From ac0a75a5462791f5091408e997fb9549b56a501f Mon Sep 17 00:00:00 2001 From: Lars Nolden Date: Tue, 1 Sep 2026 11:21:23 +0200 Subject: [PATCH] docs: dev-board requirements from experience --- docs/DEVBOARD_REQUIREMENTS.md | 35 +++++++++++++++++++++++++++++++++++ 1 file changed, 35 insertions(+) diff --git a/docs/DEVBOARD_REQUIREMENTS.md b/docs/DEVBOARD_REQUIREMENTS.md index 53db5bf..4aed7b5 100644 --- a/docs/DEVBOARD_REQUIREMENTS.md +++ b/docs/DEVBOARD_REQUIREMENTS.md @@ -28,3 +28,38 @@ pin list: devboard_signals.csv. - RF test connector (u.FL tap or directional coupler) for conducted measurements instead of over-the-air-only verification. - ESP32 BOOT/RESET buttons brought to a header for automated strapping. + +## Added after the RF experiment + +Signals and facilities that specifically proved their worth (or whose +absence cost time) during this bring-up: + +- **SX1262 sync-word register readback as a health probe.** The single most + valuable diagnostic added: a radio that lost its configuration still + reports a plausible chip mode while being deaf. On the dev board, keep + SPI accessible so this can also be verified externally with a logic + analyzer. + "I needed this because a silently unconfigured radio is indistinguishable + from a quiet RF environment." +- **A second, independent RF observer that is not a cheap USB dongle.** The + RTL-SDR drifted ~17 dB, died mid-stream without exiting, and eventually + de-enumerated. Most of the debugging time in this experiment went into + distinguishing observer faults from DUT faults. The instrumented dev + board should include either an on-board receive path that can be + cross-checked, or a documented conducted-measurement port. +- **RF test/attenuation port (u.FL tap or directional coupler).** Our + transmission reached the observer up to ~36 dB stronger than real meter + traffic and drove it into compression, which cannot be fixed in software. + A conducted tap with a fixed attenuator would have removed an entire + class of ambiguity. + "I needed this because over-the-air-only verification conflates + transmitter behaviour with receiver overload." +- **Radio power-rail current sense.** Would have answered "is the PA + actually active?" immediately instead of inferring it from decodes. +- **Always-on log UART independent of native USB.** The ESP32-S3 USB + identity changes between the application and the ROM bootloader, and the + early-boot window is invisible; a dedicated UART would have shown the + first `radio_init` result without a reboot-and-capture dance. +- **FEM CSD brought out to a header.** It doubles as the board-revision + detect (pull level) and is the single control that decides whether the + receiver has ~41 dB of front-end gain or not.