Eliminate thermoelectric EMF in low-ohm measurements

John Wynne

EDN

When two different-metal conductors connect together in a loop and one of the junctions is at a higher temperature than the other, an electrical current flows through the loop. The magnitude of this current depends on the type of metals involved and the temperature differential of the junctions. When you open such a loop, a thermoelectric voltage appears across the open ends. Again, the phenomenon depends on the type of metals involved and the temperature differential of the junctions. The junction of the two metals forms a thermocouple, so the thermoelectric voltage is a thermocouple voltage.

When you try to measure small voltages or low impedances, thermal EMFs (electromagnetic fields) are likely to cause errors in the readings. A standard way that digital-multimeter manufacturers deal with the problem is to initially take one reading and then to reverse the excitation and take a second reading. Reversing the current excitation through a low-value resistor or thermocouple junction reverses the polarity of the desired signal but does not affect the polarity of the unwanted EMF voltages. Subsequently averaging the two readings eliminates the thermoelectric EMFs from the final result (Reference 1). Figure 1 shows IC₁, an AD7719 ADC measuring a low-value resistor, R_LOW. The diagram also shows two thermoelectric EMFs, EMF₁ and EMF₂, representing summations of all the thermoelectric EMFs on the way from and to the ADC and the resistor. These EMFs would normally cause an error if you were to take a single measurement of R_LOW.

Figure 1.	In this configuration, the excitation current flows in RLOW from top to bottom.

However, you can program each of the AD7719's two current sources, IEXC₁ and IEXC₂, to appear at either of the package pins, I_OUT1 and I_OUT2. This feature allows you to reverse the excitation current through the low-value resistor. You can thus take two measurements and eliminate the effects of EMF₁ and EMF₂. To increase the excitation current and thereby increase the measurement sensitivity, the two internal 200-µA excitation currents appear in parallel with each other. Thus, a single 400-µA current source makes up the excitation current, I_EXC, in this design. Transistors Q₁ and Q₂ steer the excitation current through the reference resistor, R_REF, to ensure that the same polarity reference voltage always appears, regardless of the excitation-current direction. Port pins P₁ and P₂ (not shown) of the AD7719 drive the transistors in antiphase mode. A value of 6.8 kΩ for R_REF is suitable and ensures a typical ratiometric reference voltage of 2.5 V.

Figure 2.	In this configuration, the excitation current reverses, and flows from the bottom of RLOW to the top.

Figure 1 and Figure 2 show the current flow in each phase of a measurement. During Phase 1, the excitation current flows out of I_OUT1, through R_LOW, and through R_REF via Q₂ to ground. During Phase 2, the excitation current flows from I_OUT2, through R_LOW, and through R_REF via Q₁ to ground. During Phase 1,

V_DIFF(PHASE1) = V_AIN1 – V_AIN2 = V_EMF1 – V_EMF2 + I_EXC×R_LOW.

You now switch the current sources. During Phase 2,

V_DIFF(PHASE2) = V_AIN1 – V_AIN2 = V_EMF1 – V_EMF2 – I_EXC×R_LOW.

You now combine the two measurements in software to cancel the thermoelectric EMFs:

V_DIFF = V_DIFF(PHASE1) – V_DIFF(PHASE2) = I_EXC×R_LOW.

Finally, you need to turn this ratiometric measurement into an absolute one. You achieve this result by measuring a known voltage using the "unknown" ratiometric reference voltage. Taking a reading of this known voltage but with an unknown reference allows you to infer the unknown reference value and, hence, the absolute value of V_DIFF on pins A_IN1 and A_IN2. An absolute voltage reference, such as an ADR420, supplying 2.048 V output, provides the known voltage. It connects to the second pair of differential inputs, A_IN3 and A_IN4, and into the main 24-bit ADC channel.

Reference

Materials on the topic