Input Shaper issues? weirdness

[SuperCow]

Basic Information:

Printer Model: flsun Super Racer delta
MCU / Printerboard: robin nano v3.1

A few days ago I klipperized my SR, added a LGX lite and 5015 fans.
Using this mod: thing:5138588
Now i was at the adxl345 sensor, and i cant wrap my head around it.
Ive done many things now and i cant see any trees anymore in the forest.

Important settings?
Steppers:
microsteps: 256
run_current: 1.138
stealthchop_threshold: 0.01
interpolate: false

The linear rails are good greased up.

My X and Y looked like this when i started:
pic 1 and 2

Tweaking the belts etc did help some (factory method, and frequencies method using a app at 60hz)

Then i found out the belt method trying to isolate the 3 towers.
“TEST_RESONANCES AXIS=0,1 OUTPUT=raw_data
TEST_RESONANCES AXIS=-0.866025404,-0.5 OUTPUT=raw_data
TEST_RESONANCES AXIS=0.866025404,-0.5 OUTPUT=raw_data”

First time:
pic 3

doing belt tweaks:
pic 4

But the next problem was… it wasn’t consistent at all, if i do back to back runs, always somethings it was completely different.
I was thinking maybe the 5015 fans are flapping around, so put a rubber band around it, it helped a but next turn it showed something else again.

Maybe something is loose? Ive tighten every screw there was in the machine, and some i could turn 1/2 to 1 turn, and again, it looked a little better again.

Then i removed the fans from the head as a test, it was worse again argh!
got a tip to lower the microsteps to 64, that help a little but then i ended here:
pic 5.

Messing around again, the next tip i got to enable stealthchopper = 999999.
Well this did alot! the power spectral when a factor 4-5 down.

pic 1

Changed the current of the stepper to 0.8 microstepping to 16 and commented #interpolate: false out.
Also changed my stiffer rubber feet to the stock ones, and more improvement.
and running the tests back to back yielded the same results! yeah i can work with that,

pic 2

And with everything the same but stealthchopper off again:
pic 3

the power spectral density 4-5 times higher again.

But i though lets do the X-Y input shaper i must see improvements now yes?
pic 4 and 5

Nope! worse

I suspect mechanical issue, or i am clueless and don’t know what i do < very true btw.

Sorry for the many pictures, i don’t know how to explain it.
Can somebody see what i’m doing wrong or maybe something wrong with the printer

[SuperCow]

I wanted to add a update from all the mess in my first post.
Ive gotten the graphs alot cleaner now, but still many issues.
Maybe somebody can shed some light on these graphs?

X has 2 peaks?
and Y has a hill and peak? hehe

The blue is the back pillar (90 degree)
Orange is left pillar (210 degree)
Green is Right pillar (330 degree)

My guess is the orange en green should be the same, but green(right) looks dampening?

What have i done? tightened all screws, relubed the linear rails and balls from the effector and rods. swapped the arms around.
Where should i look next?

Thanks in advanced!

[Sineos]

Hard to judge for a delta since I have no experience with it.
On a Cartesian or CoreXY graphs like this would strongly point to a binding axis.

[SuperCow]

You are right, I took the linear rails out, and 1 (the green line) is binding not a bit…it doesn’t want to move only with force.
Here is a video about it:

update linear rail super racer

Just finished the 3e rail, its even more bad lol, 2 out of 3 bad rails, nice flsun.

[Sineos]

Thanks for the confirmation. I would love to see a comparison when you have fixed the axis.

[SuperCow]

I wanted to update my topic again, to see where i am now.
And what the thought are from the forum members.

Shaper:

Its looking much cleaner now (still with the stock linear rails)
I also found a error in the printed housing to add the LGX lite extruder, i’ve used a other design now.

Im getting very exiting again, it looks much cleaner, but i do wonder what is the low hz noise in the graphs?

Al screws are tight, and all the wires are zip-tied, and taped there is no slack.

[Sineos]

Looks a lot better now. Also see the “Power Spectral Density” (left y axis), which increased by a factor of 10. This is an indication that your axes are running freely now.

[SuperCow]

Yes, its much better, do you have a idea what the low HZ noise is? it there something i can do about it, or are the rails to blame for this?
No shaper type is able to get rid of it, and i can see little waves like the see on prints

[Sineos]

Low frequency resonances are the most detrimental in terms of ringing.

Unfortunately from the graphs you can only tell that they are existing but not from where they come. I have no experience with deltas, so I cannot really tell what’s normal and what’s not.

Generally speaking, the most important thing is your mechanics. If there is any binding, hard going, out-of-angle etc then you will have issues (not only with the input shaper).

[Blackstump]

There is not much to the mechanics of a delta printer, I have had one in one form or another for about 8 years starting with a rostock, tho haven’t used the current one for a year or so voron is now pride of place.
The places I would look is belt tensions, diagonal rods then possibly tower resonance.
Diagonal rod pivots points might also be a point of interest.

[SuperCow]

A new update again, thanks for the suggestion, ive still learning.
I always have suspected my linear rails as a problem, so i did something naughty.
Ive honed/de-burred my rails(using knife sharping stone), to make it more smooth and having less preload.

I have to redo it, as at Z height 20 it isnt great yet, but half way the Z is much nicer:

Im getting there!