Thermocouple data acquisition means measuring the very small voltage a thermocouple produces, tens of microvolts per degree, and converting it to temperature. That takes four things a general-purpose input lacks: a high-resolution, low-noise millivolt input; cold-junction compensation for the temperature at the terminals; linearization to the standard tables for the thermocouple type; and good noise rejection, usually including isolation. The final accuracy is the sum of the thermocouple’s own tolerance, the cold-junction error and the input’s error, so it is best treated as a budget.
This article is part of our complete guide to data acquisition (DAQ), and expands on the thermocouple row of our signal conditioning article.
How a thermocouple works
A thermocouple is two wires of different metals joined at the measuring point, the hot junction. Because of the Seebeck effect, a voltage appears that depends on the temperature difference between the hot junction and the point where the thermocouple wires connect to the copper of the measuring instrument, called the cold or reference junction. A thermocouple therefore never measures temperature directly. It measures a difference, and the instrument must know the cold-junction temperature to work out the hot-junction temperature.
Thermocouple types
| Type | Conductors | Approximate range | Approx. sensitivity near 25 °C | Typical use |
|---|---|---|---|---|
| K | Chromel / alumel | −200 to 1,250 °C | 41 µV/°C | General purpose, the most common type |
| J | Iron / constantan | 0 to 750 °C | 52 µV/°C | Older equipment, plastics, reducing atmospheres |
| T | Copper / constantan | −200 to 350 °C | 41 µV/°C | Low temperatures, food, labs; good accuracy |
| E | Chromel / constantan | −200 to 900 °C | 61 µV/°C | Highest output, cryogenic and low-signal work |
| N | Nicrosil / nisil | −200 to 1,250 °C | 26 µV/°C | High-temperature stability, alternative to K |
| R and S | Platinum / platinum-rhodium | 0 to about 1,450 °C | About 6 µV/°C | High temperatures, furnaces, calibration |
| B | Platinum-rhodium alloys | About 600 to 1,700 °C | Very low below 600 °C | Very high temperatures |
Ranges and sensitivities are approximate and depend on wire gauge, sheath and atmosphere. Sensitivity also changes with temperature, which is why linearization matters.
Cold-junction compensation (CJC)
Cold-junction compensation corrects for the temperature at the terminals where the thermocouple connects to the DAQ. The input measures the terminal temperature with a separate sensor, typically a thermistor, RTD or IC sensor mounted on the terminal block, converts it to the equivalent thermocouple voltage, adds it to the measured voltage, and converts the total to the hot-junction temperature. Historically, the reference junction was held in an ice bath at 0 °C; CJC replaces the ice bath with a measurement.
Why CJC accuracy matters so much
Any error in the cold-junction temperature appears one-for-one in the final reading: if the CJC sensor reads 0.8 °C high, every thermocouple on that terminal block reads about 0.8 °C high. Common CJC error sources:
- The accuracy of the CJC sensor itself.
- Temperature differences between the CJC sensor and the actual terminals, caused by airflow, nearby heat sources or warm electronics. Good terminal blocks are designed to be isothermal for this reason.
- Warm-up: CJC readings settle as the enclosure reaches operating temperature.
- Extra junctions: a splice in copper wire or a non-thermocouple connector creates a new, uncompensated junction.
Linearization
Thermocouple voltage is not a straight line against temperature. Standard reference tables and polynomials (based on the ITS-90 temperature scale) define the voltage–temperature relationship for each type, and DAQ software uses the inverse polynomials to convert voltage to temperature. Using a single microvolts-per-degree factor is only acceptable over a narrow range; across hundreds of degrees it introduces errors of several degrees.
What the DAQ hardware needs
- A millivolt input range. A Type K thermocouple at 1,250 °C produces only about 51 mV, so thermocouple inputs typically use a range of roughly ±80 mV.
- High resolution. 24-bit delta-sigma converters are standard, because one degree is only tens of microvolts.
- Mains noise rejection. Temperature changes slowly, so inputs sample slowly (often 1 to 100 S/s) and filter or integrate over whole 50/60 Hz cycles.
- Built-in CJC sensors on the terminal block, with a stated CJC accuracy.
- Open-thermocouple detection, so a broken wire reads as a fault instead of a plausible wrong number.
- Isolation, ideally channel-to-channel, especially with grounded-junction thermocouples or sensors mounted on electrically live or separately grounded equipment.
Worked example: an accuracy budget
Consider a Type K thermocouple with special-limits wire measuring about 200 °C, connected to an isolated 24-bit thermocouple module:
| Error source | Value |
|---|---|
| Thermocouple tolerance (special limits, Type K) | ±1.1 °C |
| Cold-junction compensation accuracy | ±0.5 °C |
| Module measurement error at this range | ±0.3 °C |
| Noise after filtering | ±0.05 °C |
| Worst case (sum) | ±1.95 °C |
| Root-sum-square (more realistic) | About ±1.25 °C |
The thermocouple itself dominates. Buying a more accurate DAQ module would barely change the result. If you need better than about ±1 °C, calibrate the thermocouple and DAQ together as a system, or use an RTD.
Standard tolerances (ASTM E230)
| Type | Standard limits | Special limits |
|---|---|---|
| K and J | ±2.2 °C or ±0.75%, whichever is greater | ±1.1 °C or ±0.4% |
| T | ±1.0 °C or ±0.75% | ±0.5 °C or ±0.4% |
| E | ±1.7 °C or ±0.5% | ±1.0 °C or ±0.4% |
Percentages apply to the measured temperature in °C and are for temperatures above 0 °C. Check the standard and your supplier’s data for the exact range.
Grounded, ungrounded and exposed junctions
| Junction | Construction | Response | DAQ consideration |
|---|---|---|---|
| Exposed | Bare junction | Fastest | Not protected; electrically connected to whatever it touches |
| Grounded | Junction welded to the sheath | Fast | Connects the signal to the sheath and process ground: use isolated inputs |
| Ungrounded | Junction insulated inside the sheath | Slower | Electrically floating, easiest for the DAQ |
A grounded-junction thermocouple on a non-isolated input is a classic source of ground loops: noise, offsets and readings that change when nearby equipment switches on.
Wiring and installation
- Use thermocouple extension wire of the same type all the way to the terminals, and thermocouple connectors of the matching alloy. Any copper section creates extra junctions.
- Observe polarity. In the North American color code, the red lead is negative.
- Route thermocouple wire away from power and motor cables, use twisted pairs, and use shielded wire in noisy areas.
- Ensure good thermal contact and enough immersion depth (a common rule of thumb is ten times the sheath diameter) so the junction is at the temperature you care about.
- Remember response time: a thick sheath may take many seconds to follow a fast temperature change, regardless of the DAQ’s sample rate.
Thermocouple, RTD or thermistor?
| Sensor | Strengths | Limitations |
|---|---|---|
| Thermocouple | Wide range, rugged, inexpensive, fast when small | Lower accuracy, needs CJC, tiny signal |
| RTD (e.g. Pt100) | High accuracy and stability | Narrower range, needs excitation, slower and more fragile |
| Thermistor | Very sensitive, inexpensive | Narrow range, strongly non-linear |
Thermocouple data logger or DAQ system?
A standalone thermocouple data logger is the simple choice for a few channels recorded unattended. A DAQ system makes more sense when you have dozens of channels, need temperatures alongside pressure, strain, vibration or digital signals on one timebase, or need the data to drive control, alarms or pass/fail decisions. Our DAQ requirements checklist helps decide.
Common thermocouple measurement mistakes
- Ignoring CJC accuracy, which adds directly to every reading.
- Splicing thermocouple wire with copper or using the wrong connectors.
- Connecting grounded-junction thermocouples to non-isolated inputs.
- Using a linear scale factor over a wide temperature range.
- Expecting a better accuracy than the thermocouple’s own tolerance allows.
- Mounting terminal blocks near heat sources or in drafts, creating CJC gradients.
- Sampling too fast without mains rejection, adding 60 Hz noise to a slow signal.
Temperature is part of most data acquisition systems Sciotex builds, from multi-experiment R&D software that measures electrical, temperature and vision data on different time bases, to production monitoring systems.
Need temperatures you can trust?
Sciotex designs data acquisition systems with the right thermocouple inputs, compensation and isolation for every channel, built and supported in the US since 1997.
Frequently asked questions
Cold-junction compensation measures the temperature where the thermocouple connects to the instrument and adds the equivalent thermocouple voltage, so the system can calculate the true temperature at the measuring junction.
Type K suits most general-purpose work up to about 1,250 °C. Type T is better for low temperatures and accuracy, Type J for older equipment, Type N for high-temperature stability, and Types R, S and B for very high temperatures.
A Type K thermocouple with standard-limits wire is accurate to about ±2.2 °C or ±0.75%, whichever is greater, before adding cold-junction and instrument errors. Special-limits wire halves that. For better accuracy, calibrate the system or use an RTD.
Only if it has a millivolt input range, enough resolution and cold-junction compensation. A general-purpose ±10 V input cannot resolve thermocouple signals usefully; dedicated thermocouple modules are the usual choice.
Isolation is strongly recommended for grounded-junction thermocouples, sensors on separately grounded or electrically live equipment, and noisy industrial environments, because it prevents ground loops.
