Cancel sensor-wiring error with bias-current modulation

The approximately -2mV/°C temperature coefficient of diode junctions is a popular means of temperature measurement, especially in cryogenic applications (Figure 1). Diode temperature sensors are compact, stable, robust, sensitive, and inexpensive, and, unlike thermocouples, they require no reference junction. All of these benefits help explain the durable popularity of this—to use the polite term—“mature” technology.

A complicating factor and potential error source affecting these sensors arises from their need for bias-current excitation, however. The resulting contribution of ohmic IR (current/resistance)-voltage drop in the wiring and the connectors' resistance to the sensor's output voltage create spurious and temperature-sensitive voltage offsets. These offsets can introduce unacceptably large measurement error. This situation is especially likely when you use small and, therefore, high-resistance-gauge wire for sensor cabling, such as in cryogenic applications. In those cases, designers prefer exceptionally fine-gauge wire to minimize thermal conductivity and leakage.

The usual solution to the IR problem is to employ four-wire “Kelvin”-interconnection topologies, in which one pair of conductors carries the sensor's bias current and a separate, independent pair differentially senses the sensor's output voltage. This approach prevents corruption of the sensed voltage by IR drop in the bias pair. This traditional fix works well but complicates the wiring and doubles undesirable thermal leakage due to the extra wires, thus defeating much of the point of using fine-gauge cabling in the first place.

Figure 2 illustrates a circuit that implements a different approach. It cancels the wiring-resistance error and needs only two conductors in the sensor cable. It takes advantage of the fact that IR-voltage drop is directly proportional to current, but the sensor voltage is mostly constant. It works by alternating the magnitude of the excitation current, I_B, between two values, I_B1 and I_B2, where I_B1=2I_B2. The ac component of the resulting signal is thus approximately I_BR_W, where R_W is the total wiring resistance plus a minor contribution from nonzero sensor impedance.

The clock for both I_B1/I_B2 excitation modulation and synchronous demodulation of the resulting response is the internal oscillator of the LTC1043, which you set to approximately 500 Hz by connecting the external 0.01-µF capacitor to Pin 16. The resulting toggling of the excitation ballast resistance between 1M and 1M+1M=2M creates the 2-to-1 current modulation and an ac-signal component proportional to wiring resistance: I_BR_W.

The other side of the LTC1043 synchronously rectifies the IBRW ac component, storing the IB1RW=VC1 phase on C₁ and the I_B2R_W=V_C2 phase on C₂. Op amp A₂ buffers V_C1 and inputs it to the resistor network and A₁, which subtracts it from the average sensor signal, producing an output voltage independent of cabling-resistance offset. One downside of the technique is that, due to sensor-impedance effects on the order of 20mV, the thermometric diode usually requires custom temperature calibration.

Click here for the illustrations:

Figure 1, Figure 2

Caption
Figure 1: The typical -2mV/°C-voltage-versus-temperature coefficient of diode sensors is large and nearly constant over a wide range of temperatures.

Figure 2: This circuit cancels the wiring-resistance error inherent in diode temperature sensors and requires only two conductors in the sensor cable.