Here is a much simplified test program using the values output from my
routines and test.
If someone can help me figure how to make the tool follow the path shown
in the preview, I would very much appreciate that !
I tried on the actual mill, with G61 enabled causes lots of stop and go
but the trajectory is still completely wrong and goes way off to the
side. Same as on the sim.
Bruno
;;;;;;;;;;;;;;;;;;;;;;;
;
#<mill_number> = 5
( some init )
G21 (Unit in mm)
G90 (Absolute distance mode)
;G64 P0.005 (Exact Path 0.001 tol.)
G61
G17
G40 (Cancel diameter comp.)
G49 (Cancel length comp.)
G54
G0 X0 Y0 Z0
( select endmill )
T[#<mill_number> ] M6
G43 ( enable tool length compensation )
G0 Z5
F300
;;;;;;;;;;;;;;;;;;;;;
G0 X[[0.000000 - 0.000000] * 0.943750 + 0.000000] Y[[40.000000 -
0.000000] * 0.943750 + 0.000000]
G5.2 X[[40.000000 - 0.000000] * 0.975000 + 0.000000] Y[[40.000000 -
0.000000] * 0.975000 + 0.000000] P[356.124661] L3
X[[40.000000 - 0.000000] * 0.975000 + 0.000000] Y[[0.000000 -
0.000000] * 0.975000 + 0.000000] P[356.124661]
G5.3
M2
On 5/30/14 8:03 PM, [email protected] wrote:
> ------------------------------
>
> Message: 2
> Date: Fri, 30 May 2014 17:29:28 +0200
> From: bruno<[email protected]>
> Subject: [Emc-developers] bug with nurbs ?
> To:[email protected]
> Message-ID:<[email protected]>
> Content-Type: text/plain; charset=ISO-8859-1; format=flowed
>
> Hello,
>
> I wrote a quadrangle pocketing routine in G-Code which uses nurbs
> (G5.2). The motivation for this is that I dislike the current ngcgui
> routine (qpocket) for a couple reason. When qpocket cuts from one quad
> to another, "spiraling out", it actually cuts a scaled quad, then adds
> one edge where it slowly cuts toward the next scaled quad. That cut is
> really slow and is superfluous if the cut was really following a quad
> spiral instead. Second, and most important for me, the cut fully digs in
> on each corner (depending on the angle of the corner of course, but
> generally it is pretty big).
>
> So the width of cut goes from the overlap set for each pass, to almost
> the full endmill width. Maybe not important on a big mill, but I have a
> mini-mill and I like to try to cut at full depth if possible and with
> very small overlap, on many fast passes, it gives (I think ) better
> finish and wears the mill more evenly. (kind of like with adaptative
> clearing). So the way corners are cut is really a problem with qpocket.
>
> Here is the fruit of my experiments using nurbs, pasted below. The nurb
> weight are scaled with an exponential as the path comes closer to the
> final pocket shape, to (un)round the corners progressively. The function
> is parametrable.
>
> The preview looks great, however, when running it, I get a path that
> does not follow the previewed path ! And I see no reason for the massive
> divergence from the programmed path. See a screenshot here:
>
> http://imagebin.ca/v/1OCw5gNzXQBF
>
> http://ibin.co/1OCw5gNzXQBF
>
> notice how the purple path has "wobbles" in some sections, which go way
> aside the previewed path.
>
> I tested with the sim on linuxcnc 2.5.3
>
> At first sight it really seems to be a bug as the preview looks correct.
> Could anyone test this on their setup ? Is there an explanation for this ?
>
> Thanks !
> Bruno
>
------------------------------------------------------------------------------
Time is money. Stop wasting it! Get your web API in 5 minutes.
www.restlet.com/download
http://p.sf.net/sfu/restlet
_______________________________________________
Emc-developers mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/emc-developers
