Heatsink Selection Guide: How to Calculate Thermal Resistance & Size a Heat Sink

Why Heatsink Selection Is More Than Just "Pick a Big One"

Every power component generates heat. Voltage regulators, MOSFETs, RF power amplifiers, LED drivers — they all turn electrical energy into thermal energy, and that heat has to go somewhere. Every one of these components has a maximum junction temperature ($T_{J(max)}$) stamped on its datasheet, and if you exceed it, reliability goes straight off a cliff. The heatsink's job is simple: keep that junction temperature safely below the limit. But here's the thing — choosing the right heatsink means actually understanding the complete thermal path from the silicon die all the way out to the air around it.

I've seen this go wrong in two directions. Some engineers slap on a massive heatsink just to be safe, wasting cost, weight, and precious board space. Others undersize it, then discover the problem during thermal testing. Or worse — they find out in the field after units start failing. The math to get this right isn't complicated. You just have to actually do it instead of guessing. That's why the Heatsink Selection Calculator exists — it handles the calculation in seconds so you can focus on whether your design actually works.

The Thermal Resistance Chain

Heat flows from the semiconductor junction through a series of thermal resistances. Think of it like resistors in series — each interface adds resistance, and the total determines how hot your junction gets. The complete thermal resistance from junction to ambient air is:

Breaking this down:

• $\theta_{JC}$ is the junction-to-case thermal resistance. You'll find this in the component datasheet, usually buried in the thermal characteristics section.
• $\theta_{CS}$ is the case-to-heatsink thermal resistance. This depends entirely on how you mount the part and what interface material you use between the package and the heatsink.
• $\theta_{SA}$ is the heatsink-to-ambient thermal resistance. This is the spec you're actually solving for when you pick a heatsink.

The fundamental equation that ties everything together is:

Where $T_J$ is your junction temperature, $T_A$ is the ambient temperature around the heatsink, and $P_D$ is the power being dissipated. Rearrange this to solve for the maximum allowable heatsink thermal resistance:

This is the equation that matters. If you can't find a heatsink with $\theta_{SA}$ at or below this calculated value, you've got a problem. Your options at that point are to reduce power dissipation, lower the ambient temperature somehow, improve your thermal interface material, or add forced airflow with a fan.

Worked Example: Linear Regulator Dissipating 5 W

Let's walk through a real example. Say you're using a TO-220 linear regulator to drop 12 V down to 5 V at 700 mA. First, calculate the power dissipation:

Linear regulators are simple but they turn all that voltage difference into heat. Now check the datasheet for thermal specs:

• $T_{J(max)} = 125\,°\text{C}$ — this is the standard rating for most commercial-grade parts
• $\theta_{JC} = 3.0\,°\text{C/W}$ — typical for a TO-220 package

You're planning to use a silicone thermal pad as the interface material, which gives you $\theta_{CS} = 0.5\,°\text{C/W}$. The worst-case ambient temperature inside your enclosure is $50\,°\text{C}$ — not room temperature, because your box will have other components generating heat and it might sit in the sun or a hot equipment rack.

Plug everything into the equation:

So you need a heatsink rated at $11.5\,°\text{C/W}$ or lower. A standard stamped aluminum TO-220 heatsink in the 8–10 °C/W range would work here and give you some margin. Let's verify the actual junction temperature if you use a heatsink rated at $10\,°\text{C/W}$:

That gives you a thermal margin of:

Is 7.5 °C enough margin? Depends on your application. For a benign commercial environment with controlled temperature, probably yes. But if your design sees vibration, altitude changes, occasional solar loading, or extended operation at high ambient temps, you'd want more headroom. Many engineers derate to $T_{J(max)} = 100\,°\text{C}$ in those situations, which would require either a significantly better heatsink or a fundamental design change — maybe switching to a buck converter instead of burning 5 W continuously.

Understanding the Temperature Rating Options

The calculator offers three common junction temperature limits, and picking the right one matters more than you might think:

125 °C (standard) is the most common rating for commercial and industrial-grade components. This is where you start for most designs. It's what the manufacturer tested to, and it's what they'll guarantee. 150 °C (high-temp) shows up on automotive-grade parts and some military-spec components. This gives you more thermal headroom, which sounds great, but don't just assume you can use this number. Check your specific part's datasheet — not all devices are rated for 150 °C even if they're in a high-temp package. 100 °C (derated) is a conservative engineering choice that pays off in reliability. Many reliability guidelines, including MIL-HDBK-217 and Telcordia standards, recommend derating junction temperature by 25 °C or more below the absolute maximum. Why? Because running cooler dramatically improves mean time between failures. As a rough rule of thumb, every 10 °C reduction in junction temperature can double the component's expected lifetime. If you're designing something that needs to run for years without failure, this derating isn't optional — it's cheap insurance.

Choosing the right $T_{J(max)}$ is fundamentally a design decision based on your reliability requirements, not just whatever the datasheet lists as an absolute maximum rating.

Common Pitfalls

Ignoring $\theta_{CS}$ is probably the most common mistake I see. The interface between the component case and the heatsink is not zero resistance. A bare metal-to-metal contact without any thermal compound can easily be 1.0–2.0 °C/W for a TO-220 package. Thermal grease brings this down to 0.3–0.5 °C/W, and a dry thermal pad might be 0.5–1.0 °C/W depending on thickness and quality. Always account for this resistance in your calculations, because it's not negligible when you're trying to squeeze performance out of a marginal design. Using free-air $\theta_{JA}$ instead of $\theta_{JC}$ will completely wreck your calculations. That $\theta_{JA}$ number on the datasheet assumes no heatsink and a very specific test board layout with defined copper area. It's essentially useless for heatsink sizing. Always use $\theta_{JC}$ when you're mounting a heatsink, because that's the actual thermal resistance from the silicon junction to the component's case or mounting tab. Forgetting that ambient temperature isn't 25 °C in the real world. Datasheets test everything at a comfortable room temperature. Your actual enclosure, on a summer day, with other components generating heat nearby, might easily hit 50–70 °C. I've seen designs that worked perfectly on the bench fail in the field because nobody accounted for a hot equipment rack or direct sunlight on an outdoor enclosure. Always design for your actual worst-case ambient temperature, not the lab conditions. Neglecting the effect of airflow is leaving performance on the table. Heatsink $\theta_{SA}$ ratings are almost always specified for natural convection — meaning still air. Adding even gentle forced airflow at 1–2 m/s can cut $\theta_{SA}$ in half or better. If your design already includes a fan for other reasons, make absolutely sure you're using the correct heatsink rating curve for forced convection. The difference between natural and forced convection performance is enormous, and using the wrong number means you're either over-designing by a huge margin or under-designing dangerously.

When the Numbers Don't Work

Sometimes you run the calculation and the required $\theta_{SA}$ comes out ridiculously low — say, under 2 °C/W — and no reasonably sized heatsink can hit that number in natural convection. At that point, you're not picking a heatsink anymore, you're redesigning something fundamental. Your options are:

Add forced airflow to dramatically improve heatsink performance. Even a small fan can make a 5 °C/W heatsink perform like a 2 °C/W heatsink in still air. This is often the cheapest fix if you have the space and can tolerate the noise and power consumption. Reduce power dissipation at the source. Switch to a buck converter instead of a linear regulator. Use a MOSFET with lower $R_{DS(on)}$. Redesign the circuit to operate at lower current. Sometimes the thermal problem is telling you that your circuit topology is fundamentally wrong for the power levels you're trying to handle. Spread the heat across multiple devices or use the PCB copper as a heatsink. Modern power components in exposed-pad packages can dump a lot of heat directly into the PCB if you design the copper area properly. This won't replace a heatsink for high-power designs, but it can significantly reduce the heatsink requirement. Use a higher-rated part with lower $\theta_{JC}$ or higher $T_{J(max)}$. Larger packages generally have better thermal performance. A TO-247 will outperform a TO-220. A component rated for 150 °C instead of 125 °C gives you 25 °C more headroom. Sometimes spending an extra dollar on a better component is cheaper than the mechanical complexity of a massive heatsink.

The calculator makes it easy to explore these trade-offs quickly. Change the power dissipation, adjust the ambient temperature, try different junction temperature limits, and immediately see what heatsink thermal resistance you need. It's much faster than doing the algebra by hand every time you want to try a different scenario.

Try It

Stop guessing at heatsink selection. Plug in your actual power dissipation, your real worst-case ambient temperature, and your thermal resistance values. See instantly whether your heatsink choice has enough margin or whether you need to rethink the design before you commit to a prototype. Open the Heatsink Selection Calculator and run the numbers. It takes maybe 30 seconds and can save you an entire board respin when you discover the thermal problem before manufacturing instead of after.