Page 1 of 3

Arc Junctions, corner error max? linear error max? ??

PostPosted: Sat Nov 24, 2018 11:13 pm
by Robertspark
Confused....

I've got some big arcs, and I'm trying to understand the feedrate fluctuation between the junctions and which settings may adjust (improve / reduce / stop) this.

My (demo) machine has an velocity of 10400, acceleration of 3000, and X & Y steps per unit of 20.

I've uploaded a youtube clip here.
https://youtu.be/iCDamhQs1Ys, yes I know the sound is annoying.... kids left a wepage of FRIV in the background :roll:

Attached is the CAD file (rename it to DXF from txt and it will open just fine)
arcstestDXF.txt
(214.98 KiB) Downloaded 756 times


Also attached is the simple gcode file.
arcstestGCODE.txt
(699 Bytes) Downloaded 794 times


Corner Error Max is for the junctions between two linear moves (and maybe a linear move and an arc??? or an arc and a linear move????) ...... what is the factor which changes feedrate between arcs?

Thanks

Re: Arc Junctions, corner error max? linear error max? ??

PostPosted: Sun Nov 25, 2018 4:21 am
by ger21
Confused....

I've got some big arcs, and I'm trying to understand the feedrate fluctuation between the junctions and which settings may adjust (improve / reduce / stop) this.


I didn't have time to check, but I'm assuming the arcs are all tangent?

If so, I believe it's a problem with the current trajectory planner. It should not slow down at the transition between tangent arcs (or lines and tangent arcs), but it frequently does.

Re: Arc Junctions, corner error max? linear error max? ??

PostPosted: Sun Nov 25, 2018 7:40 am
by Robertspark
Hi Gerry, yes it's a test code, first bit is a circle with the code in 4 lines which are the 4 quadrants, the last quadrant leads on to two more in an S shape (with the semi circle divided into 2 quadrants), at the end of the S, it leads on to two more S shapes all divided into quadrants.

Re: Arc Junctions, corner error max? linear error max? ??

PostPosted: Sun Nov 25, 2018 11:41 am
by Derek
Looks to me like the same issue I'm having with the planner. I think UCCNC requires a ton of acceleration to smooth out the transitions. Keep slowing your feed rate down and at some point the pauses will go away. Conversely you can keep raising the acceleration until you get smooth movement. On my machines, generally speaking, the fastest I can cut without pauses between arcs and arcs to lines is 40 ipm. This is on a machine with 120 IPM rapids.

My machines can't handle a ton of acceleration before they fault the drives so if I want smooth movement and even chip load I have to cut way slower than I would like.

If I go to Offline Mode and raise the acceleration the issue goes away. It seems UCCNC has a minimum velocity/acceleration ratio that it wants to see for smooth movement and sadly none of my 4 machines can pull the freight. Still better than Mach though:)

Re: Arc Junctions, corner error max? linear error max? ??

PostPosted: Sun Nov 25, 2018 1:06 pm
by ger21
I did look at the .dxf, just didn't verify they were tangent. :)

Re: Arc Junctions, corner error max? linear error max? ??

PostPosted: Sun Nov 25, 2018 1:46 pm
by Robertspark
I hand coded them at first....(and failed :( ) Then used sheetcam.... They are posted from sheetcam just cut out the other non XY code

I've got a long draft email I'm working on with math as a brain storm session, maybe cncdrive could do something with it

Re: Arc Junctions, corner error max? linear error max? ??

PostPosted: Sun Nov 25, 2018 3:06 pm
by ger21
Hopefully they already no how to fix this when they start on the new planner soon.

Re: Arc Junctions, corner error max? linear error max? ??

PostPosted: Sun Nov 25, 2018 3:15 pm
by Robertspark
I was wondering about the math such as this, (ignore the formula, just the three intersecting circles ...

Say circle b was the corner error max parameter.

With your intended motion being a and c but not in an s shape, say the m shape.

Image

Re: Arc Junctions, corner error max? linear error max? ??

PostPosted: Sun Nov 25, 2018 3:56 pm
by Robertspark
The motion planner should imho ( but it doesnt) test for segments of the same circle and not apply and junction arc or slow down.

Take two g02 moves.

I would think that the motion planner would be able to calculate the arc origin by X+I and Y+j, then at the next g02 it checks X+I and Y+j, if they are the same in both the current and next g02 it knows no deceleration is required between the moves

__________________
Another one

If you have two g02 moves, the junction arc will always be an g03 and visa versa

Re: Arc Junctions, corner error max? linear error max? ??

PostPosted: Sun Nov 25, 2018 3:58 pm
by ger21
OK, with the attached code, here's what I found.
Only two CV settings have an effect.
1) Exact stop at angles > must be set to 180, to effectively disable exact stop mode, to get the two arcs to blend. At 179°, it runs in exact stop mode.
2) Maximum Corner error. If set very low, the tool will follow the path. As you increase the Corner Error, the blend will increase, but only up to a certain point, which is maybe 1", maybe a little less. But it still slows way down.

The expected behavior would be to go as fast as possible while staying within the corner error. But it doesn't go anywhere near the corner error.

Acceleration plays a role. But even when set very low, the blending stays the same. And it gets tricky when the accel is too low to reach the commanded feedrate. If you are trying to maintain velocity, how much do you blend when you can';t reach your target velocity at all?
At 300ipm, the accel needs to be about 9 to be able to reach the 300ipm on the first arc.

Note that when you set Exact Stop at angles to 180°, it adds the undesirable affect of blending the plunge and retract moves at the beginning and end.
It might be nice to be able to exclude the Z axis from these calcs.