You are on page 1of 2

REVIEW OF SCIENTIFIC INSTRUMENTS

VOLUME 75, NUMBER 3

MARCH 2004

LETTERS

Comment on Implementing of a precision fast thermoelectric cooler using a personal computer parallel port connection and ADN8830 controller Rev. Sci. Instrum. 74, 3862 2003
A. W. Slomana)
Sophiaweg 135, Heiliglandstichting, 6564AB, The Netherlands

Received 13 October 2003; accepted 21 December 2003 Eli Flaxer has described a feedback controlled circuit to drive a Peltier junction to control the temperature of a specimen in the range 0 to 50 C in a room temperature environment Rev. Sci. Instrum. 74, 3862 2003. The amount of heat transferred per unit current by a Peltier junction varies substantially with the temperature difference across the Peltier junction. Flaxers circuit does not provide any mechanism to compensate for this variation. This means that if the proportional-integral-differential control loop he uses is optimized at any particular temperature differential, the control loop with be over-damped at lower specimen temperatures, and under-damped at higher specimen temperatures. A circuit using a second thermistor to monitor the temperature of the exhaust side of the Peltier junction, and a digital control loop, can minimize this problem A. W. Sloman, Paul Buggs, James Molloy, and Douglas Stewart, Meas. Sci. Technol. 7, 1653 1996. This circuit has the incidental advantage of offering ten times better temperature stability. 2004 American Institute of Physics. DOI: 10.1063/1.1647699

Flaxer1 describes a circuit to drive a Peltier junction to control the temperature of a specimen in the range 0 to 50 C in a room temperature environment. His circuit is based on the Analog Devices ADN8830 Thermoelectric Cooler Controller.2 This incorporates an analog proportional-integral-differential PID controller, and generates a pulse-width-modulated PWM output. This integrated circuit, and the similar Linear Technology LTC1923 High Efciency Thermoelectric Cooler Controller,3 are both admirably compact and economical solutions to the problems of driving a Peltier junction. The combination of sensitive analog circuitry with a fastswitching PWM driver does appear to limit the temperature stabilities that can be attained with these devices to no more than 0.01 C, while the analog PID controllers can only be tuned by adjusting physical capacitances and resistances. These limitations do not seem to present a problem in the market for which these devices were designed. In applications where there is more space available for the control electronics, published circuits using general purpose integrated circuits can offer much better temperature stability. In 1990, Bradley, Chen, and Huet4 describes a fully analog system built around a standard instrumentation amplier and a number of conventional op amps driving a discrete high current linear output stage which attained the respectable stability of 300 C.
a

The problem of tuning the control loop in such a circuit is that the effectiveness of a Peltier junctionin watts of ` re of heat transferred in or out of the specimen per ampe current driven through the junctionchanges markedly as the temperature difference across the Peltier junction changes. This variation changes the gain around the temperature control loop, making it impossible to tune an analog PID controller for critical damping over an extended range of operating temperatures. This is mentioned in both the Linear Technology data sheet for the LTC1923 and in the associated application note AN-89.5 Unfortunately, neither document tells you how to calculate the changing effectiveness of the Peltier junction. A circuit in which the control loop is implemented in the digital domain can offer much more exible tuning of the control loop. In 1993, my colleagues and I at Afnity Sensors, Bar Hill, Cambridge CB3 8SL UK, developed a fully digital temperature controller, using a PWM output to drive a Peltier junction, that offered a stability of 1 mC over a range of 4 to 38 C. This was subsequently published Sloman, Buggs, Molloy, and Douglas 19966. Our article includes an expression for calculating the temperature difference across a Peltier junction, T b
2 T b 2.I . T max / I max I 2 . T max . 1 R . Q max / T max / I max

Q . T max . 1 R . Q max / T max / Q max , where I is the current through the Peltier junction in amperes, Q is the heat owing into the input side of the Peltier
788 2004 American Institute of Physics

Electronic mail: bill.sloman@ieee.org

0034-6748/2004/75(3)/788/2/$22.00

Downloaded 05 Jan 2009 to 210.187.49.65. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/rsi/copyright.jsp

Rev. Sci. Instrum., Vol. 75, No. 3, March 2004

Letters

789

junction in watts, and Q max is the maximum heat transferable by the Peltier junction at zero temperature difference in watts. I max in amperes is the maximum current that can usefully be driven through the Peltier junction, T max is the maximum temperature difference that can be generated across the Peltier junction at zero thermal load in degrees Kelvin, and R is the sum of the thermal resistance between the element whose temperature is being controlled and the input side of the Peltier junction, plus the thermal resistance from the output side of the Peltier junction to ambient. The parameters I max , Qmax , and T max are available in the manufacturers data sheets for single-stage Peltier junctions. The aggregate thermal resistance R can be estimatedwith adequate accuracyfrom the various manufacturers data for the thermal coupling compounds and heatsink used. The expression is not particularly accurateit does not capture the variation in the effectiveness of the Peltier junction with changes in the absolute temperature of the junction. It is sufciently accurate to be used to generate worthwhile corrections to the proportional, integral, and differential gain terms in a PID controller as a function of the temperature difference across the Peltier junction. Because the system we developed was fully digital, it was relatively easy to adjust the gain terms for our PID controller on the y. The processing capacity required for such a system is quite modestwe used a Siemens SAB80C517A eight-bit single-chip microcontroller for all the computation involved. This microcontroller included a ten-bit analog-to-digital converter ADC which we used to monitor the output of a thermistor mounted on the exhaust side of the Peltier junction. The thermistor on the specimen whose temperature was being controlled had to be monitored more precisely, and we used a 20-bit sigma-delta ADC, which delivered 20 conversions per second to the microcontroller with a nominal resolution of about 100 C. Neither the sigma-delta converter nor the micro-controller were expensive components back in

1993, and their equivalents are even cheaper now. The microcontroller chosen was probably more powerful than we neededthe part was already used elsewhere in our product, and was cheap enough that we could not justify spending time looking for a cheaper part. Our system would seem to provide a better basis for a wide-range general-purpose Peltier-based temperature controller than the system described by Flaxer. It is worth noting that Flaxer has failed to discuss one of the disadvantages of the fast pulse-width modulated output stageelectromagnetic interference generated by the higher harmonics associated with the fast switching edges. This is discussed in the Linear Technology applications note and data sheet, and the Linear Technology LTC1923 offers the option of slowed-down switching edges. The wound inductor Flaxer uses to lter the PWM output has a self-resonant frequency of 45 MHz, which means that it looks like a 2.7 pF capacitor at frequencies above this, while the electrolytic capacitor has an equivalent series resistance of 35 m, which dominates its impedance above 200 kHz up to the frequency not available from the manufacturers data where the inductance of the internal connections starts presenting a signicant impedance. We used a nonwound inductor ferrite bead in series with our much higher inductance wound inductors to maximize the series impedance at high frequencies, and by-passed our much higher capacitance electrolytic capacitor with a 100 nF ceramic disk to minimize the parallel impedance at high frequencies.
Eli Flaxer, Rev. Sci. Instrum. 74, 3862 2003. Analog Devices, ADN8830 Datasheet http://www.analog.com/ 3 Linear Technology, LTC1923 Datasheet http://www.linear.com/ 4 C. C. Bradley, J. Chen, and R. G. Huet, Rev. Sci. Instrum. 61, 2097 1990. 5 Linear Technology, Application note AN-89 http://www.linear.com/ 6 A. W. Sloman, Paul Buggs, James Molloy, and Douglas Stewart, Meas. Sci. Technol. 7, 1653 1996.
1 2

Downloaded 05 Jan 2009 to 210.187.49.65. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/rsi/copyright.jsp

You might also like