Check Thermistor Self-Heating in a PCB Temperature Divider

A thermistor divider measures the temperature of the thermistor body. That body can be warmer than its surroundings because the measuring current dissipates power in it. A stable, repeatable ADC reading can therefore contain a systematic error that calibration at one excitation level fails to reveal.

Calculate the power across temperature

For a thermistor resistance R and a fixed series resistor S driven by voltage V, the divider current is V divided by R plus S. Thermistor power is V squared times R divided by the square of R plus S. Evaluate this over the intended resistance range rather than only at room temperature.

As an illustrative example, a 10 kilohm thermistor in series with 10 kilohms across 3.3 V carries 165 microamps and dissipates about 0.272 milliwatts. That calculation does not determine the temperature rise until the thermal conditions are known. With fixed V and S, the power expression reaches its maximum when R equals S.

TDK’s NTC technical information relates electrical dissipation, thermal behavior and self-heating. Use the selected sensor’s dissipation information under relevant conditions; a value measured in one medium or mounting arrangement should not be treated as universal.

Review what the sensor is coupled to

Identify whether the intended measurement is board, air, liquid or surface temperature. A PCB-mounted thermistor near a regulator can follow conducted board heat even with negligible electrical self-heating. Copper area, airflow and enclosure contact may dominate the difference between the sensor and the desired measurement location.

Keep those mechanisms separate during debugging. Reducing divider voltage tests electrical heating. Moving the sensor or changing nearby load tests thermal coupling. Changing both at once makes an improved reading difficult to explain.

Evaluate switched excitation carefully

Pulsing the divider can reduce average dissipation, but the ADC must sample after the electrical network settles. Include any filter capacitor, switch resistance and ADC input behavior. Also ensure that the excitation and reference arrangement match the conversion calculation. A nominal duty cycle alone is not enough to establish accuracy.

Compare readings at more than one excitation power while keeping the external temperature stable. Allow for the sensor’s thermal response and the measurement system’s electrical settling. A power-dependent temperature shift is evidence worth investigating, not a reason to apply an arbitrary software offset.

Retain the resistance model, excitation timing, power calculation and mounting description with the design. Recheck them when a sensor package, divider resistor, sampling schedule or enclosure changes, because each can change the relationship between the measured temperature and the intended one.