When driving high-current loads like stepper motors, halogen lamps, or heavy relays, your power stage generates heat that can silently destroy your circuit. Semiconductor temperature sensors—like the analog TMP36 or digital DS18B20—provide the linear feedback needed for a microcontroller to implement closed-loop thermal derating. But a sensor is only as good as the thermal model it feeds. To prevent silicon death, you must calculate the junction-to-ambient thermal resistance (RθJA) of your power component, select a heatsink based on actual wattage, and interpret the manufacturer's derating curve to set your MCU's throttle thresholds.

Selecting the Right Semiconductor Temperature Sensor

Not all temperature ICs are created equal. When mounting a sensor to a heatsink or component case to monitor a power stage, you must balance accuracy, thermal coupling speed, and self-heating. Self-heating is particularly critical in low-power designs; a sensor drawing too much current will artificially inflate its own temperature reading, causing your MCU to throttle a perfectly cool MOSFET.

Part Number Type Accuracy (at 25°C) Output / Interface Quiescent Current (Self-Heating) Best Package for Heatsink Mounting
Analog Devices TMP36 Analog ±2°C 10 mV/°C (0.5V offset) 50 µA (~0.00025°C rise in TO-92) TO-92 (flat side epoxied to metal)
Texas Instruments LM35 Analog ±0.5°C 10 mV/°C (0V offset) 56 µA (~0.0003°C rise in TO-220) TO-220 (bolt directly to heatsink)
Microchip MCP9700A Analog ±2°C 10 mV/°C (0.5V offset) 6 µA (Ultra-low self-heating) SOT-23 (surface mount on thermal pad)
Maxim DS18B20 Digital ±0.5°C 1-Wire (9 to 12-bit) 1 mA active / 1 µA standby TO-92 (requires timing, high active heat)

For closed-loop analog feedback read by an Arduino or ESP32 ADC, the TMP36 or LM35 are the bench standards. The TMP36's 0.5V offset at 0°C allows you to measure sub-zero temperatures on a single-supply 3.3V or 5V ADC without a negative rail. If you are surface-mounting directly onto a copper pour acting as a heatsink, the MCP9700A's 6 µA draw virtually eliminates self-heating errors.

Thermal Path Math and Heatsink Selection

You cannot pick a heatsink by guessing its size. You must calculate the required case-to-ambient thermal resistance (RθSA) based on your maximum allowable junction temperature and actual power dissipation. The fundamental thermal path equation is:

TJ = TA + PD × (RθJC + RθCS + RθSA)

  • TJ: Junction temperature (°C)
  • TA: Ambient temperature inside the enclosure (°C)
  • PD: Power dissipated by the component (Watts)
  • RθJC: Junction-to-case thermal resistance (°C/W)
  • RθCS: Case-to-sink thermal resistance (thermal pad/paste) (°C/W)
  • RθSA: Sink-to-ambient thermal resistance (°C/W)
Warning: Never design to the absolute maximum TJ listed on page one of the datasheet. A 175°C max rating means the silicon will physically survive, but long-term reliability plummets. Always derate your target TJ to 125°C for continuous operation.

Worked Example: Sizing a Heatsink for an IRFZ44N MOSFET

Assume we are driving a 12V, 10A Peltier cooler using an IRFZ44N MOSFET in a TO-220 package. The MOSFET has an RDS(on) of 0.017Ω at 25°C, but at 100°C, this roughly doubles to 0.034Ω. Our steady-state dissipation is PD = I² × R = 10² × 0.034 = 3.4W. Wait, let's look at a heavier load: a 15W dissipation scenario (e.g., higher current or linear region operation).

Let's calculate for PD = 15W inside an enclosure where TA = 40°C.

  1. Target TJ: 125°C (derated from 175°C max).
  2. RθJC: 1.5°C/W (from the IRFZ44N datasheet).
  3. RθCS: 0.5°C/W (using a Bergquist Sil-Pad 2000 thermal insulator).

Plugging into the equation:
125 = 40 + 15 × (1.5 + 0.5 + RθSA)
85 = 15 × (2.0 + RθSA)
5.66 = 2.0 + RθSA
RθSA = 3.66°C/W

You need a heatsink with a thermal resistance of 3.66°C/W or lower. A bare TO-220 in free air has an RθSA of about 62°C/W, which would result in a junction temperature of 993°C (instant magic smoke). Selecting a real part like the Ohmite FA-T227-25E extruded aluminum heatsink provides an RθSA of roughly 2.5°C/W in natural convection, giving us a safe 1.16°C/W margin for dust accumulation or slightly higher ambient temps.

Interpreting Derating Curves and Airflow Dynamics

Every power semiconductor datasheet includes a Power Derating Curve. This graph shows the linear reduction in allowable power dissipation as ambient temperature rises. For a TO-220 rated at 75W at 25°C case temperature, the line slopes down to 0W at 175°C. The slope of this line is exactly 1 / RθJC. If your semiconductor temperature sensor reads a case temperature of 100°C, you must look at the derating curve to find your new maximum allowable wattage, and program your MCU to throttle the PWM duty cycle if you exceed it.

What Airflow and Enclosure Changes Buy You

Heatsink ratings are strictly dependent on airflow. The Ohmite FA-T227-25E might be 2.5°C/W in free air, but if you seal it inside an unvented IP65 enclosure, the local ambient air (TA) rapidly rises, effectively choking the heatsink. The air inside the box becomes the new, much hotter ambient.

Adding forced convection drastically alters the math. Mounting a 40mm Sunon MF40101VX fan (moving 8.9 CFM) across the fins of a standard extruded heatsink typically drops the RθSA by 50% to 70%. If your passive RθSA is 4.0°C/W, a 40mm fan can pull it down to ~1.5°C/W. This allows your MCU to use the semiconductor temperature sensor to implement a variable fan curve: run the fan at 20% PWM when the sensor reads 50°C, and ramp to 100% PWM at 85°C.

Failure Signatures of Thermal Stress

How hot is too hot? For continuous operation in embedded systems, 125°C at the junction is the practical ceiling. Transients can touch 150°C, but sustained heat triggers physical degradation mechanisms that a simple over-temperature shutdown cannot reverse once initiated.

When your thermal management fails, the silicon doesn't just 'melt'—it fails in specific, diagnosable signatures:

  • Electromigration: At sustained temperatures above 125°C, high current density causes metal atoms in the silicon's interconnects to physically migrate. This increases the MOSFET's RDS(on) over time, creating a positive feedback loop where the part runs hotter, migrates faster, and eventually fails open.
  • Bond Wire Lift-Off: The microscopic aluminum wires connecting the silicon die to the TO-220 pins are bonded with epoxy and ultrasonic welding. Repeated thermal cycling (heating to 130°C, cooling to 40°C) causes the silicon and the packaging materials to expand and contract at different rates (Coefficient of Thermal Expansion mismatch). This shear stress eventually pops the bond wire off the die, resulting in an instant open-circuit failure.
  • Thermal Runaway: In bipolar junction transistors (BJTs) and some power MOSFETs, leakage current increases exponentially with temperature. If your heatsink is undersized and the sensor fails to trigger a shutdown, the increased leakage current generates more heat, which generates more leakage, until the junction physically shorts and destroys the device.
Pro-Tip for Sensor Placement: Do not rely on the MCU's internal semiconductor temperature sensor to monitor external power stages. The ESP32's internal sensor measures the die temperature of the microcontroller, which is thermally isolated from your power MOSFET. Always epoxy an external TMP36 or LM35 directly to the metal tab of the power component, using a high-dielectric-strength thermal epoxy like Arctic Alumina, to ensure your closed-loop derating code reacts in milliseconds, not minutes.

By pairing rigorous RθJA math with the fast, linear feedback of semiconductor temperature sensors, you transition your design from hoping it stays cool to mathematically guaranteeing it will throttle before failure. For deeper reading on IC package thermal metrics, refer to Texas Instruments Application Report SPRA953.