Stand alone high speed K-Line logging to an sdcard using an arduino. Canbus support is a work in progress. Although no Romraider Logger code was used directly it was an essential guide to get "fast polling" of ecu data. That project can be found at https://github.com/RomRaider/RomRaider

Firmware — alpha8.3

Standalone SSM (K-Line) data logger for the Arduino Uno R4 Minima. Reads RomRaider logger.xml profiles + logger_*.xml definitions from the SD card, builds a complete per-ECU parameter dictionary, and logs to CSV via SSM fast-poll (continuous) batch reads. Auto-generates a profile when the ECU is unknown or no logger.xml is present.

• Source: firmware/sssal/sssal.ino
• Compiled binary: firmware/binaries/sssal_alpha83.bin / .hex
• Build: arduino-cli compile --fqbn arduino:renesas_uno:minima (libs: SdFat 2.3.0, RTClib 2.1.4, Arduino_CAN). USB-MSC requires two one-line core patches — see firmware/BUILD_NOTES.md.
• Flash: arduino-cli upload -p <COMx> --fqbn arduino:renesas_uno:minima (Renesas RA4M1, DFU).

★ Hardware-validated on-car (2026-10-05). First on-car run on an EJ255 (ROM 2F12515506) detected the ECU, resolved 55 params, and logged ~96k clean closed-loop rows. Earlier alphas were bench-validated only via the pre-flight tool; the logger path now has a real on-car run.

USB Mass Storage (SD-as-USB-drive) — new in v5

Plug the logger into a PC and it enumerates as a USB flash drive, exposing the SD card so you can copy logs off and drop config files on without removing the microSD. Mode is chosen at boot: a USB host present → USB-drive mode; powered from the car with no host → normal logger. This needs the two core patches in BUILD_NOTES.md (one of which fixes a real upstream core bug in the full-speed MSC endpoint descriptor).

Changes since Alpha v3

• v4: SSM-over-CAN transport stubbed in with a byte-exact ISO15765 spec (seam only, not yet active); ECU-init buffer 96→112 bytes; default profile tweaks.

• v5: USB Mass Storage (above); RTC set-on-boot via a one-shot /logger/settime.txt (YYYY-MM-DD HH:MM:SS) with a compile-time fallback when the RTC is unset; FAT directory timestamps now track the RTC (SdFat date/time callback) so file-property dates match the filename; K-line transceiver sleep pin (D2) driven awake at boot.

• alpha6: standby-recovery fix. After an engine-off stream-watchdog the logger re-enters Phase 1; the low-power idle path (Software Standby) previously never recovered when the OBD port stays powered (nothing power-cycles the board). Now re-inits the UART + transceiver on every wake before probing, and emits an IDLE_WAKE count=N status line so a log can tell "woke, found nothing" from "never woke". status.log rotates to status.1.log at 2 KB instead of being deleted, so the close/wake history survives a reboot. (Flashed; on-car verification of the standby path pending.)

• alpha7.0: RomRaider-aligned log format. Each log opens with a #-prefixed provenance remark (# Created by SSSAL <ver> | ROM <id> | <datetime> | defs <file>); the time column is now Time (msec) and is zeroed per log (first row = 0) to match RomRaider's relative timestamp; value-column headers are Name (units) [ID]. Data values keep full SSSAL resolution (uniform 2 decimals — not rounded to RomRaider's per-column format).

• alpha8.0: low-power Software Standby wake now works. The parked idle path had never recovered: the board entered Software Standby and stayed there, so with a continuously-powered OBD port (nothing power-cycles the logger) every drive after an engine-off was lost. Two real defects in the AGT1 wake-timer setup, both found by measuring registers on-car rather than by inspection:

  1. Wrong register offsets. AGTMR1/AGTMR2 were addressed at +0x18/+0x19 instead of +0x09/+0x0A (verified against the RA4M1 SVD). The names AGTIOC/AGTISR pointed at the real mode registers, so the "clear the block" step zeroed them and the intended config writes landed outside the peripheral. AGTMR1.TCK stayed 000 = PCLKB — a clock that stops in Software Standby — so the timer froze the moment the board slept.
  2. AGTMR2.LPM was set while the timer was stopped. That prevents the counter from ever starting: AGTCR.TCSTF never sets and the count stays frozen. Renesas' own FSP r_agt.c sets only the CKS divider before start, starts with AGTCR = 0xF1 (the status flags are write-0-to-clear, so 0x01 silently clears them), and waits for TCSTF == 1 before modifying any other AGT register. LPM must be enabled only after that.

  Bench-proven on hardware: a 10 s standby returned rtc_s=10 on the external DS1307 with millis_s=0 (the MCU clock did not advance — genuine Software Standby), one AGT1 underflow wake, then stable operation. Parked draw drops from ~7.5 mA to ~40 µA. STANDBY_ENABLED is a compile-time switch; set it to 0 for the fast-search build that stays awake (~7.5 mA parked) if a standby problem is ever suspected. status.log now rotates at 32 KB instead of 2 KB so the IDLE_WAKE history survives a long park. On-car verified 2026-10-12 (on alpha8.1): one boot, then 70 consecutive 2-minute standby wakes across two parks, with two separate drives detected and logged straight out of standby — no reboot, no lost drive.

• alpha8.1: stop advertising a broken disk. The MSC interface is always in the USB descriptor — __USBInstallMSD() is resolved at link time and the core assigns install_MSD during static init, before setup() runs — so a host always finds a mass-storage device even on a boot that goes to logger mode. TinyUSB only defaults to MEDIUM NOT PRESENT when no sense is already set (if (p_msc->sense_key == 0)), and sense data is sticky until a REQUEST SENSE clears it, so a stale key from any earlier failed command masked it and Windows offered to format the card holding the only copy of the logs. Every not-serving path now sets the sense explicitly (SCSI_SENSE_NOT_READY, ASC 0x3A) and tud_msc_capacity_cb reports 0/0 so READ CAPACITY stalls with the same status — the host sees an empty card-reader slot, which is what the device actually is. Read/write failures while serving now report MEDIUM ERROR with a real ASC instead of a bare -1. Also fixed: a failed enterMscMode() used to return from setup() anyway, leaving the board neither serving the card nor logging, with nothing in status.log to explain it; it now falls through to logger mode.

  Pulling logs: the MSC-vs-logger choice is latched once at boot. Plugging a laptop into an already-powered board does not produce a drive — power-cycle it so it boots with the host attached. With standby enabled a parked board is usually asleep, and USB is deliberately not a standby wake source, so it will not enumerate until its next AGT1 underflow.

• alpha8.2: code-review fixes.

  • A corrupt K-line frame no longer ends the log. Any header/checksum error used to be treated exactly like the stream stopping, so one bad byte closed the log and restarted the ECU search (and status.log said "no data" either way). The reader now matches the 80 F0 10 <N+1> E8 header byte by byte, skips bad or cut-off frames, and resyncs on the next one; the log only closes after 5 s with no valid frame. Skipped frames are reported per log as STREAM_ERRORS <file> bad_frames=N resync_bytes=M (written only when nonzero; an occasional bad_frames=1 at key-off is a frame cut off by the engine stopping).
  • Second drive in one power-up no longer gains extra columns. Hidden dependency params pulled in for calculated params were re-loaded as visible columns on the next engine start — a path that only became reachable once standby recovery worked.
  • Nothing-to-log selections (every selected param unsupported) write one NO_FETCHABLE_PARAMS line per boot and sleep, instead of creating an empty log on every start.
  • BATCHMAX= in address_map.csv is clamped to 48 (the probe's own ceiling); a hand-edited value of 84+ overflowed the request builder.
  • The K-line transceiver is put to sleep for each standby window (woken again before the probe). (On-car verified 2026-10-12: two key-on cycles detected straight out of standby.)

• alpha8.3: bug-sweep fixes.

  • E-params no longer lost for 77 ROMs. The definitions parser read lines into a 256-byte buffer; 53 of the <ecu id="..."> ROM lists in logger_IMP_EN_v370.xml are longer (up to 855 characters), so any ROM ID past the cut was never matched — 559 ROM×parameter pairs, up to 29 E-params for one ROM. The parser now has its own 1 KB buffer and reports any line that still overflows (DEFS_LINE_TRUNCATED). address_map.csv gains a FMT=2 header line, so a cache built by the old parser is rebuilt once (CACHE_REGEN reason=BUILDER_UPDATE, ~19 s).
  • A failed cache build can no longer stick. The dictionary is built into address_map.tmp and only renamed into place if the whole SSM section was read and written; previously a defs read error or power loss mid-build left an empty/partial cache that was reused forever. Failures log CACHE_BUILD_FAIL and retry at the next wake instead of halting.
  • No more 50 MB files after a power cut. SdFat's preAllocate() sets the file size to the full reservation immediately, so a log that never closed was 50 MB of rows plus stale card data. Pre-allocation is removed (at ~0.5 KB/s it bought nothing); an unclosed log now ends at the last 1 s sync.
  • Fatal errors no longer drain the battery, and their LED pattern is actually visible. The halt loop ran awake forever (~15 mA). It also blinked an LED that could not light: D13 is the SD card's SPI clock, and digitalWrite() doesn't take a pin back from a peripheral. Fatal states now flash a visible pattern for 60 s (4 blinks = config error, reason in status.log; 6 blinks = SD card failed), then sleep in Software Standby and repeat the pattern every ~4 min. An empty profile is no longer fatal — it is reported as NO_FETCHABLE_PARAMS and parks.
  • NO_FETCHABLE_PARAMS is now really quiet: later wakes that find the same ECU go back to sleep instead of re-running detection and re-writing its status lines every 2 minutes.
  • Wake-timer re-arm waits for the timer to stop before reconfiguring it, as Renesas' FSP driver does. It worked before only because every re-arm wrote identical values.