Thermocouple Cold Junction Compensation Design
A thermocouple produces a voltage proportional to the temperature difference between its two junctions, not to the temperature of the hot one. Everything the circuit does afterwards is about establishing what the cold junction temperature is, and that is where most of the error comes from.
The Seebeck Effect and the Voltage
Two dissimilar metals joined at both ends produce a current if the two junctions are at different temperatures. Cutting the loop and measuring the voltage gives a value proportional to the difference, and the coefficient is tens of microvolts per degree.
The voltage is small and non linear. A type K thermocouple produces about forty microvolts per degree near room temperature and rather less at high temperature, and the curves are tabulated in the standard for each type.
The measurement is a difference, so the absolute temperature is only known when the temperature of the reference junction is known. That is the whole problem of cold junction compensation, and it has been solved in several ways with different costs.
Where the Cold Junction Is
The cold junction is wherever the thermocouple wire meets a different metal. In a practical instrument that is the terminal block or the connector on the board, and it is usually not at a known temperature.
A compensation cable that matches the thermocouple alloy extends the junction to a place where it can be controlled. The cable is not copper, and it is one of the components that is frequently substituted with an unsuitable wire.
If ordinary copper wire is used, the junction moves to the point where the copper meets the thermocouple wire, which may be inside a hot enclosure or at the end of a long run. That is the mistake that produces an error which varies with the weather.
The isothermal block
cold junction compensation works by making all the junctions the same temperature and measuring that temperature with an independent sensor. An isothermal block is the piece of metal or copper that makes them equal.
The block is a copper area or a metal bar under the terminals, and its purpose is to conduct heat between the junctions quickly so that no gradient can develop. Its thermal resistance has to be low compared with the rate at which the ambient changes.
A thermistor or a silicon sensor is mounted on the same block, close to the junctions. The temperature it reports is taken to be the temperature of the junctions, and any difference between the two is a direct error in the measurement.
The compensation Sum
The measured voltage gives the difference between the hot junction and the block, and the sensor gives the block temperature. The two are added to obtain the hot junction temperature.
The addition is done in the non linear domain. A linear addition of a voltage and a temperature is only accurate over a narrow range, because the thermocouple coefficient varies with temperature.
A practical approach converts the measured voltage to a temperature using the tables or a polynomial, adds the block temperature, and then converts back if a voltage output is needed. The arithmetic is a few lines of firmware and it removes an error that can reach several degrees.
Accuracies and Error Budget
The block temperature sensor is usually the largest term in the budget. A sensor with a half degree accuracy contributes half a degree to the result, whatever the electronics do, so a part with a tenth of a degree accuracy is worth its price.
The amplifier offset adds directly to the voltage and therefore to the temperature. An offset of ten microvolts is a quarter of a degree on a type K thermocouple, which is why a chopper or a low offset amplifier is used.
The amplifier also has to work with the very high source impedance of a thermocouple and reject the common mode of a long cable. The layout of the input is therefore part of the accuracy and not only a signal integrity question.
Layout and Thermal Design
Place the terminal block, the block sensor and the first amplifier stage close together and on the same thermal island. A copper area underneath all three makes them share a temperature and slows changes.
Keep the block away from the power supply and from the display. A few degrees of local heating is a few degrees of measurement error, and the error appears only after the product has warmed up.
Avoid air movement over the block. A fan or a vent produces a fluctuating gradient, and the effect appears as a slow oscillation in the reading that is easily mistaken for a drifting thermocouple.
Open Circuit and Fault Detection
A broken thermocouple reads as an open input, and an amplifier with a bias current then drives the input to a rail. The firmware sees a large off scale value, which is a useful symptom but not a reliable detector.
An intentional pull up resistor at the input, combined with a bias current, produces a defined voltage above the normal range when the circuit is open. The firmware can then report a fault rather than a temperature.
The detection should be checked at production. A channel that reports a plausible temperature with the sensor disconnected is dangerous, because the control system may act on it.
Cable Resistance and Noise
The resistance of a long thermocouple cable is tens of ohms, which is small compared with the input impedance of the amplifier and harmless in itself. What matters is the interference that the cable picks up on its way to the instrument.
A twisted pair with an overall screen, earthed at the instrument end only, keeps most of the interference out. Earthed at both ends, the screen would carry the ground current of the building and couple it straight into the pair.
A filter at the input removes what remains. Its corner has to be low enough to reject the mains frequency and high enough not to slow the response of the channel beyond what the process needs.
Verification and Faults
Verify the channel with a millivolt source at the terminals, using the standard tables to calculate the expected reading. A source is more convenient than a hot bath for checking the electronics, and it isolates the electronics from the sensor.
Check the block temperature with an independent sensor placed on the block itself. A difference between that reading and the one used for compensation is a direct error in the result.
A reading that is correct in the laboratory and wrong in the field usually has the wrong extension wire or a junction outside the isothermal block. The release checks that keep such a channel consistent are collected in our PCB design release checklist, the assembly points in judging PCB quality, and the layout measures for a low level input in our guide to mixed signal board design.
FAQ
Why is cold junction compensation needed? The thermocouple measures a difference between two junctions. Without knowing the reference junction temperature, the absolute temperature cannot be calculated.
Can I use copper wire to extend a thermocouple? No. It creates a new junction wherever it meets the thermocouple alloy. Use the matching extension cable.
How accurate is the block temperature sensor? It sets the accuracy of the channel. A part with a tenth of a degree accuracy is a reasonable choice for a general purpose instrument.
