346 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 38, NO. 1. APRIL 1989

A New Isothermal Multijunction Differential Thermal Element Provides Fast Settling AC to DC Converter

FRED L. KATZMANN, SENIOR MEMBER, IEEE

Abstract-An isothermal multijunction differential thermal con- verter has been reduced to practice by constructing two separate in- sulated heaters of a vacuum thermal element and placing them in ther- mal contact. One heater carries the input signal while the second heater circuit provides negative feedback to maintain the dual heater struc- ture at a constant temperature. The mass of the heaters and thermo- couple temperature sensors is small and the temperature variations are limited to millidegrees Celsius. Therefore, the response time of this thermal converter is a fraction of a second which permits ac-dc con- version 5-10 times faster than conventional thermal devices. Measure- ment accuracy is comparable to conventional hot wire single junction thermal converters and widehand frequency capability beyond 100 MHz makes this new thermal converter applicable to precision ac-dc transfer standards and digital ac voltmeters.

I. INTRODUCTION UCH WORK has been done over the past 30 years M to improve thermal voltage converters used as the

basis for ac-dc transfer difference and ac voltage mea- surement. Hot wire thermal converters have been used in ac measurement for a century. Investigations by Widdis [l], Hermach [2], and Inglis [3] gave insights into the functional theory and performance of thermal converters. Most of these basic studies dealt with single thermocouple junction hot wire thermal converters.

The need to extend the high frequency capability of sin- gle junction thermal elements (SJTE) created the planar UHF pattern enclosure which permits applications be- yond l GHz. Ac-dc transfer differences and dc reversal errors caused by Peltier and Thomson effects resulted in the development of multijunction thermal elements (MJTE). The multijunction thermal elements as described by Wilkins [4] and Klonz [SI increase the number of heat sensing bimetallic (Seebeck effect) thermocouples along the heater. This reduces dc reversal error and increases the dc output voltage since the thermocouples are con- nected in series. The MJTE provides good ac-dc differ- ence measurement performance and represents an im- provement over conventional SJTE. Unfortunately, the MJTE frequency capability is sharply curtailed above 100 kHz. The MJTE is expensive and has found application mainly in primary standards of national laboratories.

Thermovoltaic sensing of heater temperature by See- beck effect thermocouples has limitations which have to

Manuscript received June 10, 1988. The author IS with Ballantine Laboratories, Inc., Boonton, NJ 07005. IEEE Log Number 8826020.

U U

Fig. 1. IMJDTE UHF pattern.

be overcome to improve the accuracy and speed of ac measurements. Thermistor temperature sensing was ex- plored by Widdis [6] and Richman [7]. Thermoresistive sensing of heater temperature using an insulated wire with a high temperature coefficient of resistivity resulted in dy- namic range of greater than 10: 1 [8]. Neither of these temperature sensing methods have been aggressively ap- plied, since thermistor or thermoresistive temperature sensing is affected by ambient temperature and requires housing the devices in an oven with good temperature control.

Infrared sensing of heater temperature appears to have merit, since the measurement would sense infrared radia- tion along the length of the heater. This may minimize the errors associated with point contact temperature sensing using thermocouples. Unfortunately, the stability of available infrared detectors in the 50°C- 150°C region limits their usefulness for accurate ac-dc transfer mea- surements. An added limitation involves the infrared filter characteristics of the glass which houses the thermoele- ment. Only the shorter wavelengths above the visible or- ange region are transmitted so that infrared sensing has only been used for heater overload detection [9].

Thermal transfer devices using semiconductor temper- ature sensing have been described by Burr-Brown in 1972, Fluke in 1984, and recently by Linear Technology on their Model LT1088. These devices deposit a resistive heater

0018-9456/89/0400-0346$01 .OO O 1989 IEEE

KATZMANN: DIFFERENTIAL THERMAL ELEMENT 341

I 1 STANDARD PATTERN I I FREQUENCY 1 IHJDTE AC-DC DIFFERENCE* I

1 kXz 10 iHz 50 kHr

100 kEz 500 kHz

1 WHz 10 MHz 30 nHz 50 WHz

100 I H Z

Performance +10 ppm

-8 P P ~ +12 ppm

-7 PPm -10 ppm +80 ppm +350 ppm +1010 PPb +6250 ppm

* Compared to coaxial thermal converter with SJTE

Fig. 2 . IMJDTE frequency response.

Fig. 3. IMJDTE operating circuit.

close to one or more diodes or transistor base-emitter junctions. Their advantage appears to be higher dc output voltage and lower mass with faster settling time. Defi- ciencies include: 1) a limitation in signal-to-noise ratio in the semiconductor junction when compared to Johnson noise in thermocouples and optimized low noise ampli- fiers exterior to the thermal device, 2) lower frequency capability compared to hot wire devices, arid 3) indeter- minate long term stability. These devices generally ex- hibit limited ac-dc transfer linearity over a wider operat- ing range and require dynamic correction which mandates computation and several successive measurements to ob- tain accurate readings in approximately 6 s . The Fluke Company has successfully used semiconductor converters in the Model 8506 digital voltmeter which reports uncer- tainty of 120 ppm below 2 kHz.

11. BASIC PRINCIPLE OF THE DUAL HEATER CONVERTER

Fig. 1 shows the UHF pattern embodiment of the iso- thermal multijunction differential thermal element (IMJDTE). The UHF pattern IMJDTE is usable beyond 250 MHz. It has also been configured into the standard pattern where a usable frequency range of more than 50 MHz has been noted.

Fig. 2 lists the frequency response characteristics of a standard pattern IMJDTE mounted in an aluminum block heat sink. Each heater has a nominal resistance of 70 Q. Current in the signal input heater was limited with a 2000 Q resistor. The comparison was made using a high quality 10-V coaxial thermal converter as the ac/dc difference standard. Further improvements in response may be im- plemented in the UHF configuration by frequency flatness compensation.

The new thermal element uses two insulated wire heat- ers. The heater wires each have a diameter of 12.7 pm (0.005 in) to minimize mass and are insulated with high temperature insulation such as Dupont PY R-ML enamel. The heater wires are made of 80-20 Nichrome alloy and are twisted together at approximately 7 turns per centi- meter to assure maximum heat transfer through distrib- uted contact. Two chromel-constantan high output ther- mocouples are attached to the heaters with insulating ceramic cement. The thermocouples have a total resis- tance of 28 Q and are connected in series for maximum dc output.

Fig. 3 shows the simplified diagram of the operating circuit for rms to dc conversion. The heater receiving the rms input signal is independent of all other elements of the IMJDTE. Isolation is excellent at dc and decreases

348 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL 38. NO 2. APRIL 1989

Fig. 4. IMJDTE settling time with feedback circuit gain Fig. 5. IMJDTE operating circuit waveform.

somewhat at higher frequencies due to a coupling capac- itance of nominally 1 pF. The second heater is used to apply negative feedback to the dual heater structure. The circuit is initially biased so that the feedback heater car- ries the rated heater current when the signal input heater carries no current. As the signal input current begins to flow, the temperature of both heaters is temporarily in- creased and raises the dc output from the two thermocou- ples. This increased dc output provides signal to the low noise op-amp and consequently decreases the current to the feedback heater. Once the input heater current ampli- tude becomes constant, the feedback system returns the dual heater structure to its initial temperature. The loop gain is as high as the system stability permits. The loop incorporates integration to limit noise effects. A phase lead network is also provided to assure fastest settling com- mensurate with critical damping and system stability. The dc output from the rms to dc converter settles within 300 ms to 100 ppm of final value after a step of input heater current is applied to the input heater. Settling times are commensurate with noise reduction and settling times of less than 80 ms have been observed.

Fig. 4 shows an oscillogram where the upper trace de- picts the normal IMJDTE response using adequate gain to the feedback heater. The lower trace shows the ex- tended settling time when less gain is applied in the feed- back loop. Controlling the settling electrically simplifies the compromise between operating speed at higher fre- quency and accuracy at very low frequencies to 10 Hz. The signal applied to the input heater comprises two dc amplitudes alternately switched within a 500-ms period.

111. IMJDTE OPERATION The IMJDTE operates on the principle of constant (iso-

thermal) heater temperature. In the commonly used single heater, single thermocouple thermal element (SJTE) the heater temperature can rise by 150°C from zero heater

' . m m ' 1.428 '! 78.335 +7 .593

200 : 2 . 8 5 7 '; i6.988 ' I S 0 ~ 2 . 1 A i *'

' 400 j 5 . 7 1 4 " 1 2 . a?. 3 ~ 450 I 6.428 'i -13.1068

-7.616 500 , 7 . 1 4 3 ,; -2 .112 550 ; 7.85 '1 ,;

Fig. 6 . IMJDTE transfer characteristic

current to the usual 5-mA full scale current. The SJTE may take up to 10 s to reach thermal stability and provide a constant dc output. In Wilkins type multijunction ther- mal elements the heater temperature may rise only 30°C. The mass of the heater and multiple thermocouple struc- ture of the Wilkins type of multijunction thermal con- verter is about 3 times greater. Therefore, the settling time may be 40 s or longer to reach dc output stability after applying a step of heater current. The new IMJDTE op- erates in a manner where the summed total power input of both heaters remains constant as measured by the ther- mocouples attached to the two twisted heaters. Constant heater power infers constant operating temperature. Therefore, as the signal input heater current increases, the second reference, or feedback heater must have its current decreased proportionately. Fig. 3 shows the simplified circuit diagram where active negative feedback with high loop gain ( >40 000) rapidly maintains constant heater temperature within a few millidegrees centigrade. Al- though the structural mass of the IMJDTE is greater than the mass of the standard SJTE, the IMJDTE dual heater sustains insignificant net temperature deviation during ac- dc conversion. Settling time of experimental IMJDTE was observed to be approximately 60 times faster than a con- ventional thermal element.

KATZMANN: DIFFERENTIAL THERMAL ELEMENT 349

DC-SDIITLO

Fig. 7. IMJDTE plot of input/output transfer characteristic.

tal IMJDTE was observed to be approximately 60 times faster than a conventional thermal element.

Fig. 5 shows an oscillogram where the upper trace de- picts the noisy waveform at the input to the IMJDTE feed- back heater and the integrated waveform at the output sig- nal port described in Fig. 3. The upper trace on the oscillogram of Fig. 5 shows the signal applied to the IMJDTE feedback heater. It has a rapid rise time, exhibits some overshoot, and is noisy. The output signal is the lower trace and shows the noise reduction due to the filter and integrator in the slow loop.

The operating circuit uses two feedback loops as first described by Cox and Kuster [l 11. A dc amplifier con- sisting of 4 low noise LT1028 amplifiers, connected in parallel, statistically reduces the effects of shot noise and 1 /f noise. The output from these amplifiers is summed and then applied to another summing amplifier. This rep- resents thefust loop to assure rapid settling. The summed output from the 4 low noise input amplifiers is also ap- plied to a 3 pole active low-pass filter with cutoff at 5 Hz. The filter output is applied to an integrator which provides further filtering and very high dc gain. The noise reduced output signal is taken at the output of the integrator. It is also applied through an inverting amplifier to the sum- ming amplifier which drives the feedback heater through a phase lead network.

The IMJDTE has a dc output voltage of nominally 12-

14 mV and a source impedance (thermocouple resistance) of 26-36 a. The hot junction of the thermocouple oper- ates at approximately 135°C (408°K) and the dc output amplifier bandwidth is approximately 5 Hz. Conse- quently, the noise of the 3 6 4 source may be calculated as 2.01-nV rms and 8.52-nV peak-to-peak. Therefore, with a nominal output voltage of 12 mV, the signal-to- noise performance permits resolution of better than 0.7 PPm.

Fig. 6 shows a listing of measurements to obtain the input/output characteristics of the IMJDTE using the feedback circuit described later in this report. Fig. 7 shows a plot of this data on a semilogarithmic scale indicating the square law characteristics of the new device. In ad- dition, the observed dc reversal difference was nominally 5 ppm. This is believed to be due to reduced Thomson effect in the constant temperature heater structure and by using more than one thermocouple to sense temperature at several points distributed along the heater.

The new IMJDTE achieves less than 1-s settling time. It provides good square law conversion, frequency capa- bility greater than 50 MHz (standard pattern), lowest noise, and appears to have the same long term stability established for vacuum thermal elements. It adds en- hanced capability to an established concept which has been proven in precision ac-dc standards used by all na- tional standards laboratories throughout the world.

350 IEEE TRANSACTIONS ON INSTRUMENTATlON AND MEASUREMENT, VOL 38. NO 2. APRIL 1989

IV. CONCLUSION RFERENCES Providing ‘Onstant heater temperature by using a [I] F. C. Widdis, “The theory of Peltier and Thomson effects errors in

thermal transfer devices,” Proc. Inst. Elec. Eng . , vol. 497, p. 328, heater structure with negative feedback to the reference heater permits substantial improvement in the response time of ac-dc The well established precision stability wide frequency range and long term reliability

Jan. 1962. 121 F. L. Hermach and E. D. Williams, “Thermal converters for audio-

frequency voltage measurements of high accuracy,” IEEE Trans. In- strum. Meas , vol. 1M-15, pp 260-286, 1966.

proven by several years of testing appear comparable to conventional single heater vacuum thermal elements. The new IMJDTE permits a thermal device to have improved measuring speed compared to logarithmic circuits or pres- ent semiconductor sensing thermal converters. The IMJDTE has been applied to ac voltage standards [ 121. It is also proposed for ac-dc transfer standards, digital ac voltmeters, ac null meters, and precision thermal con- verters where the new device permits improved measure- ment accuracy due to shorter measurement cycles and the application of computer-aided techniques.

ACKNOWLEDGMENT The author thanks D. Stollery and S. Colsen for build-

ing the new isothermal elements in both UHF and stan- dard configuration. Special thanks are also extended to A. Harrison and I. Luck for constructing the circuits and per- forming the tedious measurements.

[3] B. D. Inglis, “A method forthe determination of ac-dc transfer errors in thermoelement,” IEEE Trans. Istrum. Meas . , vol. IM-27, p. 44, Dec. 1978.

141 F. J. Wilkins et al., “Multijunction thermal converter,” Proc. Inst. Elec. Eng., vol. 112, pp. 794-806, Apr. 1965.

[5] M. Klonz, “Ac-dc transfer difference of the PTB multijunction ther- mal converter in the frequency range of 10 Hz to 100 kHz,” IEEE Trans. Instrum. Meas . , vol. IM-36, pp. 320-329, June 1987.

[6] F. C. Widdis, “The indirectly heated thermistor as a precise ac-dc transfer device,” Proc. Inst. Elec. Eng. , vol. 103, part B, no. 12, Nov. 1956.

[7] P. L. Richman, “A new wideband rms to dc converter,” IEEE Trans. Instrum. Meas. , vol. IM-16, pp. 129-134, June 1967.

[8] F. L. Katzmann, “A thermoresistive ac-dc transfer element,” IEEE Trans. Instrum. Meas. , vol. IM-35, pp. 580-584, Dec. 1986.

[9] R. Gerr and F. L. Katzmann, “Automating wide-band acldc transfer measurements,” IEEE Trans. Instrum. Meas . , vol. IM-25, pp. 533- 537, Dec. 1976.

[IO] Model 8506A Thermal True RMS Multimeter. Instruction Manual, P/N 638858, Para. 3-31, John Fluke Mfg. Co., Everett. WA.

1111 L. G . Cox and N. L. Kuster, “An automatic rms/dc comparator,” IEEE Trans. Instrum. Meas . , vol. IM-23, pp. 322-325, Dec. 1974.

[ 121 Model 6400A AC Calibration Standard. Technical Data Literature and Instruction Manual. Ballantine Laboratories, Inc., Boonton, NJ.