more Re: tool comp

Hoyt,

Another source for the type of info you are looking for is OLD cnc
machine programming manuals. Many had detailed info about how the
cutter comp was handled. In short, to achieve cutter radius
compensation you need to look ahead at least TWO blocks. Many of the
old controls looked ahead five blocks in case there were "non-XY
motion" blocks interspersed, which would have otherwise broken the
cutcomp calcs. One thing that is easy to forget is the velocity
changes necessary to keep the feedrate consistent with the
programmed "f" word, when performing cutcomped moves.

As for EMC, the cutcomp is in the 274NGC interpreter section. This is
the part which "reads" the gcode and outputs canonical commands to
the rest of the system. The canonical commands will be easily
translated into IX-LPT "motor" calls... There is a lot of
interaction going on between other parts of the EMC system, but for
your usage, much of it can be ignored. (You will want to understand
these other modules later as you begin to implement CVV/ trajectory
planning) Trust me, you WILL get there eventually<G>

Hope this helps,

Ballendo

P.S. Looking for IJK to discern arc when you have a "2" or "3" (from
G02, or G03 seems like backtracking to me. The structure of your
array should allow for these modals to be "kept" until replaced. IMO,
this is a good place for a UDT. This UDT can then be used in an array
of sufficient depth to accomodate any lookahead needed.

--- In CAD_CAM_EDM_DRO@y..., batwings@i... wrote:
> At 10:35 AM 4/2/02 -0800, you wrote:
> >Try EMC's source code. I think it might have it. Yes, I'd like to
> >study some examples of tool comp, later...
>
> I looked in past and never figured it out. I'll try again. If
anyone has an
> idea of what module it's in I'd appreciate a shout. I can tell a
priori, it
> will require look-ahead. I suppose one would read in several lines
of code
> into an array on X Y Z I J K, check by presence of the IJK whether
an arc
> is involved or a linear move, check what quad is involved by slope
etc,
> figure up offsets on toolpath in XY or add/subtract radius for arcs,
> compute new ending points for move, send command according to new
coords
> ... go to next line. There's always an economical way of laying
this out so
> you don't crunch more than you need.
>
> >What are "polar canned" routines"?
>
> Things like bolt circle and spiral drilling patterns, polygon,
helix of
> course, lately I've used polar routine to do spiral pattern outfeed
for
> circ-pocket also.
>
> >Rectangular to polar transforms?
> >Liner and/or circular interpolation?
>
> It iterates its way around the circle, doing the right things at
each
> point. For bolt circle it drills at each station, for spiral it
adds a
> radial increment before computing each station and drilling, for
helix it
> generates arcs and down-steps, for polygon it feeds to depth and
then feeds
> between stations, etc. All of that requires sin/cos or rect-to-polar
> transforms. I use same kernel to compute for all of above, just
flag in/out
> of different parts of the prelim data handling and same on output,
to make
> the right thing happen.
>
> >Check out TurboCNC list.
>
> You have subscription info?
>
> Regards, Hoyt
>
> Belfab CNC - http://www.freeyellow.com/members/belfab/belfab.html
> Best MC Repair -
http://www.freeyellow.com/members/batwings/best.html
> Camping/Caving -
http://www.freeyellow.com/members/batwings/caving.html
> Hamas vows revenge against perpetrators of the USS Stark incident