For 90% of embedded PCB thermal management tasks, the TI TMP117 (a digital I2C sensor) is the definitive type of temperature sensors to use. It delivers ±0.1°C accuracy without the ADC drift, self-heating, and lookup-table headaches inherent to analog thermistors. However, picking the right sensor is only step one. If you place a highly accurate sensor in the wrong thermal node, your microcontroller will still throttle or your power stage will still melt. Effective thermal design requires calculating the junction-to-ambient thermal path, interpreting power derating curves, and matching the sensor's physical packaging to the heat source.
1. The Thermal Path: Why Your MCU's Internal Sensor Isn't Enough
Most modern MCUs (like the ESP32-S3 or STM32H7) include an internal silicon die temperature sensor. This sensor is excellent for preventing the CPU core from melting, but it is virtually useless for monitoring external power components. The internal sensor measures the MCU's junction temperature ($T_J$), not the temperature of the MOSFET or voltage regulator sitting 20mm away on the same PCB.
To manage external heat, you must calculate the thermal path using the standard thermal resistance model. Heat flows from the silicon junction (J) through the case (C), into the heatsink or PCB (S), and finally to the ambient air (A). The governing equation is:
T_J = T_A + P_D × (RθJC + RθCS + RθSA)
- RθJC (Junction-to-Case): Fixed by the silicon manufacturer (e.g., 1.5°C/W for a TO-220 package).
- RθCS (Case-to-Sink): Determined by your thermal interface material (TIM). A standard silicone pad is ~0.5°C/W; bare metal-to-metal with thermal paste is ~0.1°C/W.
- RθSA (Sink-to-Ambient): Determined by your heatsink geometry and airflow. This is the only variable you can easily engineer.
If you rely solely on the MCU's internal sensor, you are blind to the $R_{\theta SA}$ bottleneck of your power stage. You need an external sensor physically coupled to the component case or the local PCB copper pour to close the thermal feedback loop.
2. Which Type of Temperature Sensors Fits Your Embedded Design?
The market is flooded with thermal sensing ICs and discrete components. Choosing the correct type of temperature sensors depends entirely on your required accuracy, bus interface, and physical mounting constraints. Below is a decision matrix to terminate your part selection process.
| Application Condition | Sensor Category | Concrete Part Pick |
|---|---|---|
| Need ±0.1°C accuracy on I2C without MCU ADC calibration | Digital IC | TI TMP117 (Default for PCB/component monitoring) |
| Monitoring Li-ion battery pack bulk temperature on an analog pin | NTC Thermistor | 10kΩ 3950 B-value (e.g., Vishay NTCLE100E3103) |
| Measuring >300°C exhaust, soldering irons, or industrial ovens | Thermocouple | K-Type + MAX31855 breakout |
| High-vibration industrial motor stator or long cable runs | RTD | PT100 Class A (Requires 3-wire RTD-to-digital IC) |
For standard embedded projects—like a 3D printer controller board, a high-power LED driver, or a motor controller—the TI TMP117 wins. It eliminates the need for precision voltage references and complex Steinhart-Hart equation math in your firmware. You simply read a 16-bit register over I2C. For a deeper look at why analog alternatives require so much overhead, review the Adafruit thermistor guide, which highlights the self-heating and non-linearity pitfalls of NTCs.
3. Heatsink Selection and Derating: A Worked Example
Let's apply the thermal math to a real-world scenario. You are using an IRLZ44N logic-level MOSFET in a TO-220 package to switch a 12V, 10A resistive load via PWM from an ESP32.
At 25°C, the datasheet lists the $R_{DS(on)}$ as 22mΩ.
Power dissipation ($P_D$) = $I^2 × R_{DS(on)} = 10^2 × 0.022 = 2.2W.
We must design for the worst-case 4.4W. Our goal is to keep the MOSFET case temperature ($T_C$) below 85°C to protect the PCB solder joints, assuming a worst-case ambient enclosure temperature ($T_A$) of 40°C.
Calculating Required RθSA:
$\Delta T = T_C - T_A = 85°C - 40°C = 45°C$
$R_{\theta SA} = \Delta T / P_D = 45 / 4.4 = 10.2°C/W$
You need a heatsink with a thermal resistance of 10.2°C/W or lower. A standard Aavid Thermalloy 513002B02500G (designed for TO-220 packages) offers a natural convection $R_{\theta SA}$ of roughly 7.2°C/W. Because 7.2 is less than 10.2, this specific heatsink will safely keep the component under the 85°C threshold without requiring forced air. Always verify the manufacturer's thermal sensor and management documentation to ensure your feedback loop matches the physical hardware limits.
4. Enclosure Airflow and Failure Signatures of Thermal Stress
How hot is too hot? Silicon junctions are typically rated to 150°C or 175°C before intrinsic carrier concentration causes the semiconductor to lose its blocking capability. However, designing to the silicon limit is a catastrophic mistake in embedded systems.
The Real Limits:
- FR4 PCB Glass Transition ($T_g$): Standard FR4 begins to soften and lose mechanical rigidity around 130°C to 140°C.
- Solder Joint Creep: Lead-free SAC305 solder experiences accelerated fatigue and creep above 85°C due to the Coefficient of Thermal Expansion (CTE) mismatch between the silicon die, copper leadframe, and PCB.
- Electrolytic Capacitors: The lifespan of nearby aluminum electrolytic capacitors halves for every 10°C rise above their rated temperature (usually 85°C or 105°C).
What Airflow Buys You:
If your enclosure is sealed and natural convection fails, adding forced air dramatically alters the math. Pushing just 100 LFM (Linear Feet per Minute) of airflow across a finned heatsink typically drops its $R_{\theta SA}$ by 40% to 60%. A 15°C/W passive heatsink instantly becomes a ~7°C/W active heatsink with a cheap 30mm 5V brushless fan.
Failure Signatures to Watch For:
- Parametric Shift: Before a part melts, it drifts. RC oscillators inside MCUs lose timing accuracy, causing I2C bus collisions or UART framing errors.
- Brownout Resets: As $R_{DS(on)}$ climbs, the voltage drop across the power stage increases, starving the downstream 3.3V LDO and triggering the MCU's brownout detector (BOD).
- Physical Outgassing: The distinct smell of rosin flux or off-gassing FR4 resin indicates you have crossed the 120°C threshold on the PCB surface.
5. The Final Verdict: What to Buy and Where to Put It
Stop relying on analog NTC thermistors and MCU internal sensors for power-stage thermal management. The default, no-compromise choice for the type of temperature sensors in embedded designs is the TI TMP117 on a breakout board.
| Parameter | TI TMP117 Specification |
|---|---|
| Accuracy | ±0.1°C from -20°C to +50°C |
| Interface | I2C (Address selectable via 1 pin) |
| Resolution | 16-bit (0.0078°C) |
| Supply Current | 3.5 μA (ideal for battery-backed systems) |
Placement Rule: Do not just place the sensor anywhere on the board. Mount the TMP117 breakout less than 5mm from the hot component's case or the primary thermal via array. Use a small dab of thermally conductive, electrically insulating silicone pad (like Bergquist Sil-Pad) to bridge the sensor to the copper pour. Configure the TMP117's internal high-limit alert pin to trigger a hardware interrupt on your ESP32 or STM32, immediately cutting the PWM duty cycle to 0% if the case temperature breaches 85°C. This creates a bulletproof, hardware-level thermal shutdown loop that survives even if your firmware crashes.
