I've been slacking. I wanted to write this entry weeks ago but issues with the latest revision have somehow kept me from writing about it.

It's a been a crazy 1.5 months. You may have already heard it but the project on the 2019 Hackaday Prize for Best Design :)
Which was totally awesome, I was at Supercon and showed of my prototype and received the award in person. It was a great experience and it was lovely to talk to so many of you. The enthusiasm around the project is just tremendous.

Now about the latest PCB revision. As I already mentioned it involved a lot of design changes. The biggest being that I switched to the production module of the Jetson Nano SOM and that I switched to a reverse mount SO-DIMM connector.
The PCBs are gorgeous, I use JLCPCBs impedance controlled JLC2313 stackup which allows for really tight diff. pairs.
It was a really large order as I revised not only the motherboard but also all the peripherals I made so far.
A big thank you to JLCPCB for their continuing to sponsor this project, while they are crazy affordable, all these 4 layer ENIG boards and stencils amount to quite a bit of money over time and having a PCB sponsor here definitely helped the pace of the project a lot.
Here some pictures of the PCB and pre and post reflow. Reflow was really good this time, only 2-3 minor shorts.

Unfortunately this special stackup is only available in green. Understandable given the few orders they probably get for these.
Apart from a reversed tantalum cap (which you can see in the last pic) and a missing pullup for the (now slightly smarter) 3.3V LDO it booted up just fine but to my great disappointment, none of the display outputs worked, yet again. Too an even larger dismay, none of the USB ports worked either and I had no clue why. The pinout of the new connector seemed right, it was booting and I could connect to the debug UART of the Jetson.
So I started to suspect a device tree issue originating from the change from the pre-production to the production Jetson Nano SOM.
I had a many weeks long back and forth with the Nvidia support to find the culprit. I learned a lot about Linux device tree and the specifics of the Jetson in this regard to flash changes in the device tree to the SOM. Which was time well spent but this weekend it turned out that this wasn't the root cause at all.
The footprint of the reverse mount SO-DIMM connector was wrong! You might think that this was the first thing I checked, which I did. All seemed well everytime I checked it but my culprit was the way I was checking the footprint. I verified that Pin 1 and 2 and 259 and 260 were in the correct spot and I verified that the power pins line up the way they are supposed to, same with the UART, which they did (otherwise I would have gotten some smoke too).
So what on earth was wrong? Turns out that the provider of the footprint, SamacSys had switched the pin numbering half way through! After the key notch they flipped it and changed back again when arriving at pin 1 and 2.
So right to Pin 2 was no Pin 4 but instead Pin 3! So everything was upside down and all the ports I was testing, USB, DP, HDMI were all on this faulty part of the footprint!

Just to be clear, both footprint and schematic are correct but the pin assignment was wrong half way through!

So back to the drawing board, I fixed the footprint (SamacSys was also very quick to reply and get on the issue btw. except for this instance their footprints were perfect so far, 90% of the footprints on this board are from them). I rerouted the broken parts and ordered revision 4, lets hope that now, finally everything will be all right!