Continuous rotation on a manual_stepper: proper timing issues

Developers
1

Hi all, here with a request for assistance.

I’m trying to get a manual_stepper rotating continuously and smoothly.

I managed to get it accelerating, cruising indefinitely, and braking on request. :partying_face: HOWEVER the joints between movement segments are “bumpy” (as can hopefully be seen in the plot, and the very interesting subset of data below).

image

   start        end           I       c      a duration i_final i_next
11 113441695023 113441760705  65682   1      0    65682   65682 112292
12 113441760705 113441872997 112292   1      0   112292  112292  64000
13 113441872997 113465360997  64000 367      0 23488000   64000  64000
14 113465360997 113465424997  64000   1      0    64000   64000  79750
15 113465424997 113465568997  79750   2 -15500   144000   64250  64000
16 113465568997 113490976997  64000 397      0 25408000   64000  64000
17 113490976997 113491040997  64000   1      0    64000   64000  95750
18 113491040997 113491200997  95750   2 -31500   160000   64250  64000
19 113491200997 113514624997  64000 366      0 23424000   64000  64000
20 113514624997 113514688997  64000   1      0    64000   64000  79879

Those sharp peaks correspond to unwanted single/double steps, with irregular intervals, that are inserted between adjacent segments of the move. Those produce audible “bumps” regularly during rotation, and will stall the stepper if rotating fast enough.

I’m passing apparently sensible print_time s to trapq_append, but I’ve no clue how the downstream C code might be implicated.

I would really appreciate some pointers on how to appropriately time the moves with trapq_append , generate_steps , and trapq_finalize_moves to get smooth motion.

Just in case, the relevant “extras” module is here: klippy/extras/manual_spinner.py · pipetting · pipettin-bot / Forks / Klipper · GitLab

Best!
N.

2

Just in case you are still working on this, if you are using a TMC stepper driver you can actually just get continuous rotation with VACTUAL using SET_TMC_FIELD. Not sure about the acceleration though, I assume it’s done internally by the driver.