Design Ideas: December 21, 1995
Accurate temperature measurement in single-supply, low-power applications presents a difficult signal-conditioning problem. In many applications, the temperature sensor may be located a fair distance from the signal-conditioning circuitry, and, as a result, the measurements may suffer contamination from external noise sources. A resistance-temperature-detector (RTD) sensor is often used in a high-end application because of this sensors accuracy and repeatability. However, the devices low to moderate absolute value, low-temperature coefficient, and inherent nonlinearity make accurate temperature measurement a challenging task. The primary difficulty in using an RTD to measure wide temperature ranges is correcting its inherent nonlinearity. A second challenge is achieving high performance with a single, low-voltage power source. Figure 1 shows a circuit that operates from a single 5V supply and conditions a three-wire, 1-k ohm RTD from 0 to 400×C to an accuracy of ±0.3×C.
The circuit uses a 2.5V, micropower precision-voltage reference; a Darlington-connected pnp transistor; and a low-power, single-supply, rail-to-rail precision quad op amp. The circuits special features are its correction of the RTDs inherent curvature using a current source with positive feedback, its single 5V-supply operation, and its compensation for wiring voltage-drop errors through the use of constant-current excitation. RTD-curvature correction comes from the current source (IC1Cs circuit), connected in parallel with the RTD. A 200-µA constant current feeds this parallel combination. IC1D amplifies the voltage developed across it to provide a 10-mV/ºC output that an ADC can process directly or that a 4½- or 5½-digit DVM can read.
The 200-µA constant-current source consists of op amps IC1A and IC1B, an MPSA63 transistor, and current-sense-resistor R1. IC1A boosts the output of the 2.5V reference to 4V. (The current source is bootstrapped against this potential.) The current-source circuit could have been bootstrapped directly to the 5V supply but would then have suffered from poor supply rejection. Amplifier IC1B provides the base drive for the pnp transistor and maintains the potential at the emitter equal to the REF191s output voltage. Thus, the voltage across R1 remains fixed and accurate, setting the collector current of the pnp transistor to 200µA. Finally, potentiometer P1 compensates any residual error in the current source.
To span the range of 0 to 400×C, you need to linearize the RTDs response. For example, if the output voltage of the circuit were linearly scaled to the RTDs low-temperature coefficient of 3.85 ohm/8C, at 200×C the error would be 3×C. If the temperature were increased to 400×C, the error would be 20×C. Thus, with a constant 200 µA supplied to the RTD, the output voltage of the circuit is lower than expected because the RTD exhibits a downward-oriented, concave response. Ideally, you should inject current into the RTD as its temperature increases to correct its inherent curvature.
IC1Cs circuit provides the curvature correction required to linearize the RTD. R5 and R6 set IC1Cs closed-loop gain at +2. Any incremental voltage change at IC1Cs noninverting input is doubled at its output. This change, in turn, results in an incremental change in current through R4 and, therefore, into the RTD. For positive input voltages at IC1Cs noninverting input, current through R4 is flowing from node A to node B; hence, the circuit behaves like a resistor with a value of R4. Now, R4 appears in parallel with the RTD, and the voltage produced at the input of IC1D is given by VB=RP·200 µA, where As long as the resistance connected to IC1Cs noninverting input is lower than the value of R4, circuit instability is not a problem because of positive feedback. Linearizing the RTDs response, then, is a matter of choosing the correct value for R4, such that RP over the temperature range of interest mitigates the RTDs nonlinearity. To find the right value of R4 for lowest nonlinearity requires forcing RPs resistance-temperature curve to have an inflection point roughly in the center of the measurement range. You then determine R4s value by evaluating the second derivative of RP with respect to temperature and equating the result to zero. The result of that operation yields the equation for R4:
With the inflection point set at 200×C, the value returned for R4 is 25 k ohm (24.9 k ohm is the closest 1% value).
The change in RP over the temperature range determines the gain of IC1D. From 0 to 400×C, the resistance change is 1.699 k ohms. Using a 200-µA constant current, the full-scale voltage change is 339.8 mV; therefore, to produce a 4V full-scale output requires a gain of 11.8, adjusted at full-scale with P3. The REF192s 2.5V output and R9 provide an offset current for RPs initial resistance of 1046V at 08C. R8 and P2 provide the adjustment for this offset and IC1Ds input offset voltage. RN and CN effectively reduce noise pickup along the wires to the remotely located RTD. Typical values are 249 ohms for RN, 1.5 k ohm for RP~, and 1 µF for CN. You can adjust the filters cutoff frequency by changing CN. The 0.1-µF capacitor across the RTD also helps cut RFI pickup.
To calibrate this circuit, you can use a set of precision resistors or a precision decade resistance box in place of the RTD. First, substitute 1000.00 ohm for the RTD at 0×C and adjust P1, such that VB reads 208.6 mV. This trim adjusts both the constant-current excitation and the curvature-correction circuits. Next, adjust P2 such that the output of IC1D reads 0.1V. Last, substitute 2470.4 ohms for the RTD at 400×C and adjust P3 until the output reads 4.10V. The P2 and P3 trims are interactive, so the sequence requires repeating to fix the two end-points. Note that a 0.1V scale offset allows a precise 0×C trim. You can easily remove this offset by referring the ADC inputs common terminal to 0.1V. (DI #1803)
RTD = resistance of the RTD at temperature T, given by
R0=value of RTD at 0×C (1000 ohm for Pt1000 detectors); A=linear detector constant=3.908E-3(×C-1); and B=quadratic detector constant= 5.802E7(×C-2)