CV mode jitter when G1/G2 or G1/G3 alternates

[cnv]

Symptoms:
Axis movement gets jittery when G1 and G2 or G1 and G3 segments (curves and straights) follow each other in short intervals.
However the toolpath is a smooth curve, motion is not smoothed out, it behaves as if the machine was in ES mode.
Having only G1 OR only G2 or G3 segments does not cause issue, motion is smoothed out as it should be.

Steps to reproduce
Generate any G-code with close tolerances - those usually contain such alternating G1/G2 segments.

Environment and setup
UCCNC 1.2047 (earlier versions also behaved like this)
UC100 FW 3.0110, HW 2.0000
API 2.1320
UCCNC set to CV mode
G code does not contain G61 directive (ES mode)
G code may or may not contain G64 directive, it has no effect

[Robertspark]

Would it be possible for you to post an example gcodefile and your machine profile.

Its impossible to replicate your issue without being able to see your setup and the file

Steps to reproduce
Generate any G-code with close tolerances - those usually contain such alternating G1/G2 segments.

means nothing..... "close tolerances".....you need to define what a close tolerance is.

Also you need to define what the feedrate was at the time of the issue becoming apparent. (this via your profile will advise how many steps were being sent to the drives and how notchy the motion was {they are (normally) stepper drives.... and move in a certain number of steps....}

what was the corner error max setting (your profile will advise).

This is close tolerance:
https://www.youtube.com/watch?v=sFrVdoOhu1Q

[cnv]

True, I was not very specific, sorry.
By close tolerance I meant the CAM software POST is set to close path tolerance, like in this example 0.0002 mm.
This results in a toolpath made of very short linear and curve segments, so accelerations/decelerations (unwanted stops at segment ends) become more apparent.
But this is only the symptom.

Example G-code files and my profile file are attached.

In this example it occurs at feed rate 600 mm/m, but it can be reproduced at lower rates as well (only less apparent).

A short video showing the jitter (from 9 sec):
https://www.youtube.com/watch?v=LNxNxJITsPA

Sorry, volume is low (my steppers are quiet).
Notice that the first pass (long segments) moves smoothly.
In the second pass, arcs made out of short G1/G2 segments are noticeably jittery.
When the same profile is made out entirely of (however short) G1 segments, it also does not jitter.

[Robertspark]

I presume that the machine is setup in mm, given you made reference to "mm"

I can't see how this is going to be achieved....

By close tolerance I meant the CAM software POST is set to close path tolerance, like in this example 0.0002 mm.

Your post processor has output co-ordinates to 3 decimal places
..... so its not possible to have 4 decimal place precision....
extract from jitter (offending) example post.

Code: Select all

G1 Z-1.104 F200.
X6.812 Y-7.253 Z-1.151
X6.815 Y-7.248 Z-1.198
X6.82 Y-7.241 Z-1.244
G3 X6.827 Y-7.23 Z-1.289 I-0.505 J0.324

Also your machine uses 1600 steps per unit (X,Y & Z axis) .... again... presuming that its "mm", that gives a single step size of 0.000625mm
.... its not possible to ask the machine to step to 0.0002mm precision

maybe you've got an extra zero in that "0.0002mm precision".... .... so it should be 0.002mm.... that would tie in with 3 decimal point precision and also tie in with the motor resolution.....

ALTHOUGH, there is one little catch..... given your machine has 0.000625mm per step, and the smallest increment is 0.001, you have a bit of a rounding error (1 step = 0.000625mm, 2 steps = 0.00125, 3 steps = 0.001875mm etc), uccnc will keep track of this error, but its worth bearing in mind,

Also you have no backlash distance set for X axis,
backlash enabled for the Y axis (0.01mm)
and backlash enabled for the Z axis (0.02mm)

..... and you're still looking for potentially 0.0002mm precision..... good luck on that one....

backlash compensation will make motion look very odd sometimes when the backlash is very large relative to very small moves, because the machine will try to follow the cut line as closely as possible.

You have a linear error of 0.01mm
A corner error of 0.01mm

I think you're asking a bit much from that setup.

cncdrive may be able to offer some guidance, but I suspect that its the backlash compensation that is doing it.

What does the finished cut look like (does the jitter really matter? If the finished cut is fine then all well & good)

[Robertspark]

suggestion to try.....

remove the backlash settings from the axis (or not enable backlash).

and change the linear and corner error max to 0.001

then re-run the code and have a look how they compare.

[cncdrive]

You need to setup the CV parameters on the General config properly.
Especially concentrate on the
Tolerance:
- Linear error max.
- Corner error max.
And the path control strategy:
- Linear addition length
- Linear unify length

The software will go off the path with a maximum of the Tolerance setting and it will not buffer in longer segments than the linear addition length and will not buffer in longer distance in total than the linear unify lenght. So if those values are short can cause a slowdown and exact stops too.
For more info on the parameters please read the UCCNC manual.

[Derek]

UCCNC CV is very sensitive to acceleration. If your acceleration is low you will get the kind of motion you are seeing.

Try checking all of the settings others have suggested but in the end you may find your steppers aren't up to the task of delivering smooth motion when working with line segments.

[cnv]

..... and you're still looking for potentially 0.0002mm precision..... good luck on that one....

Robert, with all due respect, I think you are missing the point.
I never said anything about achieving 0.0002mm (or any other) precision, it's only toolpath accuracy in CAM.
Usually I'm happy with about 0.01mm actual precision on the workpiece, which can be achieved with some fine tuning.
But my post was about jittery movement, not accuracy.

What does the finished cut look like (does the jitter really matter? If the finished cut is fine then all well & good)

The bigger problem is that it needlessly slows down machining and causes unnecessary accelerations/decelerations.
In aluminum or thermoplastics it also causes occasional rubbing, as it does not maintain consistent feed, even though the toolpath itself should allow it easily.

[cnv]

UCCNC CV is very sensitive to acceleration. If your acceleration is low you will get the kind of motion you are seeing.

Try checking all of the settings others have suggested but in the end you may find your steppers aren't up to the task of delivering smooth motion when working with line segments.

Derek,
The problem is not with linear segments.
If an arc or circle or other curve is made up entirely of linear (G1) segments, it moves correctly, motion is smoothed out by CV as it should.
Pure arc segments (G2 or G3) are also smooth.
But if linear AND arc segments alternate (G1 then G2 then G1 again, in short intervals), then comes the problem.

[cnv]

You need to setup the CV parameters on the General config properly.
Especially concentrate on the
Tolerance:
- Linear error max.
- Corner error max.
And the path control strategy:
- Linear addition length
- Linear unify length

The software will go off the path with a maximum of the Tolerance setting and it will not buffer in longer segments than the linear addition length and will not buffer in longer distance in total than the linear unify lenght. So if those values are short can cause a slowdown and exact stops too.
For more info on the parameters please read the UCCNC manual.

Hi,

I re-ran the same Gcode with some pretty forgiving settings:

- Comm. buffer 0.5 sec
- Stop angle 89 deg
- Lookahead 2000 lines
- Linear error max. 0.1mm
- Corner error max. 0.1mm
- Linear addition length 20mm
- Linear unify length 40mm

I saw exactly the same behavior.

But again, it only appears then G1 and G2 alternates.
Never happens with pure G1 segments.