Axis jerk, servo tuning and capturing data

[GabrielR922]

Hi Tom.

I still have some jerk motion in my g-codes.

Before I had some play in the X axis, but solved it with a preload in the ballscrew.
I still find some jerk in the axis movement and can't find the source of the problem.

I captured some motion with the CaptureXYZPosDestToFile example, but don't know what to do with the data.

Do you have a tutorial or can give me a help to solve this ?
Also, how can I generate the path with the data from the capture?

I can supply you with all the data you need, just request it please.

There are 3 example with different g-codes in the KFLOPdata files attached.

Thanks Tom!

[TomKerekes]

Hi Gabriel,

Do you know Microsoft Excel?

Attached is an Excel Spreadsheet to import the data.

Here is a plot of the XY Path

Here is the X Velocity of the Trajectory vs time:

Here is the Following Error vs time:

[GabrielR922]

Hi Tom.

That was exactly what I needed.

I tried but wasn't reaching that final result.

It seems that the profile and the following error are good, do you agree?
For woodworking at least.
30~ counts following error, 17000 counts / inch, total: 0.0017" or 0.05mm

So it must be coming from mechanical backslash?

Attached some photos of the "steps" in the final parts.
Theses steps are a little bit small, there are some other that I run with double speed that are concerning big. i think around 0.2mm.
The ones on the photos may be 0.1mm.

The data collected was at a high speed, around 8000mm/min.
The cutting that you see in the photos was done around 2000mm/min
The ones that had big steps were done at 4000mm/min

Thanks!

[TomKerekes]

Hi Gabriel,

Sort of hard to say. The following errors are +/-0.05mm. Then you say the errors in the part are 0.1mm. But then you say those are at a lower speed?

[GabrielR922]

Hi tom.

Will cut another piece with fast speed and take some photos.

Also will capture data with the same data for us to compare.

Thanks

[GabrielR922]

Hi Tom.

Can you take a look at these captured motions ?

I analyzed them and found just the acceleration looks strange. It looks like accelerate and decelerate very fast.

They give a lot of jerk when the helix end and the radial motion starts.

the 3 data captured are from the same g-code, but with increasing speed and then increasing acceleration.

KFLOPdata7000 is captured with 7000mm/min and so on.

Thanks

[TomKerekes]

Hi Gabriel,

The acceleration goes to ~53 inches/s^2 which is under 60 inches/s^2 so I believe that is correct.

[GabrielR922]

Hi Tom.

The jerk seems to be where the helix end and the radial motion starts. (flat, not with Z movement).

Raising the Tau from 0.01 to 0.02 seems to do a lot better.

Now the machine is simple stopping in random places in linear one axis movement. (G1)

Without outputing a console error. photo attached.

Thanks.

[TomKerekes]

Hi Gabriel,

It seems something is disabling the Axes. You might have checked the Axis Screen to verify.

Do you have any code running that disables axes?

[GabrielR922]

Hi Tom. I have a code that disable axis all axis if one axis is disabled due to following error.

I had the Y axis following error set to 10.000, lowering it to 100 solved the problem. (my system is tuned for around 50 counts).

The Y axis seems to be doing very fine, but the X axis, at ~4000mm/min gives some marks on the finish, if you hold your hand in the axis, you can feel that it runs with some bumps. (even without load)

At 6000mm/min this can't be seem.

I'm starting to tune X axis again. Can you give me a help with filters?

Check the attached photo. i need to lower the Gain to raise D. what can i use? is ok to use for example a low-pass filter of 300 ?

Thanks.