The Thermal Reality of RTD Measurements: Self-Heating and Rθ Math
A resistance temperature detector RTD is only as accurate as its thermal coupling to the target and its isolation from its own excitation current. When monitoring high-power semiconductors like IGBTs or MOSFETs, you are fighting two distinct thermal gradients: the junction-to-ambient path of the power device, and the self-heating thermal path of the RTD probe itself.
Let us establish the thermal math for the target device. Suppose you are monitoring an Infineon FF450R12ME4 IGBT module dissipating 150W in a motor drive. The datasheet specifies a junction-to-case thermal resistance ($R_{\theta JC}$) of 0.11 K/W. If you mount the RTD on the exterior baseplate, you are measuring the case temperature ($T_C$), not the junction temperature ($T_J$).
The junction temperature is calculated as:
$T_J = T_C + (P_D \times R_{\theta JC})$
At 150W, the internal gradient is $150W \times 0.11 K/W = 16.5^\circ C$. If your RTD reads 85°C on the baseplate, the silicon junction is already at 101.5°C. If you ignore this $R_{\theta JC}$ offset and set your thermal shutdown at a baseplate reading of 130°C, your junction will hit 146.5°C, dangerously close to the 150°C absolute maximum.
An RTD requires an excitation current to generate a measurable voltage. This current causes $I^2R$ self-heating. A standard Pt100 probe in still air has a dissipation constant of roughly 5.0 °C/mW. If your interface circuit pushes 1mA through a 100Ω Pt100, it dissipates 0.1mW. That yields a self-heating error of 0.5°C. In a vacuum or low-flow gas environment where the probe's thermal resistance spikes to 20 °C/mW, that same 1mA excitation creates a 2.0°C false reading. Always verify your ADC's excitation current and match it to the probe's dissipation constant.
How Hot is Too Hot? Derating Curves and Failure Signatures
Knowing the limits of both the sensor and the power device dictates your thermal trip points. For the RTD itself, "too hot" depends on the assembly materials, not just the platinum element.
| Component | Absolute Max Temp | Continuous Derated Limit | Failure Signature |
|---|---|---|---|
| Pt100 Ceramic Element | 850°C | 600°C | Platinum grain growth causing permanent positive resistance drift |
| PFA Insulated Lead Wire | 260°C | 200°C | Insulation melting, creating leakage paths and erratic low-resistance readings |
| Fiberglass Braided Wire | 500°C | 400°C | Embrittlement and fracturing under vibration |
| MAX31865 Interface IC | 150°C (Storage) | 125°C (Operating) | Silicon latch-up, SPI bus lockups, and ADC reference drift |
When interpreting the power device's derating curve, look at the Safe Operating Area (SOA) and the linear power derating slope. Most silicon devices are rated for maximum power up to a case temperature ($T_C$) of 25°C or 80°C, after which the allowable wattage drops linearly to zero at $T_{J(max)}$ (usually 150°C or 175°C). If your IGBT is rated for 300W at 80°C and derates to 0W at 150°C, the slope is -4.28 W/°C. At a case temperature of 105°C, your maximum allowable dissipation is only 192W. Your RTD firmware must implement this exact derating curve to throttle PWM duty cycles before the hardware fails.
Failure signatures of thermal stress on the RTD assembly manifest in two ways. First, repeated thermal cycling (e.g., from 30°C to 120°C daily) causes the platinum wire to strain against the ceramic bobbin, resulting in a permanent positive drift of 0.2Ω to 0.5Ω (reading 0.5°C to 1.2°C high). Second, if the lead wire transitions through a steep thermal gradient without strain relief, the bond wires inside the transition head will snap, causing intermittent open-circuit faults that the ADC reads as a maximum-scale error (e.g., 32767 on a 15-bit register).
Heatsink Selection and Airflow for the Power Stage
To keep the IGBT within its derated limits, we must size the heatsink based on the thermal path from the case to the ambient air ($R_{\theta SA}$). Returning to our 150W IGBT example, we want to limit $T_J$ to 125°C for long-term reliability in a 50°C ambient enclosure.
The total allowable thermal resistance from junction to ambient is:
$R_{\theta JA} = (T_J - T_A) / P_D = (125 - 50) / 150 = 0.50 K/W$
We subtract the known resistances:
- $R_{\theta JC}$ (Junction to Case) = 0.11 K/W
- $R_{\theta CS}$ (Case to Sink, using 0.05 K/W for high-performance thermal grease like Arctic Silver 5)
$R_{\theta SA} = 0.50 - 0.11 - 0.05 = 0.34 K/W$
You need a heatsink with a thermal resistance of 0.34 K/W or lower. A natural convection heatsink of this size would be massive and heavy. This is where airflow changes buy you significant thermal headroom. Introducing forced air drastically lowers the boundary layer resistance.
Adding just 100 CFM of directed airflow across a standard finned extrusion can drop its effective $R_{\theta SA}$ by 30% to 40%. For our 0.34 K/W target, select the Wakefield-Vette 164-254AB-100 extruded aluminum heatsink. In natural convection, it sits around 0.8 K/W (insufficient). But at 2 meters/second airflow (easily achieved with a 120mm server fan like the Noctua NF-A12x25), its $R_{\theta SA}$ drops to approximately 0.28 K/W, safely clearing our 0.34 K/W requirement and giving you a 4°C thermal margin.
Mount the RTD probe directly into a threaded hole in the Wakefield heatsink base, as close to the IGBT mounting footprint as possible, using thermally conductive epoxy (like 3M TC-2707) to ensure the probe reads the sink temperature without an air-gap lag.
Decision Tree: Picking the Right RTD and Interface
Selecting the correct sensor and interface IC depends on your wiring distance, ambient electrical noise, and required precision. Use this decision path to lock in your bill of materials.
| Condition / Constraint | Selection Path |
|---|---|
| If wiring distance is < 1 meter and space is tight | Choose Pt100, 2-wire. Lead resistance is negligible (~0.1Ω). |
| If wiring distance is 1m to 10m in a noisy VFD panel | Choose Pt100, 3-wire. The third wire compensates for lead resistance. |
| If wiring distance is > 10m or high precision (±0.1°C) is required | Choose Pt1000, 4-wire. Higher base resistance negates lead wire impact; 4-wire eliminates it entirely. |
| If interface IC must reject 50/60Hz mains noise | Choose an IC with integrated sigma-delta ADC and 50/60Hz rejection (e.g., MAX31865). |
| If operating in high-vibration environments (motor housings) | Choose a mineral-insulated (MI) metal sheath probe over a ceramic bobbin to prevent internal wire fatigue. |
The Default Recommendation: For 90% of high-power inverter and motor drive applications where the RTD runs 3 to 5 meters back to the control board through a noisy cabinet, terminate your search here: Buy a Pt1000 Class A, 4-wire, 316 stainless steel sheath probe (such as the Omega PR-13-2-100-1/4-6) paired with the Analog Devices MAX31865 interface IC. The Pt1000 allows you to drop the excitation current to 100µA, virtually eliminating self-heating errors, while the 4-wire configuration ensures that the 5 meters of copper lead wire adds zero offset to your measurement.
Wiring, Compensation, and Final Bench Verification
Once the hardware is selected, proper wiring and bench verification are mandatory before deploying the system to the field. The MAX31865 requires specific reference resistors ($R_{REF}$) depending on the RTD type. For a Pt100, use a 400Ω reference resistor; for a Pt1000, use a 4000Ω reference resistor. Tolerance on this resistor must be 0.1% or better, as any drift here maps directly to a temperature reading error.
When routing the RTD cables, never run them parallel to high-voltage PWM motor leads. The high $dV/dt$ of the IGBT switching edges will capacitively couple into the RTD leads, causing the ADC to read erratic, spiking temperatures. Route the RTD cables at a 90-degree angle to power conductors, or use a shielded twisted-pair (STP) cable with the shield grounded at the control board end only (to prevent ground loops).
Bench Verification Step: Before applying power to the IGBT, submerge the mounted RTD probe and a calibrated reference thermometer (like a Fluke 1523 with a 5615 probe) into an ice-water bath (0.00°C) and a heated oil bath (80.00°C). Record the MAX31865 raw ADC counts. If the Pt1000 Class A reads outside the ±(0.15 + 0.002|t|) °C tolerance band at these points, check your $R_{REF}$ solder joints and verify that your microcontroller's SPI clock polarity (CPOL/CPHA) matches the MAX31865 datasheet requirements. Only after the bench calibration curve is verified should you rely on the sensor to protect your 150W power stage.
For deeper mathematical models on RTD linearization and Callendar-Van Dusen coefficients, refer to the Analog Devices RTD measurement guide and the NIST ITS-90 temperature scale documentation.
