NTC failure modes and how to derate around them

1. Self-heating — the silent accuracy killer

An NTC has a finite dissipation constant — typically 1-5 mW per °C in still air, more in flowing fluid. Whatever current the measurement circuit pushes through the bead is dissipated as heat, and the bead temperature rises above the medium temperature by P_diss / DC.

At a 100 µA measurement current through a 10 kΩ NTC: 100 nW of dissipation, 0.05 °C error in still air. Negligible.

At a 1 mA measurement current through the same NTC: 10 mW, 5 °C error in still air. Not negligible.

This is the most common “mystery temperature offset” on a precision NTC build. The fix is to keep the measurement current well below the self-heating threshold for the application’s dissipation environment. Two practical approaches:

• Pulse the measurement — a 1 ms read every 100 ms gives the bead 99 ms to cool between measurements. Average dissipation drops 100×.
• Use a higher-resistance NTC — 100 kΩ instead of 10 kΩ reduces the dissipation 10× for the same excitation voltage.

2. Thermal runaway — when self-heating goes positive feedback

Because the NTC has negative temperature coefficient, self-heating reduces its resistance, which (in a voltage-source circuit) increases the current, which increases the dissipation. If the dissipation exceeds the heat removal from the bead, you have thermal runaway and the part destroys itself in a few seconds.

The textbook fix is to never drive an NTC from a stiff voltage source. A modest series resistor (or, better, a current-source excitation) gives the NTC a current limit that holds the dissipation bounded regardless of bead temperature. Modern ADC reference designs all include this; legacy or custom analog circuits sometimes do not.

3. Long-term resistance drift

Even at zero electrical stress an NTC’s resistance drifts slightly over time as the ceramic micro- structure stabilises. Typical drift for a modern NTC is around 1 % per year for the first year, then about 0.1 % per year indefinitely. In temperature terms this is < 0.3 °C in the first year and < 0.05 °C per year thereafter.

For consumer-grade applications this is below the noise floor. For metrology-grade applications (calibration references, medical thermometers) the right answer is either (a) a periodic re-calibration in firmware, or (b) a Pt100 RTD instead — platinum drift is 5-10× lower.

Manufacturers can “accelerated-age” NTCs by storing them at 150 °C for several weeks before shipping, which compresses the first-year drift into the factory floor. Ask for aged parts if your application’s initial accuracy matters.

4. Mechanical — lead-pull and bead crack

The bead itself is a small sintered ceramic block, mechanically fragile. The wire leads exit the bead through a glass seal or a conductive epoxy fillet. Two stress modes will break this connection:

• Lead pull — pulling on the lead axially can fracture the seal at the bead. NTC datasheets specify a maximum allowable lead-pull force (typically 5-10 N for a radial bead); installation processes should never exceed this.
• Lead bend — bending the lead at < 2 mm from the bead body, repeated through installation cycles, fatigues the seal. The standard fix is to support the lead with a strain-relief sleeve or to specify a leadframe NTC variant with mechanical strain relief built in.

In an over-moulded probe build (like our IM-series floor sensors, see floor sensor failure mode article) both of these stress modes are eliminated because the bead and leads are fully encapsulated.

5. Vibration / shock

The bead in a leaded NTC will vibrate with respect to its housing if not constrained. Continuous vibration above ~10 g RMS will eventually crack the bead-lead seal. Solutions:

• Pot the bead in a thermally conductive RTV adhesive (typical for automotive thermistors).
• Use a glass-encapsulated bead variant (Jianlu’s CWF glass-bead is rated to 200 °C and inherently shock-resistant).
• Use a SMD NTC mounted directly on a PCB — no leads, no fatigue.

6. Moisture ingress

Bare leaded thermistors are not waterproof. Even small amounts of condensation on the bead will create a parallel leakage path that drops the apparent resistance — reading low (warm) temperatures. In humid environments use a moulded or coated probe build, never a bare bead.

For underfloor heating, see the dedicated article on floor sensor failure modes — the over-moulded build is specifically designed to eliminate this attack path.

7. Chemical attack on lead jacket

PVC lead jackets are not compatible with all environments. Hot oils, fuel vapours, some refrigerants and many cleaning solvents will degrade PVC over time. The bead itself is usually fine, but the lead jacket can fail open-circuit or short-circuit.

In aggressive environments specify silicone (high-temperature, oil-resistant), Teflon / FEP / PFA (universal chemical resistance), or fluoropolymer-encapsulated probe builds.

8. ESD damage

NTCs themselves are extremely ESD-robust (no semiconductor junctions, no thin oxides to puncture). ESD damage is almost always to the surrounding circuit, not the NTC. A grounded probe and a TVS diode on the controller input handle production ESD without difficulty.

Derating summary

For high-reliability builds Jianlu’s CWF series NTC thermistors are available in glass-bead, over-moulded probe, SMD chip, ring-tongue and brass-housed variants, with PVC, silicone, Teflon or PFA cable jackets, and accelerated-aged-stock options for applications that need first-year stability without field re-calibration.