The familiar thermocouple (TC) is probably the oldest species of electronic temperature sensor extant. TCs consist merely of two dissimilar conductors joined at the temperature measuring point. They’re simple to fabricate and look (at first glance) equally easy to use.
Unlike other temperature sensors, TCs are intrinsic voltage sources and therefore require no excitation or power supply. They can be miniaturized to fit inside needle points and made to function in temperatures ranging from cryogenic to incandescent. So it’s easy to understand the durable popularity of these ancient devices. Unfortunately, several annoying quirks make TCs significantly harder to apply than their simple construction would imply. One such complication is cold-junction compensation (CJC).
TCs work via a phenomenon called the Seebeck EMF (VS) that’s generated whenever different metals interface at different temperatures. In any TC, two wires drawn from two precisely formulated alloys (i.e., iron/constantan or platinum/rodium, or other combination based on the TC type) are welded together at the spot where temperature is to be measured. At the other end, they’re attached to a suitable voltmeter circuit. The voltmeter then observes the following temperature-dependent Seebeck voltage:
VS = S(TH − TC)
Here, S is the Seebeck coefficient for the particular TC type, while TH indicates the temperature of the TC “hot” junction (the temperature of interest). TC represents the temperature of the “cold” junction of the TC wires with the copper conductors of the voltmeter.
So, unless TC is rigidly held at a known value (traditionally 0°C, typically achieved by literally dunking the cold junction in an ice-water bath—not very convenient), two temperature measurements are necessary to get TH = (VS + S × TC)/S. These measurements are TH from the TC, and TC from a separate, so-called “cold-junction compensator” sensor (CJC). The job of the CJC is to generate a compensating signal (VC = S × TC) and add it to VS. This produces the composite voltage (VS + VC), which is then ready for 1/S scaling and subsequent TH = (VS + VC)/S temperature readout.
Due to the ubiquity of TCs, many designs for CJCs and at least one commercial integrated circuit CJC (Linear Technology’s LT1025) are available. Figure 1, however, presents a new CJC that’s unique for two reasons. One is the device’s ability to generate the negative output associated with TC < 0°C without a negative power supply. The other is that it conveniently accommodates grounded TCs (Fig. 2). This sensor is based on the ΔVBE principle (see “Low-Cost Precision Thermometry,” Electronic Design, Aug. 21, 1995). Consequently, it can use virtually any ordinary small-signal npn transistor as precision cold-junction temperature sensor Q.
ΔVBE derives from a classic bipolar junction I/V relationship that predicts that for an ideal transistor, the DVBE corresponding to a change in collector current (I1/I2) is proportional to absolute temperature (T + 273.1°C):
ΔVBE = 86.17 µV × (T + 273.1°C) × ln(I2/I1)
I2/I1 = (R1 + R2 + 462)/R1. Therefore, picking the right R1 and R2 will match the temperature coefficient of the amplitude of the ΔVBE square wave to the Seebeck coefficient of any TC type. The problem then remaining is subtraction of the 0°C offset voltage = 273.1°C × S.
In Figure 1, this is accomplished by generating and summing an oppositephase signal −(R3 × I2) with Q’s ΔVBE square wave. Also, R3 is trimmed so that VC = 0 at TC = 0°C (or R3 can simply be adjusted for (VS + VC) = S × T at any convenient temperature T). To produce the desired VC = S × TC, the composite ac waveform is synchronously rectified by S2 and filtered by C2. R1 and R2 values (to the nearest standard 1%) appropriate for the common TC types are listed in the table.