#emc-devel | Logs for 2009-10-14

[06:24:49] <micges_work> hello
[06:25:07] <alex_joni> hi micges_work
[13:09:06] <CIA-82> EMC: 03jepler 07master * rd082eb048399 10/src/emc/kinematics/rotatekins.c: Fix forward kinematics calculations (SF#2870683)
[13:13:21] <CIA-82> EMC: 03jepler 07v2_3_branch * r490125ea82e1 10/ (debian/changelog src/emc/kinematics/rotatekins.c): Fix forward kinematics calculations (SF#2870683)
[14:27:21] <CIA-82> EMC: 03jepler 07v2_3_branch * r76f31443c2a6 10/ (debian/changelog src/emc/usr_intf/stepconf/stepconf.py): save direction setup value when leaving page (SF#2870131)
[14:41:29] <CIA-82> EMC: 03jepler 07master * r3b1d663dd782 10/src/emc/usr_intf/stepconf/stepconf.py: save direction setup value when leaving page (SF#2870131)
[18:04:20] <cradek> I'm dealing with setting up a probe that requires you to turn it on before using it. Ideally you also turn it back off afterward. I wonder what the most generic hal-ish way to support this kind of probe would be.
[18:11:47] <skunkworks> jr behaving?
[18:16:25] <cradek> yep, more or less. having big trouble getting spindle feedback, so I'm moving on to the probe stuff, which will be really fun
[18:18:54] <skunkworks> belt connection not helping? or something else
[18:19:22] <cradek> the resolver on there is 50 pole or some such ridiculous thing, and needs 50-60v to drive it, so it won't work with jon's board
[18:19:33] <skunkworks> oh - yeck
[18:19:55] <cradek> so I thought I'd put a kit encoder on there, but they have a 100kHz count rate limit (??!?)
[18:19:56] <skunkworks> isn't required though now - right - so you could replace it? or is it pretty deep in?
[18:20:08] <cradek> yeah I actually had it apart once
[18:20:40] <cradek> so I have to choose between smooth tapping motion (500 rpm) or still counting at full speed (8000 rpm)
[18:20:59] <skunkworks> is the 100khz limit mesa?
[18:21:08] <cradek> no, the us digital encoders
[18:21:14] <skunkworks> oh - duh
[18:21:19] <cradek> the mesa can count at several to many MHz I'm pretty sure
[18:21:23] <skunkworks> I thought so.
[18:21:57] <cradek> I'd like to have a 1000-2000 line encoder which would give me 1MHzish at full speed
[18:22:22] <cradek> it has to fit in a 3" dia x 1.5" (?) tall spot around a 20mm shaft
[18:22:36] <cradek> all these requirements add up to no solutions that I can see
[18:22:46] <skunkworks> heh
[18:23:06] <cradek> I'd make a new housing, but the spindle cooling fan is sort of part of it, and I wouldn't want to mess up the airflow
[18:24:26] <skunkworks> no way to attach it to the actual spindle?
[18:25:32] <cradek> can't go above (drawbar plunger)
[18:26:03] <cradek> can't practically do 1:1 off the side - spindle is about 10" around
[18:26:12] <skunkworks> heh
[18:40:20] <cradek> hm, maybe I could do 1:1 off the 20mm shaft to avoid having the extra height
[18:40:35] <cradek> I have some extra resolvers around that are pretty small
[18:41:02] <cradek> I think they have NO problem with the high speed
[18:44:59] <robh> cradek, u need something like these http://www.motec.co.uk/encoders_lb-minicoders_247.htm motor encoders
[18:46:14] <jepler> "frequency range: 0..25kHz" and they look very low counts/revolution as well
[18:46:31] <robh> pritty exspenive tho whats they put in most new spindle motors for feedback i looked into them when i was doing out mill
[18:47:10] <robh> they do other ranges also low count as im guessing spindle drives cant count that fast
[18:48:11] <jepler> cradek: fwiw I notice that demo_mazak has 1440 counts per revolution
[19:47:36] <SWPadnos> a 1000 line (= 4000 count) encoder at 8000 RPM would be ~130 kHz
[19:48:07] <SWPadnos> I'd bet that a 360 line (=1440 count, like the Mazak) would be fine, and that would be well within the encoder speed
[19:48:40] <SWPadnos> they generally specify cycle count as the bandwidth of each phase, not the x4 count rate we get with quadrature
[20:17:05] <jepler> SWPadnos: I think cradek's concern is as follows: consider a 500cycle/rev encoder (2000 counts/rev) threading at 500RPM. If it's perfectly at speed that's 16.666.. counts/ms. If in one cycle you get 16 counts and in the next cycle you get 17 counts, that's a 6% change in speed in 1ms (likely followed by a -6% change the next ms)
[20:17:16] <jepler> suddenly your whole shop hums at 500Hz and the machine falls apart
[20:17:51] <SWPadnos> clearly he should be using interpolated-position for threading ;)
[20:18:01] <jepler> hostmot2 doesn't have that
[20:18:03] <SWPadnos> (unless he's doing rigid tapping)
[20:18:13] <jepler> and ye, this is for rigid tappint
[20:18:14] <SWPadnos> oh, then Seb should get to work
[20:18:20] <jepler> s/int/ing/
[20:19:45] <SWPadnos> I wonder if there's a way to do some smart averaging, based on the speed and resolution
[20:20:27] <SWPadnos> and some speed tolerance
[20:20:36] <SWPadnos> and the servo cycle period