Re: [CAD_CAM_EDM_DRO] High Speed Machining Look ahead EMC

Leslie M. Watts wrote:

>Does EMC have look ahead? Well, it's supposed to. It has a cubic
>interpolator which should require a 4 position look ahead. Paul
>Corner and I had a look at the source code and I can't really say
>I know exactly what it is doing. There are many different ways to
>interpolate and it may be optimized for relatively low speeds.
>
>

The program is supposed to reduce feedrates in the trajectory planning
stage, and is supposed
to have a horizon of 200 blocks of G-code, although the complexity of
the blocks
may cause the horizon to become a little shorter under some conditions.
What it does
is reduce the feedrate in advance of a move where a sharp change in
direction is
required, so that it can make the move without exceeding the
acceleration of the machine.

EMC cannot produce precisely sharp corners when the blend mode (G64) is
on. It begins
interpolating in the next move when the current move reaches the
deceleration point.
If you use G61 (exact stop) then it will come to a full stop on each
move before starting
the next one.

>
>This I think is a trajectory planning issue, not a servo loop one.
>Note that EMC does circular motion G2/G3 commands totally smoothly
>at any speed. Trajectory planning can be causal however; since the
>machine code is known it can be done independently of servo calcs,
>and perhaps is.
>
>For the time being I will watch the machine sometimes stutter around smooth
>corners sounding like a machine gun even though it could be perfectly
>smooth. The machine can do it. It probably puts unnecessary
>wear on some of the components though.
>
>

These smooth corners are a string of G01 moves, not a circular arc (G02
or G03) move?
It really shouldn't stutter even on a string of G01 moves, though. I
don't see that on my
servo system. If the moves are insanely short, however, like .0001"
then it may have
so many blocks to process before a decent move that the progress through
the blocks
is slower than the mechanical movement. That is essentially bad G-code
produced by
the CAM package. I have a freebie engraving package that makes rounder
letters like
that, and you have to tune the step resolution to make it perform properly.

>Although accused sometimes of writing c code I am mostly a math guy
>and could easily generate some equations for what I have mentioned.
>It's not complicated. I'm sure others on the list could do this as well.
>I do need to know what's going on in the interpolator source code though.
>
>

Well, how about writing a C program that accepts strings of G01 moves
and fits circular
arcs to them where possible, reducing an arbitrary number of G01 moves
to a single arc
move. As long as n points lie on a circle within a settable tolerance,
then if the next point
also lies on that circle, just keep moving the endpoint of the arc move
to that point.
This would be a really neat program to have for this kind of cleanup.

(Sorry you couldn't make NAMES, but there's always next year! Hope your
recovery
is going well.)

Jon