Thermocouple Tutorial V2b

Brian O'Connell Practical Thermocouple Theory and Technique Part A. Thermocouple Theory The measurement of component temperatures is not a simple practice, and the principles of thermocouples may not be obvious. The construction and application of thermocouple (t/c) material requires the considerations of physics and craftsmanship. In my physics text, I note that temperature is "A measure of the average kinetic energy of a system of particles". Temperature is a scalar quantity. Temperature measurements are recorded for a small surface area of a larger mass, for a single instance in time. Do not think of temperature measurements as vector quantities, and do not consider a temperature measurement to represent a 'constant' characteristic, and do not think that component temperature is solely dependent on I2R (see references 1 and 2). The three effects involved in a t/c circuit determine thermocouple performance: the Seebeck, Peltier, and Thomson effects. The Seebeck effect determines the voltage potential that results from the temperature gradient along the wires between two junctions. The change in voltage, due to temperature difference between the two junctions, is proportional and is determined by the Seebeck coefficients. The Peltier effect describes the temperature difference generated by voltage potential and is the 'reverse' of the Seebeck effect. The Thomson effect is the most interesting (but not the dominate effect), and is reversible. Heating or cooling of a single material can be controlled by the direction of current flow. The Thomson effect is interesting in that some metals will develop a positive potential at the hot end, and others will develop a positive potential at the cold end. For the purpose of TCA measurements, we can consider the Thomson effect as a source of thermal or electric noise of any current-carrying conductor (except lead), with a temperature difference between two points (I am not certain if Johnson noise and the Thomson effect are of similar origin). The important bit of physics to note about thermocouples is that the resultant DV is from a temperature gradient. That is, the thermocouple junction can be considered the reference point for the thermal gradient, but the t/c junction itself does not act like a voltage cell – the t/c bead does not produce a deterministic voltage on each 'side' of the t/c junction. Thermocouples DO NOT MEASURE AT THE JUNCTION. It is not possible to have a temperature gradient at a point. The E-field strength (i.e., volts/meter) at such an

impossible condition would be infinite, sufficient to tear the materials apart, and perhaps warp the fabric of space-time. Destruction of space-time continua is poor engineering practice. Another important principle is the Law of Intermediate Metals. This 'law' says that if a third metal is introduced into the thermocouple circuit, it will not adversely affect the reading if the two junctions of the third metal are at the same temperature. Remember that the Seebeck effect requires that the junctions be at different temperatures to induce a potential. If the two connecting junctions of the third metal are not at the same temperature, non-linear errors will occur from this second thermocouple (see item 5 in the discussion of t/c error sources). Thermocouple voltages vs. t/c temperatures are well documented, and according to the NIST table in the Omega temperature handbook (a highly recommended read), the type K t/c will produce 12.2 mV at 300°C. But you cannot directly connect a meter to the t/c leads, because the connection of the voltmeter leads will make another thermocouple junction. Additional t/c junctions are compensated by using cold junction compensation. By convention the voltage values from the NIST thermocouple tables assume an ‘ice point’ reference (0°C), which results in the TC 'hot' end minus TC 'cold' end, minus the cold junction. The hot end minus cold end concept is important, because a t/c in a 150° component connected to your instrument does not generate the mV table equivalent of 150°. That T/C generates the mV table equivalent of 140°, because the 150° hot end minus the 10° cold end is 140° The other point of question is the fact if you short the t/c cable you will probably read ambient field temp at the instrument, i.e. approx 25°C. Back to the cold junction. All standard thermocouple tables allow for this second ('reference') thermocouple junction by assuming that it is kept at exactly 0°C. Traditionally this was done with a carefully constructed ice bath (hence the term 'cold' junction compensation). For an 'electronic' CJC, the actual temperature at the point of connection of the thermocouple wires to the instrument is measured. And any mechanical contact assemblies up to the CJC must be isothermal (see Item 7 in the error source discussion). Typically, cold junction temperature is sensed by a precision thermistor or some other electronic temperature sensor that is in good thermal contact with the input connector assembly of the measuring instrument. This second temperature reading, along with the

reading from the t/c itself is used by the measuring instrument to calculate the temperature at the t/c tip. A note about electronic sensors: a cold junction compensation performed by a semiconductor temperature sensor is considered less accurate unless the error curve of the device is used to calculate CJC temperature. Any error in the measurement of the cold junction temperature will be seen as error in the t/c measurement. The conversion from the measured t/c voltage to engineering units (temperature) is a somewhat messy 5th to 9th order polynomial, depending on t/c type and required accuracy. The selection of t/c type, for TCA, is limited to 'J', 'K', or 'T'. For any of these t/c types, the voltage gradient across a thermocouple pair is on the order of 100s of microvolts or perhaps 10's of millivolts,

Part B. Thermocouple Measurement Issues and Error Sources 1. The most common source of large temperature errors are caused by unintentional junctions, which results from any place the two different t/c wires may touch prior to the ‘bead’, and where exposed t/c wire touches another surface prior to the ‘bead’ junction. When t/c wire forms a junction by a mechanical crimp or by twisting the t/c wire ends together, several unintentional junctions will be formed that will cause unreliable temperature measurement. 2. For some instruments, the trade-off between lead resistance and t/c signal integrity can be significant. The least thermal shunting (see Issue #4) and best response times are provided by thin t/c wire. But the smaller the wire gage, the higher the lead resistance, which makes it somewhat more sensitive to noise and causes some error depending on the input impedance of the measuring instrument. In general, smaller gauge t/c wire is more susceptible to contamination, annealing, strain, and shunt impedance. Small gauge t/c wire is not a significant issue with a good instrument and careful test layout. A typical exposed junction t/c with 30 AWG wire will have a resistance of approx 15 Ohms/m. The Agilent 34970 data logger input impedance of > 100M will have an error of less than 0.01% for the ridiculous length of 12 meters. The Omega t/c meters have much lower input impedance, and are more prone to lead resistance errors. 3. T/C signals are very susceptible to noise. And by the term 'noise', I mean both electrical and acoustical/mechanical disturbances. The Agilent 34970 data logger is very good at rejecting most of the common mode noise, so noise can be minimized by twisting the insulated portion of t/c wires together as close as possible to the t/c junction. Also, the 34970 uses an integrating analog to digital converter which helps average any noise that does not cause a signal offset. If electrical noise pickup is suspected, switch off power to the test unit and see if the reading immediately has a significant change. Of course, the observable 'immediate' change is limited by the time constant of the measuring instrument. Because we typically install a t/c in an extremely noisy environment, careful consideration must be given to t/c cable routing. T/C cable must not be allowed to 'flap in the breeze' because mechanical t/c motion could result in electrical noise. Acoustic-frequency disturbances, which may induce mechanical motion not seen by the human eye, can cause electrical noise. Note that I have had HALT chamber t/c measurement problems because a t/c was not properly mechanically secured.

Another source of noise are metal parts that have a low Z to an earthed reference. A typical example of an 'earthed' junction would be measuring the temperature of a grounded chassis with a non insulated/non isolated thermocouple. If there are any poor earth connections, several volts may exist between the chassis and the earthed reference of the measuring instrument. These signals are again common mode so will not cause a problem with the 34970, but will cause much weirdness with the Omega meters and with the YEW data logger to a lesser degree. 4. Thermal Shunting can be the dominant measurement error for some components, but its impact for most power-supply measurements is negligible. Thermal shunting is when one mass transfers, via direct contact, energy to another body. All thermocouples have some mass. Heating this mass takes energy, which could effect the temperature measurement. Also, there is some amount of thermal shunting due to the Thomson effect. In general, thermal shunting as the dominant source of error occurs only when the mass of the component if very small, is lead-mounted, there is minimal air flow over the measured component, the t/c wire is a relatively large gauge, and is Type J or Type T (because iron and copper have relatively low thermal R). Several engineering guides claim that thermal shunting can result from airflow over a length of stripped t/c wire, but I have never been able to identify this as a significant error source. A t/c with thin gauge wires may help, as it will cause a steeper gradient of temperature along the thermocouple wire at the junction between test component and ambient air. If thermocouples with very thin gauge wires are used, consideration must be paid to lead resistance. The use of a thermocouple with thin wires connected to much thicker thermocouple extension wire can offer a compromise, but consideration should be given to reduced accuracy problems from the additional connection. 5. Solder/Weld technique of the t/c junction (see Appendix A). If you solder the leads of the thermocouple and the two metals touch only through the solder, as long as the soldered section is all at the same temperature, the solder will have no effect (remember the Law of Intermediate Metals). If, however, the soldered area is relatively lengthy and in a high thermal gradient area where the temperature changes rapidly with very small distances, the temperature readings will be significantly effected by this second thermocouple. Similar error will result from a welded junction that is not sphere-shaped and/or the bead composition is not even mixture of the two t/c materials. 6. Separation of the t/c bead from the test component. This, as of late, appears to be the second-most common source of error. The cyan-acrylate glue. that is used to attach the t/c bead to a component, is also a thermal insulator. So the thickness of any cyan-acrylate between the t/c bead and the component surface will increase thermal response time geometrically with the thickness of the intervening glue layer. In the presence of any

amount of air flow, the cyan-acrylate insulator will never allow the t/c bead to reach the temperature of the component. 7. Isothermal routing of t/c wires is essential. That is, each t/c conductor must be subject to the same thermal gradient. I have observed 10° errors where one of the stripped t/c wires was touching a component while the other t/c conductor was spaced away from any other material contact. Another source of non iso-thermal routing is differential airflow rates; that is, one t/c conductor is exposed to more airflow than the other t/c wire. Again, t/c wire should only be used as a twisted pair. 8. Galvanic action, for the typical temperature measurement is a 'red herring'. Galvanic action should be a very unusual source of error for the typical TCA temperature measurement. This is discussed only for mitigation of problems encountered during use of the environmental and HALT chambers (because I set min humidity for 25% during thermal Type testing, and because of the rapid temperature changes during HALT result in constant phase changes of any material dissolved in the ambient atmosphere). Some insulating materials contain dyes that form an electrolyte in the presence of water and heated water vapor. Other electrolytes can be dissolved in liquid that precipitates on the unit, and from whatever weird stuff that remains from the manufacturing process. The electrolyte generates a galvanic voltage between the leads, which in turn, can produce output signals hundreds of times greater than the net open-circuit input V to the data logger. Thus, good installation practice calls for shielding exposed thermocouple wires from high humidity to avoid this error source. In any case, I do not consider the formation of conductive precipitates a significant issue outside of HALT or safety Type Tests. 9. Adhesives must be applied consistently, using a 'systematic ritual'. Use only cyanacrylate that has been 'vetted' for t/c application. One of the problems with cyanoacrylates is discussed in issue #6. Other problems are caused by using so much cyanacrylate that the component is surrounded by a thermal isolative blanket; moving the t/c before the cyan-acrylate has polymerized; not gluing the t/c cable in additional places for strain relief; contaminating the catalyst; and applying the cyan-acrylate to a contaminated surface - which can make the bond line very hygroscopic – a big issue for strain gages.

Part C. Component Selection and Test Configuration There are two considerations for the location of thermocouple and the selection of components for temperature measurement - to indicate compliance with design (DVT) and component manufacturer’s rating; and to indicate compliance with safety and EMI requirements (yes, I do mean both safety and EMC). These two areas of considerations have significant overlap, so instrumenting for a DVT unit will probably require just a few more t/c channels. The EMC requirement is to not exceed the limits of any magnetic materials under the least favorable rated operating conditions. If the temperature limits of magnetics can be exceeded, then the unit may not comply with CISPR/FCC emission limits under all operating conditions. The safety requirement is driven by the requirement that no safety-critical component shall exceed the rated limits, and that no safety-critical insulating material may exceed the safety-rated temperature limit. What is a 'safety-critical' component? In general, any component that is required for the safe operation of the unit, any component that limits current, voltage, or power to the end-use device, or any component that can prevent the unsafe operation of the unit is a safety-critical component. Specifically, a safety-critical component is whatever the agency or I say it is (and take all complaints to Tom). And what is a 'safety-critical' insulating material? Any material that provides protection from electrical shock, any material that is used where the standard requires Basic, Double, or Reinforced insulation, any material that supports a primary-side component, any material that is used to limit current, voltage, or power. Specifically, safety-critical insulating material is whatever the agency or I say it is (and take all complaints to Tom). There are two test condition considerations for producing data that can also be used for inclusion in safety reports - the least favorable normal operating condition, and the component's hot-spot. Typically, there will not be a single least favorable operating condition that is applicable to all components. An infrared temperature sensor can identify the component hot spot. Specifically, the least favorable normal operating condition is whatever the agency or I say it is (and take all complaints to Tom). Safety Type Tests will also include temperature measurements during abnormal operating conditions.

Part D. Thermocouple Materials We have three t/c wire types available. The choice is determined by the operating environment and the available instrument. For our use, the temperature range of t/c wire is limited by the number of polynomial terms required to calculate an accurate reading, and the temperature range that yields a decent sensitivity. Type J (Iron/Constantan) Engineering guides warn that J types should not be used above approx 400°C as an abrupt magnetic transformation will cause permanent decalibration; this is NOT an issue, simply because our component temperature measurements do not exceed 200°C. Also, do not use J for sub-0°C measurements. As we all have experienced, the of use of a ferrous metal in magnetic fields can result in bogus temperature measurements, so as a general rule, do not use type J on or near magnetics. Type J has good sensitivity (50uV/°C) Type K (Chromel/Alumel) Type K is a good general-purpose t/c and, for our use, has a very wide temperature range. Sensitivity is good (40 µV/°C). In general, use type K unless you have a good reason not to. As the type K material is 'mildly' ferrous, care should be taken for use in very strong magnetic fields. Type T (Copper/Constantan) Do not substitute the negative wire intended for Type T with the Type J wire. As both conductors are non-magnetic, type T is very good for use on any type of transformer or choke. Sensitivity is ok (30-40 µV/°C) if connected to a good instrument. Do not use T for measurements over 300°C, which should not be an issue for our use. Thermocouple Connectors Connectors intended for extending t/c wire or for instrument input must be either rated for the particular t/c type, or be a junction that thermally isolated and referenced to an intermediate metal or a temperature sensor. Do not join t/c wire with normal signal connectors or terminals. Do not expose t/c connectors to an area with a high temperature gradient.

Appendix A – Welding Thermocouples 1. Do not weld t/c wire that has been exposed or has been previously ‘worked’. Use wire pairs from the same production batch. 2. Be consistent in technique, materials, and settings for each type and gauge of t/c wire. 3. Evenly twist the t/c pair for approx 2 to 4 mm. The ends of the twisted pair should be cut to the same length. 4. A single capacitive discharge is recommended for welding t/c material. Use a hi-current DC supply, connected to a capacitor through a switch. Connect the output of the capacitor to a small copper sheet that is attached to a heat sink, and the other end to a pair of ‘welding pliers’. 5. Adjust the charge voltage and close the switch to the capacitor. 6. As a starting point for Type K t/c wire that is between 26 and 32 AWG, weld with 38Vdc at 40,000µf. Adjust coulomb and voltage to your technique and materials. 7. Charge the capacitors and open the switch to the power supply. 8. Using the welding pliers, touch the end of the twisted pair to the copper using a gradual approach until the arc discharges the cap. If the bead is stuck to the copper, do not use – cut and repeat weld. 9. Examine the bead for uniformity of shape and material mixture - typically indicated by even texture and color of bead. The bead should be close to spherical, and the diameter should not exceed 4x the wire thickness. 10. Separate the twisted portion of the wire so that the t/c wires do not make contact before the bead.

Appendix B - References

1. http://hyperphysics.phy-astr.gsu.edu/HBASE/Kinetic/kintem.html 2. www.calcware.com/cwnmcalc2.htm 3. www.omega.com/temperature/Z/zsection.asp 4. http://me.queensu.ca/courses/MECH215/documents/215Temperature.pdf