In my previous log I already hinted that the thermally improved layout does quite well. So here's the data! The graph below shows the battery voltage of the LiFePO4wered/Pi when starting to charge from empty, with the load Raspberry Pi 3 running (idle) and WiFi connected:

The thermally improved layout does better than the heat sink! I hadn't really expected that, but I guess it makes some sense. The heat sink pulls the heat from the chip, but the way it is sandwiched between the top and bottom boards is not ideal for convection. So a lot of the heat stays around the chip. The improved layout on the other hand seems to pull the heat away from the chip, into the PCB tail.

So next I wanted to see how the system would perform with the improved layout combined with a heat sink. Unfortunately, that's when I ran into a problem. The WiFi on my Pi 3 stopped working! After fighting with it for a while, I noticed this:

Could that little nick out the the WiFi chip's edge be the problem? It is right where the antenna connections seem to be coming out...

Anyway, I decided to do the testing for the improved layout plus the heat sink with an Ethernet connection. This of course results in extra load so the charge times can't be compared to the earlier WiFi tests, but they can be compared to each other.

Adding the heat sink to the new layout or omitting it doesn't seem to make too much of a marked difference. It's almost in the noise of external factors that could affect the test, such as ambient temperature.

Since I'm very happy with the performance of the new layout, I went ahead and created a panel:

It's on order through Dangerous Prototypes' beta PCB store, which seems to be a collaboration with Dirty PCBs service that I've used successfully in the past.

Another thing to note is that my improved layout prototypes were made by OSH Park and were using the standard 1 oz copper weight. My panels will be made with 2 oz copper, which hopefully will make their thermal performance even better!