2026-10-25: setpoint loops, the e-stop release and the watchdog lease

Status: complete. Every step passed. Expected values come from the host tests and the PRs.

Verifies: on main at 718a41f, #84 (SET_SETPOINT names its loop, #57) and #85 (the commands-only watchdog lease), with bread-crumbs-contracts 0.11.0 (#82). Also the position-mode half of #30 (e-stop release), which needed a board with an encoder. By: @CameronBrooks11

Setup

Firmware	718a41f, env gen2_nanoevery. bread-slice-core 0.1.0, DCMotorServo 1.2.0, LMD18200 0.1.5, CRUMBS 0.16.2, bread-crumbs-contracts 0.11.0
Board	gen2 Nano Every (ATmega4809), with the D10/D11 motor-2 rework
Controller	Raspberry Pi 4 Model B, I2C at 50 kHz, address 21, CRUMBS Python binding 0.16.2 and contracts 0.11.0, via dcmt_bench.py (this record’s copy adds sp5, wdl, raw and watch)
Power	backplane (I2C and 12 V), one DCMT on the bus
Outputs	an encoder motor on M1; M2 empty. E-stop: a jumper from D2 to GND

Power-cycle before S2 and S4, so the encoder starts at 0: GET_STATE positions saturate at ±32767 (2026-10-24 notes).

Steps and Results

Step	Expected	Observed	Result
S1 baseline: version, caps, state, wdog	module 2.0.0; caps level 3, flags 0x7FF (0x1FF plus DCMT_CAP_SETPOINT_LOOP and DCMT_CAP_WATCHDOG_LEASE); mode 0, PWMs 0, estop 0; watchdog disarmed	CRUMBS 1602, module 2.0.0; caps 0103ff070000: level 3, flags 0x7FF; mode 0, PWMs 0, pos=0,0, estop 0, count 0; watchdog disarmed	pass
S2 a position target is not a speed request (#57): after a power-cycle, brake 1 1, mode 0, sp 5000 0, mode 2, brake 0 0, watch 3; then brake 1 1, mode 0	sp acked status 1; in speed mode GET_STATE sp=0,0 and M1 stays still (pos within a few counts). Before #84, 5000 became the speed request (host-tested)	all acks status 1; in speed mode sp=0,0, spd=0, m1_pwm=0, and pos=0 before and after the 3 s with the brakes released	pass
S3 the speed form: sp5 2 60 0, mode 2, brake 0 0, watch 5, brake 1 1, mode 0; then sp5 3 1 1 and raw 0x04 010203040506	sp5 2 acked status 1; in speed mode sp=60,0 and spd near 60 RPM (51 to 68 on 2026-10-24). sp5 3 acked status 2 reason 4 (UNDEFINED_VALUE); the 6-byte frame status 2 reason 1 (MALFORMED)	sp5 2 60 0 acked status 1; in speed mode sp=60,0; pos 495 to 2799 in 4.0 s, about 43 RPM at 800 counts/rev (spd read 30, 45, 60 in 15 RPM steps), m1_pwm 30 to 37. sp5 3 acked status 2 reason 4; the 6-byte frame status 2 reason 1	pass
S4 e-stop release in position mode (#30): after a power-cycle, mode 1, sp 2000 0, brake 0 0, watch 4; sp 8000 0 and jumper D2 while it moves; remove the jumper, watch 2; turn the shaft by hand about a quarter turn, brake 0 0, watch 2; jumper and release again, sp 1000 0, brake 0 0, watch 4; brake 1 1	M1 drives to about 2000 and holds. On the press it brakes (brakes=0b11, estop 1); after the release it does not drive toward 8000. After turning it by hand, the brake release holds it where it is: sp follows the encoder until then. After the second hold, the brake release drives it to 1000	not power-cycled first (pos 2997, inside ±32767). M1 drove to 2007 and stopped. The first press landed after it had settled: brakes 0b11, estop 1, sp followed to 2007, and sp 8000 0 during the hold was refused reason 3. After the release, the brake release and sp 8000 0 started the move; the second press landed at pos 3136: brakes 0b11, m1_pwm=0, sp followed the coast to 3252 (count 2 to 3: the jumper bounced). Turned by hand to 2977 with sp following; the brake release held at 2977 for 2 s. After a third press and release, sp 1000 0 was applied with the brakes on (sp=1000, pos 2977) and the release drove M1 to 1004 in about 1 s	pass
S5 the lease: brake 0 0, mode 0, open 60 0, wdl 1000 1, watch 3; wd 0, clear (disarm first: under the lease clear refreshes nothing, so an armed watchdog would trip again at once); brake 0 0, open 60 0, wd 1000, watch 3; wd 0, brake 1 1; wdl 1000 2, wdog	with the lease, GET_STATE polls alone let it trip about 1 s after wdl: brakes=0b11, m1_pwm=0, GET_WATCHDOG tripped 1. With the 2-byte wd 1000 the same polls keep it alive: no trip in 3 s. wdl 1000 2 acked status 2 reason 4, and wdog reads disarmed	wdl 1000 1 applied at 22:31:18.881; the poll at 22:31:20.083 read brakes=0b11, m1_pwm=0, and GET_WATCHDOG 01 e803 01 01. wd 0 then clear: 00 0000 00 01. With wd 1000 the motor ran through the 3 s of polls, GET_WATCHDOG 01 e803 00 01. wdl 1000 2 acked status 2 reason 4, and wdog read 00 0000 00 01	pass

Raw Output

The commands behind each step and their output: raw-session.txt.

Not Checked

• Other boards: the gen2 Nano and gen1 boards were not run. #84 and #85 change the handlers only, which build the same for each MCU apart from the speed loop, and the speed loop is on the Nano Every.
• Motor 2: no motor fitted.
• How soon the lease trips: the polls are 200 ms apart and print only on a change, so the trip is bracketed between 0.3 s and 1.2 s after wdl.

Notes

• S1’s first queries timed out. They landed during the boot wait; the next run, a few seconds later, answered.
• Closed-loop speed runs below the target. 60 RPM requested, about 43 RPM measured from position, as on 2026-10-24 (42 to 56). #84 changes which setpoint is used, not the speed loop.
• The e-stop count went up twice on one press (2 to 3 within 0.2 s) when the jumper was placed by hand: the line bounced, and each LOW after a release longer than the 25 ms debounce counts. The hold behaved the same.
• S4 was not preceded by a power-cycle. Its targets (2000, 8000, 1000) stayed inside ±32767 from the starting position 2997.