Negative Temperature Coefficient Thermistors Part IV: The Temperature Calibration Standard

The first three parts of this article, appearing in the May, June, and July issues of Sensors, addressed characteristics, materials, and configurations of the NTC thermistor; means of determining the level of uncertainty; and the temperature-controlled bath. The photo, figure, and reference numbers in Part IV are continuous from the preceding articles.

Photo 4. Thermometer readouts from entry-level systems to a primary standards model can achieve an accuracy of ±0.001ºC. Also shown are a quartz-sheathed SPRT, a secondary PRT, and fixed-point cells used in the calibration of standards probes. (Photo courtesy of Hart Scientific.)

The temperature calibration standard consists of a temperature probe and a readout instrument. Several types are available, but considerations of cost and ease of use limit this discussion to platinum resistance thermometers (PRTs) and super-stable thermistor probes (see Photo 4). The choice of a specific standard will depend on your specific requirements-overall test system uncertainty, temperature range, and capital expenditure budget. Keep in mind that the cost of the standard increases as the uncertainty decreases.

If your thermistor application calls for testing over a relatively wide temperature range, e.g., -40ºC to 200ºC, a good secondary reference grade PRT with a temperature coefficient of 0.003925//ºC may be necessary. This type of probe normally has a nominal resistance (Ro)of 100 at 0ºC; a usable temperature range of -200ºC to 500ºC; a standard uncertainty of ±0.01ºC; and a stability of ±0.01ºC/yr. The probes are usually supplied with calibration tables and have a price range of $700-$1400, depending on the manufacturer and the type of calibration service provided. The readout instruments used with 100 PRTs typically cost $1000- $4000, depending on the uncertainty level specified. PRT probes are available with higher Ro resistance values (500 or 1000), which may allow use with a less expensive readout instrument ($500-$1000) but which will show a compromise in the uncertainty and stability specifications. PRT probes with lower R0 values (25) may have better uncertainty and stability specifications than the 100 PRT but require instruments costing $4000-$12,000.

The required expanded uncertainty of the total system will guide you in selecting a PRT/instrument combination. One way to improve the uncertainty of a temperature standard is to have the probe and the readout instrument calibrated together as a unit, instead of calibrating them separately and then combining the uncertainties mathematically. Some manufacturers and calibration labs will accommodate this type of calibration and others will not. The calibration lab's decision most likely would be based on its policies and procedures and the type of PRT and readout instrument to be calibrated. You should request this type of service during the equipment evaluation proc-ess, and it should be a matter to consider before deciding on a purchase.

Although PRTs are very useful standards for calibrating temperatures in the -200ºC to 500ºC range, their sensitivity to mechanical vibration and shock can be a serious limitation in some applications. Over the many years that PRTs and standard platinum resistance thermometers (SPRTs) have been used as thermometers, PRT manufacturers have developed and used various configurations of "strain-free" platinum wirewound elements [17]. This construction allows for thermal expansion and contraction of the platinum wire without constraint from its support. Because the element is loosely supported, however, mechanical shock or vibration can create a strain on the wire and increase the PRT's resistance. For many PRTs, simply tapping the probe on a table can shift it out of calibration by as much as 0.01ºC-0.02ºC. The probe would afterward need to be annealed and recalibrated, which could be a $400-$700 exercise, depending on the type of PRT.

As previously noted, for most thermistor applications the requirements for testing are primarily for temperatures in the 0ºC- 100ºC range, and other applications may require testing down to -40ºC or up to 120ºC. For these, a super-stable calibration standard probe offers several advantages over the PRT. For example, the thermistor sensor has an inherent high sensitivity that provides temperature resolution to 0.0001ºC when used with the appropriate readout instrument. Thermistor temperature standards are available (see Photo 5) with standard uncertainties of ±0.01ºC to ±0.002ºC and stability specifications of better than ±0.01ºC to ±0.005ºC/yr., depending on the type of probe [3,4,5]. Furthermore, the way in which the thermistor element is constructed makes the thermistor probe virtually immune to normal mechanical shock and vibration.

Photo 5. For the same price as an entry-level secondary PRT system, a thermistor standard can deliver an accuracy to ±0.005ºC over a narrow range. (Photo courtesy of Hart Scientific.)

Thermistor temperature standards cost about the same as PRTs, $700-$1500, but offer much better uncertainty and stability specifications. In the temperature range of 0ºC to 100ºC, some thermistor standards can actually approach the uncertainty levels of some SPRTs that are used as primary temperature standards. Thermistor probes normally can be used with readout instruments that cost about the same ($1000 to $3000) as those used with 100 PRTs, but achieve better expanded uncertainties than the PRT systems.

Additional improvements to the expanded uncertainty of the temperature standard can be made by calibrating the thermistor probe and readout instrument as a unit. For less exacting test uncertainty requirements, lower cost thermistor probe/instrument combinations can be obtained ($1000 to $1500 for the system), but with some compromise of stability (±0.02ºC/yr.) and system uncertainty (±0.03ºC).

3. W.R. Siwek et al. 1992. "A Precision Temperature Standard Based on the Exactness of Fit of Thermistor Resistance-Temperature Data Using Third Degree Polynomials," Temperature, Its Measurement and Control in Science and Industry, Vol. 6. James F. Schooley, Editor-in-Chief, AIP, New York, NY:491.

4. S.D. Wood et al. 1978. "An Investigation of the Stability of Thermistors," J Res NBS 83:247.

5. W.R. Siwek et al. 1992. "Stability of NTC Thermistors," Temperature, Its Measurement and Control in Science and Industry, Vol. 6. James F. Schooley, Editor-in-Chief, AIP, New York, NY:497.

17. J.L. Riddle et al. Apr. 1973. "Platinum Resistance Thermometry," NBS Monograph 126, U.S. Gov't. Printing Office, Washington, DC:8.


Reading the Calibration Data

J. Randall Owen, Hart Scientific
In addition to a readout instrument for the standards probe, an instrument is necessary for reading the thermistors or probes being tested. If several probes are being calibrated simultaneously, a multiplexer is also needed. Although a high-resolution digital multimeter (DMM) with a scanner is a popular choice for this data acquisition role, its abilities are limited to reading resistance or voltage because it cannot convert readings to temperature.

Figure 5. State-of-the-art thermometer readouts can also fulfill the role of sensor data acquisition, thus reducing from three to one the number of instruments required.
Some of the more recently developed thermometer readouts can reduce the requirement from two or three instruments to a single meter that can provide the temperature standard reading and the high-precision data acquisition for the thermistors being tested. The Hart Scientific 1560 (see Figure 5) is a modular thermometer that can be configured to read any standards probe whether it's a thermistor, PRT, or SPRT. By adding 2-channel input modules, a separate standards probe can be placed in each temperature bath being used. Eight-channel scanner modules can be added for reading the thermistors being calibrated. Further flexibility is available by adding modules for scanning PRTs or thermocouples.

The standards probes and the probes being tested can all be read on the 1560's graphic LCD. An RS-232 interface is included to permit the use of an optional software package to automate the entire process of reading the standards and sensors under test and completing the documentation of the calibration process.

The readout accuracy of the 1560 is ±0.005ºC for the SPRT module and ±0.002ºC for the standards thermistor module. The scanning modules read thermistors to ±0.005ºC and PRTs to ±0.01ºC. The use of an instrument designed specifically for high-accuracy thermometry can reduce uncertainties and in some cases reduce the cost and number of instruments required to calibrate temperature sensors.

J. Randall Owen is President and Chief Operating Officer, Hart Scientific,
799 E. Utah Valley Dr., American Fork, UT 84003-9775;
801-763-1600, fax 801-763-1010.