Back
[00:28:06] <KGB-linuxcnc> 03Robert Ellenberg 05feature/spindle-sync-overhaul-2.7 6b358c9 06linuxcnc New branch with 21 commits pushed, 1017 files changed, 03643(+), 04317(-) since 2.7/332beb8
[07:29:25] <andypugh> I found out yesterday that the remap G71 sometimes double-cuts the same pass.
[07:29:53] <andypugh> This might be an interesting problem to solve.
[07:39:10] <archivist> is that an artifact of -0 being different to +0
[07:57:47] <skunkworks_> andypugh,
http://electronicsam.com/images/KandT/testing/Screenshot-36.png
[07:59:58] <andypugh> One of those is hard to explain, the others are difficult to avoid
[08:00:24] <skunkworks_> heh
[08:01:06] <skunkworks_> I assume it is because you don't look back far enough to see if a previous segment will intersect? Similar to cutter comp only seeing 1 or 2 segments ahead.
[08:02:31] <andypugh> The retract is a straight-line cut to where the retract-line touches the profile. So the retract end-point will be on the profile, but the retract itself might cut through the profile.
[08:03:13] <andypugh> However, the one in the right-middle doesn’t even fulfil that brief.
[08:03:31] <andypugh> (This is a _roughing_ cycle you know!)
[08:04:09] <andypugh> middle-right is what all retracts look like in a normal Type-1
[08:18:20] <andypugh> skunkworks_: OK, that funny cut is because the retracts are limited to 45 degrees to avoid this:
https://ibin.co/350zxg9Rp0js.png
[08:55:52] <skunkworks_> ah
[08:57:47] <skunkworks_> andypugh, sorry if I am just dense..
http://electronicsam.com/images/KandT/testing/Screenshot-37.png
[08:57:52] <andypugh> I wonder how I can pesuade the code to tell the difference between the two situations?
[08:58:15] <skunkworks_> notice my G72 line - why isn't it offsetting both ways?
[08:58:35] <andypugh> Thats a J offset on a profile where J is inappropriate for G72
[08:58:52] <andypugh> Because J doesn’t offset both ways.
[09:00:35] <skunkworks_> ah
[09:02:50] <andypugh> To offset both ways with J the code would basically need to re-create all the arcs, and would have to work out which direction each segment is pointing. That’s beyond the scope of the cycle.
[09:03:14] <andypugh> If you were actually machining that part then J0 L10 would work fine.
[09:04:24] <skunkworks_> then there is this
http://electronicsam.com/images/KandT/testing/Screenshot-39.png
[09:04:33] <skunkworks_> otherwise the profile would work
[09:23:18] <andypugh> The reason for the previous problem, incidentally, is a flaw in your profile.
[09:24:17] <andypugh> Zoom _right_ in on the corner at X0.6487 Z-1.0266
[09:29:41] <andypugh> Have you also noticed that your profile ends with 20 moves of about 0.0003”
[10:27:03] <Tom_itx> zlog
[10:27:54] <Tom_itx> ok new server up
[14:06:36] <andypugh> skunkworks_:
https://ibin.co/352YOxzg2ieX.png
[14:25:15] <skunkworks_> andypugh, awesome! did you change the profile or fix the code?
[14:28:41] <andypugh> Both
[14:29:10] <andypugh> But the fixed code probably works with the unfixed profile.
[14:32:01] <skunkworks_> neat!
[14:49:49] <skunkworks_> zlog
[15:08:57] <KGB-linuxcnc> 03andypugh 05andypugh/g71type2remap b445d61 06linuxcnc 10configs/sim/axis/g71/python/remap.py G71 / G72: Use curve tangent for retract path * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=b445d61
[15:08:57] <KGB-linuxcnc> 03andypugh 05andypugh/g71type2remap 69b4ab3 06linuxcnc 10configs/sim/axis/g71/lathe.tbl 10configs/sim/axis/g71/python/remap.py 10nc_files/G71test1_L2R.ngc 10nc_files/lathe_pawn_G71.ngc G71 / G72: Use a more intelligent retract strategy * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=69b4ab3
[15:10:24] <andypugh> bpuk: I _think_ that is it now. Unless TypeII retracts use the R-word to pull back out at the end of the ramp.
[15:17:04] <skunkworks_> andypugh, did you say why this was happening at the beginning of the shape?
[15:17:06] <skunkworks_> http://electronicsam.com/images/KandT/testing/Screenshot-39.png
[15:19:09] <bpuk> andypugh: awesome
[15:19:30] <bpuk> andypugh: my time to work on it at the minute is a bit limited - the next week won't be much better. Soon!
[15:24:08] <andypugh> skunkworks_: That should be fixed now, But it comes down the profile being impossible to machine from left to right.
[15:24:15] <skunkworks_> ah
[15:24:29] <skunkworks_> ok (I had just pulled your latest)
[15:24:34] <skunkworks_> I will play around with it
[15:25:44] <KGB-linuxcnc> 03andypugh 05andypugh/g71type2remap f8dc860 06linuxcnc 10configs/sim/axis/g71/python/remap.py G71 / G72: Avoid dividing by zero * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=f8dc860
[15:52:54] <KGB-linuxcnc> 03Sebastian Kuzminsky 052.7 f4c9802 06linuxcnc 10docs/man/man1/mb2hal.1 docs: fix cut-n-paste bug in mb2hal manpage * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=f4c9802
[21:31:35] <seb_kuzminsky> solvespace looks like a promising new foss cad package
[21:32:01] <seb_kuzminsky> https://github.com/solvespace/solvespace
[21:32:22] <seb_kuzminsky> it's in debian stretch but not jessie i think
