Cold-Junction Compensation and Calibration in Thermocouple Modules

When calibrating a thermocouple module, numbers sometimes refuse to add up. Often, a small, unglamorous sensor buried on the circuit board becomes the accuracy bottleneck instead of the thermocouple wire itself. This post examines why this occurs and details six methods to address it, ranging from a five-minute procedural fix to a permanent hardware modification.

This guide is written around a real 8-channel module, referred to here as the AFEA-TC-08, but nothing discussed is exclusive to that board. Every thermocouple input module on the market, from low-cost boards to laboratory-grade acquisition systems, operates on the same principle and inherits the same fundamental weakness.

1. How a thermocouple actually works

A thermocouple consists of two different metal wires joined at one end. Heating that joint while cooling (or leaving at ambient temperature) the other ends produces a small voltage across them—typically a few tens of microvolts per degree of difference. This is the Seebeck effect, and it is the entire principle behind every thermocouple.

The critical concept is difference. A thermocouple does not produce a voltage that corresponds to its own absolute temperature. It produces a voltage dependent on the temperature difference between its two ends. Placing a probe in a 100 °C oven while leaving the other end on a workbench means the resulting voltage is a product of both the oven and the bench – not the oven alone.

2. Why a module needs a second sensor

Because a thermocouple only reports a difference, the measurement module must independently determine the temperature of the second end in order to calculate the temperature of the first.

This second end is not a deliberate design addition; it appears automatically. An instrument’s internal wiring is ordinary copper, while a thermocouple’s wires are specific alloys chosen so the pair produces a predictable voltage. Wherever that alloy wire ends and copper begins, a second, unplanned thermocouple junction forms. On almost every module, this occurs at the screw terminals where the probe is connected.

Therefore, every thermocouple measurement genuinely involves two junctions:

• The deliberate junction placed at the actual measurement point.
• The automatic junction that forms at the terminals.

To recover the temperature at the first junction, the module needs to know the temperature at the second. This requirement is the basis of cold-junction compensation.

3. The measurement chain, step by step

Every thermocouple module performs the same four fundamental steps:

1. Measure the voltage the thermocouple pair produces. Because this voltage is tiny, an amplifier raises it before an ADC digitises it.
2. Measure the terminal temperature independently, using a standard sensor (a thermistor, an RTD, or a small temperature IC) placed near the terminals.
3. Convert that sensor reading into an equivalent voltage, using a published reference table for the specific thermocouple type in use.
4. Add the two voltages together and convert back to temperature, using the same table run in reverse.

Steps 1, 3, and 4 are fixed by physics and international standards, varying little between products. Step 2 is where designs diverge and it determines a module’s real-world accuracy far more than the ADC’s resolution.

Cold-junction approach	Typical accuracy	Where it is found
A temperature IC on the PCB, some distance from the terminals	±0.5 to ±2 °C	Low-cost modules, including AFEA-TC-08
A temperature IC deliberately placed beneath the terminal block	±0.3 to ±0.5 °C	Mid-range modules
A thermistor or RTD embedded in a thermally isolated block	±0.1 °C	Laboratory-grade acquisition systems
An external ice-point reference cell	±0.01 °C	Metrology-lab standards

The AFEA-TC-08 features an instrumentation-amplifier front end, a 16-bit simultaneous-sampling ADC, and an on-board temperature sensor supplying the cold-junction reading. That sensor is specified at roughly ±0.3 °C on its own. However, its actual contribution to measurement error is usually worse than that specification suggests. The critical metric is not the sensor’s inherent accuracy, but how perfectly its reading matches the exact temperature at the terminals a few centimetres away.

4. Turning voltage into temperature: the reference table

To convert a thermocouple’s voltage into a temperature, a reference table is required. Standards bodies publish these tables for each thermocouple type, detailing the exact voltage produced when one junction is held at precisely 0 °C (historically, an ice bath) and the other is at a variable temperature T.

This article designates that table as E(T), and its reverse-lookup (voltage in, temperature out) as E^-1(V).

A few approximate values for Type E, used as the running example throughout this text:

T	E(T)
−100 °C	−5.237 mV
0 °C	0.000 mV
25 °C	1.495 mV
100 °C	6.319 mV
200 °C	13.421 mV

This table is rarely referenced manually—the module’s firmware executes the lookups. However, understanding what the table represents is crucial for the next section.

A note on trust: The EMF values quoted anywhere in this article are approximate. Before applying them to critical calibrations, they should be verified against the NIST ITS-90 thermocouple database. An incorrect digit in a reference table is nearly impossible to spot from the final results alone.

5. Why real junctions aren’t at 0 °C, and how the table still works

A common source of confusion is that neither of the two real junctions in a typical setup is ever at 0 °C. The probe tip might be at 100 °C, and the terminal block at 23 °C.

So how does a table built around 0 °C function? The concept is similar to how UTC operates for global time zones. No one in Tokyo or New York is actually on UTC time, but because both cities’ clocks are defined as an offset from UTC, the time difference between them can be calculated by comparing each to that shared reference. 0 °C does not need to be a physical temperature in the system; it only needs to be a fixed mathematical reference point.

Because voltages between temperatures are additive, the voltage between any two temperatures can be found by routing the calculation through 0 °C: