Skip to main content

Hardware

Findings from the first articles (v4.9.0, articles 0003 and 0004) that the next fabrication should fix. The full record of what is broken, and of what must not be "fixed", is docs/known-issues.md in the repository.

ItemFoundFix in the respin
Wi-Fi antenna buried (top priority)first article, 2026-09-29: RSSI about −90 dBmantenna over the edge or all-layer keep-out, SD lines rerouted, gate fixed — see Wi-Fi
R37: display FPC connector J4first article, 2026-09-11/12use a top-contact J4 for the panel's FPC tail
R38: PDM carrier hissfirst article, 2026-09-12see Audio
J3 has no "+" markingfirst articleadd the "+" silkscreen next to pad 1 (BAT_IN)
Speaker pads unmarkedarticle 0004: the first "no sound" was the speaker wires on the wrong padsadd SPK+ / SPK− silkscreen

Candidate for the respin: the 32 MB module (not decided)​

The module swap studied in Plan D (ESP32-S3-WROOM-2-N32R16V, same footprint) would fit the same respin as the antenna fix. What the software ran into on the current module, measured on 2026-10-04:

LimitMeasured
Flash for apps64 KB free outside the app partitions, about 18 KB in the launcher's: no new app without removing one
Flash for the arcade game cache4 MB, one game at a time: changing arcade game costs 20 to 28 s of rewriting, against 9 to 13 s for the same game again
PSRAM on the largest CPS1 gamesStreet Fighter II has about 70 KB free in play

One point to check against the module's datasheet before adopting it, not covered in Plan D: with the octal flash at 1.8 V, GPIO 47 and 48 are in the 1.8 V domain, and BTN_X and BTN_B sit there with 10 kΩ pull-ups to +3V3.

Open verification​

ClaimStatusNext step
IP5306 gives a stable VOUT from VIN with no cell attached (CLAIM-005)unverified past its deadline; the verify_claims_ledger gate is redmeasure it on a board with the battery unplugged and record the result in the claims ledger

Proof: each fix is checked at first-article inspection of the respin (/first-article-check), and the claim closes with a measurement.