Just this once .. here's the log: http://www.linuxcnc.org/irc/irc.freenode.net:6667/emcdevel/2009-07-27.txt
skunkworks: how's the garage?
almost to the point of siding the last 3 sides
any progress is good
I'm finally going to have a half hour or so to play with zenbot
X and Z seem to do OK with the same tuning parameters, at least they don't ferror with the wide FERROR = .050 MIN_FERROR = .010 ..
Y, the axis I'd tuned before, I have on FERROR = .005 MIN_FERROR = .001
Z had faulted before with these settings, but seems to have stopped now that I moved the current jumper from the 1A position to the 3A position
another thing I need to test: does the position feedback stay good when running the spindle
I dunno the right way to test, but I did two short tests with the spindle running: first, with servo amps off, I looked for 1-count position changes. I didn't see any.
jepler, what are you using for the servo drives?
second, with servo amps on I did a series of moves and checked whether it returned to the mark where it started. That went OK too, but the test was only dozens, not hundreds, of inches
LawrenceG: 7i30 (quad "100W" brush servo amplifier), they're small motors
cool.. if you see any more of those motors come available, please let me know!
jepler, there are slim pickin's at electronic goldmine, all electronics. All does have some nice bigger dc motors but adding encoder is non trivial
LawrenceG: unfortunately, I think skunkworks got the last of 'em
[02:07:35] <jepler> http://skycraftsurplus.com/index.asp?PageAction=VIEWPROD&ProdID=1877
is a nice servo diagram I found on the net that supports multiple servo modes
nice price... cheap encoder even if one doesnt use the motor
What's "E sub G" compensation gain?
(that's a hard term to google, all the hits are (e.g., compensation))
is it related to back emf?
like emc ff1 I think
emc uses positions as the pid inputs, while this uses velocities .. otherwise, it looks substantially similar to emc's PID
(you achieve the switches by changing hal connections or using muxes..)
emc's ff1 is a constant times commanded velocity .. at face value, this diagram shows a constant times feedback velocity
the latter is more clearly related to back emf
yes... all pid's are pretty close to the same... the interesting option is that Mode 0 alls one to directly command the power stage to different levels... this is very handy for testing hardware without pid tuning being an influence
commanded vs actual velocity.... not sure which one gives better control... if everything was perfect, they would be the same
Nice PID chart
got a question about the size of the 5i23
from a servo site, Elm servo I think... Hi Peter
(it's nice and small :) )
pcw_home, I'm going to have to design a bracket to hold the 5i23 in a riser card. making my own PC enclosure, for the most part
I have measured the hole spacing, but wouldn't mind checking it with the source :)
Looks like actual BEMF would giver truer torque mode
(not that thats a good thing)
[02:27:57] <LawrenceG> http://elm-chan.org/works/smc/report_e.html
Hole spacing should match PCI specs
I can send a drawing Monday
is there a spec for the holes for the bracket?
Not that anyone follows it
ok, no wonder :)
I think I've even seen cards that had the bracket attached solely with the connector screws
Sure, thats common with wide connectors
here's what my dual loop looks like: http://emergent.unpy.net/index.cgi-files/sandbox/emc-dual-loop.png
the hal's written, but neither loop is tuned
(my earlier report that all 3 axes were working are with single loop)
it's easy enough to hook up another signal at vel-cmd. I don't have home switches, so I plan to set something up where I can home in emc, then have fake limit switches based off joint-pos-fb (emc subtracts the joint offset off the motor position) so I estop if the motor is going where I don't want it
er, I don't have *limit* switches
does that make sense?
hmm. maybe it does
since RT/HAL can be fully functional with userspace out to lunch
it should be plenty good enough for my needs -- by no means is it a real safety measure that anyone should use in his own system..
pcw_home, ok - I see that I measured well - I got 3.36 between holes (I may have assumed it was a round number)
I see that the actual number is 3.362
I'll send you the the design if you like. it's suitable for use with any mini-ITX board, assuming you use a standoff of the appropriate length for your riser card.
SWPadnos: Sure that would be nice. There was a typo in one of the PCI specs drawings that
had 3.632 instead of 3.362, naturally thats the one I used on our first cards...
It would be good to modify the normal PID loop so it would take HostMots2s velocity estimate
instead of DV/DT, though i would expect the velocity estimate to need some debugging...
pcw_home: yeah, jmkasunich and I have talked about that too
jepler's dual-loop setup uses the encoder velocity, since it's using a separate velocity loop
two fests ago, I think
SWPadnos: yes, that's true as well
in Wichita last year, I think - over Thai or Indian food
A good thing to try is setting up a PID loop with only KD, and a small motor
and feeling for software bugs through reversals...
(by moving the shaft by hand)
Still sunny here...
"An integrated, patent pending 32-level first in, first out (FIFO) buffer can be used to store data to minimize host processor intervention"
boy, it's 2009 and you can still patent the FIFO?
you can apply at least
I was a little worried about the counts/time velocity estimator and sure enough there are patents
in the 200x range for such estimators, but theres really good prior art in a (expired) 1987 patent
it really bugs me that we have to spend the time to look for that kind of thing
Its really silly especially for those things that _are_ obvious if you look at the problem for 5 minutes
I had no idea I was infringing NCR's patent when I wrote some code - in my spare time in high school - to XOR data on/off the screen
actually, my drawing program might have been prior art. I don't remember when they got their patent
I think we got an infringement notice on that, I think they eye-grepped all the trade journals for VGA
oops. I'm confusing GIF and XOR - it wasn't NCR
Looking up the XOR patent, someone claims it killed Amiga...
heh. I just read that :)
We just circular filed the notice...
according to an email post I just read, a lot of people did that, and more or less weren't bothered afterwards
Nope, never heard from them again.
Reminds me of one of my first computer related jobs
I suggested using XOR for checking for I/O port changes.
My boss got all excited and said we should patent the method
(this was an CPD1802 sometime in the 70's)
good night all
jepler: I'd have to ask for a 2.3.4 :/
got few minutes?
[Global Notice] Hi all, we appear to be having some routing issues causing the network which services reside on to be unreachable, we're looking into it now -- please be aware that services (chanserv, nickserv etc) are unresponsive currently. Please set yourself +w for further information as we look into this. Thank you for using freenode.
I'm a bit confused: canon code uses X..W, motion is based on joints 0..n but TP is using also X..W so some info must be stored on motion about axes besides joints info
or maybe TP only use one info about each axes max vel/acc and thats all it's need?
sorry I'm baffling (coffee..)
alex_joni: already? I just did 2.3.3 yesterday!
micges: everything is in terms of axes until after the trajectory planner. the trajectory planner plans axes, otherwise a G1 move wouldn't be a straight line in space..
jepler: I know :/
I borked the cooltool configs
I removed references for emc.nml from them (like I did on master)..
is there a limit to the size of a g code file or the number of lines?
or is the size limited by the amount of memory on a computer?
EMC: 03jepler 07v2_3_branch * r6f82811c3070 10/debian/changelog: add way to get path to Makefile.modinc
EMC: 03jepler 07v2_3_branch * r81dda51b46f6 10/src/emc/usr_intf/stepconf/ (stepconf.glade stepconf.py): Make stepconf usable with 800x600 display
EMC: 03jepler 07v2_3_branch * r8c140a44ad0c 10/src/emc/usr_intf/stepconf/stepconf.py: change stepconf to close expander widget
EMC: 03jepler 07v2_3_branch * r5209bdc1df3f 10/src/emc/usr_intf/stepconf/stepconf.py: get rid of image, it forces window to be too big
EMC: 03jepler 07v2_3_branch * r1c23496e1736 10/debian/changelog: note bugfix
alex_joni: aha, and ugh
alex_joni: feel free to push your fix for that one onto v2_3_branch, since I have no way of evaluating it anyway
it was broken in 2.3.2 too, I guess
BigJohnT: the interpreter and task are carefully written so that the only limit on part program size is disk capacity. tkemc and axis both load the program text to display it to the user, though, so RAM becomes a limiting factor. In AXIS, because the preview plot is drawn by default, the redraw time also becomes a practical limit on program size.
BigJohnT: practically it is about 500k lines on P4 2.8Ghz, 1GB ram, geforce 2
but it may vary on different pc's
Hi Ho Hi Ho it's off to work I go bbl
Just this once .. here's the log: http://www.linuxcnc.org/irc/irc.freenode.net:6667/emcdevel/2009-07-27.txt
Lerman_______ is now known as Lerman
EMC: 03alex_joni 07v2_3_branch * r6f4d1165b3ba 10/configs/ (8 files in 3 dirs): add missing emc.nml reference to the inis
EMC: 03alex_joni 07v2_3_branch * rc34d6370a9f5 10/debian/changelog: note cooltool config fix
* alex_joni wonders why 3 dirs..
git says you changed some files in plasma-thc and plasma-thc-sim
or maybe you just removed them?
configs/plasma-thc-sim/emc.nml | 47 -------------------------------------
configs/plasma-thc/emc.nml | 47 -------------------------------------
well.. I saw in the commit log that I removed them
and the change you introduced has dos-style line endings
but I have no idea why
jepler: all the files have dos-style line endings
*bam* *bam* *bam*
leave that wall alone
it's a good thing these IBM clicky keyboards are so well built
I see this:
$ git checkout v2_3_branch
Previous HEAD position was ce45e1... release 2.3.3
Switched to branch "v2_3_branch"
I bet at some point you did a "make clean"
I see the emc.nml is wrongly part of 2.3.1..2.3.3
in the plasma-thc and plasma-thc-sim folders
and those dirs are part of COPY_CONFIGS in src/Makefile
yeah -- so they were added by mistake, removed by you when you did a 'make clean', and the the change commited when you did 'git commit -a'
ok, so no harm done
(except the wrong commit message)
EMC: 03jepler 07v2_3_branch * r342b1385352d 10/configs/cooltool/ (9 files): convert from DOS to UNIX
whee.. pictures of cradek's "junior" :)
[Global Notice] Hi all, one of our sponsors is having connectivity problems resulting in a loss of services (ChanServ, NickServ etc). Please keep with us while we investigate. Thanks.
* skunkworks_ looked in all the usual places..
[20:39:23] <cradek> http://geektrap.com/bill/01248711886
well - I didn't think to look there... duh
that is one sweet looking machine
how high are your garage doors?
1/4" too high, from the description ;)
with the head all the way down, it's 6'10 tall
so on 2x4s, it still fits
I am happy I went with 10ft ceilings and 8ft doors. Although I am finding out I cannot buy the cheaper garage door openers because they only work with 7ft doors
yep that sounds nice...
crap. do I need anything else from McMaster?
this is the question that has plagued me every day for the last week
I may need to raise the ceiling (in one spot over the head) an inch or two - not sure yet
(as I was about to submit my daily order)
lower the floor? ;)
SWPadnos: so now you know the answers were all yes?
I should rephrase the question
what else do I need from McMaster?
was that from stuart?
he has one just like it still running, so he got the control parts he needed off this one
hmmm. hand tap or spiral point?
through hole, in aluminum
spiral point, no contest
either by hand or in a tapmatic (4-40 thread)
or form tap if using the tapmatic
hmmm. I'll try a form tap too - might as well
spiral point is more flexible - you can use it for hand stuff too
oh well. I know I'll need more stuff to hold the hard drive, but I don't know what yet
cradek: second from the bottom picture - analog meter?
cold dead hands, etc
that's the last and greatest of the fet-voms that radio shack sold long ago
it has a nice continuity test buzzer that responds instantly
you can just brush the lead down a row of pins and instantly pick out the right one
I've used digitals that have the buzzer but take .5 - 1 second to respond (really!)
I was never allowed to use dads analog meter. He was too afraid I was going to screw up the meters needle :)
(Measuring 110v on the 10v setting)
or the like
I've blown a few fuses in meters...
cradek: so bill doesn't like the colro?
guess not - we want it bright purple instead
at least that way you won't trip over it the night
I think it's more about the paint being nasty and peeling off
you should have seen the table - wow
the T slots were fully impacted
not sure what you'd call the material that came out
amorphouse metal-bearing goo?
shavings turned to rust?
there's probably a yiddish term with a z in it, like "shmultz" or something similar
will the drives run on single phase?
but the coolant and lube pumps won't (easily)
so I might give up and keep using my convertor (which is extremely reliable)
cradek: Ingrid votes for purple as well
I think he might do blue and white
I voted for "whatever color the primer is"
tonight I will see if my lube pump repair works
cradek, color A400FF
skunkworks_: the plunger cylinder mounted with three tabs, all broken off
I made a funky sleeve thing to repair it
but first, I melted one broken-off tab with the propane torch on the off chance that it would solder :-)
I think it's just zinc
I wonder how we'd get SPI to work reasonably easily with hostmot2
d = design
sticky matter of design
maybe having a "load_settings" pin would make sense
you set all the parameters (CPOL, CPHA, clock rate, number of bits), then hit the load bit
and the input/output are 32-bit values, which have to be externally processed to be useful
plus read and write strobes