PCB Thermal Design: Rules That Move Heat Out of the Board

Heat leaves a board by three paths: conduction through copper, conduction through the laminate, and convection or radiation from the surface into the surrounding air. The laminate is the weakest of the three by a wide margin, which is why a PCB thermal design that relies on the base material alone rarely achieves its target. Once that is understood, the measures that work follow directly.

Why the Laminate Cannot Do It Alone

The substrates in general use are copper-clad epoxy glass, phenolic glass, and a smaller quantity of paper-based material. They have good electrical and processing properties and poor thermal conductivity. As a path for heat from a dissipating component, the resin contributes almost nothing, which leaves the surface of the package as the only route into the air.

That was acceptable while components were large and power densities low. It is not acceptable now. Surface mount packages such as quad flat packs and ball grid arrays put the majority of their heat into the board through the solder joints, and with the package surface area small, the board has become the primary heat path. The consequence is that improving the thermal performance of the board, rather than of the component, is where the design effort belongs.

Heat Spreaders and Heatsinks

The simplest cases are those with a few hot devices. Where fewer than about three components dominate the thermal load, a heatsink or a heat pipe attached to those components is usually enough, and where that is still not sufficient, a fan can be added to raise the convection coefficient.

When more than about three devices dissipate significant power, a single larger solution is better than several small ones. A spreader plate can be cut or formed to follow the heights of the components on the board and placed over the whole component side, so that each device contacts it. The weakness of this approach is that the assembled heights of surface mount parts vary, so contact is uneven and the thermal path is worse than the drawing suggests. The usual remedy is a soft, phase-change thermal pad between the components and the plate, which conforms to the uneven surface and closes the gaps.

<img src="https://www.gopcba.com/wp-content/uploads/2026/06/智能无人机PCBA-1.jpg" alt="PCB thermal design with copper area and thermal vias under a power device” />

Copper Area and Thermal Vias

Inside the board, the useful conductors are copper, not resin. Heat travels along traces, pours, and plated holes far more readily than through the laminate, so the two levers available to a designer are the fraction of copper that remains on each layer and the number of plated holes that carry heat between layers.

Calculating the actual performance means working with an effective thermal conductivity for the stack, since the composite of copper and resin behaves as neither material alone. In practice the design decisions are simpler: keep copper under and around every dissipating device, connect that copper to internal planes with a via farm, and give the device a copper area large enough to spread the heat before it has to cross the laminate. The same copper that carries the current also carries the heat, so the two requirements usually point in the same direction, and the width used for the current calculation can serve as the starting point for the thermal one.

Placement for Airflow

Where the enclosure is cooled by free convection, the arrangement of the parts is part of the thermal design. Devices should be laid out in lines along the direction the air will move, either lengthwise or crosswise, so that the heated air can rise and leave without being trapped. The air itself follows the path of least resistance, so a board with a large open area will draw the flow through that area rather than across the components that need it.

That has two consequences. The first is to avoid leaving a large empty region on the board, because the air will use it instead of the intended path. The second is to consider the whole enclosure rather than the single board: when several boards sit in one chassis, the same reasoning applies to their spacing and orientation, and a stack of boards that blocks the natural path from the bottom to the top will run hotter than the individual boards suggest.

Zoning by Heat Output

Devices on one board should be arranged by how much heat they produce and how well they tolerate it. The components that generate little heat, or that are sensitive to temperature, such as small-signal transistors, small-scale logic, and electrolytic capacitors, belong at the upstream end of the cooling airflow where the air is coolest. The parts that generate the most heat or tolerate it best, such as power transistors and large-scale integrated circuits, belong further downstream, where the air has already been warmed.

The same logic applies vertically. Components sensitive to temperature should be placed in the coolest region available, which in a typical enclosure is the lower part, and they should never sit directly above a heat source. Where several of them have to be grouped, staggering them horizontally rather than stacking them vertically keeps each one in cooler air.

Edge and Corner Placement

Position matters in two directions. Horizontally, a high-power device is best placed close to the edge of the board, because that shortens the conduction path to the chassis or to a mounting feature; the mechanical aspects of that interface are described in board outline and mounting design. Vertically, it is best placed near the top of the board, so that the air it warms rises away from the rest of the circuit rather than across it.

Corners and edges are the wrong place for a device with no dedicated cooling provision. A component in a corner is surrounded by material on two sides, sees the least airflow, and has the least copper area available to spread heat into. Where a board has several high-power parts, they should be distributed with the heat in mind rather than packed together, and power resistors should be chosen in a physically larger size than the electrical rating alone would require, so that they have the surface area and the layout space to dissipate what they produce.

<img src="https://www.gopcba.com/wp-content/uploads/2026/09/280-2.jpg" alt="Heat spreader plate contacting components over a thermal pad” />

Metal Core Substrates

Where the heat load exceeds what a normal laminate can carry, the answer is a different substrate rather than a different layout. Radio frequency power amplifiers and LED boards commonly use a metal base, with the circuit built on an insulating layer over an aluminium or copper plate. The metal base conducts heat laterally and spreads it across the whole assembly, which is far more effective than any arrangement of copper pours on a resin board. The general trade-off between solid and gridded copper is discussed in copper flooding, mesh or solid.

Avoiding Hot Spots

A hot spot is worse than a uniformly warm board. A region of high power density raises the local temperature above what the average calculation predicts, and that local temperature is what sets the junction temperature of the device inside it. Spreading the power across the board keeps the surface temperature more even and removes the peaks.

Achieving strict uniformity is difficult, and it is not the objective. The objective is to avoid any region where the power density is high enough to create a peak, and then to verify the result rather than assume it. Thermal analysis included in layout tools can estimate the distribution across the board and identify the areas that need attention, and the analysis is worth running whenever the power density is high. Where large currents are involved, the same copper sizing rules that govern the traces are described in trace width and current calculation.

FAQ

Why can the laminate not be relied on to carry heat away? Because the resin is a poor conductor. Heat travels along copper, pours, and plated holes, and the board material contributes very little.

What should sit at the upstream end of the airflow? The components that tolerate heat least, including electrolytic capacitors and small-signal devices. The hottest parts belong downstream, where the air has already been warmed.

Why is a corner a bad place for a hot device? Because it has the least airflow, the least surrounding copper, and the longest path to any cooling feature, so the local temperature is higher than the same device would see elsewhere.

When should a metal core board be used? When the heat load exceeds what copper pours on a resin laminate can spread, as happens with radio frequency power amplifiers and high-power LED arrays.