Klipper FW + Input Shaper + more tuning.

[lowfat]

Might as well make my own thread to post results.

More info on the resonance testing.

Measuring Resonances - Klipper documentation

www.klipper3d.org

Pic of the accelerometer installed on the carriage. If your bed moves on the Y axis you'd also need to mount it to the bed as well.

First runs. From what I understand the power spectral density is the strength of the vibrations. Tried to google what it actually means and it is way above my head. My results actually seem decent. The PSD is pretty low. But I don't like the 2nd hump in the frequency curve. Maybe belt tension can fix it?

First comparison is too see how much of an impact the fans have on vibrations. You can see another small hump for the X axis.

But the Y axis the impact is huge. The fan adds more vibrations than the printer's motion system.

So I went to tightening the belts a bit more. Tightened about 3 teeth more, which was pretty damned tight. This looks rather bad. Over 4 x the strength of vibrations and a much wider frequency range. This absolutely would have an impact on print quality.

Needed to print a part as my motor mounts are being bent out of shape from belt tension. So more testing after I get that resolved.

 

[danmitch1]

Jesus... so does it print better.. lol

 

[lowfat]

Jesus... so does it print better.. lol

No idea. Hot a hot end jam I can't clear. Have to pull the entire carriage apart. And I sure didn't design it to be easily taken apart. It is a right hassle to get it back together. Too many 'captive' nuts, that like to fall out, causing me to restart the entire process....

 

[lowfat]

Spend all day tuning. One sick part of Klipper is that it has a 'tuning tower' build in. You can enter a command before the print starts so that it automatically increases/decreases a value every so many mm.

So I did one for acceleration. Although it is hard to tell, when the gap starts to widen is where I need to limit acceleration.

And then on to pressure advance. Similar to linear advance, but it seems to do a much better job. And w/ the tuning tower it was really easy. You can see at the bottom of the part the corners buldge, then it eventually they become good, and then too thin. So I can set it to where it is just right.

And lastly I did it for temperature. Starts @ 210C, and then every 1mm it decreases 1C. Didn't make it very far before the corners started to suffer. So I only dropped it to 205C.

Wanted to see how fast I could print a benchy. This took about 20-21 minutes. Obviously not perfect. My extruder can't keep up w/ pressure advance due to my print speeds + bowden. Not sure it is a limitation that can be resolved unless using direct drive.

The speed settings I used for the print.

As for the Input Shaper. I'm not sure it has done anything to be honest. Yet others are saying its the best feature in years. I'm currently trying to do another acceleration test w/ it enabled. Maybe I can increase it beyond 4000 without prints suffering.


EDIT: Reran the acceleration tower but this time w/ Pressure Advance enabled. It was suggested to turn it off before. This time I could see an area around 6500 acceleration where there was less ringing. So I went ahead and printed a benchy with 6500 acceleration including walls. Printed it in under 17 minutes now. 40 minutes faster than my Prusa can print it. Although it ain't as pretty. At that speed I can't cool such a small part fast enough. The bow started to sag. But for larger prints it won't matter at all.

Printed the Voron cube @ 6500 acceleration including walls. Boggles my mind how good this looks for a print so fast.

Surface finish ain't great. But this is easily fixed w/ little impact on speed.

Vid of it in action. It has garbage DRV8825 drivers, so don't expect it to be quiet....

https://imgur.com/XUoolVb

Think I'm going to go ahead and try to install Klipper on my Prusa tomorrow. I'm sold.

 

[danmitch1]

Holy Fkn shit man, look at that thing move!

That 20 min benchy doesnt look that bad.. I mean 20 min... I dont think i ever printed one below 45 min.

So are you more sold on the klipper firmware, or both the accelerometer and the firmware?

 

[lowfat]

Holy Fkn shit man, look at that thing move!

That 20 min benchy doesnt look that bad.. I mean 20 min... I dont think i ever printed one below 45 min.

So are you more sold on the klipper firmware, or both the accelerometer and the firmware?

Well the sensor was like $4 from Aliexpress. It really isn't a lot of work to get working. You just need to make the cable as short as possible. The first cable I made did not work. And you can use the same sensor for every printer you have (if you get more).

I'll be moving to Klipper for all my printers I believe. Pressure Advance tuning was so easy on it, and it seems way better than Linear Advance on Marlin.

 

[danmitch1]

I might jump on the band wagon the next time I need to update my firmware.

 

[lowfat]

Teaching Tech did a vid on Input Shaping today.

EDIT: He didn't show the graphs.

 

[Bartacus]

I completely missed this somehow, but good work Lowfat! I'm shocked that you've done all that with a bowden setup and haven't set up direct drive yet.

 

[lowfat]

I completely missed this somehow, but good work Lowfat! I'm shocked that you've done all that with a bowden setup and haven't set up direct drive yet.

Prusa uses DD. And whenever I get around to building tool changer, all will be direct drive. This printer may actually be converted to DD as well. There is currently a jam in hot end. And I designed the mount to be small and light. To remove the HE is a huge PITA. So I have to redesign it. Might just throw in the Orbiter + Dragon HF. Then for the tool changer buy the new E3D Revo hotends.