LuxDMX RDM Simulator & Analyzer (RP2350)

A cycle-accurate, multi-fixture RDM (E1.20) simulator and, longer term, a standalone DMX/RDM analyzer, built on the Pimoroni Pico Plus 2 W (RP2350B). The goal is a bench tool that's better than anything on the market: put a whole rig's worth of virtual RDM fixtures on one RS485 bus, behave (and misbehave) like real hardware, and measure exactly how a controller — specifically LuxDMX's ESP32-S3 — copes.

Prototyping now, built as a real product. Lives in this dmx-analyzer/ subfolder so it can be pulled out into its own repo later.

Why RP2350

PIO gives exact 250 kBaud break / MAB / turnaround (and can mangle it on demand), which an MCU UART can't; dual Cortex-M33 + 8 MB PSRAM hold hundreds of fixtures + an on-wire capture; the RM2 radio drives the web UI and the eventual RDM-over-network work.

Hardware / wiring

• Pimoroni Pico Plus 2 W (RP2350B, 8 MB PSRAM, WiFi/BT, 16 MB flash, USB-C).
• MAX3485 (3.3 V) half-duplex RS485 transceiver:
  • GP0 → RO (RX) GP1 → DI (TX) GP2 → DE+RE (low = receive)
  • A/B on the bus with the DUT, common DMX ground, 120 Ω termination.
• The transceiver's RO/DI can be wired either way — auto-detect is on the roadmap (see below).

Web UI (WiFi)

The sim serves a self-contained dashboard — no internet needed, styling is embedded.

• First run / no WiFi set: it brings up its own access point — SSID LuxDMX-RDM-Sim, password luxrdm-sim — and serves the dashboard at http://192.168.42.1/. Open the Settings tab, enter your network under WiFi, and Save; it reboots and joins your LAN (its IP is printed on the serial [boot] line). Creds are stored in flash (EEPROM); "Forget" clears them and drops back to the AP. If it can't join, it falls back to the AP so the page stays reachable. (You can also pre-seed a default network by copying src/wifi_secrets.example.h to src/wifi_secrets.h, which is gitignored.)
• Tabs: Dashboard (live metrics + RDM activity feed + bus-health bars), Live Bus (live 512-channel DMX view as bars/grid with per-channel hover, plus a decoded, colour-coded, filterable DMX+RDM packet log you can record and download), Fixtures (table + bulk generate), RDM Monitor (discovery-tree log + found devices), Analyzer (DMX framing-error counter + refresh/slots/break, Swisson-style, plus RDM integrity + E1.11/E1.20 reference), Stress/Fuzz (fault injection), Settings (turnaround, count, WiFi).
• JSON API: GET /api/status|/api/metrics|/api/fixtures|/api/log|/api/dmx|/api/capture|/api/wifi, POST /api/config|/api/unmute|/api/fuzz|/api/analyzer/reset|/api/wifi|/api/wifi/forget.
  • /api/dmx = the last complete frame (start code, channel count, refresh, framing, hex of the values).
  • /api/capture?since=<seq> = decoded packet log since a cursor (DMX markers + every RDM request/response), newest-only so nothing repeats.
  • /api/wifi GET = current mode/SSID/IP; POST {ssid,pass} saves + reboots; /api/wifi/forget clears creds.
• WiFi runs on core 0; the RDM engine runs on core 1 with its own PIO state machines (claimed before the radio inits), so the web UI never disturbs RDM timing — verified on hardware.

Console (USB serial, 115200, one command per line)

t<us>   responder turnaround delay (E1.20 window 176..2000 us)
f<n>    active fixture count (1..200), regenerates the virtual bus
r       un-mute all fixtures      v  toggle per-DUB depth trace     ?  status
m       metrics (reAsk, resp, rdmReq, badCsum, rxStall, muted)
a       DMX analyzer stats (framing errors / refresh / slots / break);  ar  resets the window

The sim auto-reports each discovery sweep: [disc#N] discovered X/Y (…missed) responses=… S3-dropped-replies=… turnaround=….

Status

• M0 bring-up (USB, activity LED) — done
• M1.1 PIO DMX receive (channel-accurate) — done
• M1.2 decode RDM requests (custom PIO RX + message-length framing) — done, verified on wire
• M1.3 reply within the turnaround: DISC_UNIQUE_BRANCH response + DISC_MUTE ACK — done
• M2 many fixtures + real binary-search discovery — working (controller walks a deep tree and discovers virtual fixtures). Crowded-bus reliability: root-caused to ~30% RDM request corruption at the half-duplex turnarounds (the S3 catches our responses fine). See below.
• M3 GET/SET PID set (DEVICE_INFO, SOFTWARE_VERSION_LABEL, DEVICE_MODEL_DESCRIPTION, MANUFACTURER_LABEL, SUPPORTED_PARAMETERS, DMX_START_ADDRESS, IDENTIFY_DEVICE; NACK others) — done
• M4 fault injection (drop % of replies, late turnaround, corrupt checksum, flaky mute) — core done, wired to the web Stress tab; more quirks (runt breaks, baud drift) on the list
• M6 WiFi AP + self-contained web UI + JSON API — done, RDM-timing-safe (verified on hardware)
• DMX framing-error analyzer (issue #64) — the sim now counts DMX framing errors on the wire like a Swisson tester (dedicated pio1 SM samples 8 data + stop bit; break/frame from the inter-slot gap). Built to test the "core-separation" fix for #64 — firmware + rig done, A/B measurement is the next bench step. See docs/ISSUE_64_CORE_SEPARATION.md.
• Reliability investigation — docs/RDM_S3_RX_RELIABILITY.md. Headline: no S3 RDM-RX bug (the S3 catches its responses, reAsk ≈ 0). ~30% of requests arrive corrupt at the sim = ~9% rig physical layer (bias present, controller reset both ruled out) + ~21% the sim's own half-duplex turnaround. To close: logic-analyzer capture (S3 serial = COM5) + a real PCB front end for the sim. Do NOT loosen the checksum-validating decoder to hide it.

Roadmap

• Close the reliability finding: logic-analyzer capture of a turnaround; add fail-safe bias and re-run the crowded-bus sweep; also re-check the physical RS485 wiring.
• M5 bus-load model + DMX latency/refresh/flicker meter (how much does RDM traffic hurt DMX?).
• M4+ more fuzz quirks: runt/short breaks, baud drift, preamble-length variation, NACK storms.
• M7 RDM over Art-Net (ArtRdm) / sACN (E1.31 level + E1.33).
• Analyzer standalone DMX/RDM analyzer product (same PIO capture + PSRAM ring buffer core).
• Web UI next: per-fixture inline edit, sensor sim, timing-scope (µs break/MAB/turnaround view), record/replay of a controller's transaction stream, CSV/PCAP export. (The Live Bus tab now covers the live DMX channel monitor + a decoded DMX/RDM packet log with text download.)

Feature vision (why it beats what's on the market)

Bench RDM testers today are either expensive closed boxes or a USB dongle + PC app. This aims to be:

• A crowded bus in one box — dozens–hundreds of fixtures with independent UID / address / footprint / personality / sensors / mute state, from a single node, no rack of real fixtures.
• Deliberately hostile — every out-of-spec timing and malformed-packet trick real fixtures pull, toggleable, so a controller gets stress-tested, not just happy-path.
• Measuring, not just talking — direct metrics (dropped replies, turnaround histogram, discovery tree depth, DMX jitter under RDM load) instead of "seems to work".
• Standalone + networked — runs headless on the bench or driven from a web UI; bridges to Art-Net/sACN so you can test network→wire RDM.
• Open — reproducible, scriptable, and hackable, unlike closed testers.

Backlog of feature ideas (grow this list continuously)

• Scenario/record-replay: capture a real controller's transaction stream and replay/mutate it.
• Per-fixture fault injection UI (drop N% of responses, add jitter, wrong checksum, NACK storms).
• Sub-devices and proxies (DISC binding UID, managed proxies) for deep discovery tests.
• Sensor simulation with live-changing values + status messages / queued messages.
• Full E1.37-1/-2 PID coverage (dimmer curves, lock states, power, IP settings).
• Discovery collision realism: true overlapping collision waveforms (not just single-responder).
• Timing scope view: break/MAB/turnaround measured to the microsecond, pass/fail vs E1.20 limits.
• DMX side: framerate/flicker analysis, per-slot value graphs, break-length + MBB measurement.
• Export: CSV/JSON transaction logs, PCAP-like capture, shareable test reports.
• Conformance suite: automated E1.20 controller test checklist with pass/fail + evidence.
• Multi-node: several of these boxes on one bus for very large / multi-manufacturer rigs.
• USB or network control API so it slots into CI for the LuxDMX firmware.

Build & flash

pio run -e pico_plus_2w -t upload      # auto-resets into BOOTSEL

Serial console 115200 (assert DTR — the RP2350 USB CDC gates output on it).