[AlkaMotors]

No description provided.

[AP-Review]

Deprecated — see below for the updated review.

Previous review (2026-11-05)

Automated review note — AI-generated (Claude+Codex), validated against the live diff. Please sanity-check before acting.
Verdict: REQUEST CHANGES

Reviewed at head 58a8ce3ee6.
Full report: https://firmware.ardupilot.org/Tools/APReview/DevCallReviews/PRReviews/am32-firmware/am32/455/1.html#prAM32-455

This PR moves sine startup onto the COM timer, but a throttle excursion into the changeover band and back can leave sine mode stalled, which needs fixing before merge.

Blocking

1. Src/main.c:2428: if input goes into 137..200 and then drops back below 137 before phase A reaches 0, this DISABLE_COM_TIMER_INT() turns the sine timer off but leaves sine_timer_active set. The restart at line 2437 is then skipped and the PWM compares stay at 0, so the motor stops under throttle until input goes <=48 or >200. Reproduced in SITL: 8 of 8 dips stall, against 0 of 8 on main. Calling stopSineTimer() in the do_once_sinemode block fixes it.

Low

• Src/main.c:957: the comment says COM_TIMER runs at 2 MHz on all MCUs, but AT32F415 TMR11 uses div=75 (TMR9/TMR10 use 143 for 1 MHz), so it ticks at about 1.89 MHz and sine speed is about 5% slow there. Fix the comment or set the ARR per MCU.

Minor (1)

• Src/main.c:946: timer sine adds about 600-720 B of flash on non-E230 targets (F031 +604 B, now 23232/27424 B). Everything links, but trimming could help on the 32K parts. E230 shrinks by about 3.8 KB.

Not fully reviewed: V203 builds (no RISC-V toolchain) and anything on real hardware, including timer and IRQ behaviour and the E230 register-level conversion.

[AP-Review]

Deprecated — see below for the updated review.

Previous review (2026-11-06)

Automated review note — AI-generated (Claude+Codex), validated against the live diff. Please sanity-check before acting.
Verdict: COMMENT

Reviewed at head 711ba35cb8.
Full report: https://firmware.ardupilot.org/Tools/APReview/DevCallReviews/PRReviews/am32-firmware/am32/455/2.html#prAM32-455

No blocking problems remain at 711ba35; one low-severity comment fix is still worth doing.
Since 58a8ce3 the cancelled-changeover stall is fixed: stopSineTimer() at the start of the do_once_sinemode block lets sine restart after a throttle dip out of the 137..200 band, which checks out in SITL (0/8 dips stall, was 8/8) and in an extracted-code test.

Low

• Src/main.c:957: F415 COM_TIMER is not 2 MHz (still open). TMR11 runs at 144/76, about 1.9 MHz, so sine startup is about 5% slow there. Fix the comment or set ticks/us per MCU.

Minor (1)

• Src/main.c:946: timer sine flash cost (still open). About +600-700 B on non-E230 targets (F031 +604 B), while E230 shrinks by about 3.8 KB; everything links. For information only.

Not fully reviewed: V203 builds (no RISC-V toolchain) and anything on real hardware, including the periodic COM timer's reload and IRQ behaviour and the E230 register-level conversion.

[AP-Review]

Automated review note — AI-generated (Claude+Codex), validated against the live diff. Please sanity-check before acting.
Verdict: COMMENT

Reviewed at head 9f301d1862.
Full report: https://firmware.ardupilot.org/Tools/APReview/DevCallReviews/PRReviews/am32-firmware/am32/455/3.html#prAM32-455

No blocking problems at 9f301d1; two low-severity fixes are worth making.
Since 711ba35 the branch was rebased onto main and gained 9f301d1, which clears demag state on sine entry and at changeover; the earlier cancelled-changeover fix still holds after the rebase (extracted-code harness 718/718, SITL dip test 8/8).

Low

• Src/main.c:973: F415 COM_TIMER is not 2 MHz (still open). TMR11 runs at 144 MHz/76, so the 50 us sine tick is about 53.3 us and sine startup is about 5% slow there. Fix the comment or set ticks/us per MCU.
• Src/main.c:1005: sine_ccr_offset/sine_ccr_scale are only computed in startSineTimer(), so changing SINE_MODE_POWER over DroneCAN during sine startup has no effect until sine restarts (the old advanceincrement() used the live value). Recompute them when the parameter changes, or document that it needs a restart.

Minor (1)

• Src/main.c:962: timer sine flash cost (still open). About +600-880 B on non-E230 targets (F031 +648 B, V203 +880 B), while E230 shrinks by about 3.8 KB; every image links. For information only.

Not fully reviewed: anything on real hardware, including the periodic COM timer's reload and IRQ behaviour on each MCU and the E230 register-level conversion, plus live DroneCAN parameter changes.