Integrating an RTD and an
Combining the two technologies compensates for the drawbacks associated with RTDs—and gives you high performance at low cost.
Robert Schreiber, Texas Instruments
Measuring with an RTD
Let’s look at common RTD styles and the measurement methods used with them. RTDs can come in Style 1 (two-wire), Style 2 (three-wire), or Style 3 (four-wire) configurations (see Figure 1). Style 1 is the simplest and least expensive; however, the lead resistance (RL) introduces significant measurement errors. For example, a common 100 PRTD has a temperature coefficient of about 0.00385 //ºC. A lead resistance of 8 introduces about a 20ºC error in the reading. In terms of absolute temperature measurement, this configuration would not be a good choice. For this style, voltage measurements would include lead resistance errors and would not produce a valid absolute temperature measurement.
If you were measuring the differential temperature between RTDs, Style 1 would be adequate as long as the lead length resistance of the RTDs is kept the same and the differential temperature measurement is kept ratiometric with regard to the voltage reference.
For all styles, the RREF resistor sets the value of the differential voltage reference, which will always be close in magnitude to the supply voltage magnitude because the lead resistances and RTD resistances are negligible compared with the magnitude of RREF (see Figure 2).
As the supply voltage varies, so does the current through RREF and the RTDs. Because the same current flows through all components, the voltages across all components will increase or decrease with the change in supply voltage. Thus, variations in the supply voltage are canceled out so that the conversion result is ratiometric.
The voltage across RTD2 compared with the voltage across RTD1 provides an accurate measure of the differential temperature between RTD2 and RTD1 as long as RL is the same for both RTDs. At 0ºC, both RTDs will have an ideal resistance of 100 ; thus, the voltage difference between RTD1 and RTD2 will be 0 V, indicating a 0ºC temperature differential. If you vary the temperature at both RTDs, you can easily calculate what will happen. If the temperature at RTD2 increases to 400ºC, the resistance of RTD2 increases to 254 . If the temperature at RTD1 increases to 100ºC, the resistance of RTD1 increases to 138.5 (see Table 1).
Now drastically vary the supply voltage (see Table 2).
Notice that the differential voltages change with the supply voltage, but the actual A/D output code and differential A/D output code stay the same. In fact, any combination of differential temperatures between RTD1 and RTD2 that equal 300ºC will give a differential conversion result of 41,229. This is true for any value of VDD, as long as RREF = 47 k and the lead resistances for all RTDs in the system are equal.
Style 3 (four-wire) solves the problems encountered in Style 1, but at a price. The difference between Style 1 and Style 3 is that the two extra leads act as sense leads. This allows accurate voltage measurement across the RTD for both absolute and differential temperature measurement. The error introduced by lead resistance is removed from the measurement. The drawback to Style 3 is that it requires more wires and is also more expensive than either the Style 1 or Style 2.
The Style 2 can also give you absolute temperature measurements, if designed properly. Figure 2 shows a Style 2 configuration that gives absolute and differential temperature measurements and is less expensive than a Style 3 configuration.
The A/D Converter
For example, when the input multiplexer is configured for the AIN0-AIN1 differential pair (see Figure 2), the lead resistance for RTD1 can be measured directly. The input impedance of the A/D converter prevents significant current from flowing into the device, so the conversion results in a voltage drop caused by lead resistance. When the input multiplexer is configured for the AIN1-AIN2 differential pair, the voltage across RTD1 is measured along with the other leg of the lead resistance. If you keep the RTD1 lead lengths the same, the voltage drop caused by the lead resistance is the same for both legs. The result from the AIN0-AIN1 conversion is subtracted from the AIN1-AIN2 result to give the voltage across RTD1. You can make this operation almost transparent by subtracting the AIN0-AIN1 result from the A/D converter offset calibration register. This automatically subtracts the lead resistance value from the conversion result, so the AIN1-AIN2 result is an accurate measure of the differential voltage across RTD1.
For differential temperature measurements, the process is much the same. If the lead resistances for both RTDs are the same, you simply subtract the AIN3-AINCOMMON reading from the AIN1-AIN2 to get the differential voltage. If the lead resistances are different for the RTDs, then the individual lead resistances can be removed as described previously. The lead resistance for RTD1 would be AIN0-AIN1, and the lead resistance for RTD2 would be AIN2-AIN3. With both these techniques, the conversion is not affected by variations in the supply voltage for the same reasons that were described for the Style 1 configuration.
One advantage of the ADS1240 implementation is that it doesn’t require exact matching of lead resistances for all RTDs being measured. Exact matching is typically required when using a traditional 2-channel differential A/D converter.
Another advantage of this device is that it makes more efficient use of pins. To achieve the same performance with a traditional multiplexer scheme, you would need additional analog channels. You would have to use either an A/D converter with more input channels or an external multiplexer. Either way, the overall cost would increase. Using the ADS1240 further simplifies the circuit by eliminating the need for an external current source and the extra signal line needed for a four-wire RTD.
Two other important considerations are the accuracy of RREF and the resolution of the converter. In the techniques described here, changes in RREF affect the ratiometric measurement. For instance, a 5% tolerance on the 47 k RREF could result in a 14ºC temperature measurement error, but a 1% tolerance on the 47 k RREF would reduce the error to about 3ºC. This is true whether the supply voltage or a current source is used to generate VREF. Therefore, you should either use a precision, low-drift resistance for RREF or perform a simple calibration. You can accomplish the calibration by providing a known voltage at the A/D input and comparing it with the expected value.
An actual output code from the A/D converter that is higher than expected implies that RREF is lower than expected. Table 4 shows the condition for a change in RREF to 46 k.
The digital output code changes from an expected value of 35,695 to 36,471. The ratio of the expected code to the actual code is the correction factor that should be used on the RTD conversion results. Table 5 shows that the actual output code from the A/D converter is the same regardless of a change in supply voltage.
So if RCAL1 and RCAL2 are known precise values, then the only parameter affecting the actual output code is the value of RREF, and you can correct for this.
Now apply this correction technique to the first example for a Style 1 configuration. Table 6 shows the effect of a 46 k RREF resistor on the output code.
For a 300º‚ temperature difference, you expected a differential A/D output code of 41,229 on the basis of data from Table 1. But with a 46 k RREF resistor, you actually have a differential A/D output code of 42,125. If you multiply the result by the correction factor ratio of 35,695/36,471, you get an accurate result of 41,229.
If you use a PGA = 128, you would have an rms resolution of about 104 nV and a noise-free resolution of about 660 nV. The full-scale range of the A/D converter with a PGA = 128 would be limited to about ±20 mV, which exceeds the input signal range of 27 mV. But you could use the offset D/A converter to subtract 10 mV from the input signal, which would move the RTD signal back into the full-scale range. You would then have rms temperature resolution of about 0.003ºC and flicker-free temperature resolution to about 0.02ºC.
By analyzing the important considerations of a temperature measurement system, you can easily implement a cost-effective, high-performance solution with the increased integration and unique features of a high-performance A/D converter.
Robert Schreiber is Strategic Marketing Engineer, Data Acquisition Products, Texas Instruments, 6730 S. Tucson Blvd., Tucson, AZ 85706; 520-746-7883, fax 520-746-7220, firstname.lastname@example.org.