mach 3 cv mode stalling

[atypicalguy]

Is anyone thinking about NURBS controllers? Seems like EMC2 has a 2d version; a couple of guys seem to be developing 3D implementations. The anti-jerk and acceleration look ahead algorithms are pretty smart and easier to implement. I haven't seen one running in 3D though.

[lenne0815]

Is anyone thinking about NURBS controllers? Seems like EMC2 has a 2d version; a couple of guys seem to be developing 3D implementations. The anti-jerk and acceleration look ahead algorithms are pretty smart and easier to implement. I haven't seen one running in 3D though.

Thing is mach3s implementation isnt bad at all, if ure running 3d in a different mode as zigzag with irregular stepover it works really well, problem is that zigzag with non 90degree stepover is really the thing ull want to do as soon as ure doing more 3d stuff.

Maybe art itself could offer some suggestions to this Problem ?

[ihavenofish]

Thing is mach3s implementation isnt bad at all, if ure running 3d in a different mode as zigzag with irregular stepover it works really well, problem is that zigzag with non 90degree stepover is really the thing ull want to do as soon as ure doing more 3d stuff.

Maybe art itself could offer some suggestions to this Problem ?

machs countouring is actually very sophisticated, and id say better than many "low cost" vmc's abilities, like haas or fanuc mates from what ive seen. its only this one bug that holds up up. im fairly sure art know about it, but its likely incredibly hard to deal with. my understanding is its a logged bug. but in case im wrong, and he doesnt know, someone should point him to this thread

i spent almost an hour on the phone with DMG this morning. im just about ready to ditch the hobby things and go straight to guaranteed performance. i have a new project coming and basically, a $150k 5 axis dmg is "cheaper" than almost anything else per part. go figure. i guess thats the difference between "fun" and "production".

[ger21]

Fwiw, Art retired from Mach3 several years ago.

A new "Known Bug List" was just started on the Mach forum. PM Andrew with the bug, as he's the official bug tester.
Known Bugs

[ihavenofish]

Fwiw, Art retired from Mach3 several years ago.

A new "Known Bug List" was just started on the Mach forum. PM Andrew with the bug, as he's the official bug tester.
Known Bugs

ah, cool, thanks for that info.

[AndrewM]

Hi Guys,
The Known Bug list has actually been moved to the website. So if you could please email me with the, or a, file you see the issue on I will test it here at the shop.

Thanks
Andrew

[ihavenofish]

Hi Guys,
The Known Bug list has actually been moved to the website. So if you could please email me with the, or a, file you see the issue on I will test it here at the shop.

Thanks
Andrew

i can probably get you a file next week. it may be difficult to reproduce on a different machine though.

[AndrewM]

That's fine, Just email me with it whenever and I will test it.

No worries, I will give it a go anyways ^_^
Thanks

[lenne0815]

today i did experience the problem again and found the solution aswell, if i let my cam software produce arcs to smoothen the cut stalls will occur if i do the same toolpath just using straigt moves it runs perfectly, in between the corners which stall theres no speed difference from tiny arcs / lines.

This bit of gcode wasnt 3d though, it was just a 2d Pocket operation

[ger21]

I read the other day that some work was done on this issue in the latest development version of Mach3, 3.043.058 It's not listed in the release notes, but Brian said he made some changes.

[lenne0815]

I read the other day that some work was done on this issue in the latest development version of Mach3, 3.043.058 It's not listed in the release notes, but Brian said he made some changes.

I did read that in the changelog aswell, but appearantly it didnt really work its not too much of an issuue though, as long as ure aware of it and make sure ur cam doesnt produce arcs. ( as mentioned before, the feed is pretty much exactly the same )

[ger21]

I've not run into this problem, but was at a friends the other day and he showed me some code that has serious issues.
He creates his code with Visual Mill, and it all works fine. He was cutting some parts for someone with code created in BobCAD, and it stalled and lost steps through the entire program.

So I guess the "quality?" of the toolpaths makes a big difference.

I did read that in the changelog aswell

The changes aren't listed in the change log, so maybe we're talking about different things?

[lenne0815]

I've not run into this problem, but was at a friends the other day and he showed me some code that has serious issues.
He creates his code with Visual Mill, and it all works fine. He was cutting some parts for someone with code created in BobCAD, and it stalled and lost steps through the entire program.

So I guess the "quality?" of the toolpaths makes a big difference.

The changes aren't listed in the change log, so maybe we're talking about different things?

I Just rechecked the changelog and ure right its not in there, dunno where i read it, probably from the same source as u did.

I did notice a diffent thing aswell, if ur gcode is "too long" Mach3 isnt able anymore to look far enough into the future to predict sudden changes in the toolpath, i did set this to the highest setting mach3 allows already but i still have to set my cam to its lowest setting, i think that may be the error u guys encountered.

[BanduraMaker]

I've not run into this problem, but was at a friends the other day and he showed me some code that has serious issues.

That would be me.

I think I sort of agree with whoever it was that said that Mach 3 is violating the acceleration parameters in certain moves.

My machine has a top end of about 400ipm-ish on the long axis. For most tool paths, I can also cut at 400ipm but this one file that was made in BobCAM was causing it to massively stall at just around 300ipm. It was a 3D roughing tool path and I wonder if somehow Mach doesn't get lost when doing 3D changes in direction?

That matches the scenario that ihavenofish shows above (which looks like it was made in VM - the same software I use).

I made a tool path by hand that did air cuts back and forth and ran it at over 450 on my long axis with no troubles. You could see the machine slowing and speeding back up as it changed directions (I think I'm using an accell of 45 but I've had it higher before, it's just some tool paths that give me trouble). With some of the 3D paths, it sure seems that the accel is much faster than what's programmed in.

[ihavenofish]

That would be me.

I think I sort of agree with whoever it was that said that Mach 3 is violating the acceleration parameters in certain moves.

My machine has a top end of about 400ipm-ish on the long axis. For most tool paths, I can also cut at 400ipm but this one file that was made in BobCAM was causing it to massively stall at just around 300ipm. It was a 3D roughing tool path and I wonder if somehow Mach doesn't get lost when doing 3D changes in direction?

That matches the scenario that ihavenofish shows above (which looks like it was made in VM - the same software I use).

I made a tool path by hand that did air cuts back and forth and ran it at over 450 on my long axis with no troubles. You could see the machine slowing and speeding back up as it changed directions (I think I'm using an accell of 45 but I've had it higher before, it's just some tool paths that give me trouble). With some of the 3D paths, it sure seems that the accel is much faster than what's programmed in.

O_o

o_O

so, setting up another new xzero raptor. skipped mach, went to emc, seems they have a new trajectory planner id liked to see if it works better than the old.

anyhow. the person making it took some interesting "scope" plots of mach3 in cv mode. sure enough, there are seemingly random (although we know they aren't truly random) SPIKES in the velocity. lots of them. good visual representation of this bug. I assume that in any given program there are many many spikes, but it will depend on the machine setup weather each spike will cause a stall. id also imagine a servo would "smooth" over the spike.

Hmmmmm


5th image down.
LinuxCNC Documentation Wiki: NewTrajectoryControl


FIX IT!!!!!!!!!!!!!!!!


HUH? WHA?

[BanduraMaker]

After a couple of ruined expensive parts, I bought a Kflop and will be switching over the KmotionCNC as soon as I can get it running. KmotionCNC is based on a newer version of EMC so hopefully it will get the new planner soon too.

I plan on adding encoders and eventually running the steppers in closed loop servo mode.

[samco]

Actually - the small spikes are probably not as big an issue (they are very short).. The bigger issue is the acceleration constraint violations... (the mach config was set to 30in/s^2 acceleration) I was seeing peaks of 100in/s^2...

http://electronicsam.com/images/Kand.../MachViol1.png

This is as you increase the velocity of mach. I was running programs at 500ipm testing linuxcnc.

O_o

o_O

so, setting up another new xzero raptor. skipped mach, went to emc, seems they have a new trajectory planner id liked to see if it works better than the old.

anyhow. the person making it took some interesting "scope" plots of mach3 in cv mode. sure enough, there are seemingly random (although we know they aren't truly random) SPIKES in the velocity. lots of them. good visual representation of this bug. I assume that in any given program there are many many spikes, but it will depend on the machine setup weather each spike will cause a stall. id also imagine a servo would "smooth" over the spike.

Hmmmmm


5th image down.
LinuxCNC Documentation Wiki: NewTrajectoryControl


FIX IT!!!!!!!!!!!!!!!!


HUH? WHA?

[samco]

oh - and in case someone says - 'that is a screenshot of linuxcnc'.. I am using linuxcnc to plot the motion of mach. (in effect using mach as 'encoder (step/dir) feedback to linuxcnc)



linuxcnc is calculating the acceleration from the step/dir info.

sam

[lenne0815]

Any News from the Mach Team on this ? this still severly criples the speed of my router, i resorted to running very low maximum speeds on my steppers ( Acelleration settings dont really matter because they are disobeyed anyway )

[ger21]

I believe that no more work is being done on Mach3, and no more bugs will be fixed.

Just about every Mach3 bug fix in the last several years has introduced new bugs, due to the complexity of the software. I have a feeling that this issue was too difficult to fix without breaking a lot of other things.

Hopefully, this issue won't exist whenever Mach4 gets released. Because Mach4 has taken so long, they better get it right, or a lot of users are going to look at other options.