PCB thermal analysis predicts where a board will get hot, how hot, and whether any part will exceed its rated temperature. You can do it at three levels: a hand calculation from datasheet thermal resistance, a fast layout-aware estimate, or a full 3D simulation. The right one depends on how much power you dissipate and how much a wrong answer costs you. This guide walks through each level, the inputs that matter, and the mistakes that most often make a thermal result look better than the hardware behaves.
TL;DR
- Start with a hand calculation: junction temperature is ambient plus power times thermal resistance. It takes five minutes and tells you whether you need to go further.
- Datasheet theta-JA values come from a standard test board. Your board, copper area and enclosure will give a different number, sometimes a very different one.
- Size copper traces for current first. Vias mostly act as heat paths, not heat sources, so piling on via counts is rarely the fix.
- Use a quick estimate to compare options early. Use full FEA or CFD, then a measurement on a real prototype, before you sign off a design.
- Our thermal map in Circuit Center gives an early 3D estimate from the watts you enter. It is for comparing options, never for sign-off.
Table of Contents
- What thermal analysis actually answers
- Level one: the hand calculation
- Level two: sizing copper and placing vias
- Level three: simulation
- The inputs that decide whether your answer is right
- Where our thermal map fits
- Our perspective: thermal work belongs before layout, not after
- A simple workflow from schematic to prototype
- Common mistakes
- FAQ
- Sources
What thermal analysis actually answers
Three questions cover almost every case.
- Will any junction exceed its limit? Every semiconductor has a maximum junction temperature in its datasheet. Staying under it, with margin, is the pass/fail test.
- Where does the heat go? Into the copper planes, through vias to the other side, into a heat sink, or into the air. Knowing the path tells you what to change.
- What happens to the neighbors? A regulator running at a safe temperature can still cook an electrolytic capacitor or shift a sensor reading two centimeters away.
Power is the first filter. A board of microcontrollers and sensors drawing a few hundred milliwatts rarely needs a simulation. A motor driver, a buck converter at several amps, an LED array or an RF power stage almost always does.
Level one: the hand calculation
The basic relationship is simple: junction temperature equals the ambient temperature plus the power dissipated times the thermal resistance from junction to ambient. Thermal resistance is measured in °C per watt. Chip makers define the common terms (theta-JA for junction to ambient, theta-JC for junction to case, theta-JB for junction to board) in application notes such as Renesas's guide to thermal characterization of packaged semiconductors, which follows the JEDEC JESD51 family of standards.
A worked example. Say a linear regulator drops 12 V to 5 V at 0.4 A. It dissipates (12 − 5) × 0.4 = 2.8 W. If the datasheet gives a theta-JA of 50 °C/W and your enclosure reaches 45 °C ambient, the junction sits at 45 + 2.8 × 50 = 185 °C. That is above the usual 125 °C or 150 °C limit, so you already know the design fails without any software. Your options: switch to a buck converter, add copper and a heat sink, or drop the input voltage.
Why the datasheet number can mislead you
Theta-JA is not a property of the chip alone. It depends on the test board. The ROHM application note for the TO252 package and the research behind IEC 63378-6:2026 show the spread: for a TO252-3 package under JEDEC JESD51-2A conditions, theta-JA is about 132.2 °C/W on a single-layer board but about 23.2 °C/W on a 4-layer 2s2p board. That is a factor of nearly six for the same silicon.
The lesson is practical. When you compare two parts by theta-JA, check that the numbers were measured on the same board type. And when you use a datasheet figure for your own design, ask how close your copper area and layer count are to the test board. If your board is a two-layer design with a small pad, the real number may be closer to the worst case than the headline.
The same standard also introduces a model for transient behavior. IEC 63378-6:2026 specifies a thermal resistance and capacitance model, called DXRC, to predict transient junction temperature. That matters when your load pulses: a part that survives a one-second burst can fail under continuous load, and the reverse.
Two more habits help:
- Use psi-JT (junction to top) with a thermocouple on the case when you measure a prototype. It is easier to relate to a measured case temperature than theta-JA.
- Do not trust a single datasheet number without considering the layout around the part. System-level thermal behavior also depends on the board's thermal conductivity and the enclosure cooling, as Siemens notes in its electronics cooling overview.
Level two: sizing copper and placing vias
Before you simulate, get the copper right. Trace temperature rise depends on current, copper weight, trace width, whether the trace is internal or external, board thickness and how close it sits to a plane. The IPC-2152 standard replaced the older IPC-2221 charts for this reason; a summary is in this guide to determining circuit board current carrying capability, which notes the standard was published in August 2009 and includes supplementary charts for metric units, parallel conductors, etched coils and vias.
A rough rule from that work: internal traces do not necessarily run hotter than outer ones, and a nearby copper plane pulls heat away. So a narrow trace next to a ground plane can carry more than an old nomograph suggests.
Vias are cooler than you think
A common instinct is to add dozens of vias under a hot part and call it done. Measurement work reported in Printed Circuit Design & Fab suggests a via is often cooler than the traces it connects, because it acts as a heat sink for those traces rather than a heat source. A related article in Signal Integrity Journal makes the same case.
What to do with that:
- Size the connected traces and pours for the current first.
- Use thermal vias under exposed pads to move heat to a plane or the opposite side, in a number your fab can reliably fill or tent.
- Check the effect with an estimate rather than adding vias by habit. Past a point, extra vias add cost and routing difficulty without lowering temperature much.
A quick layout checklist for hot parts
| Question | What to look for |
|---|---|
| Is there copper under and around the hot part? | Large pours on the layer the part sits on, tied to planes |
| Is the heat path continuous? | Exposed pad, vias, then a plane, then ideally a chassis or heat sink |
| Are hot parts clustered? | Spread them out or give each its own copper area |
| Are heat-sensitive parts nearby? | Electrolytic capacitors, crystals, sensors, batteries |
| What is the airflow, if any? | Sealed enclosures need a conduction path to the case |
| What is the worst-case ambient? | Use the enclosure's hottest realistic temperature, not room temperature |
Level three: simulation
When the hand calculation leaves you in the gray zone, simulation comes in. There are roughly three kinds of tool.
Layout-aware estimators. These read your copper layers and spread heat across them quickly. They are fast enough to run after every placement change. The open-source KiCad Thermal Sim plugin is an example: a lightweight KiCad plugin that estimates heat spreading across copper layers. Its documentation positions it for comparing options and iterating, not as a replacement for a full solver.
Open-source FEA workflows. If you want to go further for free, KiCad can export geometry that feeds tools such as ElmerFEM, FreeCAD and GMSH. One hobbyist walkthrough of open-source PCB thermal simulation with KiCad and ElmerFEM describes the process as tedious and spread across several programs, though far easier than it used to be. Expect to spend an afternoon on your first run.
Commercial 3D FEA and CFD suites. These handle airflow, radiation, enclosures and transient behavior. They also demand clean inputs. The Ansys documentation on PCB thermal analysis explains that the thermal conductivity of a board varies considerably because of its layered construction and uneven copper distribution. In practice that means your stackup, copper weights and component models must be accurate before the solver gives you anything trustworthy, and a badly defined stackup is a common reason a run fails to converge.
Which level do you need?
| Situation | Suggested approach |
|---|---|
| Low-power board, parts well under their limits | Hand calculation only |
| Choosing between two regulators or two package options | Hand calculation, then a quick estimate |
| Compact power board, new layout | Layout-aware estimate during placement, prototype measurement later |
| Sealed enclosure, mixed power parts, tight margins | Full 3D simulation plus measurement |
| Safety-critical, high-reliability or regulated product | Full simulation, measurement and a formal review |
The inputs that decide whether your answer is right
Whatever the tool, the output is only as good as the inputs. Check these before you trust a number.
- Real power, not nameplate power. Efficiency losses are what heat the board. A converter delivering 20 W at 92% efficiency dissipates about 1.7 W, and that figure moves with load and input voltage.
- Worst-case ambient. Use the hottest realistic temperature inside the enclosure, which is often well above the room.
- Stackup and copper weight. A 1 oz and a 2 oz board behave differently, and so does a 2-layer versus a 4-layer one, as the TO252 numbers above show.
- Heat sinks and mounting. A part screwed to a metal chassis has a very different path to ambient than one on bare FR4.
- Temperature-dependent losses. Resistance rises as copper warms, so power-delivery losses grow at operating temperature. A cool-bench estimate can understate the hot case.
- Source of each value. If you cannot say where a thermal number came from (datasheet, measurement, a guess), mark it as a guess and test its sensitivity.
Where our thermal map fits
Circuit Center's thermal map is built for the early stage, when you are still choosing parts and board layout and want to know which option runs cooler before you commit. You enter the watts for each part, and it estimates how hot each part and the board will run, shown in 3D. Every thermal value shows its source, heat sinks and chassis mounting are built in, and a parts table exports for the mechanical engineer who has to fit the board into a housing.
We want to be plain about its limits. It is an early estimate for comparing options. It is not a sign-off, and it does not replace a full solver or a measurement on a prototype. If your margins are tight, use it to narrow the field, then verify the finalists the slower way.
Because the thermal map sits next to the rest of the workspace, you can open a KiCad project in the Circuit Designer, look at the schematic, board, 3D view and stackup, and price the bill of materials from the catalog without exporting files between tools. Design files stay in your browser unless you choose to save them to your account. When a hot part turns out to need a swap, the Circuit Center parts directory compares price and stock across distributors at 1, 10, 100 and 1,000 pieces, and says whether each price comes from a live distributor feed or an older listing.
Our perspective: thermal work belongs before layout, not after
Most thermal problems we see described by engineers are not exotic. They are a regulator chosen for its price and package without checking the dissipation, a small copper pad under a power part, or a prototype that passes on the bench and fails in a closed box. These are cheap to fix on paper and expensive to fix after boards are built.
That is why we put the thermal estimate in the same workspace as part search and the BOM. Choosing a part is a thermal decision as much as a price decision: a cheaper linear regulator that dissipates three watts can cost you a heat sink, a bigger board and a respin. When the part, the price, the stock and the heat estimate sit on one screen, you make that trade-off while it is still free.
We also think estimates should say where they came from. A number with no source invites false confidence, which is why each value in our thermal map shows its origin, and why we call the result an early estimate rather than an answer. Use it to rank options, then spend your simulation time and your prototype measurements on the one or two designs that survive.
— Matthew
A simple workflow from schematic to prototype
- List your power parts. Regulators, drivers, MOSFETs, LEDs, power resistors. Note the dissipation of each at worst-case load.
- Run the hand calculation. Use datasheet theta-JA, then adjust your expectations if your board is smaller or has fewer layers than the test board.
- Choose parts with thermal headroom. Compare packages as well as prices. An exposed-pad package with a good thermal path can beat a cheaper one that needs extra hardware.
- Draw the copper. Size traces for current using IPC-2152 guidance, add pours and a via array under exposed pads.
- Run a quick estimate. Compare layout options and heat sink choices before routing is final.
- Price the build. Upload your KiCad, Altium or Eagle export to the BOM tool to match every line to real parts and see a recommended supplier per line at your quantity.
- Verify on a prototype. Measure case temperatures under worst-case load in the real enclosure, and compare them with your estimate. Adjust your model with what you learn.
If a part is missing and your schedule is tight, Parts for Pickup maps nearby suppliers with a pickup counter, so a swapped component does not cost you a week of shipping.
Common mistakes
- Testing on an open bench. Free air is kind. Test in the enclosure, at the highest ambient you expect.
- Comparing theta-JA across different test boards. Check the board type before concluding one part is cooler.
- Over-trusting one number. The datasheet value is a starting point. Your layout is the real variable.
- Adding vias instead of fixing the path. Look at the whole route from junction to ambient, and fix the biggest bottleneck first.
- Ignoring pulsed loads. Average power can hide peaks that matter for transient junction temperature.
- Skipping the measurement. Every model is an approximation. A thermocouple or infrared camera reading on a real board is the final word.
FAQ
What is the difference between PCB thermal analysis and thermal simulation?
Thermal analysis is the broader practice of predicting and checking board temperatures, and it includes hand calculations, estimates and measurements. Simulation is one way to do it, using software to solve heat flow across the board and its surroundings. Many designs need only a hand calculation and a prototype check.
Can I do PCB thermal analysis for free?
Yes, at the early stage. You can calculate junction temperature by hand from datasheet values, use a lightweight plugin such as the KiCad Thermal Sim, or set up an open-source workflow with ElmerFEM, FreeCAD and GMSH. Circuit Center's thermal map is also free to use for early estimates. Full commercial solvers cost money and need clean stackup data.
How accurate are quick thermal estimates?
They are good at ranking options and spotting obvious problems, and less good at predicting an exact temperature. Their accuracy depends on your inputs: real dissipation, copper area, stackup and ambient. Treat them as a comparison tool and confirm with a full simulation or a prototype measurement when margins are tight.
Do more thermal vias always lower the temperature?
No. Vias help move heat from an exposed pad to a plane or the other side of the board, but beyond a certain count the gain is small. Research on via temperature suggests vias are often cooler than the traces they connect, so sizing the copper correctly matters more than a large via count.
Is a thermal estimate enough to sign off a design?
Not on its own. Sign-off should rest on a validated simulation or on measurements of a real prototype at worst-case load and ambient, and on any requirements of your industry or customer. If your product is safety-critical, involve a thermal or reliability engineer.
Sources
- Renesas: Thermal characterization of packaged semiconductor devices
- ROHM: TO252 package thermal resistance information
- IEC 63378-6:2026 thermal resistance model for semiconductors
- Siemens: electronics cooling simulation
- Determining circuit board current carrying capability (IPC-2152)
- Printed Circuit Design & Fab: How hot is my via?
- Signal Integrity Journal: Vias are cooler than we think
- KiCad Thermal Sim plugin on GitHub
- Open-source thermal simulation of PCBs with KiCad and ElmerFEM
- Ansys documentation: thermal analysis of printed circuit boards