PCB Thermal Design: Placement, Reliefs and Symmetry

Heat is generated in a few small places on a board and has to leave through the whole of it. Thermal design is the business of arranging those places so that the heat finds a path out, and arranging the parts that cannot tolerate heat so that they are not standing in the way.

This article sets out the placement and copper rules that a thermal design applies, from airflow and heatsink placement to the thermal reliefs that keep a pad solderable and the pad symmetry that keeps small parts in place.

Airflow, Position and the Direction of Escape

On a forced-air product, the hot components are placed where the air is coolest, which normally means near the inlet and away from the corner where the flow recirculates. A part placed behind another part in the same air path is in the other part’s exhaust, and the temperature it reaches will be higher than the airflow figures suggest. Where the layout cannot avoid that, the second part has to be given more margin.

Tall components belong where they do not block the path: a row of tall capacitors placed across the flow turns a duct into a wall. Heatsinks follow the same logic and are oriented so that their fins align with the airflow rather than across it. None of this is visible in the electrical schematic, which is why it belongs in the placement review.

Keeping Sensitive Parts Away From Sources

Components whose parameters change with temperature have to be kept at a distance from anything that runs hot. Electrolytic capacitors are the common case, because their life falls as their internal temperature rises, and the distances are usually specified as a minimum rather than a preference.

A practical rule for a source with a temperature rise above thirty degrees is to keep temperature-sensitive parts at least two and a half millimetres away under forced cooling, and at least four millimetres away in natural convection. Where the space does not allow it, the assembly has to be measured and the parts confirmed to be operating within their derated temperature range, which is a smaller margin and a documented one.

Thermal Reliefs and When Not to Use Them

A pad connected to a large area of copper is difficult to solder, because the copper conducts heat away faster than the iron or the oven can supply it. The standard solution is a thermal relief: the pad is connected by a small number of narrow spokes instead of being attached around its whole perimeter.

The relief reduces the heat path, which improves the joint and increases the resistance of the connection. For a pad that has to carry more than about five amperes, that resistance is not acceptable, and the pad is connected solidly to the plane and soldered by a process that can supply the heat. The choice is therefore a trade between solderability and current capacity, and it is made per pad rather than per board, with the current each pad will carry written down rather than estimated.

Pad Symmetry for Small Chip Components

A small chip component sitting on two pads of unequal thermal mass will not stay where it was placed. As the solder melts, the pad that heats first pulls the part toward it, and if one end of the alloy is still solid the component lifts onto its end. This is the familiar tombstoning defect, and it is a thermal design problem before it is a process problem.

The control is symmetry. Both pads of a chip component should be connected to copper that is similar in area and arranged similarly, and where a connection has to be made to a plane on one side only, the connecting neck should be narrow enough to limit the heat flow. A width of about a third of a millimetre on the asymmetric side is a common starting point. The mechanisms behind the movement are collected in the notes on component shift during reflow.

When Heat Density Exceeds What the Leads Can Carry

A device that dissipates a large amount of power in a small volume cannot be cooled through its own leads alone. Beyond roughly four tenths of a watt per cubic centimetre, the design has to provide a deliberate path: a copper area on the board, a heatsink bonded to the package, or a bus bar soldered alongside the current path.

Bus bars are the traditional answer for high currents, and their connections matter as much as their cross section. Several attachment points rather than one, riveted and soldered in the wave, give a lower resistance and a mechanically sound joint. A long bar also has a coefficient of expansion different from the board, so it has to be free to move during soldering or it will pull the laminate out of shape as it heats. The behaviour of the board under that kind of thermal load is described under dimensional stability and expansion.

Thermal Vias and Copper Areas

A device with a thermal pad on its underside is cooled through the board. The usual construction is an array of small vias under the pad, connecting it to a copper area on the far side or to an internal plane, so that heat can travel away from the package into the mass of the board.

The vias have to be filled and capped before the pad is soldered, or the paste will drain into them during reflow. Their diameter is a compromise between thermal resistance, which favours larger holes, and manufacturability, which favours smaller ones, and the geometry is governed by the same via design rules that apply elsewhere.

Verifying the Thermal Design

The numbers used in a thermal design are estimates until the assembly is measured. A thermocouple on the device case, a thermal image of the board under load and a measurement of the temperature rise of the sensitive parts are the standard checks, and they are performed on a prototype rather than on a simulation.

Where the measured result exceeds the derated limit, the remedies are the same ones the design began with: move the source, move the sensitive part, add copper, or add airflow. The calculation of trace width from current appears in the same review, because a conductor that is undersized contributes its own heat to the problem.

FAQ

Should every large copper area use thermal reliefs? No. Reliefs help soldering and hurt current capacity. They belong on pads that will be soldered by hand or by a wave, and not on pads carrying high current.

How much copper is needed under a thermal pad? Enough area on the far side or an internal plane to spread the heat, connected by an array of filled vias. The limit is usually the size of the area available rather than the copper itself.

Is a heatsink always the answer to a hot device? It is the last answer. Moving the device to a cooler position, spreading the heat into the board and improving the airflow are usually cheaper and more reliable than adding a part that has to be attached and then has to survive vibration, handling and rework.