So, backstory: I setup a tuning print for Klipper input shaper that makes use of matched speed projected onto the opposite axis at junctions, and high instantaneous junction velocity (scv), to allow them to be taken in a way that only one axis experiences an impulse. The idea here was to get measurement noise from the other axis's resonances out of the measured ringing. And that seems to have worked fine. But I also found something very interesting:

The magnitude of ringing, and even whether it appears at all, is dependent on the direction of the acceleration/deceleration impulse. When the motion of the axis is away from the motor and abruptly stops, there is essentially no measurable ringing. But when the motion of the axis is towards the motor and abruptly stops, there is strong ringing.

Is this normal? It seems plausible via one or more mechanisms:

  • Belt run all the way around the idler and back to the motor, which is doing the pulling in this direction, is far longer and has more elasticity
  • Idler is untoothed, so maybe the teeth compress against the smooth idler and act as a spring
  • ... ?

The printer is an Ender 3 original with lots of modifications, but none of them to the motion system. The observations above apply to both the X and Y axis, but the X axis is so light that it requires extreme levels of acceleration to see the issue anyway, so I'm not concerned about it. The Y axis is where the problem is.

Seeing how good I have this in one direction (towards the motor) makes me really want to find a solution that gets equally good results in both directions, as it would allow me to up my outer wall print acceleration for high quality up from 500-2000 to 5000+ and max acceleration up to 30000+ - the limiting factor is the bad ringing from one direction on this axis.

What approaches should I try to mitigate this (and confirm it's the issue)? I'm thinking of:

  • Going dual-Y-motor with the second Y motor mounted at the front of the printer, so both directions are "motor pulling on short segment of belt".
  • Replacing belt. It's the original Creality supplied one and some folks have suggested a genuine Gates belt would not have this issue.

Here is a photo of the test print:

3D print of a Z-direction-extruded rectangle with a trapezoidal segment subtracted from the middle of the long side

The extrusion path was clockwise around the part (right to left on the side facing the camera and front of the printer/bed). Acceleration, scv, and input shaper settings were adjusted manually during the print to give different views of the ringing. The important detail is that there's almost no ringing in the middle flat section, except at the very top where input shaper was turned off completely.

  • $\begingroup$ how thick is the part? Have you tried the latest prusa slicer with Arachnae enabled? $\endgroup$
    – Trish
    Jul 30, 2022 at 16:02
  • $\begingroup$ @Trish: It's a single-wall test piece. Latest Cura with Arachne, but Arachne vs legacy isn't relevant because the wall path is exactly the inset outer profile I designed the model to yield. $\endgroup$ Jul 30, 2022 at 16:07

1 Answer 1


No, at least not to the extent the above result suggests. I had the idea to try rotating the piece 180° around the Z axis, and got nearly the same effect: low ringing in the middle segment, strong ringing on the final segment. So what is the explanation?

The impulse intended to start the ringing for the middle segment is exactly opposite the impulse that ended the first segment, and generates a time-shifted ringing of opposite sign, cancelling out with the tail of the ringing from the previous impulse.

Overall, the ringing does seem to be somewhat stronger in this rotation, so I think my original hypothesis that there's an effect related to belt length or the idler is still plausible, but it's nowhere near as dominant as I thought.

Further investigation of this seems to call for a better-designed test piece without cancelling impulse responses.

  • 1
    $\begingroup$ I'm actually starting to doubt this explanation somewhat because there should be cancelling impulses on the last segment too. But hopefully a new test model will shed some light on this. $\endgroup$ Jul 27, 2022 at 23:51
  • $\begingroup$ Print speed seems like an important variable that I don't see mentioned. Consider running the test with the print speed increased and/or decreased by a small amount, maybe 5 or 10%. $\endgroup$
    – Caleb
    Aug 10, 2022 at 14:40

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .