Since I spent a bit of time yesterday seeing how the Pi holds up in a sealed case in hot circumstances, I thought I'd post my findings here for others to benefit from too. I don't have any fancy pictures like those found over there, however, so you might find that post more visually pleasing.
- Raspberry Pi B (we haven't got any model A's sitting around yet)
- Plastic shipping case from RS (the pink one)
- A DS18B20 temperature sensor for measuring the air temperature inside the case (just jumpered onto P1, including a pull-up)
- A heat source (in this case, an incandescent desk lamp at very close range)
I configured the Pi to log the temperature every few seconds both from the DS18B20 and the on-chip sensor (via /sys/class/thermal/thermal_zone0/temp). That done, it got sealed into the plastic case using electrical tape, making it effectively air-tight. The only connection to the outside world was the power lead. While it would have been nice to have had a remote console available to it to monitor the temperature in real time, I wanted to avoid loading the LAN chip at all as we're expecting to be using model A's anyway, and that chip has a considerably smaller temperature range (see the FAQ).
The Pi was then left to run for ~1h under the heat of the lamp. The outside temperature went up to 85C on the side facing the lamp, and ~40-45C on the opposite side, measured using an IR thermometer. This was more than I had hoped for/expected, and got me somewhat concerned for the welfare of the Pi. (Side note: 85 degree plastic isn't very pleasant to touch - guess what made me go looking for the IR thermometer...)
After the hour or so, the lamp was switched off and the Pi left to cool for a bit (and give it time to sync its logs to disk) before it was switched off. The resulting log was the plotted:
As can be seen, the air temperature inside the case went up to about 65C, and the core temperature nudged 98C, but seemed to stabilise at that.
A second run was then performed, this time with the CPU fully loaded:
As can be seen, the temperature is actually noticeably lower this time around. Presumably the distance between the lamp and the Pi was slightly greater here, though the surface temperature again reached 85C. It's not all that surprising to not see an increase in the core temperature though, as others have already concluded that the CPU load does not really contribute much to the heat load (see link earlier in this post, for example).
Conclusion: even in the middle of summer here in Australia (where it can get up to 45C in the shade in my area), the Pi seems unlikely to face any issues, even if stuck in a sealed box outside, especially if it's not exposed to direct sunlight. Long-running tests would be needed to tell whether repeated heating/cooling would cause damage over time, but for now the results are looking good.
Conclusion #2: Don't use hot-glue in the case. The 65C was enough to seriously soften it back to a sticky state. (Side note #2: guess what else I touched?)