Re: [CAD_CAM_EDM_DRO] High Speed Machining Look ahead EMC

Jon, Keith, Fred, Byron,

I wanted to study this a little before I answered.

I think it really gets back to the old problem of cubic interpolation
that we have discussed much in the past.

Smarter programs like Fred S. mentioned can help, but I think
the real issue is the discontinuous acceleration of the cubic.

It's a lot better than nothing for sure- but it becomes a big
issue for those cutting softer materials at higher speeds.

EMC is just doing what it's told as many even very high end
cam programs are going to generate a lot of G1 moves for
curves. An optimizer (circular interpolator) would help loads, but
it still will have hard corners- just fewer of them. Any corner
at any angle (even shallow ones) will cause a command to
go to MAX_ ACCELERATION... Paul and I saw this clearly
enough on the scope. Only the pulse width varies. And that's
as it should be... the program is just doing what is commanded.

For me typical speeds one might use to cut metal on a vertical
mill result in very smooth motion. Here's the problem- the
v^2/r accel requirements for a corner. At 10 in/min things are
nice and smooth, but go to 100 in/min with the same corner radius
and the accel at corners goes up 100 times!! Go to 200 and it will
be 400 times! A lot of us have machines that can do this... after all
many can create forces of 1000 lb+. But I for one would rather
not see these at 20 Hz on what is just otherwise a lazy smooth
path curve.

That I think is what we are seeing, and with cubic we are going
to have a MAX_ACCELERATION pulse on every hard corner.
If it happens to be a rapid sequence of corners that coincide with
a system pole things can get downright hurky jerky. Setting chordal
tolerance to a low value (as we must for smooth finishes) tends to do
just that.

I'm not one that really wants to add to EMC bloat, but I really REALLY
think we need to get quintic interpolation implemented. We have talked
about it for years. Fred P. wrote something a long time ago but it must
have had a problem.

As the user base has increased dramatically I think this is the one thing
that would most improve the most installations, both stepper and servo.

I sat and stared at cubic.c and cubic.h for a while. It appears to be a
Hermite
interpolation around two points with given values and derivatives. It of
course
needs a four point buffer. I think it is global and probably called by
emcmot.

As you know quintic will have accel continuity at corners with bounded jerk
values.

It needs a 6 point buffer.

I wrote down the equations but didn't solve the system. It needs to be
closed
form, but it is doable and probably already worked out in a gazillion books.

I note that Fred P. had some buffer initialization stuff and what looks like
some convex hull things or something. I don't understand everything going
on there.

The structures look like they could be easily expanded. There may be some
problems with the calling program- not sure.

But I sure wish we could do this somehow. It's a little disconcerting
when the silly MS spell checker wants to replace "quintic" with
"quixotic" :^) but hopefully that isn't the case!

Les

Leslie Watts
L M Watts Furniture
Tiger, Georgia USA
http://www.alltel.net/~leswatts/wattsfurniturewp.html
engineering page:
http://www.alltel.net/~leswatts/shop.html
Surplus cnc for sale:
http://www.alltel.net/~leswatts/forsale.html

----- Original Message -----
From: "Keith Rumley" <DSCadCam@...>
To: <CAD_CAM_EDM_DRO@yahoogroups.com>
Sent: Saturday, May 03, 2003 9:58 PM
Subject: Re: [CAD_CAM_EDM_DRO] High Speed Machining Look ahead EMC


> > In my case I often get quite jerky motion on paths that really could
> > be nice smooth curves.
>
> Les,
>
> Interesting that you should mention this. I've seen this, too. I just
> figured it was because my servo amps have a 3ms latency between voltage
> command and signal out. However, now that you mentioned it...
> In my case it appears after about 1k or so lines into a contouring
> program, at speeds above 40 or so IPM. It'll run great at 90 ipm for a
> minute or so (G64), then stutter to beat the band. Reducing the feedrate
> percentage seems improve it temporarily (i.e. set ipm to 100, reduce by
20%,
> stutter reduces for a bit)