time Tx', as a value provided by (m) times of addition, almost the same value (n) as the case of input integration time Tx can be provided. In this case, the time of a minimum control time interval (instruction execution time) at a controller is synchronized with a reference time interval to be measured.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to securing the conversion accuracy of a double integration type A / D converter, and more particularly to the same integrator by switching a plurality of types of output voltages obtained from two or more different physical quantities with a switch. And A in double integration
Effectively maintains the accuracy of the / D conversion device.

[0002]

2. Description of the Related Art Generally, in order to convert a load into a voltage value, the load is received by a load cell to which a strain gauge formed in a bridge shape is attached, and the output of the load cell is connected to a differential amplifier to amplify the load. A voltage Vx is obtained, and a digital value proportional to Vx is obtained by a double integration type A / D converter as shown in FIG.

That is, when the amplified voltage Vx is connected by SW to the resistor R1 whose other end is connected to the minus terminal of the integrator, a current Ix = Vx / R1 flows through the integrator. The output Vo of the integrator whose + terminal is connected to the reference voltage (0 V) is
It decreases with the current Ix. When the output of Vo is connected to the + terminal of the comparator and the-terminal of the comparator is connected to 0 V, the output of the comparator changes from Hi to Lo as soon as Vo reaches 0 V from a positive voltage.

The control device detects that the comparator has become Lo, and starts counting the output of the reference time generator and the number of pulses of fCL from zero. After counting the fixed number N of pulses, the control device switches SW to -VR and simultaneously counts the number of pulses of fCL from zero. Assuming that the time from when the comparator becomes Lo to when the switch is switched (hereinafter referred to as input integration period) is Tx, and the time of one pulse of the output of the reference time generator and fCL is tCL, the input integration period Tx is N *
It is represented by tCL.

When SW switches to -VR, a current IR = -VR / R1 flows through the integrator, and the integrator output voltage (V
o) goes up. When the integrator output voltage (Vo) reaches 0 V from minus, the comparator output changes from Lo to Hi.
Changes to The control device detects this, and stops counting the number of tCL. The number of pulses measured at this time is defined as n.

After switching SW to -VR, Vo becomes 0V
Is defined as TR (hereinafter, a reference integration period), the reference integration period TR is represented by n * tCL. As a result, the amount of charge (Qx) Tx flowing through the integrator during the time Tx
* Ix is represented by Tx * Vx / R1, and the amount of charge (QR) TR * IR flowing from the integrator during the TR time is TR * (-VR)
/ R1. On the other hand, Vo is Qx + QR = 0 since the Tx start time and the TR end time are equal, and (N * tCL
* Vx / R1) + (n * tCL * (-VR) / R1) = 0.
Rearranging this equation, n * tCL * VR / R1 = N * tCL *
Vx / R1 and this equation gives n = N * Vx / VR. Here, since N and VR are predetermined constant values, n becomes a value proportional to Vx, and a digital value n proportional to Vx is obtained.

However, a measurement error occurs here. For example, if the input integration time is 0.1 second and tCL is 4 μs, N
= 25,000. Further, when no load is applied, n is n ₀ = 2,
000, setting n when the weighing load in such a way that nw = 12,000, the amount of change becomes nw-n ₀ = 10,000 count, which is converted to the load value, is displayed.

Here, it is assumed that the control device counts up at the rise of the output of the reference time generator.
As shown in (c), when the comparator changes from Hi to Lo and the input integration is started, the input integration time (Tx) is set to tC.
An error (δ) less than L occurs. Therefore, the maximum error of the input integration time is tCL / Tx = 1 / N = 1 / 25,000.

When fCL has a sufficiently small time interval, fx
Is obtained by dividing the frequency from
When the frequency divider is reset at the moment, the input integration time synchronized with the output of the comparator is obtained. In this case, Tx = N × tCL, and the error (δ) disappears. Among microcomputers with built-in counters that are generally used as such, there are those that do not have such a frequency divider and those that cannot reset the frequency divider. Similarly, when the comparator changes from Lo to Hi and the reference integration period (TR) ends, an error (Δ) smaller than tCL occurs. In the above example, since the difference between the counts when no load is applied and when the weighing load is applied is 10,000 counts, an error of up to 1 / 10,000 will occur.

The integrator output voltage (V) at the end of the input integration time
ow) That is, the minimum voltage of the integrator output is: Vow = Vx *
Tx / (R1 * C), and even when Vx is the maximum (weighing load), the integrator output voltage must be within the output capability range of the integrator and is a finite value. If the value of R1 is increased, the charge / discharge current is reduced and C can be reduced. However, the error current called the bias current of the integrator (OP1) and the leak current of the substrate due to humidity cannot be ignored, and the effect on the performance is affected. give.

On the other hand, a capacitor has a phenomenon called dielectric absorption. Unlike an ideal capacitor, the difference between the output of the integrator and the ideal capacitor increases as the capacitor is charged, resulting in an error. If one measurement time is shortened,
The charge / discharge time is reduced and C can be reduced, and the degree of dielectric absorption is reduced. However, since the reference time interval t by the control device is finite, the ratio of the measurement time error increases, and the measurement accuracy also deteriorates.

That is, in the above example, if N is 1/10, 2500 to make Tx 1/10, then n ₀ = 200, n w =
1,200, and the maximum error of the input integration time is 1 / 2,500,
The error of the count value of the reference integration time becomes 1 / 1,000 and measurement accuracy cannot be obtained. In the case of a device having a display device, the measurement time interval is too fast, so that the display flickers too quickly, which makes it extremely difficult to read.

[0013]

The integration capacitor C of the double integration type generally has a large value. This is because when the current flowing through the integrating resistor R1 is charged and discharged (integrated) for a long time (about 0.1 to 1 second) (integrating operation), the output voltage of the integrator must be within the performance compensation range of the operational amplifier OP1. It is because it does not become. If the value of R1 is increased, the charge / discharge current is reduced, and C can be decreased. However, it is not preferable to increase R1 because an error current called a bias current of OP1 and a leak current of the substrate cannot be ignored. Further, a capacitor has a phenomenon called dielectric absorption. Unlike an ideal capacitor, the integrated output increases in error as the battery is charged, and affects the measured value. If one measurement time is shortened, the charge and discharge time is reduced and C can be reduced,
By reducing the degree of dielectric absorption and reducing the capacitance of C, the cost can be reduced and the mounting area can be reduced. However, since the ratio of the measurement time error increases, the measurement accuracy deteriorates. Further, in the case of a device having a display device, the measurement time interval is too fast, so that the display flickers too quickly and cannot be read. It also has a plurality of sensors, and the voltage obtained from each sensor is S
In the case where a digital value is obtained by one double-integration A / D converter switched by W, the voltage is transiently disturbed when the sensor is switched, and a situation occurs in which an accurate value cannot be obtained. An object of the present invention is to obtain an accurate and inexpensive A / D conversion of a double integration method in which an error generated in the above A / D conversion is minimized.

[0014]

Means for Solving the Problems The time of the minimum control time interval (command execution time) of the control device is synchronized with the reference time interval to be measured, the end of the reference integration time of the double integration method, and the input integration time. Match the start of. The input integration time of each time is set to a sufficiently small value, the count value of the reference integration time of each time is sequentially added and stored, and the added value of the count value of the set number is converted to a weight value as a count value of one measurement and displayed. I do. When A / D conversion is performed on a plurality of types of output voltages obtained from two or more different physical quantities, only the count value of the first predetermined number of the set times for one measurement is not added.

[0015]

Since one measurement is divided into sufficiently short input integration times, a conversion error due to dielectric absorption of the integration capacitor C and an error due to a bias current of the integrator and a leakage current of the substrate caused by increasing the integration resistance R1. Can be reduced, the command execution time of the control device is synchronized with the reference time interval to be measured, and the end of the reference integration time of the double integration method is coincident with the start of the input integration time. Time errors can be eliminated, and an increase in the ratio of measurement errors that occurs when one measurement is divided into short input integration times can be suppressed.

[0016]

DESCRIPTION OF THE PREFERRED EMBODIMENTS A reference time tCL for A / D conversion is synchronized with a reference time t of the control device, and a point at which SW is switched to Vx is defined as the end of the reference integration time and the start of the input integration period. The input integration time (Tx ') is determined with the number of pulses (N') of the output fCL of the reference time generator during each input integration period (Tx ') of the double integration type A / D conversion as N / m. , The count number (n) of each reference integration period (TR ')
(a)) is added to m times from a = 0 to a = m−1 to obtain a count number for one measurement. When performing A / D conversion on a plurality of types of output voltages obtained from two or more different physical quantities, addition is not performed only for the first few count values out of m addition times.

Hereinafter, the present invention will be described in detail with reference to the accompanying drawings.
The input integration time Tx 'of each time is Tx' = N '× tCL = (N ×
tCL) / m, which is 1 / m of Tx. Assuming that the count value (measurement count value) in the a-th reference integration period is n (a), the input integration time is 1 / m, so the reference integration time is also 1
/ M, TR 'is 1 / m of TR, and n (a) is n /
m.

The processing is performed as shown in FIG. a
= 0 to m-1 count value, n (0), n
(1),... N (m−1) are sequentially added, and the result (n ′) is
n (0) + n (1) +... + n (m-1), and n (a) = n / m even with a small input integration time Tx '. A value n substantially the same as when the integration time is Tx is obtained.

The individual measurement count values are described above (paragraph 000).
As described in 8), the measurement accuracy is reduced, but the error is arbitrarily varied, and the error is averaged by addition.

According to the above-described method, the A / D conversion time can be reduced to 1 / m without causing a significant decrease in accuracy, thereby reducing C and reducing the influence of dielectric absorption. be able to. In addition, in order to obtain a count for each A / D conversion m times per measurement and perform an arithmetic process,
Even in the case where the display device is provided, the count value is displayed at the same display interval as in the related art, so that a state in which the display cannot be read too early does not occur.

However, the error at the time of AD conversion does not always vary arbitrarily, and an undesired error may occur even in the case of a balance requiring accuracy even by the addition processing. As described above (paragraph 0009), the error of the input integration time is obtained by resetting the frequency divider that generates tCL.
Although the accuracy may be ensured, the error at the end of the reference integration time cannot be eliminated.

The error at the time of measurement is caused not only by tCL but also by the relation with the control device. A control device such as a microcomputer requires a certain time for each processing / control. This is because the control device performs processing / control for each t in accordance with the reference time unit t generated by the oscillator. t is called “instruction execution time”. Therefore, the integrator output Vo becomes 0 V, and the comparator outputs Hi.
The reference integration period ends at the moment of switching from to Lo,
Cannot move on to input integration period.

Since the control device cannot recognize the time less than t, an error Δ less than t occurs from the moment when the signal is switched to Lo until it is detected. This t is generally small as an error with respect to the reference time tCL for the A / D conversion, but exists as an error. Further, depending on the way of processing (software) of the control device, the number of t may be several. An error occurs and cannot be ignored.

Further, it takes time until the process of stopping the counter for counting tCL and the process of storing the count value n and the process of connecting the SW to the Vx side are completed, and Vo at this time is higher than 0V. I will go to. Then, by confirming that the comparator becomes Lo again, Vo
Is returned to 0 V, but at this stage, an error δ smaller than t occurs due to the detection of Lo again. Also in this aspect, if the measurement time is too short, the accuracy is reduced.

However, the principle of double integration is that if the voltage of the integrator output at the start and end of the measurement is the same, the sum of the charges charged / discharged during that time is zero. It is not always necessary that Vo = 0V.

Here, as shown in FIG. 2B, Hi of the output of the comparator is detected, and the counter for counting tCL is stopped.
The count value n (a) is stored, the SW is connected to the Vx side, and the counter is reset / started (starts the input integration period), while the reference time t of the control device is synchronized with tCL. This synchronization can be obtained, for example, by using the same clock generator for the control device and the measuring device and using outputs from frequency dividers having the same frequency division ratio as t and tCL, respectively.

The error Δ in the portion for detecting Hi of the comparator output is the same as that of the conventional example, but there is no portion for detecting Lo of the comparator output during the input integration period, so that no error δ occurs.

According to this method, the output of the integrator at the start of the input integration time departs from 0 V due to the processing time of the control device.
The time ΔT spent by the controller before connecting W to the Vx side
Is the amount of charge flowing into C during this period is -ΔQ, ΔV = −ΔQ / C = −ΔT *
IR / C = −ΔT * VR / (R1 * C), and VR, R
Since 1, C is a constant value, if ΔT is constant, ΔV is also constant. 2B, FIG. 3 and FIG.
It is easy to perform processing so that the required time ΔT from 6) to (2-0) is constant, and ΔV can be constant. That is, the integrator output voltage Vo at the time when SW is connected to the Vx side.
(a) is always a substantially constant value, falls within the error range for one clock, and the voltage of the integrator output at the start and end of measurement is the same, so that there is no practical problem.

On the other hand, since the reference integration period TR 'is up to the point of time when the SW is connected to the Vx side, it is larger by ΔT than the measured value n (a)', but ΔT is a constant value. Because of this, the true value n (a) is obtained by n (a) = n (a) '+ ΔT / tCL, and it is sufficient to consider that a constant value is always added, so that there is no practical problem.

Similarly, when the SW is connected to the Vx side,
If the time until the start of the measurement count in the input integration period is processed so as to be a constant value ΔTx, Tx ′ will always be constant. The true count number of the input integration period at this time (N ')
Is N "+ [Delta] Tx / tCL. Since the point at which SW is switched to Vx is the end of the reference integration time and the start of the input integration period, Tx 'causes an error like δ in the above paragraph (paragraph 0008). Therefore, the total Tx when the double integration is repeated m times is m * Tx '= m * N' / tCL, and m
Even if it is repeated twice, no error occurs in the sum of the input integration periods.

In the conventional example, if the end of the reference integration period TR is "time when the comparator becomes Hi" which occurs asynchronously with the obtained count value n, an indefinite tCL which cannot be removed in the calculation. However, in this method, the “time when SW is connected to the Vx side” is referred to as the reference integration period T.
Since the end of R 'and the synchronization of t and tCL are synchronized, the obtained count n (a) and the end of TR' are synchronized, and TR '= n
(a) * tCL, and the reference integration period is proportional to n (a) without error.

For this reason, the amount of charge Qx flowing through the integrator during the entire input integration period after m double integrations is m * N '* Vx / (R
1 * tCL), the amount of charge QR flowing from the integrator during the entire input integration period is m * (n (0) + n (1) +... + N (m-1)) * (-V
R) / (R1 * tCL). On the other hand, the integrated output voltage Vo
Is that the integrator output voltage Vo (0) at the start and the integrator output voltage Vo (m) after m repetitions are the same within the error range of Δ, so that Qx + QR = 0 and n ( 0) + n (1) +
.. + N (m-1) = m * N '* Vx / R1. That is, the input integration time is set to 1 / m of the normal value, and the integration of the count values obtained by m A / D conversions is the same as the count value obtained in the normal input integration time without error.

As described above, the reference time tCL for A / D conversion
And the reference time t of the control device, and switch SW to V
x is the end of the reference integration time,
Assuming that the time is at the start of the input integration period, the error is within a maximum of one count even if the integration is performed m times. Where S
The processing at the time when W is connected to Vx only needs to be a fixed time, so that t may be an integral multiple of tCL as long as the processing is synchronized.

The present invention is particularly effective for those having a plurality of sensors. For example, a four-point scale having four load sensors or a health meter with a fat meter having an impedance measurement sensor is often used by switching one sensor with one A / D converter. In this case, since the voltage Vx changes when the sensor is switched, a time period in which Vx is unstable often occurs transiently. The data including the transient state at the time of switching is inaccurate and needs to be discarded. However, in the conventional example, since one measurement is long, the time wasted during measurement increases. In such a measuring device, there are many cases where it is desired to obtain the values of the respective sensors almost at the same time. For this reason, measures have been taken to prevent measurement in a transient state on the circuit, but this leads to an increase in cost and a reduction in accuracy due to that.

In the present invention, since one A / D conversion time is short, it is possible to perform the measurement in the minimum necessary time only by discarding the measurement of the part where the transient state occurs. As shown in FIG. 5, if the number of A / D conversions is considered to be less than or equal to the number of times A / D conversion is considered to occur, the measured count value may not be added. Therefore, quick and easy measurement is possible without correction in the circuit.

[0036]

The conversion error due to the dielectric absorption of the integrating capacitor C, the error due to the bias current of the integrator and the leak current of the substrate are reduced, and the conversion error of the A / D conversion by the double integration method is reduced. In particular, it has a plurality of sensors, and the voltage obtained from each of them is switched by a switch to one of two sensors.
In the case of obtaining a digital value by a multiple integration type A / D converter, it is possible to easily remove a transient voltage disturbance generated at the time of sensor switching and obtain an accurate value.

[Brief description of the drawings]

FIG. 1 is an explanatory diagram of a conventional example. (A) Circuit diagram (b) Flow diagram (c) Output waveform diagram

FIG. 2 is a flowchart of an embodiment (a) a main part of the flowchart of claim 2; and (b) a flowchart of claim 1.

FIG. 3 is an output waveform diagram according to the embodiment of claim 1;

4 (a) is an output waveform diagram, and FIG. 4 (b) is a partially enlarged view of FIG.

FIG. 5 is a flowchart of an embodiment of claim 2;

FIG. 6 is an explanatory view of the third embodiment. (A) Flow chart (b) Output waveform chart

Claims (3)

[Claims] 1. A mechanism for converting a load into a voltage, an A / D converter of a double integration type, and a control device for controlling the A / D converter. In the electronic balance for obtaining the count value, the time of the minimum control time interval (command execution time / t) of the control device is synchronized with the reference time interval (tCL) to be measured, and the reference integration time of the double integration method is An electronic balance characterized by matching the end with the start of the input integration time. 2. A mechanism for converting a load into a voltage, an A / D converter of a double integration type, and a control device for controlling the A / D converter, wherein the A / D converter is controlled so as to be proportional to the load. In an electronic balance for obtaining a digital value, a storage device for setting an input integration time of each time of an A / D converter by a double integration method to a sufficiently small value and sequentially adding and storing count values of a reference integration period for each time. The electronic balance according to claim 1, wherein an addition value of the count value for the set number of times is converted into a weight value as a measurement value of one measurement. 3. The input integration time of each time of the A / D conversion by the double integration method is set to a sufficiently small value, the count value of the reference integration time of each time is sequentially added and stored, and the count value of the set number of times is calculated. In an electronic balance that converts the added value to a weight value as a count value of one measurement, the electronic balance is not added only to the count value of the first predetermined number of times of the set number of times for one measurement. Item 3. The electronic balance according to item 2.