Building a Japanese mechanical split keyboard Pt.5
OK, this should be the one where I fix this thing, right? So we have a couple of keys that don’t work and a bit of an LED Party Mode going on. Come find out what’s wrong and watch as we get this Corne Chocolate split keyboard finally working!
This was streamed on Twitch on Dec 30, 2020.
Special thanks to Jesse Vincent of Keyboardio for loaning the Corne Chocolate and sending the Keyboardio Atreus!
00:00:00 Stream Starts
Hello Everyone! Happy Wednesday!
Scotty welcomes regulars and newcomers alike. Interacts with chat and lets people get in the stream before he starts on the KEyboard.
Something with the LEDs is freaking out, we’ll see in a minute.
Scotty has been shooting a shop tour video for YT!
Baldengineer raids with a party of 29 ( https://www.twitch.tv/baldengineer )
Scotty recaps the keyboard project.
00:02:05 Keyboard status
Switching to the top down camera, Scotty shows us what’s up.
There are 2 keys in the same position on both boards that don’t work.
Initially it looked like there were 4 keys however Scotty has found out 2 of the 4 are not configured / unbound so these are actually fine, just the testing software didn’t pick them up.
Connecting the keyboards to USB, LED Partymode is enabled on the left board. Scotty believes a LED on of the first keys has a soldering issue as it looks like data is getting corrupted.
00:03:08 Let’s take this apart!
To the dismay of chat, Scotty announces “Party mode is not intended” the LEDs should be solid colours, not flashy and weird.
Scotty installed VIA ( https://caniusevia.com/ ) and it was during this control that partymode started.
Scotty takes the screws out of the backplate stand offs and then figures he may have to take out the key switches.
Explaining his thought process, Scotty decides he may not have to as all the soldering is on the back side the PCB anyway and is accessible.
00:05:11 Troubleshooting PartyMode
Scotty opens the documentation ( https://github.com/foostan/crkbd/blob/master/corne-chocolate/doc/buildguide_en.md ) to figure out what the first LED in the chain is.
He wants to check the diode directionality and LED solders.
catching up with chat, Scotty confirms 2 of the 4 LEDs that appeared unresponsive DO work, they’re just not assigned to anything.
Scotty confirms the diodes he placed are in the right direction.
Nightbot has new commands! We have !robot !project !microscope and !switcher
He checks if all diodes should have the same direction, this is confirmed to be true.
Scotty presumes that LED1 is a problem on the left board and right as he touches it, the LEDs fix themselves.
By touching it again, the partymode is enabled again.
Scotty reckons he just needs to reheat a solder joint.
00:10:47 Fixing PartyMode
Scotty prepares the soldering iron, microscope and switcher.
After quickly reheating the joints on LED1, the LEDs appear to work as intended.
Touching the LED doesn’t make the LEDs freak out any more.
The microscope and soldering iron a put away.
00:14:14 Troubleshooting the mystery not working key
Scotty starts VIA and shares his screen.
VIA shows the problem key in question as if it’s in a fully “depressed” state.
Scotty removes the keyswitch from the board and tests the socket with tweezers, shorting the socket out to check for a keypress in VIA.
Nothing notable really happens when using tweezers although admittedly, Scotty doesn’t know if his tweezers are conductive.
He gets his multimeter to check voltage drop.
Scotty works out he does see a difference in voltage across all switches (2V) vs the broken one (0.7V)
He checks the other board as it’s the same switch which is weird. It should not be a symmetrical bug.
If it IS a symmetrical bug, Scotty is buying a lottery ticket.
He removes the backplate from the right board and the switch that has issues.
Checking voltage, he sees 1.2V on the right board and 0.4V on the broken one.
Scotty is convinced it’s not the switches but tests them anyway – they work just fine as expected.
Chat and Scotty now suspect a firmware bug.
Seeing there’s a kicad file ( https://kicad.org/ ) for the keyboard PCB available and downloads kicad in the background.
He returns to checking the diodes. There are some test pads on the board but these aren’t super helpful for this particular problem.
Visual inspection of the board doesn’t reveal any obvious hints as to which diode goes where.
Scotty has some suspicions but would like to see the PCB diagram.
Once kicad installs, Scotty inspects the PCB diagram.
Scotty identifies Switch17 in Row2, Column 4 as the culprit but has no real lead on the cause.
Chat says diodes are to prevent ghosting through back-powering.
Scotty decides to pull the diode for switch 17 to check if that does anything.
Scotty prepares to pull the diode for switch 17.
Alternating melting solder on each diode leg, Scotty rotates the diode off the board and visually confirms correct directionality again.
With the diode removed, the symptom does not change.
Scotty quickly checks if his tweezers conduct before checking if shorting other key switch contacts does anything.
That works as expected but doesn’t give any more meaningful information.
Scotty changes his multimeter probes to grabbers and connects to the pins on the pro micro to check continuity and immediately gets a hit.
This sort of confirms the fully depressed state so it looks like a short, but testing other switches doesn’t get any meaningful info.
He connects the multimeter to the pins for switch 1 and checks activity there, notices he has to set the multimeter to diode mode and confirms correct functionality.
So, with new helpful information, he reconnects to the pro micro pins for the broken key and now has a proper confirmation there is a short.
Scotty removes the OLED to try to remove the MCU but realizes it’s soldered on.
He breaks out the microscope again to inspect the solders.
Immediately spots the mistake; one of the pin solders is shorting with row 2.
@DumDuck caught this earlier, thank you DumDuck!
00:58:06 Fixing the mystery not working key
A visual inspection confirms a very similar short has been made on the other board.
Scotty gets out the soldering iron and flux to clean up the pin solders that are causing the shorts.
reheating the pins mostly fixes the issue but some solder wick is used on one for good measure.
Scotty has a minor false start in thinking the fix is done and wanting to test but is reminded by chat to put the diode back he removed earlier.
01:07:01 Keyboard Re-assembly
without putting the keyboard completely back together, Scotty quickly checks the VIA Keymap test by shorting the socket with his tweezers.
Scotty puts the key switches back in place and tests both boards, All looks ok.
Scotty screws the backing plates back on while chatting with chat about loosing and looking for small components like screws and the importance of spares.
Putting the OLED Cover on, chat moves to /me chat
Scotty gets out the keycaps!
It’s going to be a surprise how it all fits together because several caps certainly won’t fit and others seem too big.
Once Scotty completes the letters he’s left with ill-fitting keys for a “normal” keyboard that don’t really fit the chocolate.
He starts using random keys but as he completes the right board, the left board goes into partymode again.
He sticks on 3 entirely random keycaps, unplugs the board and sets up to fix the LEDs, again.
Scotty removes the LED that’s giving issues and pre-emptively replaces another one while chat talks about Alexa wake words
Once the LEDs are replaced, the keyboard looks like it’s fully functional!
01:44:18 Keyboard testing
The keyboard works!
Scotty sticks rubber feet onto the right backplate, puts the left backplate back onto the board and attaches the feet to it.
The Pro Micro MCU pins on the left board interfere with the board laying flat on the desk so for one last time, Scotty removes the backplate and cuts down the pins.
During this, he talks chat through the Gas mask project and how YT content creation can sometimes be misaligned with world events and thus affects the project as a whole.
Noticing cutting the pins worked, he does the same on the right board.
Once the boards are put together and powered on again, Scotty runs through the VIA keyboard tester.
He swaps round 2 keys to give better LED passthrough and tries typing with the chocolate in vim.
Scotty goes through the VIA modes and layers a little bit.
01:58:35 Winding down
Scotty catches up on chat
Apparently requests queue isn’t showing up for Scotty and he’s been having more trouble with it lately.
While he fulfils a hydrate he starts to wind down as it’s nearing curfew.
Scotty has been playing around with making keycaps and is interested in maybe designing his own keyboard.
Chat concurs that this would be great.
Tired and hungry as he is, Scotty wraps up the stream.