[1ctoolfool]

Haas literature touts cutting feeds up to 500ipm but this seems impossible to attain accept with the simplest of toolpaths or programs designed specifically to demonstrate high speed machining, rarely applicable to "real" parts.

Realistically the upper limit for smooth finishing seems to be around 100-150 ipm in my experience.

What is the truth about the max speed attainable with finish 3D toolpaths with small straight line segments on the order of .001" - .010" line lenght?

Are there other machines/manufacturers that can attain this finish speed with a different processor or look ahead algorithm, or is this purely a function of programming?

I am talking about finish 3D contouring with like .0002" cusp height, not high speed roughing which I have been able to attain much higher feeds.

Thanks

[HuFlungDung]

"Roughing" is also machining, even though it may not be that accurate.

I'd be really amazed if a part that was finished at more than 100ipm was really right on the money, unless it was straight line cuts. My opinion is purely a function of my experience, so it could well be the case that there are better machines out there that do somehow maintain accuracy. When I want very accurate circular interpolation, I keep the feed below 50 ipm, and the rpm maxed out at 7500. Since I can obtain .0002" roundness that way, I figure that the same accuracy will carry over into real 3d paths.

But I've always said, if you rerun the finish toolpath over the part a second time, and it still takes some chip, then the first pass was inaccurate in an absolute sense.

[Geof]

...Are there other machines/manufacturers that can attain this finish speed with a different processor or look ahead algorithm, or is this purely a function of programming?...

I think it is more a function of machine stiffness; that is the mechanical stiffness of the entire structure which I think is the source of the poorsurface finish discussed in this thread.

http://www.cnczone.com/forums/showthread.php?t=36496


Also the 'electronic stiffness' of the servo control which I show in this thread.

http://www.cnczone.com/forums/showthread.php?t=22589

My example is a bit extreme because it is a large and not very stiff machine, but I have seen similar effects on a much smaller scale on a SuperMiniMill compared to a VF2. Running the same program on either of these machines gives the same ripple effect where there is a rapid change in direction but, not surprisingly, the VF2 ripple is almost imperceptible while the SMM is easily noticeable.

I think if you are driving the machine hard enough that the intrinsic flex in everything is going to show up it is not going to be easy to control it with software. Some form of dynamic acceleration control would probably work but I think this would require changes to the machine algorithms and is not possible at the G code level.

Possibly it could be done by changing the axis acceleration parameter(s). Normally you want to save time by having an axis bang up to the programmed feedrate as fast as possible but if the acceleration was reduced it may be possible to run at a higher average feed rate.

[marco antonio g]

is 1018 steel hard to machined?

[CarbideBob]

Well lets see... .001 increments at 500 IPM works out to 8333 blocks per second.
You'd need a pretty fast controller to keep up with that kind of data rate.

Of course they do make machines that will do it.

A buddy of mine has two 5 axis machines that contour within a couple of tenths at 600 IPM. Linear motors, very special controls, and just shy of 2 million bucks each

Typically these type machines will accept B-Splines for programming data.
Bob

[ckirchen]

1ctoolfool, the Haas controller with HSM can process 1000 blocks per second, which equals 60 ipm for .001" moves and 12 ipm for .0002" moves.

There's a description on the Haas site: http://www.haascnc.com/MAIN_high-prod_HSM.asp. The way my HFO explained it to me is that HSM does not make the machine go faster; instead it makes it more accurate. The gains come from the fact that you don't have to slow down the feedrate to prevent gouging. The key to the Haas HSM is that you don't make the controller to process more than 1000 blocks per second. So, to get 833 ipm, you would need .0139" moves.

You can turn off the display to squeeze a little more performance out of the machine by using M76/M77. Updating the display requires processing power, but I'm not sure if puts it over 100% or simply brings it up to 100%. Regardless, I tested it a few months back and there is a difference.

Chris Kirchen

[1ctoolfool]

Chris is absolutely right on about the block processing speed, this is the limitation in this case. There is a big difference between just getting the machine to travel at the programmed rate vs. getting the results you want at that rate. There is a lot that can be done in programming, and workflow planning, etc. to maximize productivity, but this block processing speed is a 'brick wall' that cannot be overcome so there is always the max possible feed for any given path. But this gets even more complicated when you start talking about servo tuning, communication speed, look-ahead algorithms and trajectory planning, other I.O. besides code, etc.

I also discovered some tools which will analyze your code and/or apply smoothing algorithms within a certain tolerance to imporove speed.

The answer to my question lies in this 1ms block processing speed and I got that answer today.

much appreciated.