[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: librecad

On 07.12.18 09:17, John Hasler wrote:
> Gene writes:
> > Thats a huge part of the problem, but theres another fence to
> > jump. most of these so-called cad programs cannot generate even the
> > most basic gcode.
> I can see not wanting to learn even a small part of a CAD program if all
> you want to do is draw a simple floor plan and then never use it again,
> but doing CNC without CAD is baffling.
> Freecad can produce gcode which you can inspect and modify.  The
> post-processor can be customized for your particular machine.
> Solvespace can also output gcode.  I have no idea how good the gcode
> produced by either of these programs is because I have no way to use it.

Gene would have no trouble tweaking that if it was in the ballpark,
after his years of writing the stuff from scratch. (Although
auto-generated gcode will be a bit primitive, with a lot of data points,
rather than a few optimised functions.)

While I'm the one who did the 8 house plan drawings in Postscript, I
wouldn't advocate that path for CAD for CNC. OK, Postscript can "print"
text to stdout in addition to its drawing operations, so it could spit
out gcode equivalent to what is drawn for inspection, _but_ you'd have
to hand knit the gcode to be generated by each of your hand knitted
drawing functions. After a couple of months of development you'd have a
nifty library available for re-use, but it's quite a hike to get there.

> The major feature of these CAD programs is the constraint solver, which
> understands geometry and does a huge amount of work for you.  You can
> draw a part with dozens of equally spaced holes, change the length of
> the part, and the holes will move appropriately.  Draw an angle bracket
> with a brace, change the angle, and either the bracket length or
> position will adjust appropriately depending on how you constrained it.
> Draw a gear specifying pitch and tooth count and the diameter will be
> computed. Change the pitch or tooth count and the diameter will
> change. The diameter will only change by a commensurate amount: no need
> for you to calculate allowable dimensions (unless you want
> fractional-tooth gears).

If moved to try that in Postscript, you'd have to program the
computations, and therefore decide in advance what would be
auto-adjusted, e.g. input gear pitch and tooth count, then diameter will
result. (While tolerances may appear on a drawing, gcode specifies a
precise toolpath, and it's up to the machine to do its best to follow
that. LinuxCNC will issue a "following error" and stop if it can't meet
run-time tolerance specs.)

> You can also do assemblies and specify constraints between assemblies so
> that a change in one part will cause appropriate changes to others.
> Interferences that would result in two objects occupying the same space
> won't happen.  You can animate assemblies and watch the machine run.  No
> more making cardboard models of linkages and then having to stick to the
> tested ratios.

Hmmm ... D for design, not just drawing. Powerfully tempting, if I ever
have to do more than shuffle bits of house plans around on a sheet.
(There the human utility, convenience, and comfort considerations, and
even building regulations aren't amenable to automation, so drawing
suffices, I found.)

Sounds like Gene's bitten the bullet and done the job with what he has,
by the sound of it. That's never a wrong move when it works.

(Who's gone owner-builder (swore I wouldn't do that again), and is
currently too flat out to look at Freecad.)

Reply to: