Design Ideas: January 4, 1996

Many temperature-sensing applications require a fast thermal-response time. Dipping a temperature sensor into a fluid on a production line or the protective shutdown of equipment under a runaway thermal condition are two examples. In general, semiconductor temperature sensors are limited in their response times because of the high thermal mass and low thermal conductivity of plastic packaging. Temperature settling times of a few minutes for large step changes ([Delta]T>40°C) in ambient temperatures are common. The block diagram in Figure 1 and the circuit in Figure 2 present a technique that signal conditions the output of a temperature sensor to speed a temperature measurement.

To use this method successfully requires an understanding of the temperature sensor’s thermal response. A single-pole RC time constant models the response time of IC temperature sensors. In the model (Figure 1), capacitance is equivalent to thermal mass, and resistance is equivalent to the inverse of the sensor’s thermal conductivity. You can use this information to create a correction signal for the sensor’s output. The correction signal, V_DIFF, is larger when the temperature is changing rapidly. In other words, the faster the temperature changes, the higher the final temperature value is likely to be.

Therefore, for this design to work correctly, the thermal time constant must not change. As a result, the packaging (plastic DIP vs SOIC), additional heat sinks, surrounding pc-board tracks, the type of material to be measured, and the method of measurement should all remain constant. Only thermal measurements where the environment is tightly controlled (that is, repetitive laboratory processes, industrial production lines, and biomedical applications) are candidates for this method.

As Figure 1’s block diagram shows, the method senses and then differentiates the output of the sensor. For this example using the TMP10 sensor IC, the VPTAT output equals 1.49V at 27°C with a temperature coefficient of +5 mV/°C. Next, the method multiplies the V_DIFF correction signal, which represents the rate of change of the sensor output, by -A, which is a precalibrated gain based on the system’s thermal time constant. Finally, VDIFF and VPTAT sum together to produce VPTAT_FAST.

Because VPTAT changes slowly, the differentiator needs a gain greater than 10 to generate a correction signal of useful amplitude. As Figure 2 shows, a rail-to-rail quad JFET op amp is a good choice in this application because it operates on a single-supply 5V rail, and you can provide the necessary gain using one quad op-amp IC. Also, JFET-input types exhibit low input-bias current, which reduces errors that the large feedback resistors cause. C₁ and R₂ configure IC1A as a differentiator, and R1 and C2 ensure stability and reduce noise. In general, you need to empirically determine the actual values of these four elements because of the infinite possibilities in thermal mass and thermal conductivity. However, the values in Figure 2 correspond closely to maximum thermal conductivity and minimum thermal mass.

You must invert the signal VDIFF that appears at the output of IC_1A before the signal sums with the original VPTAT signal. IC_1B performs this inversion, and IC_1C produces the sum and also inverts the result. IC_1D inverts the result a final time and outputs the final corrected temperature-dependent voltage, VPTAT_FAST.

Overall, this method provides the final temperature estimate at approximately a factor of 10 faster than the absolute temperature given by the temperature sensor’s unconditioned output. One experiment places the TMP01 sensor in a bath of slowly circulated, 23.5°C water. The temperature sensor is coated with a thin layer of cyano-acrylate to provide electrical isolation in the water bath. In less than 1 sec, the sensor was then placed in a 73°C bath of water. The 5% settling time for VPTAT was 9.5 sec vs 0.97 sec for VPTAT_FAST. The 1% settling time for VPTAT was 25 sec vs 3 sec for VPTAT_FAST.

The circuit uses a single-supply 5V and op amps in inverting configurations; thus, the circuit requires a false ground. In most single-supply applications, the false-ground level equals the midpoint between the supplies to balance the signal-swing range. However, this design sets the false ground to the expected ambient-temperature voltage (27°C or 1.49V). R₃ and R₄ generate the false ground, and the resulting voltage applies to all IC₁’s noninverting inputs.

When ambient temperatures are greater than 60°C, bipolar-input op amps may be a better choice than JFET-input types, because the bias currents of JFETs double for every 10°C rise in ambient temperature. In addition, many components have significant temperature coefficients, so mechanical considerations are important, as well. For example, placing the temperature sensor a small distance away from the rest of the circuitry minimizes the thermal mass associated with the sensor. This approach helps to make the sensor more sensitive to ambient temperature changes and the related circuitry less so. Finally, this entire circuit requires less than 0.5 sq in. of pc-board space when using SOIC components. (DI#1808)