EBB36 accelerometer orientation

Basic Information:

Printer Model: Voron 2.4/350 LDO kit (many homebrew mods), under construction
MCU / Printerboard: Pi 4 / Octopus 1.1 / EBB36
klippy.log

This is NOT an issue ! At most a feature request.

Searched about the EBB36 ADXL345 orientation, and the influence on the accelerations. Found nothing interesting…

The EBB36 is oriented at 45° (43 to 46 or so, depending on the gears adjustments - measured in CAD). Meaning that X accelerations will influence both X and Y MEMs, and Z accelerations are read by both as well. Shouldn’t have any effect on the IS recommandations. And it doesn’t. I compared with the readings from a ADXL345 attached to the hotend cardridge.

The x+y+z curves are very similar. For the individual curves (x, y and z), the result is not the same (it was expected)
Was expecting x test with readings on x and z, both being equal ; verified
No effect on the resonance frequency (obviously)
Was expecting some 0.707 (sqr(2)/2, sin(45°)) relation between the PSD peaks ; it is there

EBB36 onboard ADXL345 (I just reaffected z to y, the two others are defaults) :

The results with the accelerometer attached where it is supposed to be on a Stealthburner :

Somewhere, there was a FR for an option in calibrate_shaper.py so we could enter an angle, and recalculate the real X and Z from the rotated readings… (maybe on Github)

For the input shaper tuning, it is of no interest. But could it be interesting for results analysis, mechanical diagnosis, and belt tensioning for example ? (I have no knowledge, just starting to learn from some videos on YT)

Also quickly read calibrate_shaper.py ; unfortunately not familiar with Python (C/C++ only, for decades). It is simple math (vector rotations)… But can’t see where I could do the math… Also could preprocess the csv file in Excel… Could it be of any interest ? (rarely use Excel, but I think it could be easier)

I did extensively test around with this and TBH none of the results lead to substantial improvements or additional insights.
Some of my earlier test in the PR you are referring are documented here: adxl345: Added automatic accelerometer calibration by dmbutyugin · Pull Request #4931 · Klipper3d/klipper · GitHub

Thanks for the link.
It was this thread. I didn’t read it completely, as I was searching for something much simpler : a virtual (mathematical) arbitrary rotation around one axis, the angle being known.
Interesting work. Reading it again (quickly), I remembered the love/hate relation with this feature you mentioned in one of your posts. Have to read it a couple more times…

Any chance ACCELEROMETER_CALIBRATE will be back in the future ? Or something much simpler, in the form of a list of angles in [adxl345] section, in order to set the transformation matrix by hand ? Could be usefull to anyone designing a toolhead, the accelerometer could be oriented more freely, and compensated in software. Just had a look to adxl345.py from the PR, seems to me _calculate_axes_transform sets the transform matrix ?

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.