@pinsl i can review it on monday if you link me the files here!
Discussion
Loading...
@mntmn I just soldered another one of the PCBs and it shows the same problem. Using a shorter / better USB cable seems to help.
@pinsl i've checked your board now. schematic seems fine, but i think the issue is that you have multiple ground splits under your USB diff pair (as this is a two layer board, and you have a lot of other signals crisscrossing on the way) and no ground stitching around the USB diff pair. also the 27Rs are quite big and the diff pair is not differentially routed at the beginning near the MCU. the meander also doesn't help (meanders are not needed at these USB speeds).
@pinsl easiest fix would be to change to a 4-layer board with signal/GND/GND/signal (and power mixed with signal) or signal/GND/PWR/signal (power+gnd in PWR layer), and make sure there's GND on the opposite side always. i compromised about this a bit on the orig. kdb4 design as it is also 2-layer, but the diff pair is mostly referenced to ground along the longest stretches.
@pinsl what you could try is desolder R6/R5 and U4 and try to replace that connection with a shielded (shield connected to GND) twisted pair cable (or length of cut, thin USB2 cable) with smaller 27 ohms resistors on one end
@pinsl there are ground splits here as well as you got 3v3 and 5v routed under the diff pair. maybe this stub also creates the main issue, you could try cutting the traces off close to the USB-C connector
@mntmn That would be great, thanks! Files are here: https://codeberg.org/funnes/reform-split-keyboard/src/branch/split/reform2-keyboard4-pcb