#emc-devel | Logs for 2009-09-30

[06:50:51] <alex_joni> BJT_Shop: congrats ;)
[06:50:52] <alex_joni> https://www.ohloh.net/p/emc/contributors
[07:02:35] <micges_work> hi alex_joni
[07:26:47] <alex_joni> hi micges_work
[12:28:31] <jepler_> jepler_ is now known as jepler
[17:51:18] <micges> hi seb
[17:53:00] <micges> seb_kuzminsky: take a look at this: http://imagebin.ca/view/BpN9pD0.html
[17:54:56] <cradek> ooh, inf velocity for one cycle?
[17:55:20] <cradek> you should zoom in and see what counts did (20K/div is not possible to see)
[17:57:31] <jepler> is there only one realtime thread in this system?
[17:57:44] <micges> yes
[17:58:06] <micges> I have this too: http://imagebin.ca/view/g7PEZeW.html
[17:59:55] <micges> this is simple servo machine at servo thread on (1khz)
[18:05:07] <micges> do we have docs about halscope?
[18:06:38] <jepler> http://linuxcnc.org/docs/2.3/html/hal_tutorial.html#sec:Tutorial%20-%20Halscope
[18:06:56] <micges> thanks
[19:53:57] <jepler> yuck, position-interpolated even has reversals when the incoming pulses are irregularly spaced
[19:54:26] <jepler> http://emergent.unpy.net/files/sandbox/uneven-pulses-make-encoder-cry.png
[19:54:51] <jepler> 'stool' makes pulses that are in a regular, but not equally spaced, pattern
[19:55:56] <cradek> bleh
[19:59:02] <jepler> maybe I picked bad parameters for the wobble by luck; others I try are less overtly bad
[20:01:09] <jepler> and others are worse: http://emergent.unpy.net/files/sandbox/uneven-pulses-make-encoder-cry2.png
[20:01:37] <jepler> (I probably fiddled the scales since the other one)
[20:01:54] <jepler> (no, velocity isn't crossing over zero)
[20:05:39] <SWPadnos> does stool generate bursts of pulses with gaps between bursts?
[20:07:55] <skunkworks_> I love how these things get named.
[20:08:21] <SWPadnos> yeah - I was wondering if it was meant to be like "shit" or "some tool"
[20:09:45] <skunkworks_> we have a database application called 'lookey' ;)
[20:22:43] <jepler> actually, I was thinking about the minimum number of legs in a useful stool
[20:23:33] <jepler> stool makes a repeating cycle of 2 pulses. the two tunable parameters are the length of the cycle and the distance between the two pulses in the one cycle
[20:24:04] <jepler> so if cycle=5, wobble=2, then the output is !.!.. repeating over and over again
[20:29:56] <jepler> those pics are with generally more than 1 pulse per base period
[20:31:29] <jepler> the situation I think we're discussing is with less than 1 pulse per base period. Here are two with less than 1 pulse per period on average: http://emergent.unpy.net/files/sandbox/uneven-pulses-make-encoder-cry3.png http://emergent.unpy.net/files/sandbox/uneven-pulses-make-encoder-cry4.png
[20:37:51] <jepler> oh, and showing that interpolated- leads position by up to 1 count (that we knew): http://emergent.unpy.net/files/sandbox/uneven-pulses-make-encoder-cry5.png
[20:40:29] <jepler> but this plot shows that it leads by more than one count (which I thought was contrary to the implementation): http://emergent.unpy.net/files/sandbox/uneven-pulses-make-encoder-cry6.png
[20:43:35] <jepler> if two pulses arrived 980us ago and 960us ago, I think the speed estimate is based on a pulse interval of 20us, even though we have more information than that
[20:43:55] <jepler> information that makes a 20us pulse interval look quite unlikely
[20:49:08] <jepler> one last plot -- this is the worst I've been able to manufacture so far. http://emergent.unpy.net/files/sandbox/uneven-pulses-make-encoder-cry7.png
[20:50:35] <jepler> the closer wobble is to 2 the worse the error in position and velocity estimate (the "fast" pulse is fastest with wobble=2)
[21:36:01] <BJT_Shop> alex_joni: thanks
[22:31:00] <Guest267> Guest267 is now known as skunkworks