Input Shaping + Pressure Advance Artifact

evan38109

I've been exploring the new input shaping features in v3.4.0. The input shaping really does work! But when combined with pressure advance, there's a new artifact about 4mm before and after a turn or seam.

Below, I've printed a test piece with no features enabled, input shaping alone, pressure advance alone, and both input shaping and pressure advance combined:

Apologies for my photography skills, or lack thereof
is+pa_artifact.jpg

There's a pre- and post-ringing when both features are enabled that is not present when they are not combined.

This is perhaps more noticeable when printing a pressure advance calibration with input shaping enabled. Below, the pressure advance value increases every 5mm from zero (right side) to 0.14 (left side). Note how the pre- and post-ringing artifact is not present when PA is zero on the right-hand side:

Experimenting with different parameters, I've noticed that the artifact's distance from the turn or seam varies with input shaping frequency, but not IS damping factor, PA value, acceleration, or jerk. I haven't completed exhaustive tests, but I'm happy to perform any suggested.

The printer is a not-quite-stock Railcore II ZLT with a Duet 3 6HC running v3.4.0. Config can be found in GitHub here.

Has anyone else noticed this artifact? Any suggestions for getting rid of it? I like what both input shaping and pressure advance do for the print and I'd love to find a way to use them together.

mikeabuilder

Both Input Shaping and Pressure Advance affect the flow of the extrusion at corners, so they both need to be adjusted at the same time to get a good setting. The IS documentation suggests adjusting adjusting Pressure Advance after IS. In my case, I liked a PA setting of 0.1 without Input Shaping, and 0.08 after I selected IS parameters.

One thing that I wish I did know was what the expected effect of adjusting each of the IS parameters. For example, when damping factor is increased, what is the expected change that should be observed in the ringing tower? Good stuff to add to the M593 page.

evan38109

@mikeabuilder said in Input Shaping + Pressure Advance Artifact:

Both Input Shaping and Pressure Advance affect the flow of the extrusion at corners, so they both need to be adjusted at the same time to get a good setting. The IS documentation suggests adjusting adjusting Pressure Advance after IS. In my case, I liked a PA setting of 0.1 without Input Shaping, and 0.08 after I selected IS parameters.

Indeedy, you can see that happening in the second photo above. I first found good settings for input shaping, and then printed a PA calibration piece. That's a range of PA values with input shaping held constant. The same artifact shows up on the whole set of PA values from way too little (right) to way too much (left). But when PA is 0 (right edge), it's gone. For what it's worth, I've settled on a PA value of 0.09 with input shaping enabled, and 0.07 with it disabled.

For reference, the input shaping configuration I'm currently using is M593 P"ei3" F42.0 S0.1 .

I'll try and get some photos of a print holding PA constant but printing through a range of input shaping frequencies. The ridge is closer to the seam/turn with lower frequencies, and further away with higher frequencies. I've gone from 10Hz to over 100Hz, and it's always there. The damping parameter makes no difference to the artifact, nor does the PA value.

The above were printed with the EI3 shaper. I'll try others soon. I've primarily focused on the EI2 and EI3 shapers since I have different resonant frequencies on the X and Y axes of my CoreXY printer. With 30-something Hz resonance on one axis and 50-something on the other, I can get pretty decent results with a shaping frequency in the low 40's using EI3.

evan38109

Try as I might, I can't get rid of these artifacts with any combination of IS and PA options.

Here's how it shows up in real prints. Note the gaps where it looks like some segments of certain layers just didn't print.

Here's a comparison with input shaping enabled + pressure advance enabled on the left (M593 P"ei2" F42.0 + M572 D0 S0.09); and only PA enabled on the right (M593 P"none" + M572 D0 S0.07). Same filament, gcode, printer, etc. Printed back to back, didn't even restart or re-home the printer. The one with IS disabled is fine.

Is anyone out there using both input shaping and pressure advance at the same time with v3.4.0?

Egon.Net

@evan38109 I'm having the very same problem, you are not alone.

Egon.Net

Any news about this issue?

evan38109

@egon-net I put this one down and figure I'll get back to it. I assume I'm doing something dumb somewhere.

In the meantime, I've got an accelerometer and cabling in the mail. I've been tuning manually by printing calibration towers and such, but I figure I should try the more rigorous approach eventually. I've also been tuning other elements of my printer. If you look closely, even the "good" prints above have some artifacts and could be better.

I'll probably revisit input shaping eventually, and I hope to be in the best possible place for it to succeed when I do.

CCS86

I can't see what you could be doing wrong here.

Printing the same gcode with IS on and off is about as isolated as it can get, and it shows a very noticeable quality issue.

Did you ever look at this @dc42 ?

gnydick

@CCS86 there is another very long thread about this. I didn't realize it's been at least 10 months since the problem started

dc42

@CCS86 yes we know that the current implementation of PA isn't ideal when IS is also used. We thought we could get away with not adjusting PA to account for IS, because Klipper doesn't either. But Klipper has got it wrong too.

I hope to address this before the RRF 3.5 release.

evan38109

Awesome! Thank you!

CCS86

@dc42 said in Input Shaping + Pressure Advance Artifact:

@CCS86 yes we know that the current implementation of PA isn't ideal when IS is also used. We thought we could get away with not adjusting PA to account for IS, because Klipper doesn't either. But Klipper has got it wrong too.

I hope to address this before the RRF 3.5 release.

Glad to hear that!

Any chance that some sort of PA smoothing might be implemented as well?

dc42

@CCS86 the main reason that Klipper uses PA smoothing is that it doesn't control extruder jerk, and without applying PA smoothing it's likely that the extruder would miss steps.