tadfisher/nyx-kb

Low-profile, split, wireless, ergonomic keyboard. Lots of keys but not too many.

Hack

The PCB is a Horizon EDA project in pcb/. It requires the horizon-kb pool, so clone that repo and add it to your pools in Preferences before opening pcb/pcb.hprj.

There is an experimental e-paper adapter board in epd/, but it has not been tested as it ended up not fitting in the allotted space. This is being worked on.

Order

For the main PCBs, send off the Gerbers in pcb/gerbers/nyx.zip to your favorite manufacturing house. I used JLCPCB for the prototype build pictured above, but the board should meet design rules for Oshpark and most other houses.

You will need at least two boards for the keyboard halves. The prototype uses an additional unpopulated board for the bottom plate, which could be less expensive than having one cut from some other material.

As-is, the board requires either Kailh Choc or MX hotswap sockets.

SVGs for the plates and MCU shield can be found in case/. These are sized in millimeters, so double-check when you order that you aren’t paying for something four times the size you need, as some cutting services assume the image is sized such that 1px == 1mm. I had SendCutSend cut the top plate out of .048” 304 stainless steel, which fits Kailh Choc (PG1350) switches perfectly. Sculpteo cut the MCU shield out of 1mm acrylic which turned out just fine.

I painted the top plates by first scuffing the surface, then applying several thin coats of self-etching primer, followed by several thin coats of matte-black Krylon Fusion. This is a pretty durable finish, but it may chip slightly at the edges, so a few clear-coats might be desirable.

BOM

Keycaps

Keycaps required:

The keycaps shown are MBK Choc (as seen here), with dimensions:

Switches use 19mm spacing, so MX keycaps (18x18mm) should be compatible.

Extras

• 301230 rechargeable battery fits under the nice!nano when using sockets. You can go larger by sacrificing some thickness or lopping off the blank area of the top plate under the MCU.
• If using the JST connector, 4x ACHL crimp connectors and a crimping tool is needed. I used the Engineer PA-09 which worked with some finagling, but the PAD-11 may be a more appropriate, but more expensive, solution. You’ll also need two ACHLR connector housings.
• For standoffs and screws, I bought a few variety packs of M2 screws and two sets of knurled aluminum standoffs from Pyro Drone. I initially used nylon nuts as spacers, but you can get thinner by cutting up some rubber hard drive shock mounts. If there’s a reliable source of rubber/silicone spacers about 1.2mm long, suggestions are welcome!
• For nuts, I eventually shelled out for 4 packs of stainless steel M2 hex nuts from McMaster-Carr, but at US$8.57 per pack of 10, there are certainly cheaper options out there. I used 2mm M2 nylon nuts for spacing the bottom plate from the PCB.
• I made a non-slip bottom surface by cutting up a silicone sheet in the shape of the bottom plate and using a hole punch to make clearance for the nuts. This particular sheet attracts debris though, so suggestions are welcome for alternative materials or feet.

Solder

Tips:

• Use a high-quality flux, and lots of it. You can’t use too much flux. I highly suggest ChipQuik tack flux.
• Clean and tin your iron’s tip regularly. I love Hakko’s FS-100 tinning paste for this.
• Don’t go too low on the temperature. I did the whole board at 650°F with leaded solder. You’ll likely need a higher temperature and more heating time with lead-free solder.
• The board uses surface-mounted components to stay fairly thin. Don’t be discouraged; in fact, I find it easier to solder small SMT components than to deal with both sides of the board while soldering through-hole parts.

Start with the diodes. Moving in columns from left to right (or right-to-left if you’re left-handed), apply flux on the pads labeled D* and pre-tin them with a moderate amount of solder. Re-apply flux. Grab the diode with a pair of tweezers, ensuring it is oriented correctly; see the image below.

Heat up one pad until the solder is melted and push the diode in from the side until the pin is seated in the center of the pad. Remove the iron. While pressing down on the diode with the tweezers, heat up the other pad until the other diode leg is seated and bonded with the solder.

After all the diodes are in place, move on to the switch sockets. Again moving in columns, apply flux to the pads, tin them with a moderate amount of solder, and re-apply flux. Place the socket, jam your iron tip in the cup, and apply solder until both the pad and cup are covered and obviously bonded. These are pretty easy to do, and you can go wild with the solder if you feel like it; just make sure the bottom of the cup is bonded to the pad.

I recommend soldering the nice!nano sockets now if you’re using them; see the nice!nano documentation for a great guide on how to do that. Otherwise, solder the nice!nano itself to the board, making sure it fits in the silkscreened outline so that you’re using the correct set of pins.

The JST connector can be tricky. What helps here is a minimal amount of tinning on the pad, tons of flux, and plenty of force with tweezers so it doesn’t slide around on you. Start with the smaller pads first, then anchor with the larger pads.

Solder the reset switch like you did with the diodes.

The through-hole components are the power switch and the optional headers. Be sure to cut off the excess pin length or you may have interference issues when assembling the board.

If soldering the battery to the nice!nano, use the B- and B+ pins, making sure your polarity is correct (red = positive, black = negative).

If using the JST connector, crimp the male pins to the battery leads and insert them into the connector housing, making absolutely sure that the polarity matches the connector socket.

Building

If using the JST connector, insert the battery first, as the top plate covers the connector.

Place the appropriate-length screws in each of the screw holes in the top plate. On the other side of the plate, add the spacer of choice; I cut up some rubber HDD shock mounts. Keep in mind you have 1.2mm of space to work with, so normal 2mm nylon nuts won’t work.

With the top plate inverted, align the PCB and place it over the screws. Use the nylon nuts to loosely tie the assembly together.

Turn the board back over. Working from the center out, insert your switches of choice so that they mount in the socket and clip in to the plate at the same time. I found a thin plastic spudger to be helpful for spacing apart the plate and PCB, but a pair of tweezers could also work if you’re careful.

After all the switches are inserted, you may want to test the electronics before fully assembling the board, in case of a bad solder connection or flipped diode. See Flashing below, or use a multimeter if you like.

Tighten down the spacer nuts underneath the PCB so there isn’t play and the switches aren’t moving around when you push them in.

On the bottom plate, insert four screws into the holes surrounding the MCU. Tighten nylon spacer nuts around these screws so they don’t have play.

With the board and bottom plate in either hand, sandwich them together so all the screws protrude from the holes. Use your nice metal nuts on the bottom.

Flashing

A configuration for ZMK Firmware lives in firmware/. You can grab pre-built images from the latest CI run by clicking it and downloading the firmware artifact. Connect the left board via USB; a filesystem named NICENANO should be mounted. Copy nyx_left_nice_nano.uf2 into the root folder of that mount. Likewise, do the same for the right board using nyx_right_nice_nano.uf2 instead.

Instructions for modifying and building the firmware locally can be found in the ZMK Firmware documentation.

License

Copyright © 2021 Tad Fisher.

All firmware files located in firmware/ are available under the MIT license to match ZMK.

Except for the third-party works in doc/, all other files are available under the GNU GPL version 3 or any later version.