Terrible IS on a modified V0

[shy lintel]

This is a V0 with Tri-zero belted bed and Pandora gantry from Printer for ants:
https://github.com/zruncho3d/tri-zero
https://github.com/MasturMynd/Pandora
It has a DragonBurner toolhead with Orbiter 2.0 extruder and sensor, as well as an EBB36 with adxl.
Bed leveling uses ZeroClick
It is running a Manta M5P, CB2 and 200W PSU
X and Y steppers are from the original Siboor kit (siboor-35sth52-1204a)
Belts tensioning is achieved with the West3D tensioning tool.

The filament used for the cube is PLA and is calibrated (PA and flow rate)

[shy lintel]

I have checked the belts. They are correctly positioned on the bearings.
The bearings are a little rusty but they spin flawlessly

[worldly garden]

I would replace all your rusty parts

Any rust on a motion part will affect performance.

[shy lintel]

Will do that

[shy lintel]

I have replaced the rusted bearings and changed the MGN9 X rail to a better quality rail from Dllpdf.
The voron cube is printed from Polymaker silk pla to enhance the result. Print settings has the recommended IS:
[input_shaper]
shaper_freq_x: 98.8
shaper_type_x: ei
shaper_freq_y: 73.2
shaper_type_y: 2hump_ei

I have to say that the Y IS is scary. I've never seen one like that before. But the voron cube print doen't look that bad IRL

[worldly garden]

I think your gantry is racked based upon your y-axis.
Silk pla is one of the worst to use for this test. Please use non-silk pla

[shy lintel]

Should I comment the IS settings in the printer.cfg when I print the cube?

[worldly garden]

I would look at your y axis to see if it racked before making another print

Can you also take an updated picture of the front and the top of your printer please

[shy lintel]

The Y axis was racked by 1mm. I did correct that and rerun the tests

[worldly garden]

I would check all the screws and stuff holding the toolhead board, and toolhead in general. Make sure that everything is snug

[shy lintel]

I just disassembled the tool head, everything is tight: hotend, extruder, X carriage, can board, board mount, can cable is secured. I also checked the idlers and the steppers mount. I am re running the different test again now

22:08:01 // --> Detected axes_map: unable to determine it correctly!
22:08:01 // --> Detected gravity: 9.89 m/s²
22:08:01 // Dynamic noise level: 455.91 mm/s² -> WARNING: accelerometer noise is a bit high
22:08:00 // Machine axis Z -> nearest accelerometer direction vector: [ 0 0 -1] (angle error: 50.34°)
22:06:58 // Machine axis Y -> nearest accelerometer direction vector: [0 0 1] (angle error: 9.57°)
22:06:53 // Machine axis X -> nearest accelerometer direction vector: [1 0 0] (angle error: 9.74°)

[shy lintel]

Should I worry about: Dynamic noise level: 455.91 mm/s² -> WARNING: accelerometer noise is a bit high ?
Could this be the cause of the weird IS graphs?

[shy lintel]

Same kind of results

[worldly garden]

@shy lintel The axis map test is in beta and I woudl not rely on it to much
how are you measuring belt tension - what app/meter are you using and what is it reading

[shy lintel]

I use this belt tensioner from West3D and set the belts at 1.9 @ 150mm

not an easy task on a V0, the bests are so close from one to another and also close to the frame...

I'll have a look at the whole frame, see if nothing is loose there

[worldly garden]

@toxic falcon they are using a Tri-zero belt path. Not sure what is going on with y. Do you have any suggestions

[shy lintel]

I have just checked the rest of the frame and everything looks tight

[toxic falcon]

@worldly garden - Tri-Zero is irrelevant for IS. I think you meant to ping @inland forum for the Pandora gantry here.

Those weird values with mostly Z on Y motions suggest a loose toolhead mount or gantry or adxl. Suggest checking carriage screws and making sure they're not too long.

[inland forum]

Aside from what zruncho mentioned above, the low amplitude values (even for Y) lead me to think there's some binding going on here. Knowing that the bearings were replaced with nearly the same results leads me to think the Y rails aren't perfectly parallel to one another. I'd loosen one of the rails, move the toolhead to the back and tighten the closest screw, then move to the front ant tighten the closest screw. From there, tighten the rest of the screws, loosen the end screws and tighten them again (should help reduce stresses built up between the aluminum and steel components).

[shy lintel]

@inland forum @toxic falcon @worldly garden Thanks all lot for taking the time trying to help me. The Y rails are parallel to each other. When the belts are off, it slides very smoothly, and there are no hard spots along the rails. The X carriage is also mounted correctly. There's a very tiny play on the tool head. Almost not noticeable.

[inland forum]

what is your belt tension in Hz? Idk what those tension meters do...

[worldly garden]

1.9 is about 110hz

[inland forum]

At this point I'd use the feature to excite the axis at a specific frequency to see if you can isolate where the noise is coming from.

[shy lintel]

I just downloaded the Gates Carbon Drive app, so I mesured 117 on A and 118 on B drive, which also shows between 1.9 and 2.0 on the mecanical device

[lost wave]

Is it just the picture or do the toolhead belts look like this?

[toxic falcon]

That certainly looks off. No idea how that's even possible.

[lost wave]

possibly wires pushing belts back?

[toxic falcon]

I can't seem to find where the ADXL is here. If it's a toolhead-based one, I personally don't trust those; Nozzle or Carriage mounts are my pref.

[shy lintel]

You are correct, the toolhead cables push the belts a little

[shy lintel]

I am using the EBB36 with adxl

[lost wave]

Try adjusting this so belts run parallel to gantry, I think non parallel belts can cause changing tension as it goes to and from xy joints, causes weird is

[toxic falcon]

I don't trust anything with a sensor which is mounted on two flexible posts. That absolutely could be contributing to what you're seeing, which was mostly Z for Y moves. If you can get a nozzle mount adxl (I like the mellow fly CNC mount with USB - but there are plenty of others), you may find the measurements are much cleaner.

[lost wave]

Yeah that too ^

[inland forum]

It all makes sense...

[shy lintel]

I have an old USB ADXL laying around. I'll try to make a nozzle mount and redo the tests

Let's have a try with this

[shy lintel]

these are the results with the USB ADXL with nozzle mount, I must say it has some flex

[inland forum]

Now we're talking. Just a matter of figuring out that second peak on x

[shy lintel]

Could this come from the ADXL mount that is not very regid? It flexes a bit

I have a Bambu hotend and this is the mount I made

I didn't expect the difference to be that significant between the toolboard and the nozzle mount ADXL. 😮

[lost wave]

probably not, you have x following peaks but very low y or z movement so shouldn't be flex

I'm guessing non parallel belts

https://drmrehorst.blogspot.com/2018/08/corexy-mechanism-layout-and-belt.html

[toxic falcon]

I'm not surprised at the difference ....but It's hard to predict what you'll see with long floppy toolhead board mounts 🤷‍♂️

Did you update the axes map for the ADXL? That (nicely clean!) X is showing Z as the peak signal for X. All in all pretty good - 20K with MZV.

I'd suggest making a toolhead or carriage mount for your Fysetc PIS. If the second peak goes away, then you know it's mount flex or toolhead flex you're capturing.

My money's on the high weight of the orbiter contributing here.

[toxic falcon]

The BBL hotend style is super light, compared to heavier Dragons, which creates a weight offset around the carriage mount. That may be what you're seeing. Something other than the "belts plus toolhead combined" has a resonance, so some component has insufficient rigidity and/or an offset that's triggering the double peak.

[shy lintel]

Thanks a lot for all those tips. I removed the orbiter sensor to make the toolhead lighter. By doing so, I noticed the ebb mount wasn’t tight. It happens that the screws that are going through the extruder to the mount spacer were a bit too long, maybe a fraction of a mm. I replaced that and it’s stiffer now. I am also printing some toolhead spacers to straighten the belts. But for this to work I need to relocate the zeroclick dock from the back to the side. I will print the appropriate dock and unmount the frame to slide in some nuts. Also, as you mentioned, I did I not update the axes for this adxl. I will post the results when done. 🙏

[toxic falcon]

If you do the ADXL run after fixing the ebb mount, maybe it'll fix things. Only way to know is to test! Highly recommend rerunning the test after each change, if you can, so you know what an issue looks like. I doubt the belts getting pushed is causing X issues, unless there's rub.

[shy lintel]

There was no rub on the belts but I installed the toolhead spacers anyway because the cables prevented the sensorless Y homing to perform correctly. I will redo IS to compare.

[shy lintel]

Here are the results after I removed the Orbiter sensor and added a spacer to the toolhead so that the belts are parallel to the rail.
I'm having a hard time to have a nice graph for the belts but X is getting better

[toxic falcon]

That high peak on X is so high now! But you'd expect the dominant X peak to be just slightly higher in frequency than the Y peak, given the higher mass for motion there.

BTW, a spacer that moves the toolhead out can have a real and large effect, as the head has a longer arm to swing on. I saw a few years ago that even a 1mm Y motion can completely change the results when using a rail with any amount of play. But all that said, things look much better after the EBB mount, as for the same X vibration, you're getting a way-higher X accel with MZV.

On a V0, I'd think sensored Y would be pretty reasonable to set up. If you can get that going, you may find even better results here.

[shy lintel]

What do you mean by "sensored Y"?

[toxic falcon]

old-school endstop, as opposed to sensorless

[shy lintel]

Oh sorry. I thought it was some fancy blinky novelty 😂

[lost wave]

Did you remove the sock and clip for attach this or is it over everything? Was looking for one and remembered the one you made

[shy lintel]

I removed the sock and attached it to the hotend block but my mount wasn’t designed well enough. It had quite some flex. I ended up with the old school method and used a usb IS fixed on the front of the tool head and had better results.

[red mortar]

how are things going? any update for us?

[red mortar]

how do u want us to proceed with your post here? still help needed?