When an embedded system crashes under load, the culprit is rarely bad code; it is usually bad thermals. If you are asking what is a heat sensor in the context of PCB design, the direct answer is: a heat sensor is a transducer—typically a silicon bandgap IC, thermistor, or digital I2C chip—that converts thermal energy into an electrical signal to monitor die, case, or ambient temperatures. But simply slapping a sensor on a board does not solve thermal problems. You must use that sensor data to validate your thermal path math and make hard decisions about heatsinks and airflow.
This guide skips the abstract physics and goes straight to the bench. We will calculate a real junction-to-ambient thermal path, interpret derating curves, and terminate with a concrete cooling strategy for a high-dissipation embedded power stage.
What Is a Heat Sensor (And Which One Should You Use)?
In embedded systems, heat sensors fall into two categories: internal silicon monitors and external board-level ICs. Microcontrollers like the ESP32-S3 have internal thermal sensors tied to the RF power amplifier, but these are notoriously inaccurate (often ±5°C) and only measure the MCU die. For true thermal management, you need an external digital heat sensor placed near your highest-wattage component.
The Thermal Path: Calculating Junction-to-Ambient Math
A heat sensor tells you the temperature at a specific physical point, but to prevent silicon death, you need to know the junction temperature ($T_J$) hidden inside the silicon die. We calculate this using the thermal resistance network, measured in °C/W.
The governing equation is:
T_J = T_A + P_D × (R_θJC + R_θCS + R_θSA)
- T_A (Ambient Temp): The air temperature inside your enclosure.
- P_D (Power Dissipation): Watts burned as heat.
- R_θJC (Junction-to-Case): Fixed by the silicon manufacturer.
- R_θCS (Case-to-Sink): Dictated by your thermal interface material (TIM).
- R_θSA (Sink-to-Ambient): Dictated by your heatsink and airflow.
A Worked Numeric Example
Imagine you are powering an Arduino Nano from a 12V industrial rail using an AMS1117-3.3 LDO in a SOT-223 package, drawing 200mA.
- Power Dissipation ($P_D$): (12V - 3.3V) × 0.2A = 1.74W.
- $R_{θJC}$ for SOT-223: 15°C/W.
- $R_{θCS}$ using a Bergquist Sil-Pad 2000: 2°C/W.
- $T_A$ inside a sealed plastic enclosure sitting in a sunlit control panel: 45°C.
If we run this with no heatsink (where the bare PCB acts as the sink, typically $R_{θSA}$ ≈ 50°C/W for a small pad):
T_J = 45 + 1.74 × (15 + 2 + 50) = 45 + 116.5 = 161.5°C
The silicon will fry. The AMS1117 absolute maximum $T_J$ is 125°C. We need to lower $R_{θSA}$.
Reading the Derating Curve: How Hot Is Too Hot?
Every power IC and microcontroller datasheet includes a power derating curve. This graph shows how much maximum allowable power dissipation drops as ambient temperature rises.
How hot is too hot? While 125°C is the absolute silicon limit, operating above 85°C triggers severe failure signatures:
- Electromigration: At >105°C, metal atoms in the silicon traces physically migrate, leading to open circuits over 6-12 months.
- ADC Drift: MCU internal ADC references drift by up to 2% per 10°C rise, ruining sensor precision.
- Thermal Shutdown Hysteresis: The LDO hits 125°C, shuts down, cools to 110°C, restarts, and immediately overheats again, causing a continuous 2Hz reboot loop (brownout signature).
For our 1.74W LDO example, we want to keep $T_J$ under 100°C for long-term reliability. Let us solve for the required heatsink:
100 = 45 + 1.74 × (15 + 2 + R_θSA)
55 = 1.74 × (17 + R_θSA)
31.6 = 17 + R_θSA
R_θSA = 14.6°C/W
We need a heatsink with a thermal resistance of 14.6°C/W or lower. For a deep dive on interpreting these manufacturer curves, refer to the CUI Devices thermal management guide.
Heatsink Selection and Airflow: The Hardware Fix
Now we pick a physical part. Browsing Digikey's heatsink catalog for SOT-223 packages yields the Aavid (Boyd) 576802B00000G.
| Parameter | Bare PCB Copper Pour | Aavid 576802B00000G |
|---|---|---|
| R_θSA (Natural Convection) | ~50.0 °C/W | 12.5 °C/W |
| R_θSA (1 m/s Airflow) | ~35.0 °C/W | 8.0 °C/W |
| Cost (Qty 1) | $0.00 | $1.85 |
| Calculated T_J (at 1.74W, 45°C Ta) | 161.5 °C (Fail) | 91.3 °C (Pass) |
By clipping the Aavid 576802B00000G onto the SOT-223 tab, our $R_{θSA}$ drops to 12.5°C/W. Plugging that back into our equation yields a $T_J$ of 91.3°C. We are safely under our 100°C target.
What airflow and enclosure changes buy you: If your enclosure requires a sealed IP67 rating, natural convection is your only option, and the math above holds. If you can add a single 40mm Sunon MF40101VX fan pushing just 1 m/s of air across the fins, the Aavid heatsink's $R_{θSA}$ plummets to 8.0°C/W, dropping $T_J$ to an icy 78.1°C. Airflow is always cheaper and more effective than adding mass to a heatsink.
The Thermal Decision Tree: Pick Your Cooling Strategy
Stop guessing. Use this decision matrix based on your TMP117 sensor readings and power dissipation calculations to lock in your hardware BOM.
| Condition (Measured/Calculated) | Action Required | Concrete Hardware Pick |
|---|---|---|
| Calculated T_J < 70°C | No heatsink. Use standard PCB copper pour (2oz, 1 sq inch). | Standard FR4 PCB |
| Calculated T_J 70°C - 105°C (Natural) | Add extruded aluminum SMD heatsink with thermal pad. | Aavid 576802B00000G + Bergquist Sil-Pad |
| Calculated T_J > 105°C (Natural) | Add forced convection (fan) or switch to a switching regulator. | Sunon 40mm Fan OR replace LDO with TPS54302 Buck |
| Enclosure is sealed (IP67) & T_J > 90°C | Thermally couple the component tab to the metal enclosure chassis. | Thermal gap pad to aluminum backplate |
The Default Pick: If you are designing a mixed-signal embedded board with linear regulators or high-current MOSFETs and your TMP117 sensor flags case temperatures above 65°C, your default BOM addition is the Aavid 576802B00000G clipped over a 0.5mm thermal pad. It solves 80% of SOT-223 and D-PAK thermal bottlenecks for under $2.00, requires no active airflow, and keeps your junction temps well below the 105°C electromigration cliff.
