A Twelve-Step Sensor Selection Checklist

Selecting a device to measure a physical property, such as temperature, can be a confusing process. But shortcuts and inattention to detail can get you into a real jam. Here, in chronological order, are the steps I try to use most of the time, especially when the measurement is really important. The underlying concept is akin to the carpenter's maxim: Measure twice; cut once.

1. Establish Measurement Span or Range Requirements

Find out your measurement range and any allowable variations. Many vendors specify accuracy or uncertainty as a percentage of measurement span. Be sure you understand what those specs mean. Many suppliers play fast and loose with the definition of terms as used in their literature. Some do a better job. And yet others omit some of those critical to your application: effects of ambient temperatures, especially the residual temperature coefficient; response time constant; or effects of vibration, radio frequency interference, or magnetic fields.

2. Agree on Accuracy and Uncertainty Requirements

What roles do precision and resolution play? Do you need the temperature to within ±0.1°C, or ±1°C, or ±10°C (or °F)? You might need to look at the various sensor options and perform an error analysis on each candidate type. You might need to do some extra detective work to get realistic and realistically acceptable tolerances. Knowing the accuracy and precision you require will pay real dividends, allowing you to verify and do a reality check on what someone else tells you.

Everyone involved in the selection process, including your superiors and the operations staff, needs to be in agreement on your requirements. For instance, a thermocouple will measure just fine for a 10°F accuracy case, but certainly will not cut it for 0.1°F. On the other hand, a resistance temperature detector (RTD) will handle 0.1°F if the environmental or measurement context conditions are favorable.

If you need to use connecting wires on an RTD, for example, you have to be sure that the wire's resistance will not contribute significantly to the overall measured resistance of the detector plus lead wires. If that's likely, you might need a 3- or 4-wire configuration. That, in turn, will influence your choice of a readout, transmitter, and/or data logger unit. Which one will handle your 3-wire or 4-wire RTD connections?

3. Identify Influence of Measurement Conditions

Know all your measurement conditions and the effects they might have on your sensors. This is an area that can make or break a measurement, particularly in industrial and utility process applications. Thermocouples in particular can produce variable bias errors if connected to compensating cable that runs through large temperature gradients, or can experience very high temperatures along the heated length, as in the case of metal-sheathed, mineral-insulated devices. There are not a lot of test data on sources that influence thermocouple readings, but a visit to www.temperatures.com and www.tempsensors.net should prove helpful.

If you can't use a contact device, consider your noncontact options. Also, if ambient conditions look too severe for the sensor you like, be sure to take advantage of the accessories that vendors provide to handle them. In industrial applications, heat, cold, dust loading, and vibration are the principal bad guys, but humidity, rain, sleet, and snow can sometimes be even worse.

Keep in mind that measuring any surface temperature with both contact and noncontact sensors is a hard task with many potential error sources. Achieving good thermal contact is sometimes difficult, and failure to do so can contribute to measurement errors due to reduced speed of response of thermocouples and RTDs. IR thermometers, however, can follow surface temperature changes faithfully, assuming their response time constant is small enough, and they can be superior to contact sensors for some applications. Their major limitation (in theory) is not time response, but rather dealing with surface emissivity and reflectivity, particularly when the temperature being measured is close to ambient.

Will your application have any issues related to safety, in terms of either the environment classification for hazardous conditions or the working conditions for those installing and servicing the field equipment? Are there industry, government or local codes to meet? Some of the least well-known areas of environmental effects are in regions of high nuclear radiation levels. For temperature sensors, the staffs at NIST and Oak Ridge National Laboratory have considerable experience. Among commercial organizations, Analysis and Measurement Services Corp. (www.ams-corp.com) of Knoxville, TN, is perhaps the best versed in this area and has published many papers and reports. They are expanding into other areas of instrumentation and provide specialty testing services for nuclear power stations.

4. Make a List of Candidates

Make a short list of candidate devices or system types. There may only be one that you truly want, but several that could handle the job. Include all that seem acceptable.

By now, you will have realized that your sensor is really part of a multifaceted system. If it will interface with, say, a control system, be sure to check with the people responsible for the latter to ensure they have an access point for your signals. Whatever the hardware involved, a key part of any system is the person who or department that will be responsible for maintaining your device, even after you have moved on to another project or job. They may need replacement or spare parts on hand, such as air or water filters in cooling lines. They may also advise, as is often the case in continuous process lines, that a redundant sensor be used to increase reliability in the event of a catastrophic failure. On a 24/7 process line you may need alarms on the services supporting the device, too, unless there's a system already in place to provide such notice. Your support team will also want advice on intervals for initial calibration reverification, recommendations on sources or additional equipment, and/or training they will need to carry out such work.

5. Review Error Analyses and Trade-Offs

Review and, if necessary, perform an error analysis on sensors to be certain your candidates will reasonably meet your overall measurement and quality requirements. This can sometimes result in a loop, where you need to go back to Steps 1 and 2 and relax or tighten your requirements. Sometimes you'll have to face a trade-off, such as a smaller measurement span for better accuracy or relaxed accuracy specification to compensate for expected influences of, say, the sensitivity of electronics or the sensors themselves to ambient temperatures. You might need a custom enclosure with a heater or cooler to cope with extreme conditions.

6. Finalize the Measurement Specifications

Be sure to put in all the conditions of measurement as ambient condition requirements. Include any special devices and communications links to external monitors, computers, and data logging devices or control systems. If there are spares and maintenance or training support requirements, be sure they are part of the initial bid. It never hurts to be too detailed, but it always hurts to miss something important and find out after the fact. This is where the nit-picking pays off: a thorough and detailed specification is the document against which you can accept or reject a device or system. Finally, don't forget to insist on traceable calibration certification. It's about the only way you can reliably trace a measurement back to an approved reference.

7. Solicit Bids

Obtain quotations by submitting specifications for bids from two to four qualified suppliers. To be qualified, suppliers should meet the requirements that your organization's purchasing department has developed. At a minimum, your vendor must be a stable business with reasonable expertise as attested to by several referenced users. They must have an established measurement QA program such as an ISO 9001 certification. They should also provide evidence of compliance with appropriate ISO 17045, A2LA (www.a2la.org), or NAVLAP (ts.nist.gov/Standards/214/cfm) calibration competence.

8. Select Both Device and Vendor

It is good practice to review the specifications submitted with each candidate to see if they all comply and to examine with particular care any exceptions a bidder might note. Sometimes making your selection can involve requesting vendors to review their bids and detail the reasons for exceptions. This can entail a face-to-face meeting, which should happen on your turf. Before making a final decision, you might also want to pay a visit to the vendor's key people to gauge that company's experience and attention to its manufacturing practices and products.

9. Inspect

Inspect your sensors on delivery. Do they meet the promised specifications? You might need some special tools or simulation devices to mimic the process variables and conditions of measurement.

10. Install

This is often the easiest part of the measurement job, but can sometimes be more expensive than the cost of the device itself. Wiring costs in a process line can be unbelievably high. Wireless process transmitters have helped greatly in some areas and in large process plants, such as chemical and petroleum lines, they often tie into a plant-wide network system.

11. Commission

Not much is written about commissioning, but it can be critical. Many modern devices have multiple outputs, both digital and analog. You have to be sure that the installation wiring is done right. That means proof testing or signal injection to test links to external equipment. Are the placement, protection, and any settings such as response times, emissivity or emissivity ratio values, and other instrument specifics set up properly? Are installation drawings correct? Are equipment settings logged for the benefit of maintenance staff? They will need to verify the proper settings in the future for checking during operation after servicing. Have you agreed with your staff on a servicing and calibration check schedule and procedure? Have you identified the competent "go-to" people in the event of any future problems?

12. Verify Online

Verification can be easy or very difficult, as it is with IR thermometers, because there are few verification options. Bottom line: you need another, independent method against which to compare the measurement results of the new sensor. Sometimes all you have to do is collate measured results with physical or chemical change in the process or product being measured. Be prepared. It takes a while to collect, analyze, and compare the measurements by the two methods. Just don't be fooled by using a reference measurement method with poorer measurement uncertainty than the device under test.

Final Thoughts

If someone has already done the job you're about to work on and all you're getting is a replacement sensor, you will have a leg up. If you're starting with a blank page the exercise will be quite another story. Or perhaps your precursor was not so thorough as you might have wished, but now you have the opportunity to turn a marginal installation into a more accurate and reliable one. I hope this article has laid out a pathway that will prove helpful.

Additional Web Sites of Interest

  • 1. National Bureau of Standards and Measurement: www.nist.gov|~www.nist.gov/
  • 2. International Bureau of Weights & Measures: www.bipm.fr|~www.bipm.fr/
  • 3. "Commissioning and Verifying Radiation Thermometers," Chemical Engineering, June 23, 1986: http://www.che.com/sub/authentication.php?/inc/search_article.php?searchfile=1985-1999/textfiles/Vol93/chevol93_num12_41.html&pub_date=519883200|~www.che.com/sub/authentication.php?/inc/search_article.php?searchfile=1985-1999/textfiles/Vol93/chevol93_num12_41.html&pub_date=519883200

G. Raymond Peacock, MS, can be reached at Temperatures.com Inc., Southampton, PA; 215-325-1450, [email protected], www.temperatures.com. He is also the host of the Sensors Industrial Automation Newsletter (www.sensorsmag.com/sia).