feat(settings): raise the configurable motor pole-count ceiling to 128

[AlexKlimaj]

What

Raises the maximum motor pole count a user can configure from 64 to 128 (64 pole pairs), and gives the range names — MOTOR_POLES_MIN / MOTOR_POLES_MAX in Inc/eeprom.h — instead of the open-coded 2..64 that was repeated in three places.

Note on the 36: there is no 36-pole limit anywhere in this firmware. That cap comes from the configurators; the firmware-side ceiling was 64, set by the DroneCAN MOTOR_POLES parameter and mirrored by the two kv-derived schedules in settings.c. Both still excluded large direct-drive / hub motors, so this raises the firmware ceiling well clear of 36 and of 64. 128 stays below the 0xff an erased eeprom reads, so out-of-range detection (0 or 0xff → schedules fall back) is unchanged.

Changes

• Inc/eeprom.h — MOTOR_POLES_MIN (2) / MOTOR_POLES_MAX (128), documented against eeprom byte 27.
• Src/DroneCAN/DroneCAN.c — MOTOR_POLES parameter uses the macros, so the DroneCAN GUI now offers the full range.
• Src/settings.c — both gates use the macros. Above 64 poles the two schedules now compute instead of falling back to their "implausible pole count" behavior:
  • throttle-restriction envelope: worst case 10220 kv * 128 poles / 544 = 2405 kerpm, well inside the uint16 the levels live in.
  • advance normalization: kv * poles * 4096 / 200000 overflows uint32 above ~102 poles, so the reduced 256 / 12500 form is evaluated instead. Same rational number (both terms divided by 16) and it truncates to the same integer for every accepted kv/pole pair, but the intermediate stays in uint32 across the whole byte range (10220 * 255 * 256 = 6.7e8 against 4.29e9). Worst-case scale becomes 53372, leaving the 15/16 scale-down, the uint16 it lands in, and the hot path's (scale * centivolts) >> 12 on a 12S pack all in range.
• Mcu/SITL/sitl_params.py, Mcu/SITL/sitl_gui.py, scripts/esc_capture_gui.py — pole-count ranges that mirrored the old 64 raised to 128.

No behavior change at or below 64 poles: the gates accept exactly what they did before, and the rewritten scale expression is bit-identical over the accepted kv range (checked exhaustively over poles 2..255 against the old form).

Test

• make AM32_SITL_CAN builds clean; make check_format reports only the four pre-existing offenders under Mcu/ (untouched here).
• SITL end-to-end (AM32_SITL_CAN, heavy_13inch model at 360 kv, DShot300, throttle 900): 46-pole and 120-pole configs both arm, start, and hold closed loop — 3545 rpm and 3427 rpm respectively — so the newly enabled envelope and advance normalization do not disturb startup or run.
• The DroneCAN parameter test (Mcu/SITL/tests/test_params.py) was not run: pytest/dronecan are not installed in this container.

Not changed: nothing clamps motor_poles itself, so an erased-eeprom 0 would still divide by zero in the sine-startup step delay and the rpm speed-control path. Pre-existing and out of scope here, but worth a follow-up.

---

Generated by Claude Code