[Global Notice] Hi all. It seems one of our sponsors has connectivity problems, causing around 6,000 users to disconnect. We're investigating, sorry for the noise and thanks for using freenode!
steves_logging is now known as steve_stallings
It looks like Art Fenerty is at it again with motion planning. He has mentioned a new set of test code that he calls Tempest that is 4th order or more and limits jerk.
He references the University of New Brunswick and I suspect that he is working from the Sonja Macfarlane thesis paper here:
[14:51:23] <steve_stallings> http://www.mech.ubc.ca/~caris/Publications/sonja.pdf
I know the thesis paper is old news, but Art's work is the first real world application that I have heard about.
Discussion is occuring both on the Mach list on Yahoo and on the forum here:
[14:53:02] <steve_stallings> http://www.machsupport.com/forum/index.php/topic,12042.0/prev_next,next.html#new
* skunkworks_ points and gigles at cradek
what did I do this time?
I think you are the only one that would even attempt this at this time. (and I don't think it is even on your priority list)
nope, not me, I did the last one and look where it got us
it got us a working planner :)
I played with sonja's algorithm a bit and got some proof of concept plots but I never ironed out all the necessary goo.
I don't think it is possible anyways..
* skunkworks_ waits for jepler to impliment it.
dude, you're not supposed to say that
cradek: whats left on jr? other than guards :)
yeah, paint/body work
spindle feedback is half done
probe stuff is not started
how are you shoe horning that in?
it will be easy, I still have lots of room
for the third opto22 board, I made a hinged panel that swings out from the cpu box
are you using the existing resolver for the spindle? or installing something different?
it's on the motor - I will have to measure the pulley ratio and hope for the best
chuck an encoder in the spindle and count the 2 for a few thousand rotations.. :)
it has an orient pin so I'll just orient, note encoder counts, turn 100 times and orient, note encoder counts again, then get a high school student to help me with the math
your idea is better, if I had one I would do that
what we need is a university excited about emc - then someone could impliment the new planner for a thesis..
hm I just realized that probe compensation is easy - have a var value per axis that you add to the probe result
I think that's all it takes
you just calibrate by measuring how far the ball is from the centerline
which is just bore a hole and probe it I think
the probe move would always have to be parallel with the axis?
skunkworks_: good question
steve_stallings is now known as steves_logging
skunkworks_: that seldom works
skunkworks_: I think it doesn't matter which direction you move.
hm, but it's probably disconcerting if you are at Y0 and probe the left edge of something (probe X positive motion) and get a touch point of X1.234 Y.00123
if you probe a hole center, you definitely want that Y offset to be reported
yeah I'm back to thinking it's easy - if probing the left edge you only care about the X value anyway.
EMC: 03micges 07master * rc5a363d042cf 10/src/emc/ (5 files in 3 dirs): Remove one more obsolete AXIS_STAT varable
jmkasunich_ is now known as jmkasunich