A truly modular keyboard
C
journal for oct 21-24, real photo hero and video demo in readme, white keyword mask
Nathan75b7f525dbcc
diff --git a/JOURNAL.md b/JOURNAL.mdindex 3e7e429..7f38aa4 100644--- a/JOURNAL.md+++ b/JOURNAL.md@@ -2,6 +2,60 @@This journal is auto generated by Stasis.-->+# 10/24/2026 9 PM - It Finally Works!!++_Time spent: 4h_++So the press fit coupon came out and none of the 7 pockets were actually press fit, the magnets just dropped into all of them so the form 4 prints them a bit oversize and I'm sticking with glue. In the end I went with the normal J7 for the keybit (8 4mm x 3mm magnets sticking out 0.75mm) and the J7.1b for the keyword which uses 2 of the 4mm x 1mm magnets on collars on every side, since that one actually worked.++Then I cleaned up onshape, moved everything back into the original keybit, keyword and keybytes docs, trashed the old 4x2 magnet docs and updated the main assembly with the 1.5mm gap between cases. All the joint tests are in [`cad/tests`](cad/tests) and the final cases are in [`production/CAD`](production/CAD), all as STEP files now.++I also changed the keyword to white solder mask with black silkscreen on both sides in kicad.++And then I put it all together and it finally works!!! Here is a picture of it++++and here is the demo++[](assets/demo.mp4)++# 10/23/2026 10 PM - Gluing Magnets and Picking a Joint++_Time spent: 5h_++So I started gluing the magnets in with gorilla super glue gel (from the inside so it doesnt get on the faces) and when I put two J7 keybits together they tilted up a little. I thought I had to sand it but it turned out that glue had gotten onto the outside face of some magnets so they werent flat, after scraping it off with a razor the tilt went away. Btw acetone doesnt hurt the magnets but it does wreck the resin so dont get it on the case.++J7 was the one I liked the most so I made it the main keybit case. The magnets kept deflecting when I pushed them in against the keyswitch so I also designed a jig that drops in where the switch goes so they all stop at exactly 0.75mm.++For the keyword the plain pocket J7 tray was too tight for the 4x2 magnets (probably resin curing tolerance) but the collar one worked. I also realized that I need at least 2 magnets per side on the keyword like on the keybit, and since the collars stick out, a 4x1 magnet never reaches into the tray so it doesnt hit the side contacts or the inductors anymore. That made the J7.1 and J7.1b keyword trays and also a press fit coupon to see if I could skip the glue completely.++I also priced out a full sized keyboard (2 keywords and like 110 keybits) and its around $900 assembled, almost half of that being the pogo pins lol, so if I ever make a full one ill buy them seperately and solder them myself.++# 10/22/2026 9 PM - Magnet Joint Tests++_Time spent: 6h_++So I went back to the magnets and did some math, every joint has 4 pogo pins at around 120gf each so its about 520 to 720gf of spring force pushing the cases apart, and the tiny 2mm x 0.5mm magnets I had only give like 1gf through the case wall which is why it kept falling apart.++I tried adding pockets for 4mm x 2mm magnets but on the keyword they would hit the side contacts and the inductors, so I switched to the 4mm x 1mm ones that I already had. Then I tried a bunch of other ways of holding them together and made several cases so I could test all of them:++- J1 magnets only+- J2 bow tie key that drops in from the top+- J3 snap hooks+- J4 sliding dovetail+- J5 slip pegs+- J6 crush rib pegs+- J7 protruding magnets++For J7 I made the magnets stick out 0.75mm so theyre the length of the pogo pins - 0.5mm, which leaves a 1.5mm gap between cases. The keybit uses 4mm x 3mm magnets and the keyword had 3 versions (J7 with 4x2, J7b with 4x1 on collars and J7c with 4x2 on collars). I printed 2 of each one on the formlabs form 4 and yea++# 10/21/2026 10 PM - Pogo Pin Force++_Time spent: 3h_++So after the first magnet cases I noticed the magnets werent strong enough to hold the modules together against the pogo pins. I spent some time looking on digikey for single right angle pogo pins with a low enough spring force that a 4mm x 1mm N52 magnet could hold it. The closest thing was the Mill-Max 0967 horizontal SMT pins with the low force springs, but they need a different footprint so I would have to redo the boards, so I decided to make the magnets stronger instead.+# 10/20/2026 9 PM - Tinkering With Designs and LED Blink_Time spent: 8h_diff --git a/README.md b/README.mdindex 38e010a..56da0b7 100644--- a/README.md+++ b/README.md@@ -5,7 +5,7 @@</h1><h4 align="center">-A truly modular keyboard with rearrangeable keys, magnetic keycaps and Bluetooth!+A truly modular keyboard with rearrangeable keys, magnetic cases and Bluetooth!</h4><div align="center">@@ -20,13 +20,20 @@ A truly modular keyboard with rearrangeable keys, magnetic keycaps and Bluetooth<p align="center"><a href="#key-features">Key Features</a> •+ <a href="#video-demo">Video Demo</a> •<a href="#pcb">PCB</a> •- <a href="#case-and-keycaps">Case & Keycaps</a> •+ <a href="#case">Case</a> •<a href="#credits">Credits</a> •<a href="#license">License</a></p>-<img src="assets/render.png" alt="Keybytes Render" width="800"/>+<img src="assets/keybytes.jpg" alt="Keybytes" width="800"/>++## Video Demo++<video src="https://raw.githubusercontent.com/NotARoomba/keybytes/main/assets/demo.mp4" controls muted width="800">+ <a href="assets/demo.mp4"><img src="assets/demo-preview.jpg" alt="Keybytes demo" width="800"/></a>+</video>## Key Features@@ -34,7 +41,7 @@ A truly modular keyboard with rearrangeable keys, magnetic keycaps and Bluetooth- **CH32V003** RISC-V microcontroller on every key- **STM32WB55RGV6** main board with Bluetooth and USB- **Pogo pins on all 4 sides** of every key for power and I2C-- **Magnetic keycaps** with 2mm x 0.5mm magnets+- **Magnetic cases** to hold the modules together- **Key programming** from the main board over pogo pins- **2MB QSPI flash** (W25Q16JV) to store the key firmware- **Battery charging** with the BQ25895 and solder pads for a 1S LiPo@@ -57,7 +64,7 @@ A single key with a Cherry MX switch and a CH32V003. Every side has pogo pins fo**Layout:**-<img src="assets/journal/keybit_routing.png" alt="Keybit Layout" width="800"/>+<img src="assets/readme/keybit-layout.png" alt="Keybit Layout" width="800"/>### Keyword@@ -65,28 +72,22 @@ The main board that the keybits connect to. It is split into two PCBs that stack**Schematic:**-<img src="assets/journal/keyword_schematic.png" alt="Keyword Schematic" width="800"/>+<img src="assets/readme/keyword-schematic.png" alt="Keyword Schematic" width="800"/>**Layout:**-<img src="assets/journal/keyword_routing.png" alt="Keyword Layout" width="800"/>+<img src="assets/readme/keyword-layout.png" alt="Keyword Layout" width="800"/>**3D View:**-<img src="assets/journal/keyword_3d_view.png" alt="Keyword 3D View" width="800"/>-<img src="assets/journal/keyword_3d_view_2.png" alt="Keyword 3D View" width="800"/>--## Case and Keycaps--Custom designed cases and keycaps in OnShape, the STEP files are in the [`cad`](cad) folder.+<img src="assets/readme/keyword-3d-angle.png" alt="Keyword 3D View" width="800"/>+<img src="assets/readme/keyword-3d-bottom.png" alt="Keyword 3D View" width="800"/>-<img src="assets/journal/final_case.png" alt="Case" width="800"/>+## Case-The keycaps have internal holes for the magnets and latch on to the side latches of the key switch.+Custom designed cases in [Onshape](https://cad.onshape.com/documents/16fd731b2222007142e5340e/w/1583a55f05f5e1a2f8aaf622/e/787b8ffe055f15084c3845af), the STEP files are in the [`cad`](cad) folder.-<img src="assets/journal/keycap_design.png" alt="Keycap Design" width="800"/>-<img src="assets/journal/keycap_latch.png" alt="Keycap Latch" width="800"/>-<img src="assets/journal/printed_keycap.png" alt="Printed Keycap" width="400"/>+<img src="assets/readme/keyboard.png" alt="Keybit and Keyword cases assembled in Onshape" width="600"/>## Creditsdiff --git a/assets/demo-preview.jpg b/assets/demo-preview.jpgnew file mode 100644index 0000000..6ab1f75Binary files /dev/null and b/assets/demo-preview.jpg differdiff --git a/assets/demo.mp4 b/assets/demo.mp4new file mode 100644index 0000000..6a0a348Binary files /dev/null and b/assets/demo.mp4 differdiff --git a/assets/keybytes.jpg b/assets/keybytes.jpgnew file mode 100644index 0000000..12e8b6aBinary files /dev/null and b/assets/keybytes.jpg differdiff --git a/assets/readme/keybit-layout.png b/assets/readme/keybit-layout.pngnew file mode 100644index 0000000..05c9b72Binary files /dev/null and b/assets/readme/keybit-layout.png differdiff --git a/assets/readme/keyboard.png b/assets/readme/keyboard.pngnew file mode 100644index 0000000..bd8d6e6Binary files /dev/null and b/assets/readme/keyboard.png differdiff --git a/assets/readme/keyword-3d-angle.png b/assets/readme/keyword-3d-angle.pngnew file mode 100644index 0000000..cb65d72Binary files /dev/null and b/assets/readme/keyword-3d-angle.png differdiff --git a/assets/readme/keyword-3d-bottom.png b/assets/readme/keyword-3d-bottom.pngnew file mode 100644index 0000000..606c167Binary files /dev/null and b/assets/readme/keyword-3d-bottom.png differdiff --git a/assets/readme/keyword-layout.png b/assets/readme/keyword-layout.pngnew file mode 100644index 0000000..2b391fdBinary files /dev/null and b/assets/readme/keyword-layout.png differdiff --git a/assets/readme/keyword-schematic.png b/assets/readme/keyword-schematic.pngnew file mode 100644index 0000000..a5b91f1Binary files /dev/null and b/assets/readme/keyword-schematic.png differdiff --git a/hardware/keyword/keyword.kicad_pcb b/hardware/keyword/keyword.kicad_pcbindex 5c9b7fc..0cba3b4 100644--- a/hardware/keyword/keyword.kicad_pcb+++ b/hardware/keyword/keyword.kicad_pcb@@ -39,12 +39,14 @@(stackup(layer "F.SilkS"(type "Top Silk Screen")+ (color "Black"))(layer "F.Paste"(type "Top Solder Paste"))(layer "F.Mask"(type "Top Solder Mask")+ (color "White")(thickness 0.01))(layer "F.Cu"@@ -86,6 +88,7 @@)(layer "B.Mask"(type "Bottom Solder Mask")+ (color "White")(thickness 0.01))(layer "B.Paste"@@ -93,6 +96,7 @@)(layer "B.SilkS"(type "Bottom Silk Screen")+ (color "Black"))(copper_finish "None")(dielectric_constraints no)