Principles of Designing Equipment for Testing Electronic Voltage Meters

M. S. Roitman

(Presented at the 2nd Scientific and Technical Conference of the Faculty of Automation and Computer Engineering)

The steady increase in the accuracy of measuring various physical quantities and, in particular, alternating currents and voltages is dictated by the demands of rapid scientific and technological progress. At present in the USSR, voltmeters of class 0.1–0.2 [1, 2] are being produced in series, and intensive work is under way to create electronic meters of class 0.05–0.1. For testing the former, testing equipment with an error of 0.03–0.05% is required, and for the latter—on the order of 0.01%. Abroad, voltmeters with \(\delta \sim 0.025\%\) [3] are already being produced. Consequently, in the future even more highly accurate testing equipment will be needed. The most accurate instruments produced in series and intended for the calibration and testing of electronic voltmeters are compensation diode voltmeters of types V3-9, V3-24, and V4-11, measuring in the frequency range 20 Hz–1000 MHz with an error of (0.2–4)%. These instruments meet present-day requirements for the accuracy of testing voltage meters in the high-frequency range, but they do not solve the problem of testing in the audio-frequency region.

Thus, we still do not have serially produced AC testing equipment with the required error.

What, then, are the achieved levels of accuracy in measuring AC voltage in the audio-frequency range? In the USSR and the USA they are approximately the same and are characterized by an error of 0.01% [4]. Consequently, the accuracy requirements for lower-level testing equipment (0.01%) coincide with the existing accuracy at the highest level. This means that very intensive research is needed to reduce the error of the original methods and equipment*), and also to create high-quality testing equipment. The requirements imposed on reference measuring (testing) equipment of the highest and lower levels differ substantially. For the highest level, the time spent on a measurement, the operating conditions of the equipment, and its cost are not very important. In other words, an approach based on the principle of “accuracy at any price” is justified. It is obvious that for lower-level testing equipment such an approach is completely inadmissible. A whole series of limitations arises, largely contradictory and in strong de-

*) Such studies are being carried out at the D. I. Mendeleev All-Union Scientific Research Institute of Metrology under the direction of T. B. Rozhdestvenskaya, and there is definite confidence that in the near future the accuracy of the original reference measuring equipment will be brought to (0.003–0.001%).

makes the task of creating high-quality verification equipment difficult. What, then, are these limitations, fulfillment of which is highly desirable? In view of the importance of the question touched upon, let us consider it in somewhat greater detail.

Modern measuring technology is characterized by a very extensive range of measuring instruments. For example, in our country alone considerably more than 50 types of electronic voltage meters are produced. And this range increases with each year. Hence, even for purely economic reasons, it follows that verification equipment must be as universal as possible (within justified limits, of course). This leads to a limitation on the permissible value of nonlinear distortion of the generator*). Indeed, if the voltage harmonics are small, then the same equipment can be used to verify voltmeters of different systems—peak**), average, and rms values. Since precision amplitude voltmeters have an error of \(\pm 0.2\%\), the permissible nonlinear voltage distortion is \(\le 0.05\%\). Average-value voltmeters respond mainly to odd harmonics with a coefficient \(\frac{1}{n}\), where \(n\) is the harmonic number, and for their verification***) the coefficient of the third harmonic must be less than 0.03%.

Fig. 1.

Fig. 2.

Along with the indicated requirements for metrological characteristics, it is necessary to satisfy requirements for the minimum cost of the equipment, especially its operation. Of extraordinary importance for reference verification equipment of the lower levels are the verification time and the possibility of “self-verification.”

In the general case, to determine the basic and frequency error of a voltmeter, a highly stable generator with small nonlinear distortions, a reference voltage meter, and scaling devices (reference dividers or amplifiers) are necessary.

There are two fundamentally different variants for verifying a measuring instrument.

In a general interpretation, the first variant is the comparison of two meters, one of which is a reference instrument; the second is monitoring the response of the instrument under verification to a deliberately known quantity. The first is carried out according to the functional scheme shown in Fig. 1, where: G is the generator, OI is the reference meter, PP is the instrument under verification.

Verification consists in recording the magnitudes of the mutual deviations in the readings of OI and PP. The second variant has two substantially different modifications.

*) The use of generators of rectangular or any other nonsinusoidal voltage waveform for determining the basic and frequency error of voltmeters is associated with a number of difficulties [5] and therefore cannot be recommended.

**) In some cases, verification of peak-value voltmeters using sinusoidal voltage is sufficient, since knowledge of the influence of duty factor is not required.

***) Voltmeters of class 0.05% are meant.

In the first (Fig. 2), the voltage from the output of a stable source (I) is monitored by an indicator of the deviation of the signal level from the set value*) (ID). The required gradations of the output signal are obtained through the use of scaling devices (SD).

In the second modification (Fig. 3), the voltage from the output of the generator-calibrator**) (GC) is applied to the instrument under verification (IU).

In terms of verification speed, the second version is preferable, since it eliminates the need to take readings from the reference measuring instrument.

The second version can have very substantial advantages in terms of the attainable accuracy as well. This is explained by the fact that the accuracy of a measuring instrument operating in the vicinity of a single optimal point is higher than when it operates over a range of values.

Thus, in principle, it is easier to produce an accurate value of \(U\) than to measure it.

The foregoing is valid only if the error of the scaling devices is significantly lower than the error of the wide-range MI.

Fig. 3.

Verification both according to the first version and according to the first modification of the second requires the presence of a source whose voltage remains at least constant during the verification. Otherwise, impermissible fluctuations of the instrument readings will occur. This obvious requirement leads to the need to stabilize the generator output. But if the voltage-deviation indicator has acceptable stability, then it is expedient to use the signal from the ID output for automatic action on the generator and to create a single verification complex in the form of a generator-calibrator [6]. This makes it possible to sharply simplify the apparatus by combining the functions of the measuring transducer, speed up and facilitate verification, and improve metrological characteristics (the output resistance of the source, its fluctuations, etc., are reduced).

However, it should be borne in mind that a measuring transducer is applicable in a generator-calibrator system only if it can have an electrical output and, in its characteristics—above all its dynamic characteristics—is “fitted” into the automation system.

All that has been said can be summarized as follows:

1. In all cases where high-quality scaling devices are available, preference should be given to verification by the method of monitoring the response of the IU to a signal of known magnitude.

2. Any version of the verification circuit does not eliminate the need to have a highly stable source of alternating voltage.

Taking this and a number of other important circumstances into account, it is most preferable to construct verification apparatus in the form of a generator-calibrator*.

An indispensable part of the GC is the measuring transducer (MT). It is precisely this, in the final analysis, that determines the potential accuracy of the GC.

Excluding from consideration transducers of amplitude value, above all because of their high sensitivity to harmoni—

*) The ID is in fact a reference signal meter operating in a very narrow measurement range of the signal (in the vicinity of the “point”).

**) The term “generator-calibrator” is commonly understood to mean a source whose output value can be reproduced with a known accuracy. In the technical literature, especially foreign literature, it would be more correct to call such sources standards or voltage measures.

...to shocks and interference, let us compare the capabilities of transducers of average and rms values. The most advanced average-value transducers are converters based on amplifiers with rectifiers in the negative-feedback circuit [7–10]. Such transducers have come to be called converters.

In the low-frequency region, converters have very low instability \((<0.01\%)\) and error \((<0.01+0.02\%)\). But already at a frequency of 20 kHz the error increases to 0.05% and more.

This is due to a number of reasons. First, as the frequency increases, the effectiveness of negative feedback in operational amplifiers decreases.

Second, as the frequency increases, transient processes in the diodes and parasitic reactances in the MT begin to have an effect.

Thus, because of the significant frequency error of converters, on their basis it is possible to create highly stable and high-quality verification equipment only for the audio-frequency range.

Converters also have other shortcomings, the principal ones being sensitivity to harmonics and, most importantly, the impossibility of “direct self-verification.” Element-by-element verification can, of course, be performed. But it requires additional equipment and additional time expenditures. The accuracy of element-by-element verification is, as a rule, lower than that of direct verification.

The frequency range of verification equipment can be greatly extended by using measuring transducers by rms value. It should be especially noted that rms-value MTs are used in all metrological organizations of the world for creating standard voltage meters in the audio- and ultrasonic-frequency ranges. The initial measurement method is the method of comparing an unknown ac value with a dc value. Obviously, in this case the MTs must respond identically to one and the same value of ac or dc voltage. Special thermotransducers are used mainly as rms-value MTs. Unfortunately, they have a number of substantial shortcomings [11, 12].

1. Extremely low sensitivity.

2. Significant dependence of the thermal emf on ambient temperature.

3. The presence of transfer errors and errors due to asymmetry.

The first two factors do not yet create insurmountable obstacles to the use of thermotransducers in metrological organizations. True, the “price” of these obstacles is very high: gold connecting leads, precise thermostating, electrostatic and electromagnetic shielding, and special highly sensitive null indicators are used [12].

Transfer error and asymmetry are due to the fact that in a single-element thermoelectric transducer we measure not the heat flux released by the heater under the action of the current flowing through it, but the temperature of the heater at a certain point. Owing to the Thomson and Peltier effects, the temperature at this point, for the same rms value of current, will be different for direct and alternating currents, and also for different polarities of direct current. By special measures and by introducing corrections, the error can be taken into account to a considerable extent (but far from completely)*). To reduce this error sharply, a multielement thermotransducer was developed at VNIIM [11]. The same approach has also been used in other countries [14].

However, for multielement converters the frequency range is sharply limited from above. Owing to the indicated shortcomings of thermal converters, their use in verification equipment of lower levels (considering that it must operate outside metrological laboratories as well, and that the permissible errors lie within the range 0.01–0.05%) over a wide frequency range is highly undesirable.

All that has been said also makes it necessary to investigate the possibility of using various other measuring transducers for comparison purposes.

In our opinion, the most promising group of measuring transducers are photoelectric converters (PC), which are a combination of an element heated by the measured current and a light-sensitive element. PCs can be created on the basis of low-power incandescent lamps and photoresistors, photodiodes, barrier-layer photoelements (“solar” cells), or special devices with a cast high-ohmic microwire and highly sensitive bolometers. The merit of these converters is their high sensitivity, the small dependence of their parameters on ambient temperature, and the possibility of reducing the asymmetry error to a negligibly small value without at the same time reducing the frequency range. The latter is achieved by adjusting the orientation of the filament body relative to the light-sensitive element, or by integrating the luminous flux. It is still more rational to use the totality of the indicated measures. The principal factor that has impeded the realization of the great advantages of photoelectric converters and the reduction of the comparison error has been (and to some extent still is) their insufficient time stability.

Unfortunately, all known technical methods for reducing the influence of the time instability of measuring transducers on the measurement error are associated either with large expenditures of time or with the need for a considerable complication of the apparatus, which is justified only in the creation of reference installations of the highest level.

Indeed, in the simplest case, when the instability is a monotonic function of time, we are forced to select a second converter with an approximately similar drift characteristic (this is not very easy to do) or to introduce automatic correction of the circuit. Let us note that application of the automatic-substitution method (often called the control-signal method or the single-channel method) does not provide a substantial gain. This is explained as follows. Measuring transducers for mean and rms voltage have high inertia, since their time constant must be much greater*) than the duration of the period of the measured voltage [7]. Therefore the switching frequency cannot be chosen high. The transient processes arising during switching have to be “cut out” [14] in order to avoid large additional errors, which also limits the switching frequency from above.

If the drift is a random function of time, then the processing of a considerable number of statistically independent measurements is necessary, which entails large expenditures of time even if the process is automated. What has been said explains the importance of time stabil-

*) Under unchanged environmental conditions, the error due to the influence of the Thomson and Peltier effects is systematic and reduces the accuracy of absolute measurements, without affecting the accuracy of relative measurements.

*) The elegant solution available for increasing the speed of measuring transducers [15] is inevitably associated with a loss of accuracy.

of the instability of the IR (in most cases it would be more correct to speak of the magnitude of short-term instability) [11]. The strongest influence on the stability of the FP is exerted by vibration of the assembly [16]. This is explained by the imperfection of the contact between the filament and the crosspieces and, mainly, by the influence of the intermediate supporting crosspiece in mass-produced low-power incandescent lamps (NSM 6×20 and NSM 9×60).

It follows from what has been said that there is an urgent need for detailed investigations aimed at creating metrological lamps.

The real possibilities in this direction are very great. For example, among samples of experimental incandescent lamps developed to our order by A. I. Lysov, there are specimens with an instability of less than 0.001% per hour*) [17], which is at the level of the best thermoelectric converters available in metrological organizations. But even these data cannot be regarded as close to the limiting attainable values.

The situation is still considerably worse, for the time being, with the time stability of light-sensitive elements. The time drift of photoresistors and the strongly manifested “breathing” effect of SF2-2 and SF3-2 [18, 19] lead to the need for special selection of photoconverters, which is quite undesirable. New photoresistors (SF2-5 and special experimental ones) have substantially lower (approximately by an order of magnitude) reduced instability. As the design and manufacturing technology of photoresistors) are improved, they will possess ever greater stability. Encouraging results are being obtained in investigations of silicon photodiodes*). All that has been said permits the conclusion that photoelectric converters are very promising.

A great advantage of FPs is not only the possibility of creating on their basis high-accuracy measuring and verification apparatus, but also the possibility of greatly simplifying it [20, 21]. In other words, the use of FPs makes it possible to solve most rationally the problem of creating AC verification apparatus operating over a wide frequency range.

Of exceptionally great importance in creating wide-range AC verification apparatus are scaling and matching devices. At present, precision voltage dividers capable of operating in the range from low to high frequencies inclusive are a very weak point. In fact, we do not have such dividers, which to a considerable extent restrains expansion of the “dynamic” range of verification apparatus. In this connection, in the verification installations being created we are forced to split the functions of the precision divider and to use two dividers: one—an accurate inductive divider, operating in a narrow frequency range and making it possible to determine the principal error of the instrument being verified (it gives the absolute “tie-in”); the second—a low-accuracy divider, but one made as far as possible frequency-independent, makes it possible to determine the frequency error.

*) The lamps operated at \(U \approx 1\ \text{V}\) and \(I \approx 32\ \text{mA}\), and at a filament temperature of 900–1000°K.

**) The production of paired photoresistors on a common substrate and in a single sealed bulb is especially desirable.

***) The results of comprehensive investigations of various light-sensitive elements will be published in the near future.

References

1. S. M. Pigin. High-accuracy electrostatic instruments. Abstract of dissertation. L., 1962.

2. Yu. M. Tuz. Precision ac voltmeters. Avtometriya, No. 7, 1967.

3. Radio Electronics Abroad, No. 10 (344), 1966.

4. R. F. Aknayev, A. B. Berezovskii, T. B. Rozhdestvenskaya. The state and prospects for the development of initial methods for measuring the rms value of voltage over a wide range of measured quantities and frequencies. Abstracts of the Scientific and Technical Conference on Radio-Engineering Measurements. Novosibirsk, 1967.

5. M. S. Roitman. On calibrating voltmeters with rectangular-wave voltage. Problems of Radio Electronics, series “Radio-Measurement Technology,” issue 3, 1966.

6. M. S. Roitman, E. I. Tsimbalist, Yu. M. Fomichev. A device for obtaining stable and calibrated rms sinusoidal voltage. Inventor’s certificate No. 196114, bulletin No. 11, 1967.

7. B. I. Shvetskii. Electronic measuring instruments with digital readout. Tekhnika Publishing House, Kiev, 1964.

8. H. Salman. Average value transfer standard. Inst. and Control Systems, No. 1, 1963.

9. P. Richman. A New Absolute A. C. Voltage Standard. Session 26, IEEE Record of convention held March, 1963.

10. P. Richman. The application of electronics to instruments for standards and calibration, IEEE Intern. Con. Rec., 1964, V. 12, N 8.

11. T. B. Rozhdestvenskaya. Electrical comparators for precise measurements of current, voltage, and power. Standards Publishing House, Moscow, 1964.

12. Turgel Raymond. A comparator for thermal ac—dc transfer standards. “21 Annual ISA Conf. Proc., Vol. 21, Part 1, 1966” (translation Kut. N35, 1967).

13. F. I. Wilkins. Multijunction Thermal Converters. IEE Proceedings, Vol. 112, N 4, 1965.

14. B. A. Perminov, M. S. Roitman, E. I. Tsimbalist. An ac comparator based on photoelectric converters. Avtometriya, No. 5, 1965.

15. A. M. Kovalev. A method for measuring the rms value of ac voltage of arbitrary waveform. Inventor’s certificate No. 173844. Bulletin of Inventions, No. 16, 1965.

16. M. S. Roitman, E. I. Tsimbalist, A. I. Lysov. A photoelectric converter as a controlled element of an electrical circuit. Proceedings of Higher Educational Institutions of the USSR—Instrumentation, No. 6, 1966.

17. M. S. Roitman, V. R. Tsibulskii. A photoelectric converter with a silicon photodiode. Proceedings of TPI, vol. 171, TSU Publishing House, 1967.

18. M. S. Roitman. A sharply selective wide-band phase detector based on photoelectric converters. Automation and Remote Control, No. 5, 1967.

19. M. S. Roitman, E. I. Tsimbalist, Yu. M. Fomichev. A source of stable and calibrated ac voltage. Avtometriya, No. 3, 1967.

20. M. S. Roitman, Yu. M. Fomichev, E. I. Tsimbalist. Model source of alternating voltages GK-4. Proceedings of TPI, vol. 171, TSU Publishing House, 1968.

21. M. S. Roitman, Yu. M. Fomichev, E. I. Tsimbalist. Sources of calibrated ac voltage. Proceedings of the Jubilee Scientific Conference on Radio-Engineering Measurements (1967), in press.