Re: Contouring in EMC
Posted by
Jon Elson
on 2000-07-02 22:14:48 UTC
johndroc@... wrote:
faster
processors are available. I did some 3D surface cutting with a
radiused-nose
end mill. I used Bobcad/CAM, and had some serious problems with the
'winskin' function, such that is did not follow the contours I had given
as
a guide to build the surface. But, since it was mostly to provide a
rounded
shape that wouldn't hurt people that bumped into the part, it was not
critical,
and I cut it. I noticed that in places where the short vectors were
nearly
straight (very small curvature) then EMC followed at the prescribed
feedate,
but when the curvature increased, maybe a couple degrees change in
direction
between each vector and the next, then it slowed down suddenly, and
began
accelerating and decelerating for each move. The feedrate override no
longer
had much effect. Now, I'm not sure this is a problem with the G-code
interpreter.
In fact, it isn't the interpreter at all, because that is done (on files
less than
a MB or so) before you press 'run'. It could be the trajectory planner,
however. But, it looked to me like there is some parameter in the
blending
code that decides if the change in direction exceeds some threshold (or
if that
change requires acceleration greater than some threshold) then it starts
to
behave like it is in 'exact stop' mode, where one move must finish
before the
next one starts. If I slowed the feedrate to a crawl, it did smooth
out, however.
I was getting about 8 blocks a second in the 'jumpy' mode, and maybe 10
in
the smooth mode, when it was moving approximately straight.
I sent a message to Fred (or on the EMC list, actually) but haven't
heard
anything. I thought Fred was back from vacation, but maybe not yet.
And, Will was out sick with back trouble.
Anyway, this is with a 100 MHz Pentium classic CPU and servos,
using the 20-Dec-1999 version of EMC, which is apparently the last
version that worked right with servos.
Jon
> Tim,Well, this may be true, I'm not sure. Of course, if speed is a problem,
>
> I noticed that 3D surfacing was mentioned somewhere along the thread
> of this
> discussion. Have you (or has anyone else) done any 3D surfacing using
> EMC?
> My understanding, after talking to Fred, was that the G-Code
> interpreter
> wasn't fast enough to do any heavy-duty surfacing. Actually, I guess
> a
> better reference question would be what is considered heavy-duty
> surfacing?
faster
processors are available. I did some 3D surface cutting with a
radiused-nose
end mill. I used Bobcad/CAM, and had some serious problems with the
'winskin' function, such that is did not follow the contours I had given
as
a guide to build the surface. But, since it was mostly to provide a
rounded
shape that wouldn't hurt people that bumped into the part, it was not
critical,
and I cut it. I noticed that in places where the short vectors were
nearly
straight (very small curvature) then EMC followed at the prescribed
feedate,
but when the curvature increased, maybe a couple degrees change in
direction
between each vector and the next, then it slowed down suddenly, and
began
accelerating and decelerating for each move. The feedrate override no
longer
had much effect. Now, I'm not sure this is a problem with the G-code
interpreter.
In fact, it isn't the interpreter at all, because that is done (on files
less than
a MB or so) before you press 'run'. It could be the trajectory planner,
however. But, it looked to me like there is some parameter in the
blending
code that decides if the change in direction exceeds some threshold (or
if that
change requires acceleration greater than some threshold) then it starts
to
behave like it is in 'exact stop' mode, where one move must finish
before the
next one starts. If I slowed the feedrate to a crawl, it did smooth
out, however.
I was getting about 8 blocks a second in the 'jumpy' mode, and maybe 10
in
the smooth mode, when it was moving approximately straight.
I sent a message to Fred (or on the EMC list, actually) but haven't
heard
anything. I thought Fred was back from vacation, but maybe not yet.
And, Will was out sick with back trouble.
Anyway, this is with a 100 MHz Pentium classic CPU and servos,
using the 20-Dec-1999 version of EMC, which is apparently the last
version that worked right with servos.
Jon