The hidden thermal contributors in high power density designs

It seems that more and more features are being packaged into smaller electronic modules. Due to the ever-increasing power density of such modules, it is difficult for electronics engineers to keep things running cool. Here, I review a thermal design issue which had some unexpected results and provides a good example of the pitfalls and how to avoid them.

In this scenario, everything looks fine so far with 14 oC of headroom.

Putting a thermocouple on the device and powering up the module in a 105 oC chamber it was found that after one minute, the tab temperature was already up at 140 oC, and about a minute later it was hitting 155 oC before shutting down.

What the engineer failed to take into consideration here was an additional 15W of power being dissipated by the module. To determine how this would affect things, a thermal experiment was carried out with all of the other devices running at full power. The temperature from the part in question was measured at 25 oC with 0, 1, and 2 W of power being dissipated in the IC:

• 0 W: 58 oC
• 1 W: 65 oC
• 2 W: 72 oC

With the added power, the thermal resistance - from the standpoint of the IC - was now only 7 oC/W. The thermal resistance of the module dropped from 11 oC/W, when the IC was powered by itself, down to 7 oC/W when the module was fully powered up.

The drop in thermal resistance when more power is applied, is due to the fact that the greater the delta temperature between the module and the ambient temperature, the faster the heat comes off of the module. And the faster heat comes off of it, the more airflow it creates.

As a result, when the module was measured with only the linear regulator powered up, the delta temperature of the module was limited and gave one thermal resistance number. But when the module was fully powered up, the delta temperature was greater, and therefore the heat was coming off of it faster, creating its own airflow, giving the module greater cooling capability.

This lower thermal resistance comes at a cost though; the starting temperature is no longer 25 oC, but is now a ‘local ambient’ temperature of 58 oC. Recalculating the junction temperature with this new information we get:

At 161 oC, it is understandable why the IC was thermally shutting down. Even with no power being dissipated in the IC itself, the other components on the PCB would take the junction to 138 oC, leaving little headroom for the IC itself.

So in summary, one must take into account the effects of other heat generating components on the PCB, especially if they are located near the part you are concerned with.

The other take-away is that thermal resistance is not a constant. In fact, it is highly variable and can sometimes go in a direction that you wouldn’t expect. Factors that can change a PCB’s thermal resistance include: copper area, copper thickness, whether the copper planes are solid or broken up by traces, the number of vias, whether the vias are filled or not, heat sink design, airflow, and module orientation.

Also, when comparing the thermal resistance numbers on data sheets, be careful that you are comparing the components under similar conditions (similar Cu area and thickness). Addressing these areas will help you achieve a good electrical and thermal design.