17 Commits
Author SHA1 Message Date
William RickertandClaude Opus 5 49ad7e97cf Verify the L103/F103 pin overlay: all 48 identical
Parsed the pin-definition tables out of both datasheets - Table 2-1-1 (CH32L103,
C8T6 column) and Table 2 (STM32F103, LQFP48 column) - and diffed them pin by
pin. Zero mismatches, including OSC_IN/OSC_OUT, NRST, PA11/PA12 for USB,
PA13/PA14 for SWD, and BOOT0. Power pins differ only in naming.

This was the one unverified assumption the whole project rested on: three
boards laid out to a footprint that has to accept either MCU. It holds.

Full table recorded in Documents/pinout-L103-vs-F103.md.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
2026-09-01 10:42:07 -05:00
William RickertandClaude Opus 5 04a52b1eea Add PrtSc to the numpad, filling the nav block
Between Del and PgDn, which makes the nav block a solid 3x2 and leaves the
arrows isolated by a symmetric pair of gaps rather than three asymmetric ones.
26 keys. PrtSc earns a key because it is the screenshot key on every Linux
desktop; Insert, ScrLk and Pause stay on a layer.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
2026-09-01 10:35:24 -05:00
William RickertandClaude Opus 5 81abd43fdc Settle the numpad layout: 25 keys, no encoder
Nav block across the top with an isolated inverted-T below it, operators in a
right-hand column beside 9/6/3 where a numpad puts them, and a 2u Enter plus
+ on a stepped thumb cluster spanning the three nav columns. Six by five,
~114 x 105 mm.

The volume encoder is dropped, which makes this a completely plain QMK
keyboard - no encoder, no pointing device, no consumer control. Firmware notes
and BOM updated; the numpad board is now ~$4.11 in parts.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
2026-09-01 08:59:04 -05:00
William RickertandClaude Opus 5 e555e802e1 Left half schematic: matrix, MCU and passives placed
34 switches matching the 34-key layout, 68 diodes, 40 capacitors, the
crystal, MCU and connectors. 262 symbols total.

Committed with KiCad still open on the project - this is the last saved
state, not necessarily a finished edit.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
2026-09-01 08:27:04 -05:00
William RickertandClaude Opus 5 519c3a22e4 Fix paths after the docs/Documents merge
Every image and link pointed at docs/, which no longer exists - the README
rendered with six broken images. Repoints them at Documents/, updates the
board table to the real Schematics/Left|Right|Numpad tree, and renames
firmware/ to Firmware/ to match the capitalised convention.

Documents/README.md now describes the whole folder rather than just the
datasheets, since the diagrams and BOM live there too.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
2026-08-28 08:26:38 -05:00
William b3bff50525 Started schematic 2026-08-28 06:55:35 -05:00
William RickertandClaude Opus 5 fb124206c1 Add licensing: CC BY-NC-SA 4.0 hardware, GPL-2.0 firmware
Hardware and docs get CC BY-NC-SA 4.0 - build it, change it, share changes,
no commercial products. Firmware has to be GPL-2.0 because it derives from
QMK, and the GPL forbids adding a NonCommercial restriction, so that half
permits commercial use whether we like it or not.

LICENSE.md records both, plus the two honest caveats: NC means this is not
open source by the OSI or OSHWA definitions, and NonCommercial is a vague
term whose enforcement falls on the copyright holder.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
2026-08-27 15:15:19 -05:00
William RickertandClaude Opus 5 fd1ae511ca Make diagrams self-contained and lead the README with them
The SVGs relied on stylesheet classes supplied by the tool that rendered them
- outside it they were black boxes with invisible text. Each now carries its
own <style> block with light and dark palettes.

Also regenerates the overview, which still showed the old row-staggered 36/39
halves, and adds a licence section: nothing is chosen yet, so default
copyright applies.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
2026-08-27 15:12:17 -05:00
William RickertandClaude Opus 5 5e527bb7b9 Shift the right half's top two rows: 6 and F6 lead the right hand
Resolves the ragged outer edge on both halves at once. Left drops to 34 keys
in six columns each five deep, with no orphans. Right goes to 41: six full
columns, then F12/=/]/Enter, then Bksp and backslash sharing the outer pinky
column instead of backslash sitting alone two columns out.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
2026-08-27 15:09:21 -05:00
William RickertandClaude Opus 5 13c5ca5d5d Column-staggered right half, and record the ragged-edge decision
Six clean finger columns then an outer pinky edge carrying Bksp, ] and Enter,
with backslash alone in an eighth column. Documents why the artifact exists -
the number row and alpha rows are cut half a unit apart - and the three ways
to resolve it. Also flags the right thumb being asked to cover five keys.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
2026-08-27 15:02:32 -05:00
William RickertandClaude Opus 5 c592b9ab33 Column-staggered left half with thumb cluster
Same 36 keys, columns offset for finger length, spacebar broken into four
stepped thumb keys. Comes out ~7u wide instead of 8.5u because the function
row loses the 104's arbitrary gaps.

Also records the tenting dependency: hardware is CAD, but mounting holes are
PCB and need a case scheme chosen before routing.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
2026-08-27 14:58:05 -05:00
William RickertandClaude Opus 5 bf6388a061 Halves are a straight 104 split: 36 left, 39 right
Labelled layouts for both halves, and the overview redrawn from the real
shapes rather than a placeholder ergo core. The 36/39 asymmetry means the two
halves are separate board designs, not one file mirrored - README and BOM
updated to match.

100 of the 104 positions now have a physical key; only Insert, PrtSc, ScrLk
and Pause need a layer.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
2026-08-27 14:52:06 -05:00
William RickertandClaude Opus 5 1e335b7ba2 Add three-board overview diagram
All three boards at the same key spacing. Halves are a placeholder 6x4 plus
three thumbs, shown for scale only - that layout is still the open decision.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
2026-08-27 14:47:36 -05:00
William RickertandClaude Opus 5 f8d22252ad Settle the numpad layout: 25 keys, one encoder
Six by five. Arrows on the outer column, PgUp/PgDn beside them, no Num key.
The scroll wheel is dropped - no good physical implementation - which also
removes the pointing-device path from the firmware entirely, frees a USB
endpoint, and retires the detentless-EC11 sourcing risk. The freed 2u at the
top takes a single volume encoder with room for a full-size knob.

BOM: diodes and sockets 26 -> 25, encoders 2 -> 1.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
2026-08-27 14:43:40 -05:00
William RickertandClaude Opus 5 a9e550e065 Add STM32F103 datasheet for the pin-number overlay
st.com blocks scripted download and its CSP blocks a browser-side fetch from
leaving the page, so this came from a mirror and is Rev 1 (preliminary, 65pp).
Verified it is the right document and that the LQFP48 pinout and every pin the
swap depends on are present. Flagged in the README as unsuitable for electrical
figures - pull the current Rev 20 by hand for those.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
2026-08-27 14:26:26 -05:00
William RickertandClaude Opus 5 2a6dc40f2e Add datasheets for the parts actually in use
CH32L103 datasheet and reference manual, plus the CH334/335 hub datasheet.
Documents/README.md notes which section of each was load-bearing, and flags
the missing STM32F103C8T6 datasheet needed for the pin-number overlay.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
2026-08-27 14:17:47 -05:00
William RickertandClaude Opus 5 4467aa9433 Initial design record for the three-piece keyboard
Two mirrored halves plus a numpad that carries the USB hub. Captures the
architecture decisions and the reasoning behind them: CH32L103 with an
STM32F103 pin-compatible escape hatch, split protocol tunnelled over USB-C
SBU through the numpad, hotswap sockets bottom-side, and why crystal-less
USB was ruled out across the whole WCH catalogue.

No boards ordered, no firmware written. Layout still undesigned.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
2026-08-27 14:07:50 -05:00