[aq1018]

• the pot's slope as a velocity (shapes, rest noise against delay, the 16-row pattern, measured delay) and the carries with the skip rule
• the back-EMF speed on the same rows and the causal crossover: the back-EMF is the source wherever it is valid, the pot's slope catches up only at 50 to 70 ms of delay
• the rest verdict in two equal forms (band of 2 counts, slope under 5 x its own rest sd, held 50 ms)
• one velocity from two sources: the switch and the complementary pair graded side by side, none of record
• the kick's end: the pot against the back-EMF integral, the tag pairs, and a commutation count that says the shaft follows the pot after a hard deceleration (the back-EMF reads low under negative current) and the back-EMF integral through a kick (the pot runs about 2 % short); causes open
• no capture of its own; one servo, one board, one sitting
• Bench gate: none, offline analysis