High speed machining

I am looking to retrofit/build new machines and i am trying to decide either to use a normal cnc control ie fanuc,seimens,yasnac,mitsubishi I have not have good success getting mach 3 to run fast enough
I am looking for cutting speeds of 300-500 IPM with max feed rates 1200-2000 and all of this needs to be short lines 3d moves can linux cnc come close to achieving what I need?
I have a new okuma with the OSP control and I have used feed rates upto 800 IPM and its smooth as silk thats the kinda results i am looking for.


here is a sample of the code I am running
even a haas can run this code ( a haas can only run up to about 60 IPM before is starts getting jerky)

Hi

How fast a machine can run is completely down to the machine itself.

The humblest desktop can output instructions far faster than any mechanical machine can execute them.

If it is the actual code that causes the problem and your code is CAM generated and made up of thousands of short lines, the new trajectory planner may greatly
improve the performance.
It blends the path instead of treating each as an individual move, with acceleration followed almost immediately by decceleration to almost stop, then the next move etc.
www.linuxcnc.org/index.php/english/forum...stersprograms-wanted

Hopefully skunkworks or someone else on that thread may have a machine set up that can run your code and advise how it runs

regards

Just looking at the code, it isn't ever going to go 300-400 ipm because the feed rate is programmed to a max of F100, with lots of other F speeds set (some as low as F7).

If the code were programed to go at 300-400 ipm and if the machine were capable of the necessary accelerations to achieve those speeds in the programed movements, then Linuxcnc should be able to do it.

Other than the code slowing its self down I don't see any reason it shouldn't run. (it seems to run fine in simulation)

It looks like your CAM program is trying to be too clever, and program in decelerations to keep the code within the machines capabilities. That isn't necessary in Linuxcnc. Linuxcnc's trajectory planner is smart enough to limit speeds to the machine's max accelerations (provided you have correctly configured your machine in Linuxcnc).

On my test it runs in 26 seconds and maximum feed rate seems to be 100 IPM

if I limit the accel to 1G (but dont limit the feed rates) I get 8 seconds
(not that many machines are comfortable with 1G acceleration)

PCW thank you for that bit of info here is the same program as It would run on the Okuma and remember this is a 1/8 4f carb em running dry insteel about 30 hrc

I do run tools at cutting feeds of over 300 IPM the first program wast just all the haas can do (it cant run smoothly over 60 IPM its a haas)

Todd I am programming with solidcam I Machining thats why all the fed rates it maintains chip load on the tool (its agressive and it works)

I guess what i am asking is can linux run the code maintain its speed without getting jerky like a haas or mach3 does ( I am not even considering steppers )

so long as your using servo drives with a decent interface i.e mesa or ppc , i dont see a problem , and that the servo's are tuned etc
as Todd says , it's not Linuxcnc that is the problem , it's how you setup and interface it , and don't take cam programs defaults either , truly set it up to suit the machine

their are many machines running linuxcnc professionally in a production environment , at the end of the day it's how it's setup and the care taken to do so .

It will be as smooth as any machine with no jerk limiting can be. The biggest problem with Linuxcnc's motion system is that it only does linear acceleration.

I am not familiar enough with the controls in Haas or Mach3 to know what the limiting factors are or what the source of their jerkiness is for you. If it is line to line reading, Linuxcnc with the new trajectory planner should be better. If it is step generation limitations or max feed rates, Linuxcnc with hardware step generation or analog servo should be better. If the limitation is linear acceleration/infinite jerk, it won't be better. Infact if infinite jerk is the problem the old tool planner might give better results than the new planner (if line to line reading isn't to much of a limiting factor)
.

This brings up a question I have had for awhile now, are we talking about closeing the feedback loop through Lcnc or thru just the servo and drive? I was wondered if lets say I retrofitted a HAAS for example, could I more-or-less get the same preformance by closing the feedback loop thru the yaskawa drives (step/dir input)? Then I would not be nearly as dependant on PC latency and jitter? Or do I have than backwards?

If the drives handle the velocity part of the loop, linuxcnc's position loop rates or say 1 to 4 KHz
are fine for high performance.

At some point the chord error (from discrete velocity steps at the servo loop rate)
becomes significant but thats only for rather extreme accelerations.

As Todd mentioned, lack of a low jerk trajectory planner is probably the main limitation

There is ongoing work on a limited jerk TP however so this limitation may eventually be removed

Servo loop jitter is not terribly significant

OK I am using nema 23 steppers direct drive to a funky mm pitch ball screw its about a 5 turn per inch i max out and reliable 60 ipm iam usung pico systems USC with cheap encoders it will run the same code i run in a haas faster and smoother its still not smooth enough but its pretty good i am at 85 ipm acceleration i can get about 100 ipm but with all the weight on my table from fixture i would just tear my machine up and wont really have any time savings