Temperature Measurement Key to Plant Performance

By Gregory K. McMillan

Temperature is one of the four most common types of loops. While the other common loops (flow, level, pressure) occur more often, temperature loops are generally more difficult and important. It is the single most frequently stated type of loop of interest to users, and the concern for better control extends to the widest variety of industries.

Temperature is a critical condition for reaction, fermentation, combustion, drying, calcination, crystallization, extrusion, or degradation rate and is an inference of a column tray concentration in the process industries.

Tight temperature control translates to lower defects and greater yields during seeding, crystal pulling, and rapid thermal processing of silicon wafers for the semiconductor industry.

For boilers, temperature is important for water and air preheat, fuel oil viscosity, and steam superheat control. For incinerators, an optimum temperature often exists to ensure destruction of hazardous compounds and minimum energy cost. For heat transfer fluids, such as cooling tower water, chilled water, brine, or Therminol, good temperature control minimizes upsets to the users.

Good temperature control is important during the research, reaction, separation, processing, and storage of products and feeds and is thus a key to product quality. It is also important for environmental control and energy conservation.

Curiously, the slowness of the response of the temperature process is the biggest source of problems and opportunities for tight temperature control. The slowness makes it difficult to tune the controller because the persistence and patience required to obtain a good open- or closed-loop test exceeds the capability of most humans. At the same time, this slowness, in terms of a large major process time constant, enables gain settings larger than those permissible in other types of loop except for level.

Once a properly implemented temperature loop is correctly tuned, the control error is often less than the tolerance (error limits) of the sensor. If one considers that the accumulated error of an installed thermocouple or RTD system is about five times larger than the error limits of the sensor, one realizes that system measurement error seriously limits temperature loop performance.

Thermocouples and Resistance Temperature Detectors

In the process industry, 99% or more of the temperature loops use thermocouples or resistance temperature detectors (RTDs). The RTD provides sensitivity (minimum detectable change in temperature), repeatability, and drift that are an order of magnitude better than the thermocouple, as shown in table 1, "Accuracy, Range, and Size of Temperature Sensing Elements." Sensitivity and repeatability are two of the three most important components of accuracy. The next most important source of errors is the error and the resolution in the conversion of the sensor output in volts to a temperature in degrees. For 12-bit analog to digital (A/D) input cards, the error in percent of span may cause an error as large as 2 degrees for a full scale change and the resolution may be as poor as 0.2oC. Resolution is more problematic because the stair-step changes in the temperature preclude the use of derivative action in the controller tuning and the tightness of control is limited to the resolution. An error expressed as a percent of span is small for small changes in temperature and can be partially corrected by a setpoint change. The best way to make the resolution and other sources of error introduced in the conversion negligible is to use a smart transmitter with a high resolution (e.g., 0.02oC) with the calibration span narrowed and the sensor nonlinearity compensation customized to match the sensor. Drift is important for extending the time between calibrations. The data in this table dates back to the 1970s and, consequently, does not include the improvements made in thermocouple sensing element technology and premium versus standard grades. However, the differences are so dramatic that the message is still the same.

Table 1 includes data on thermistors, which have seen limited use in the process industry (despite their extreme sensitivity and fast - millisecond - response) primarily because of their lack of chemical and electrical stability. Thermistors are also highly nonlinear, but this can be addressed by smart instrumentation.

TABLE 1 HEAD: Accuracy, Range, and Size of Temperature Sensing Elements



Platinum RTD


Repeatability (°C)

1 - 8

0.02 - 0.5

0.1 - 1

Drift (°C)

1 - 20

0.01 - 0.1

0.01 - 0.1

Sensitivity (°C)




Temperature range (°C)

-200 - 2000

-200 - 850

-100 - 300

Signal Output (volts)

0 - 0.06

1 - 6

1 - 3

Power (watts at 100 ohms)

1.6 ×10-7

4 ×10-2

8 ×10-1

Minimum diameter (mm)





For bare sensing elements, thermistors are much faster-responding than thermocouples, which are slightly faster than RTDs. This point rarely comes into play because, for most industrial processes, a 1- or 2-second additional lag time in a temperature loop is well within the uncertainty of the loop's dynamics. Secondary process time lags can easily change by 10 to 20 seconds for slight changes in operating conditions.

Once these sensing elements are put inside a thermowell or protection tube (a closed-end metal tube that encapsulates and protects a temperature sensor from process flow, pressure, vibration, and corrosion), the fit, fill, material, and construction of the thermowell have the biggest impact on temperature measurement time lags, as noted in table 2, "Dynamics of Bare Sensing Elements," and table 3, "Dynamics of Thermowells." Protection tubes, like thermowells, isolate the element from the process, but, unlike thermowells, protection tubes do not necessarily provide a tapered or stepped wall, or a tight internal fit of the element (minimal air gap between sensing element sheath/tip and tube internal wall). Protection tubes for high temperature applications may be ceramic, which has a low thermal conductivity. The measurement lags from protection tubes are generally much larger than for thermowells.

TABLE 2 HEAD: Dynamics of Bare Sensing Elements

Bare Sensing Element Type

Time Constant (seconds)

Thermocouple 1/8-inch sheathed and grounded


Thermocouple 1/4-inch sheathed and grounded


Thermocouple 1/4-inch sheathed and insulated


Single element RTD 1/8 inch


Single element RTD 1/4 inch


Dual element RTD 1/4 inch



TABLE 3 HEAD: Dynamics of Thermowells

Process Fluid



Fluid Velocity

(feet per second)


Annular Clearance



Annular Fill



Time Constants







107 and 49





93 and 14





92 and 8





22 and 7





52 and 9





17 and 8





62 and 17





32 and 10





26 and 4





25 and 2





7 and 2





228 and 1





4 and 1


There are many stated advantages for thermocouples, but if you examine them more closely, you realize they are not as important as perceived for industrial processes. Thermocouples are more rugged than RTDs. However, the use of good thermowell or protection tube design and installation methods makes an RTD sturdy enough for even high-velocity stream and nuclear applications. Thermocouples appear to be less expensive until you start to include the cost of extension lead wire and the cost of additional process variability from less sensor sensitivity and repeatability. The minimum size of a thermocouple is much smaller than what is currently available for an RTD. While a tiny sensor size is important for biomedical applications, miniature sensors are rarely useful for industrial processes.

Thermocouples are used primarily when the temperature range is beyond what is reasonable for an RTD or the accuracy of an RTD is not required. Thus, for temperatures above 850°C (1500°F), the clear choice is a thermocouple for a temperature measurement that is in physical contact with the process material. For temperatures within the range of the RTD, the decision often comes down to whether the temperature is used for process control or monitoring trends. If you are monitoring multiple temperatures for trending and errors of several degrees are unimportant, you could save money by using thermocouples with transmitters mounted on the thermowell (integral mount) or nearby. If you are using temperature for process control, data analytics, statistical or neural network predictions, process modeling, or in safety systems, a properly protected and installed RTD is frequently the best choice for temperatures lower than 500°C (900°F). At temperatures above 500°C, changes in sensor sheath insulation resistance has caused errors of 10°C or more.


Temperature loops on gas volumes whose heat input is by combustion (e.g., furnaces and reformers) or by direct manipulation of reactant flows (e.g., fluidized bed reactors) have an exceptionally small process time constant. The largest time constant in the loop is consequently the temperature sensor lag. Generally, self-regulating process tuning rules are used for tuning the temperature controllers on these processes with the closed loop time constant (lambda) set equal to three or more dead times.

The process time constant for continuous temperature loops on volumes and columns is so large that the temperature ramps in the time horizon of interest and the process can be approximated as near-integrating. Temperature loops on batch processes have a True Integrating response. In both cases, the shortcut tuning method of estimating dynamics can be used: The initial percent ramp rate is subtracted from the maximum percent ramp rate in the right direction that occurs within the first four dead time intervals, the result is then divided by the change in percent controller output to get the integrating process gain (1/sec). The short cut method reduced the tuning test time from 10 hours to 10 minutes for a bioreactor.

The lambda integrating process tuning rules are used for tuning temperature controllers on these vessels and columns. For maximum disturbance rejection, an arrest time (lambda) slightly larger than the dead time is used giving a controller gain that is approximately one half the inverse of the product of this integrating process gain and the observed total loop dead time. For an ISA Standard Form PID controller, the reset time is approximately four times the dead time, and the rate time is equal to the secondary time constant (e.g., thermowell lag time or heat transfer surface lag). In general, a larger lambda is used to deal with nonlinearities and uncertainties. For a fast set point response with minimal overshoot, you can use either a smart bang-bang control, set point feedforward, set point lead-lag or 2 degrees of freedom (2DOF) PID structure. If set point feedforward is used, the proportional and derivative action should be on PV rather than error in the PID structure (beta = 0 and gamma = 0).

The use of lambda tuning- with lambda equal to the dead time- gives the tuning settings for maximum disturbance rejection regardless of the process type. However, unless accurate adaptive control is used, a lambda equal to at least three dead times is more realistic due to nonlinearities and uncertainties. It is important to realize that if the process time constant is greater than four times the dead time, the process is considered near-integrating and the tuning rules switch from those for self-regulating processes to those for integrating processes.

Resistance Temperature Detectors

RTDs operate on the principle that the electrical resistance of a metal increases as temperature increases, a phenomenon known as thermoresistivity. A temperature measurement can be inferred by measuring the resistance of the RTD element. The thermoresistive characteristics of RTD sensing elements vary depending on the metal or alloy from which they are made.

Wire-Wound RTD

Wire-wound RTD sensing elements are constructed by coiling a platinum (or other resistance metal) wire inside (internally wound) or around (externally wound) a ceramic mandrel (spindle). Most RTD sensors for the process industry are internally wound and sheathed for protection. A dual-element wire-wound RTD can be created by coiling a second set of wires inside or outside the ceramic mandrel. If connected to a second transmitter, a transmitter with dual-sensor capabilities, or to another distributed control system (DCS) card, a dual-element sensor  increases the reliability of the temperature measurement.

Temp Meas Img 1 
Wire-wound RTD elements are very sturdy and reliable. Their accuracy tends to be higher and their time response (how quickly the output reflects the temperature change) is several seconds faster than thin-film RTD elements. Wire-wound RTD elements work well for a variety of applications, although they may fail in high-vibration applications. Redundant, separate, single-element sensors are recommended for applications where reliability and accuracy must be maximized. A single-element sensor has a lower gauge sensing element and smaller time constant than a dual element sensor. The use of redundant sensors helps eliminate common mode failures and enables a better cross check of sensor drift than dual element sensors. Three sensors, each with their own thermowell and transmitter, and the use of a middle signal selection block in the Distributed Control System (DCS) will reduce noise and drift, and provide inherent automatic protection against a single failure of any type in the field.

Thin-Film RTD

Thin-film RTD sensing elements are constructed by depositing a thin film of resistance metal onto a ceramic substrate (base piece) and trimming the metal to specifications. Thin-film sensing elements are typically less expensive than wire-wound sensing elements because less resistance metal is required. However, thin-film RTD elements tend to be less stable over time, they typically have a more limited temperature range, and they may be more susceptible to damage from rough handling.

Extension Lead Wires

To get an accurate temperature reading from an RTD, the resistance of the RTD sensing element must be measured. Each copper lead wire that connects the RTD sensing element to the resistance measuring device adds a small amount of resistance to the measurement. If this added resistance is ignored, an error is introduced and an inaccurate temperature measurement results. This error is referred to as the lead wire effect. The longer the wire run, the greater the error, or lead wire effect, reflected in the temperature measurement. To compensate for lead wire effect, three-wire and four-wire RTDs are used instead of two-wire RTDs. Three-wire RTDs are created by connecting one additional copper wire to one of the lead wires. Four-wire RTDs are created by connecting one additional copper lead wire to each of the existing lead wires. These additional wires are used by the transmitter to compensate for lead wire resistances.

The third wire compensates for the resistance of the lead wires based on the assumption that each wire has exactly the same resistance. In fact, there is a tolerance of 10% in the resistance of standard wires. The fourth wire compensates for the uncertainty in the resistance of wires. For example, 500 feet of 20 gauge cable would add 10 ohms, which would cause a measurement error of 26°C (47°F) for a two-wire RTD. The 10% tolerance of the cable could create an error as large as 2.6°C (4.7°F) for a three-wire RTD. For high accuracy applications or long extension wire runs, a four-wire RTD, or a transmitter mounted on the thermowell (integral mount), should be used. The increased accuracy, stability, and reliability of microprocessor-based transmitters and the advent of secure and reliable wireless networks makes integral mounted transmitters an attractive option. Accessibility is less of an issue because maintenance requirements are drastically reduced. The transmitters rarely need removal, wiring problems are gone, and calibration checks and integrity interrogation can be done remotely.


A thermocouple (TC) consists of two wires of dissimilar metals (e.g., iron and constantan) that are joined at one end to form a hot junction (or sensing element). The temperature measurement is made at the hot junction, which is in contact with the process. The other end of the TC lead wires, when attached to a transmitter or volt meter, forms a cold or reference junction.

Thermocouple Types

Several types of TCs are available; the types are differentiated by the metals used to construct the element. While accuracies are better for types T and E compared to the most frequently used type J, the type selected in industry is often based on plant standards and the application temperature range, minimizing (if possible) the number of types that need to be stored for maintenance. The types of thermocouples more commonly used in industry are:

  • Type E-Chromel and constantan
  • Type J-Iron and constantan
  • Type K-Chromel and alumel
  • Types R and S-Platinum and rhodium (differing in the % of platinum)
  • Type T-Copper and constantan

Hot Junction Configurations

Junctions can be grounded or ungrounded to the sensor sheath. On dual-element TCs (two TCs in one sheath), the elements can be isolated or connected (unisolated). Each configuration offers benefits and limitations:

  • Grounded-Grounding creates improved thermal conductivity, which in turn gives the quickest response time. However, grounding also makes TC circuits more susceptible to electrical noise (which can corrupt the TC voltage signal) and may cause more susceptibility to poisoning (contamination) over time.
  • Ungrounded-Ungrounded junctions have a slightly slower response time than grounded junctions, but they are not susceptible to electrical noise.
  • Unisolated-Unisolated junctions are at the same temperature, and both junctions will typically fail at the same time.
  • Isolated-Isolated junctions may or may not be at the same temperature. The reliability of each junction is increased because failure of one junction does not necessarily cause a failure in the second junction.

The Seebeck Effect

TCs use a phenomenon known as the Seebeck effect to determine process temperature. According to the Seebeck effect, a voltage measured at the cold junction of a TC is proportional to the difference in temperature between the hot junction and the cold junction. The voltage measured at the cold junction is commonly referred to as the Seebeck voltage, the thermoelectric voltage, or the thermoelectric electromotive force (EMF). As the temperature of the hot junction (or process fluid) increases, the observed voltage at the cold junction also increases by an amount nearly linear to the temperature increase.

Cold Junction Compensation

As with RTDs, each type of TC has a standard curve. The standard curve describes a TC's voltage versus temperature relationship when the cold junction temperature is 0°C (32°F). As mentioned previously, the cold junction is where the TC lead wires attach to a transmitter or volt meter. Because the voltage measured at the cold junction is proportional to the difference in temperature between the hot and cold junctions, the cold junction temperature must be known before the voltage signal can be translated into a temperature reading. The process of factoring in the actual cold junction temperature (rather than assuming it is at 0°C [32°F]) is referred to as cold junction compensation.


The best practice for making a temperature measurement is to keep the length of the sensor wiring as short as possible to minimize the effect of electromagnetic interference and other interference on the low-level sensor signal. The temperature transmitter should be mounted as close to the process connection as possible. To minimize conduction error (error from heat loss along the sensor sheath or thermowell wall from tip to flange or coupling), the immersion length should be at least 10 times the diameter of the thermowell, or sensor sheath for a bare element. Thus, for a thermowell with a 1 inch outside diameter, the immersion length should be 10 inches. For a bare element with a 1/4 inch outside diameter sensor sheath, the immersion length should be at least 2.5 inches. This is just a rule of thumb. Computer programs can compute the conduction error and do a fatigue analysis for various immersion lengths and process conditions. For high velocity stream and bare element installations, it is important to do a fatigue analysis because the potential for failure from vibration increases with immersion length.

The process temperature will vary with process fluid location in a vessel or pipe due to imperfect mixing and wall effects. For highly viscous fluids, such as polymers and melts flowing in pipes and extruders, the fluid temperature near the wall can be significantly different than at the centerline (e.g., 10 to 30°C; 50 to 86°F). Often the pipelines for specialty polymers are less than 4 inches in diameter, presenting a problem for getting sufficient immersion length and a centerline temperature measurement.

The best way to get a representative centerline measurement is by inserting the thermowell in an elbow facing into the flow. If the thermowell is facing away from the flow, swirling and separation from the elbow can create a noisier and less representative measurement. An angled insertion can increase the immersion length over a perpendicular insertion, but the insertion lengths shown for both are too short unless the tip extends past the centerline. A swaged or stepped thermowell can reduce the immersion length requirement by reducing the diameter near the tip.

Temp Meas Img 2 

The distance of the thermowell in a pipeline from a heat exchanger, static mixer, or desuperheater outlet should be optimized to reduce the transportation delay and to minimize noise from poor mixing or two-phase flow. As shown in the figure, generally 25 pipe diameters is an adequate distance to ensure sufficient mixing after the recombination of divided flows from heat exchanger tubes or static mixer elements. For desuperheaters, the distance from the outlet to the thermowell depends on the performance of the desuperheater, process conditions, and the steam velocity. To give a feel for the situation, there are some simple rules of thumb for the straight piping length (SPL) to the first elbow and the total length to the sensor called total sensor length (TSL). Actual SPL and TSL values depend on the quantity of water required with respect to the steam flow rate, the temperature differential between water and steam, the water temperature, pipe diameter, steam velocity, model, type, etc., and they are computed by software programs.

SPL (feet) = Inlet steam velocity (ft/s) * 0.1 (s)

TSL (feet) = Inlet steam velocity (ft/s) * 0.2 (s)

Typical values for the inlet steam velocity, upstream of the desuperheater range from 25-350 ft/s. Below 25 ft/s, there is not enough motive force to keep the water suspended in the steam flow.

Temp Meas Img 3 

Best practices

  • Use integral mounted transmitters to eliminate sensor wires if the sensor connection is safely accessible, surface temperature is less than 50°C, and connection vibration is negligible. Avoid like the plague wiring temperature sensors directly to a DCS or a programmable logic controller (PLC) input card.
  • Use tip-sensitive and vibration-durable, four-wire RTDs for maximum sensitivity and accuracy, and for minimum drift when measuring process temperatures less than 400°C.
  • Use ungrounded, sheathed, premium TCs of the appropriate type when measuring process temperatures greater than 400°C.
  • Use spring-loaded compression fittings to ensure that RTD and TC tips touch the bottom of the thermowell with minimal annular clearance between the internal thermowell and sensor sheath to minimize response time from air acting as insulator.
  • Use tapered thermowells to minimize thermal conduction errors and vibration failure. For applications where accuracy and vibration is less of a concern, a stepped thermowell suffices.
  • Specify and purchase from the supplier a TC or RTD pre-assembled in a thermowell. The assembly should include an integral mounted transmitter via a pipe union and a transmitter calibration to match sensor nonlinearity.
  • Choose a location where the sensor sees a representative well mixed fluid with a transportation delay of less than 3 seconds and a fluid velocity greater than 1 feet per second (fps).
  • Install the thermowell in an elbow pointed into the flow at the pipe center line to minimize response time and errors from cross-sectional variation in the temperature profile.
Temp Meas Img 4


  1. Is the distance between the equipment outlet (e.g., heat exchanger exit) and the sensor at least 25 pipe diameters (for a single phase flow) to promote mixing (recombination of outlet streams)?
  2. Is the transportation delay (distance divided by velocity) from the equipment outlet (e.g., heat exchanger exit) to the sensor less than 5 seconds (e.g., 50 pipe diameters of 6" pipe at 5 fps)?
  3. Does the distance from a desuperheater outlet to the first elbow provide a residence time (distance/velocity) that is greater than 0.1 sec?
  4. Does the distance from a desuperheater outlet to the sensor provide a residence time (distance/velocity) that is greater than 0.2 sec?
  5. If application vibration is not excessive, is an RTD used for temperatures below 400°C to improve threshold sensitivity, drift, and repeatability by more than a factor of 10 compared to a TC?
  6. For RTDs operating at temperatures above 400°C, are length minimized and sheath diameter maximized to reduce error from insulation deterioration?
  7. For RTDs operating at temperatures above 600°C, is the sensing element hermetically sealed and dehydrated to prevent an increase in platinum resistance from oxygen and hydrogen dissociation?
  8. For TCs operating at temperatures above 600°C, is decalibration error (from changes in the composition of the TC) minimized by choice of sheath and TC type?
  9. For TCs operating at temperatures above 900°C, is the sheath material compatible with the TC type?
  10. For TCs operating above the temperature limit of sheaths, is the ceramic material with the best conductivity and design used to minimize measurement lag time?
  11. For TCs operating above the temperature limit of sheaths and in contact with gaseous contaminants or in reducing conditions, are primary (outer) and secondary (inner) protection tubes designed to prevent contamination of the TC element and still provide a reasonably fast response?
  12. In furnaces and kilns, do location and design minimize radiation and velocity errors?
  13. Is the immersion length long enough to minimize the heat conduction error (e.g., L/D > 5)?
  14. Is the immersion length short enough to prevent vibration failure (e.g., L/D < 20)?
  15. Is the process fluid velocity fast enough to minimize coating (e.g., > 5 fps)?
  16. Is the process velocity fast enough to provide a fast response (e.g., > 0.5 fps)?
  17. For pipes, is the tip near the centerline?
  18. For vessels, does the tip extend sufficiently past the baffles (e.g., L/D > 5)?
  19. For columns, does the tip extend sufficiently into the tray or packing (e.g. L/D > 5)?
  20. For TCs, is it more important to minimize noise by using an ungrounded junction or to minimize sensor element lag time by using a grounded junction?
  21. To increase RTD reliability, are dual RTD elements used except where vibration failure is more likely due to smaller gauge?
  22. To increase TC reliability, does the sensor have dual isolated junctions?
  23. For maximum reliability, are three separate thermowells and three separate transmitters with middle signal selection in the DCS used?
  24. Does the sensor fit tightly into the thermowell to minimize measurement lag from air gap (e.g., annular clearance < 0.01 inch)?
  25. Is an oil fill used that will not form tars or sludge at high temperature in a thermowell with the tip pointed down (to keep fill in the tip) to minimize measurement lag?
  26. Is premium TC extension wire used to minimize measurement uncertainty?
  27. Is four-wire RTD lead wire used to minimize measurement uncertainty?
  28. Are integral-mounted temperature transmitters used for accessible locations to eliminate extension wire and lead wire errors and reduce noise?
  29. Are wireless integral-mounted transmitters used to provide portability of measurement for process control improvement and to reduce wiring installation and maintenance costs?
  30. Are proper linearization tables used in the transmitter and calibrator?
  31. Is a spring-loaded compression fitting used to ensure that the tip of the RTD or TC is touching bottom of the thermowell to improve accuracy and response time?
  32. Are tapered thermowells used to reduce susceptibility to vibration damage, conduction error, and thermowell lag from metal mass?
  33. Is sensor matching used to improve the customization of the transmitter calibration to compensate for specific sensor nonlinearities?
  34. Is the transmitter span narrowed to improve accuracy?
  35. Is the DCS number of bits large enough, and is the module execution rate slow enough to reduce A/D chatter?
Temp Meas Img 5

About the Author

Gregory K. McMillan is a retired senior fellow from Solutia/Monsanto and an ISA fellow. McMillan contracts in Emerson DeltaV R&D via CDI Process & Industrial in Austin, Tex. and is a part-time consultant for MYNAH Technologies in St. Louis, Mo. McMillan received the ISA Life Achievement Award in 2010. For further information and other articles by McMillan, please visit the following web sites:

ISA Interchange | Greg McMillan
Control Talk Blog | Greg McMillan
Modeling and Control | Greg McMillan

Reader Feedback

We want to hear from you! Please send us your comments and questions about this topic to InTechPlus@isa.org.