The prototype is physically together! Finally got my parts in. Case from SuNPE, plate from Sendcutsend, and PCBs from JLCPCB.
First parts that came in were the case parts from SuNPe (apologies for the dirty background). I did some manual deburring, as I explicitly asked them not to deburr (as I wanted to preserve the raw as-machined surface of the aluminum, and SuNPe frequently deburrs using wirebrushes and scour pads that results in a more brushed finish)
Next parts that came in was the plate, from Sendcutsend (in a package with some panels for a different project)
I did the prototype plate in low-carbon steel to safe a bit of money, as I wasn't completely sure it was even going to fit (it fit, but there ended up being a couple of mistakes: two holes were too small for the screw heads I intended to use to fit through, and the cutout for the rotary encoder was too small to allow the rotary encoder to be installed on the PCB after the plate/switches were installed). Steel was also one of the very few materials you can get in 1.5mm thickness here in the US.
I wanted an interesting finish on the steel, and steel can develop
some interesting oxide colors when you heat it up, so I put it over a stove to try to get it to oxidize in interesting patterns.
DO NOT DO THIS WITH A STOVE AS I DID - this ended up warping my plate due to uneven application of heat, initially by quite a bit. I managed to correct it mostly by applying heat to different sides, but it ended up still being slightly warped. The side you heat up will contract slightly as it cools, which causes warpage, as any welder will tell you.
The warpage is a bit more apparent in this shot (I clearcoated the plate also with rattlecan clear spray paint):
The plate flattened out pretty well when I installed it in the upper case with screws. There was a spot in the middle (underneath around where the F5-F8 keys would be) that didn't have any nearby screw-down points, so I tried to flatten the plate when I w as soldering the switches in, using a hose clamp against the case:
In retrospect, if I was doing this again I would use an oven, though that would result in more uniform color, and getting interesting combinations of colors requires uneven application of heat, which will likely result in warpage. There might be some possible combination - the differences in temperature to get different colors are quite small, and you could preheat using an oven to get most of the way there, and then get different colors using a flame, which would reduce warpage compared to heating it over just a flame.
Anyways, the rest of the build:
Turns out soldering 120 diodes in is a chore:
Also the diode for the numpad Enter key hits the stabilizer, so I had to solder it to the underside.
On the controller front, I started with the buck controller circuitry that takes 5V down to 3.3V for the STM32 and LED backlight (I went with a buck instead of an LDO to provide more available current for the backlight, due to the 5V 500mA limit).