[OlfCNC]

Deviation control and code compression are now obsolete concepts.

Some wants to know what error the software will make on a workpiece is an obsolete idea?
You have a strange thinking about CNCs.

[RCaffin]

Some wants to know what error the software will make on a workpiece is an obsolete idea?

Not quite how I word put it.
I want to know that my software will not make any errors at all. I find that Mach3 does not make errors on the parts I am machining.

I was machining a rather long thin rod (4.8 mm diameter) on the lathe a few days ago. I could measure the deflection of the rod during machining: the diameter was changing over 60 mm by about 20 micrometres. So I put in a taper correction, and reduced the diameter variation to about 5 microns.

I don't want 'deviation control' - I want accuracy. I get it.

Cheers
Roger

[ger21]

I know Art Fenerty has been working on a more refined TP

He wrote one for Mach3 called Tempest. Unfortunately, it was never fully implemented. It works great, provided you never need to hit Feedhold. I've used it a few times when the standard TP has failed me.

[daniellyall]

Arts new program is working well Roger, I have it running a router it's a cool program with some crazy things you can do to it you can run monkey scripts in the middle of a Gcode

[RCaffin]

Tempest - yes, I know. But he has continued the development. He has a thread on it at Gearotic.

The problems with FeedHold - yeah, true. But that is really in the nature of a fast Feedhold, rather than the details of the TP. If you hit Feedhold, expect that you might have to restart from the beginning. Me, I stick M1s or M6s in the gcode during development, and knock them out slowly.

Edit - Daniel: ah, OK. it's been a while. I must go and look.

Edit #2: is it going to offer a functional alternative to Mach3 maybe? Before Mach4 is finished?

Cheers
Roger

[OlfCNC]

I don't want 'deviation control' - I want accuracy. I get it.

Then don't use G64 at all in the first place. Run everything with G61.

I was machining a rather long thin rod (4.8 mm diameter) on the lathe a few days ago. I could measure the deflection of the rod during machining: the diameter was changing over 60 mm by about 20 micrometres. So I put in a taper correction, and reduced the diameter variation to about 5 microns.

Lathe codes and complex 2.5D/3D codes are different animals. Try to run a 100 000 lines relief with G61 and you will see what I'm talking about.
And if you decide to change to G64 that will introduce errors, in Mach3 the amount of error is unknown, uncontrollable and with softwares which uses deviation control it is controllable.

[ger21]

No, Tempest doesn't have any Feedhold. As I said, it was never fully integrated into Mach3.

[RCaffin]

Theory:
"And if you decide to change to G64 that will introduce errors, in Mach3 the amount of error is unknown, uncontrollable and with softwares which uses deviation control it is controllable."

Practice:
I run G64, and my parts are (measured) accurate to within 10 microns.

Cheers
Roger

[ger21]

You're making small, straight cuts on a lathe.
Try doing large 3D machining at 400+ ipm.

[G59]

So Mach4 can't do that? What was the whole point of Mach4 then? I thought Art retired???

[daniellyall]

Art changed it so you hit stop and it will just stop then to start it again you push continue, it just works with no stupid move's it works well different way to do feed hold it does not lose position it does not on my machine , it is tempest.

it is usable just for a router, a Laser good as gold, there is no M6, M31. no canned cycles it's built as a laser controller, but you could just write the codes needed in monkey script it's quite a easy script to use, but at this time you can only use pokeys boards.

[daniellyall]

G59 Arts new program he did for his lasers, but it does run a router, this is his retirement along side gearotic, I think he is just doing it for fun last year he did a scanning program, next year he may start on a cad program. I think it's just fun for him. you can almost use it for a IOT program

M4 is newfangled /artsoft not Art

[G59]

M4 is newfangled /artsoft not Art

Ok. But Mach3 he wrote it.

Interesting bit of history.

[daniellyall]

yep a long time ago, he started it around the same time Linuxcnc started

[samco]

the trajectory issues with mach3/4 (mach4 has the same TP as mach3 - brian said they are re-writing it when they get some time...) is this

http://wiki.linuxcnc.org/uploads/Screenshotmach.png

That is linuxcnc logging the output of mach3. You can see it skipping many short line segments. (one direction skips less than the other - dependent on speed)

mach3 has CV Dist Tolerance_____Units. This is a tells mach how far from the end of the segment it should start blending. Kind of hard to relate that to actual path following.

anyway - lets take the stepconf in linuxcnc - it has a bug? well - if you submit a bug report - it would probably get fixed.

git.linuxcnc.org Git - linuxcnc.git/search

Linuxcnc is a getting fixes and improvements daily.

sam

[daniellyall]

only with lads like you doing it sam, your the only reason I had a go at it and got it working then the computer died but it was working well with 2.7.

[ihavenofish]

You're making small, straight cuts on a lathe.
Try doing large 3D machining at 400+ ipm.

done it. seemed perfectly fine to me.
https://www.youtube.com/watch?v=kRzU32HCt-k

(you can hear the hard stop of the cv bug on the lower left corner. no stall, but borderline, but thats about 450ipm programmed feed, and it didnt deviate too much)

i also did a glass casting pattern for uhmw, which required a precision surface (well, as precision as it gets with a cheap machine) and it worked flawlessly. some 600 grit paper and it was smooth as silk.

and theres this guy attached... i think 100ipm in aluminium with a 1/8" bit and 3 thou step over.

now dont get me wrong, mach3 isnt "good" and i cant use it at all anymore, but youve kinda chose the one thing i think it actually does very well to pick on

[RCaffin]

linuxcnc - it has a bug? well - if you submit a bug report - it would probably get fixed.

Which means LinuxCNC has serious attraction right there.
And let me state right now: I am aware of a number of serious bugs in Mach3 which can NOT be fixed. I work around them, following advice from others.

PS: Gerry - that's the straight bit of the shaft. The turned ends are rather more complex. Milling - definitely some serious 3D stuff. But 'engineering' stuff, not wood carving things. Lots of straights and circles, with thread mills and custom cutters as well as ordinary square end and ball cutters. It's all good fun.

Cheers
Roger

[ihavenofish]

the trajectory issues with mach3/4 (mach4 has the same TP as mach3 - brian said they are re-writing it when they get some time...) is this

http://wiki.linuxcnc.org/uploads/Screenshotmach.png

That is linuxcnc logging the output of mach3. You can see it skipping many short line segments. (one direction skips less than the other - dependent on speed)

mach3 has CV Dist Tolerance_____Units. This is a tells mach how far from the end of the segment it should start blending. Kind of hard to relate that to actual path following.

anyway - lets take the stepconf in linuxcnc - it has a bug? well - if you submit a bug report - it would probably get fixed.

git.linuxcnc.org Git - linuxcnc.git/search

Linuxcnc is a getting fixes and improvements daily.

sam

seems a bug. ill report it eventually (new machine is only partially running). i swapped video card on the machine and linux wont recognise the new one unless you install the driver FIRST. how the f can i do that the old one died? this type of crap and many other little things is what scare people away. something like pathpilot is great for pure users (preconfigged by tormach). just need a default package a bit more like that and youll overcome a big "user experience" hurdle.

you know me, im not one to say linuxcnc suck...

oh, wait, ive said that quite alot. but i am ok with 2.7 now. its peppy:

984ipm and 0.5G. no slowdown til the tightest arcs in the middle.


shame the ATC spindle had to go. far too heavy for the machine, would make it bounce when it stopped (jerk control please!). on the bright side, with the new lighter spindle i ordered it will go even faster

[ihavenofish]

Which means LinuxCNC has serious attraction right there.
And let me state right now: I am aware of a number of serious bugs in Mach3 which can NOT be fixed. I work around them, following advice from others.

PS: Gerry - that's the straight bit of the shaft. The turned ends are rather more complex. Milling - definitely some serious 3D stuff. But 'engineering' stuff, not wood carving things. Lots of straights and circles, with thread mills and custom cutters as well as ordinary square end and ball cutters. It's all good fun.

Cheers
Roger

3d milling is the hardest to do fast and precise. lathe doesnt compare. some of my programs have hundreds of segments per inch. its taxing on even the highest end industrial machines if you want to plow through it at 500+ipm.

also, when we talk about block processing, this has nothing to do with step generation. you can have a billion steps per second output and it doesnt do anything for you if the control system only processes 10 lines a second.

this is where new modern compressors come in handy. the type that make mathematical curves for your points to reduce the computational overhead - basically reconstructing your cad data inside the machine. (this is different than the control READING splines, which some can).

heres a cool test on brother speedio. this is one of the fastest machines on the market, but even it has a hard time keeping up at 500ipm in 3d contouring. it uses several user selectable strategies depending on weather you want to compromise speed or precision or have a blend of both..