NTC Thermistors: How Negative Temperature Sensors Work

What NTC Actually Means

NTC stands for Negative Temperature Coefficient. It's a material property where electrical resistance decreases as temperature increases. That's it. Simple concept, but the devil is in how you use it.

Most engineers encounter NTC thermistors in two places: temperature measurement and inrush current limiting. The first is what you probably think of when you hear "thermistor". The second is often invisible—sitting quietly on a power supply PCB, protecting against the huge current surge when you first switch on a device.

The reason NTC dominates over its opposite, PTC (Positive Temperature Coefficient), is sensitivity. An NTC thermistor can change resistance by an order of magnitude over a 50°C range. That responsiveness makes them cheap, fast, and accurate enough for most applications without needing a dedicated temperature IC.

How NTC Thermistors Actually Work

NTC thermistors are typically made from a sintered ceramic blend of metal oxides—manganese dioxide, nickel oxide, cobalt oxide, and similar materials. When you heat the material, the charge carriers (electrons) gain energy and move more freely through the crystalline structure. More carriers, lower resistance.

The relationship between resistance and temperature is nonlinear. Unlike a platinum RTD (which is roughly linear), an NTC thermistor's curve is exponential. This nonlinearity is both a feature and a gotcha.

The Math Behind It

The most common model for NTC behavior is the Steinhart-Hart equation:

where $T$ is absolute temperature in Kelvin, $R$ is resistance in ohms, and $A$, $B$, and $C$ are material constants provided by the manufacturer.

For quick estimates or when you only have basic specs, the simpler Beta model works:

Here, $R_0$ is the resistance at reference temperature $T_0$ (usually 25°C), and $\beta$ is the thermistor's beta coefficient, typically 3000–4500 K for common devices.

The Beta model is simpler but less accurate over wide temperature ranges. Steinhart-Hart is more work but handles 0°C to 100°C with excellent accuracy if you have the constants.

A Real Example: Measuring Room Temperature

Let's say you're building a home automation sensor using a 10 kΩ NTC thermistor. The datasheet gives you:

• Resistance at 25°C: 10 kΩ
• Beta coefficient: 3950 K
• Accuracy: ±1°C typical

You measure the thermistor's resistance as 12.5 kΩ. What's the actual temperature?

Using the Beta model:

So your room is roughly 20°C. Not bad. To verify this makes sense: resistance went up from 10 kΩ to 12.5 kΩ, which means temperature dropped below 25°C. The math checks out.

For production code or when you need better than ±1°C, use the Steinhart-Hart equation with three calibration points. Most modern datasheets include the coefficients.

Voltage Divider Readout

You don't measure resistance directly in a microcontroller. You use a voltage divider. Wire your thermistor in series with a fixed resistor, apply a known voltage, and measure the voltage across the thermistor using an ADC.

Rearranging to solve for $R_{NTC}$:

Choose $R_{fixed}$ close to the thermistor's resistance at the midpoint of your measurement range. If you're measuring 0–50°C with a 10 kΩ @ 25°C thermistor, use a 10 kΩ fixed resistor. This gives you the best ADC resolution.

Common Mistakes

Self-Heating

The biggest trap. Current flowing through the thermistor generates heat via Joule heating: $P = I^2 R$. This raises the thermistor's temperature above the actual ambient, throwing off your reading.

For a 10 kΩ thermistor with 1 mA flowing through it, dissipation is $P = (0.001)^2 \times 10000 = 0.01 \text{ W} = 10 \text{ mW}$. That might sound small, but if the thermistor's thermal mass is low and it's sitting in still air, even 10 mW can cause a 1–2°C error.

Keep your measurement current below 100 µA if you can. Use a higher fixed resistor value and accept slightly lower ADC resolution, or use a buffered ADC input.

Ignoring the Nonlinearity

Engineers sometimes assume the thermistor behaves linearly and try to use a simple voltage-to-temperature lookup table with equally spaced points. The error compounds fast at the extremes of the range. Use Steinhart-Hart or at least the Beta model. A 10-point calibration curve is better than pretending it's linear.

Wrong Reference Temperature

The Beta coefficient is only accurate near its reference point. If your datasheet gives $\beta$ at 25°C and you're measuring at 0°C or 80°C, the error can be 2–5%. Use Steinhart-Hart coefficients if you need accuracy outside a narrow band.

Thermal Time Constant

NTC thermistors have thermal mass. A typical 1206 SMD thermistor takes 10–20 seconds to reach thermal equilibrium in still air. If you're sampling every 100 ms, you're chasing a moving target. Either wait long enough between readings or use a low-pass filter on your ADC samples. Most engineers skip this and then wonder why their temperature reading drifts.

Not Accounting for ADC Quantization

If your ADC is 10-bit and your thermistor's voltage divider spans only 2 V out of a 3.3 V rail, each ADC step is about 3.2 mV. Depending on your fixed resistor and thermistor values, that might correspond to 0.2°C per step. Read the datasheet and do the math. If it's coarse, use a 12-bit ADC or adjust your resistor divider.

Inrush Current Limiting with NTC

There's a second major use for NTC thermistors that deserves a mention: soft-starting power supplies. When you first apply power to a device, the input capacitors appear as a short circuit. The inrush current can be 10–50 times the steady-state current, damaging components or tripping circuit breakers.

An NTC thermistor placed in series with the input acts as a current limiter. At room temperature, its resistance is high, restricing current. As current flows, the thermistor heats up (self-heating), its resistance drops, and steady-state current flows. By the time the capacitors are charged, the thermistor is warm and has low resistance, so voltage drop is minimal.

This is passive, requires no control logic, and costs a few cents. The tradeoff is power dissipation in the thermistor itself and a slightly slower power-up transient. For most consumer electronics, it's the right choice.

Typical Values and Selection

Common NTC thermistor values:

• 10 kΩ @ 25°C: Most common for general-purpose temperature sensing. Widely available, good sensitivity, reasonable ADC resolution with standard divider resistors.
• 100 kΩ @ 25°C: Higher impedance, lower self-heating, but requires higher fixed resistor and more careful PCB layout to avoid noise pickup.
• 1 kΩ @ 25°C: Lower impedance, higher current, more self-heating but simpler ADC interface. Used when you want to minimize noise or have a low-impedance ADC input.

Beta coefficients typically range from 3000 K (shallow curve, better linearity) to 4500 K (steep curve, higher sensitivity). For temperature measurement, 3950–4050 K is the sweet spot.

Accuracy depends on how well you calibrate. A bare thermistor is ±5°C typical. With one-point calibration at 25°C, you can achieve ±2°C. Two-point calibration (ice water and boiling water) gets you to ±1°C. Steinhart-Hart with three coefficients from the datasheet: ±0.5°C over a 50°C range.

Try It Yourself

If you're working with an NTC thermistor and need to convert a resistance measurement to temperature, the NTC Thermistor Temperature Calculator handles both Beta model and Steinhart-Hart equations. Plug in your datasheet values, measure resistance, and get temperature instantly. No mental math required.

For inrush current limiting design, check the Inrush Current Limiter (NTC) Calculator to size the thermistor and predict power dissipation.