New Input Shaping plugin v3.4.1-b1

chrishamm

I'm happy to announce the first beta release of our new Input Shaping plugin!

This DWC plugin is essentially the successor of the old Accelerometer plugin and it lets you predict input shaper effects on-the-fly. It also allows you to record and evaluate motion profiles consisting of multiple accelerometer runs in order to find the best input shaper for your machine.

More detailed instructions about it can be found in our new docs and the plugin itself can be obtained from here (source code is available here). Note that software version 3.4.1 is required to use it.

Please bear in mind that this version is still a beta and some things will change again. For the next versions we have the following roadmap:

Comparison of two different motion profiles
Auto-tune functionality
Display of shaper durations and possibly better highlighting of the effective vibration reduction
~~Fine-tuning of custom input shapers (will depend on the next software version)~~

oozeBot

@chrishamm I'm digging it! Just put it through the paces and it's really coming together. Thanks!

oozeBot

One thing I just realized is that the the moves are based on max speed set in M203 - which in our case is 400mm/s - way above normal printing speeds. It might be beneficial to allow setting the speed of the moves as a parameter as that definitely plays into the readings..

dc42

@oozebot ringing after travel moves also causes print artefacts.

oozeBot

@dc42 Understood, but setting the travel speed seems appropriate as travel speed is set in the slicer. And in our case, is much slower than the max speed we have set in M203.

mikeabuilder

Big KUDOS on the documentation, reflecting a lot of work to build out a tool that should be extremely useful.

Argo

Nice! Thanks a lot for the effort

oliof

@oozebot this is a bit offtopic, but either M203 should reflect the machine limits you expect and be set lower, or it should be ok to run the accelerator test at that speed because it may be reached with an appropriate setting in the slicer -- or am I missing something?

Alex.cr

Well done Duet team! This is awesome!

oozeBot

@oliof we were lowering the values in start.g and then raising them after the print completed.. but I see your point.

Fred-Y

I just went through the documentation and started to play a little with it on my Delta - It is a really good work.
I'm finding it pretty user-friendly and the documentation is really helpful.

One suggestion I would like to make: would it be possible to add a button to save the selected IS config into the config-override.g file or maybe update the M500 command.

samlogan87

Hey Everyone,

EDITED

This looks awesome. Looking at the manual online however, the recommended value as per the example confuses me a little bit. It recommends the VZD shaper type. Why would it not be the VZDD or VZDDD which has a much higher frequency range that it covers (Horizontally on the graph)? Also VZD does not go to zero on the amplitude where as the other two do Or are you wanting to make the bell of the curve as tight as possible. It doesn't really explain it.

Sam

Phaedrux

@oozebot said in New Input Shaping plugin v3.4.0-b1:

@oliof we were lowering the values in start.g and then raising them after the print completed.. but I see your point.

It's safer to set M203 to the max speed you would comfortably use for travel, that way if you use the speed factor adjustment to increase print speed it won't push your travel speeds into uncomfortable speed territory.

dc42

@samlogan87 I agree, ZVDD is typically more appropriate than ZVD. Both will go to zero if you set the damping factor to zero.

samlogan87

@dc42 ok cool. I think I see the documentation has been adjusted maybe down the bottom of the page

Argo

El1 and El2 shaper always perform differently than expected.
Is there a reason for that?

Example:
I have two ringing frequencies (one for each axis). 60Hz x axis and 40Hz Y axis. My optimal shaper frequency then is 50Hz. ZVD does a good job, ZVDD is even better and with ZVDDD I get the best results.
But with El1 and El2 the performance is worse. I wonder if those shaper work differently and can't be calculated with this method?

dc42

@argo the EI shapers accept up to 5% residual ringing in return for covering a wider bandwidth. Perhaps that us what you are seeing?

Also try setting the damping factor to zero if you haven't already.

nuroo

Hey guys,

Great work here 1st off.
I'm having a strange issue. Last night I had over 10 successful tuning runs with this plugin. I was on my way to finding the best input shaper for my setup. Around the 12th or 13th run the plugin started crashing after the 1st X-axis move. When it crashes the dwc becomes frozen and an automatic restart is triggered, similar to if I had pressed the stop button. I have uninstalled the plugin and reinstalled it, still crashes. I've tried doing a baseline run, no input shaper - but these also now crash.

In case it's my issue I will do a wire continuity test from sensor thru wiring to connector for duet board, maybe the repeated movement caused the issue. I have also tried tuning only a single axis.

In my config.g
M955 P0 C"spi.cs6+spi.cs5" I41 Q2000000
To troubleshoot I tried:
M955 P0 C"spi.cs6+spi.cs5" I41 Q1000000 <---still crashes

Railcore 2 zl 300
Input Shaping 3.4.0-b1
Duet web control 3.4.0
Board: Duet 2 WiFi (2WiFi)
Firmware: RepRapFirmware for Duet 2 WiFi/Ethernet 3.4.0 (2022-03-15)
Duet WiFi Server Version: 1.26

alt text

zapta

Documentation looks very nice. I should reconnect my accelerometer and give it a try.

Is the damping factor also something that need to be configured with M593?

dc42

@nuroo please post a M122 report taken after the automatic restart.