well, off for some dinner for me. biab
* alex_joni is gone ... but he'll be back
hmmm... guessing the root password on a system i havent used in 3 years...
i coulda swore it was pb4ugo2bed
fenn: use a floppy ?
telinit: must be superuser
can you reboot, and pass the kernel S or single as a parameter?
(or is ctl-alt-del limited to root?)
i'm about to wipe it anyway.. just wanted to see if anything cool was on the disk
the idea is to install puppy-emc2 over ethernet
wouldn't a CD be easier?
i'm hoping to hijack the redhat 7.3 so i can use my pcmcia ethernet card
no cd drive
well - then it may noe be easier :)
fenn: that's what single user mode is for
G night guys
Jacky^ is now known as Jacky^afk
ok so i've got a boot: prompt
how do i pass options to the kernel?
or linux single
or possibly linux initrunlevel=1
cool it worked
fenn you're not done yet =)
well now i can look around at least
* Jymmm dont remember the rest to reset the password though =)
rm /etc/passwd ;)
fenn and dont listen to SWPadnos
yeah - he's full of shite
(that's polite for shit)
fenn ^ from the horses smartass direct!
i can just delete the encrypted passwd field and it should let me login with no password
yeah - something like that, unless shadow passwords or PAM or something else was set up (I think)
you need to run passwd ro reset root
you need to run passwd to reset root
What's the best way to pull down emc2 to see if you could aid in the development effort?
Oh - i.e. anonymous or sign up as a developer?
Been lurking (well, the system has) since the beginning of yesterday's meeting. jtr = Jim Register - long time lurker on emc-users from the old nist list.
anonymous CVS is fine, it just lags behind by a couple of hours
cool, thanks. Been programming the last mumble years in the AMX Axcess language - don't know if I remember any c
off to sourceforge now.
hmmmn - it asks for a password, which I assumed to be my email addy - should have just been a <cr> maybe?
Scrolled down on the wiki page, saw the link for the cvs book - I'll start there.
just an update - for some reason I thought my ini file period was set to .000016. It wasn't - it was set to .000022. Now I can get 25ipm from z. If I go 1:1 on the drive I should be able to double that.
that would be fine for me.
until you're a developer or want to submit patches, CVS is pretty easy to use. "cvs co" to check out files (the first time) then "cvs update" to get changes since your last update or checkout (from time to time)
Ok - the command line at http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl/emcinfo.pl?CVS
is a little more involved.
I probably over-simplified
Seems like I read something somewhere about adding some lines to a "sources" file on my system
That might be /etc/apt/sources.list
apt is the software debian uses to fetch software updates and install new packages
bdi is based on debian
Thanks, But I think there's a lot of reading to be done first re:cvs.
I'm responding about a line or two behind, but yeah I finally got around to putting the BDI on a system.
Had to buy my bride a new one first.
I managed to mess up the BDI install - thought I knew better than X's probe what kind of video card I had.
had to go find out what kind of X it was running - XSetup weren't the answer. I took notes - maybe that should go in the wiki
somewhere where it doesn't distract from the main point, which is EMC.
jepler: thanks for the help; sorry to have hogged the channel
jtr: any kind of notes are welcome in the wiki
someone know the new mirror of paul_c's deb files? mirrors.neuron.com (listed in my months-old sources.list) doesn't seem to work these days
maybe this is what I want: http://solaris.cs.utt.ro/emc/debian/
no, that doesn't seem to work. :(
jepler: I'm trying to install axis on a mostly screwed up machine - do you know what package(s) I need to install when setup.py gives an error about _tkinter in togl_setup.py?
SWPadnos_: debian-based? python-tk redhat-based? tkinter
screwed-up BDI - partially (mostly) failed upgrade from 2.18-ish
2.18? that's so old, I don't even know
pyhton2.3-dev is installed, and I can't install python2.3-tk because of a dependency
well - it's almost 4.30 now :)
if the error is short can you paste it here or in /msg?
I'll try - it's on a different machine, so it'll take a sec
# env EMC2_DIR=/Project/emc2-latest python setup.py install
Traceback (most recent call last):
File "setup.py", line 27, in ?
from togl_setup import get_togl_flags
File "setup/togl_setup.py", line 2, in ?
ImportError: No module named _tkinter
well - I'm downgrading python2.3 from 2.3.5-8 to 2.3.5-3sarge1 so I can install python2.3-tk
jepler the solaris server has been down a couple days but it should work usually
must seen me coming!
I don't suppose anyone knows what OpenGL / Mesa packages are needed to build AXIS?
jepler and cradek might
SWP this all sounds pretty familiar
try installing the nvidia hardware acceleration :)
redhat did a terrible job packaging the opengl libraries
I'm not sure that nv acceleration would be good for a C&T video chip
and this is debian, BDI
I've got three here that built it axis today
any clue which of the 17000 mesa packages needs to be installed (or removed)?
let me switch boxes.
xlibmesa-dev xlibmesa-gl xlibmesa-gl-dev xlibmesa-glu
ok - how about opengl, and xlib?
actually - xlib or X11 would be useful (I have all the ones listed so far)
xlibs xlibs-data xlibs-dev xlibs-static-dev
I have no clue how many of those are essential or what each does.
Once I got the instructions for install it went without a hitch.
ok - thanks - some of the xlib packages are missing. I'll ask again if that doesn't fix it
I saw a goofy reaction from one of the axis menu lists but it only happened once.
looks like part of linuxcnc.org is down.
what part is down?
03jmkasunich * 10emc_HAL/farm_scripts/run_farm: changed farm scripts to save a history of good and bad compiles
I was missing images from org/images but they are there and being displayed now.
03jmkasunich * 10emc_HAL/farm_scripts/run_farm: changed farm scripts to save a history of good and bad compiles (maybe it will work this time)
Still here, working my way through compiling EMC2 on a BDI-4.30 system
apt-get install cvs... updated cvs, which is asking for the path to the repository
is that going to be exactly this? :pserver:email@example.com:/cvsroot/emc
03paul_c * 10emc2-auto/wiki/ (32 files in 22 dirs): "Auto update wiki from a cron job. Tue Nov 22 05:30:02 GMT 2005 "
just a little more to figure out and document.
SWPadnos is now known as SWP_Away
cool motion control q&a site: http://www.control.com/1026213484/
I got on here to bug les but he doesn't appear to be around
nope -doesn't look that way
you just posted to the user and dev lists, right?
I'm hoping someone will check out my last post
checking now ;)
how did you measure velocities in your testing?
using the emc timing....and checking against the wall clock to make certain emc wasn't lying
hmmm - I'd use a couple of marks on the table, and a stopwatch
I've done that too
also, you have to account for the accel and decel phases of the move
ah yes but my accel is at 20 ....
faster than most machines, if you discount les
a lot of BP's are at 4 or 6
but is the machine actually achieving that.. 20
I know that is still fractional G's but reallistic for the age of the machine...
ok - so it should only take 0.15 seconds to get to full speed
I think so...
and you (of course) tested each axis individually?
ie, no compound moves
how much axis travel do you have??
if it was struggling then the difference should show up at lower speeds
I used 20 on x , 10 on Y and 5 on z
sure, I'm just confirming that you ddid moves on each axis separately
I have about 28 on X, 15 on Y and 7.5 on z
yes, not combined moves
higher speeds would be more critial due to the longer accel decel zones
as for the max_vel settings, I think the one in TRAJ is supposed to be the fastest combined tool speed, and each axis is further limited by the AXIS_N setting
so it's effectively the lower of the two
that is my thinking
I asked Ray if anyone actually checked the Mazak at Cardinal Eng and he doesn't think so.
but if x & y are set differently your g17 circles will only be as fast as the sloest programmed axis
it may be that the moves along the faster axis are faster, making an uneven finish (I'm not looking at the code, so I don't know)
well most of my real cutting is done below 20 (actual ) ips
I'm just hoping a couple of people will go do real timing both with emc2 and with emc1 and post
accel is 20..ips... mot max speed
20 IPM, I'd assume ;)
accell is set at 20 ips/sec/sec
oh thats more realistic then...
please note that the speeds are relatively linear with commanded .... just not scaled correctly
but g02/g03 use accel/decell on 2 axes
my circles are just fine....
hmm ok then never mind
the only problem I have is chip load...if I believe the commanded velo
how well controlled is your spindle speed??
are your input and output scales correct?
(I guess tey'd have to be)
circles are round... ; (
what hardware do you have?
Mazak v5 with Servo Dynamaics amps running STG card on 2.18
JMK...you have any comments?
did you see it on emc1, or 2?
and since a lot of the code is the same I'm wondering if anyone checked actual velo on the Mazak
trying to remember
I dunno if we timed a move or not
I'd like to see more precise timing
I'm guessing not....I don't think anyone does that kind of thing until a problem surfaces.
something where only the constant velocity portion is measured, using physical switches for gating
stepper machine is trivial to test
don't need the machine, I can do it here
it's an STG in Dave's machine
guess I should plot the accel...
on emc2, it will make no diff
on emc1, I don't care ;-)
I should think this should show under sim
I can say that running under VMWare gives you significantly slower moves ;)
running RT code under VMware is a joke
yes - it's not really RT anymore ;)
in fact, it's *ReALLY* not RT
I'm working on limit switches ,etc now for bringing an old tracer mill up under emc2 ... but that is still a ways away.
you were just doing MDI moves to test?
jmk.... yep...emc1 is not your problem.
don't think auto changes it.
is G0 cutting feed or rapid
g0 is rapid
oops... and F1 = 1 ipm, not 1 ips ;-)
that is a good point tho.... I didn't try g0 to see if it bypasses the problem
yep units are important
ok, according to halscope, a 1 inch G1 move at F30 took a hair over 2 seconds
you weren't in cm mode, were you?
yeah - it shouldn't matter anyway - you would have noticed the short moves
I should hope so
jmk ... what is your accel set at?
have to check
using the stock emc2 emc.ini
so jmka, how bad is dc bias on the power lines? I'm not going to blow up lots of electronic stuff in my house am i?
which probably defaults to 20
TRAJ: default vel 0.0167, maxvel 1.2, default accel 15, max accel 20
I've modelled it with SPICE and after the inrush currents it's not that bad
icee - just a sec
axis: maxvel 1.2
max acc 20
I'm setting up a differentiator on position to get velocity to scopee
I should make a doohickey that has an encoder and a wheel, and records movement every millisecond or so, so you can use an external measurement tool
measuring with a possbly broken thing soesn't sound like the best approach to me
ok, accel from zero to ,5 inches/sec took 50ms
sounds right to me
good point - if the servo thread isn't ticking at 1ms, then scope wouldn't have a clue
no - off by a factor of two
2 IPS should be 100 ms, so 0.5 IPS should be 25 ms
something interesting - the velocity isn't a sharp ramp
15 * ((inches / second) / second) * (50 milliseconds) = 0.75 inches per second
there is some rounding at beginning and end
vel has an S-shape?
that's PID, with no FF, right?
yeah, at best the middle third is linear
rounding isn't so strange
I'm looking at the command out of the motion controller, before the PID
let me try a faster speed
oh, i didn't think the motion controller did any jerk limiting
I didn;t either, thats why I'm confused
ok, rounding is the same in absolute terms, when I doubled the speed the linear section got longer
the first 20mS is rounded
is that an interpolation artifact?
in 20mS it gets to about .200 ips
then it takes 60mS to go from .2ips to .8ms
40ms to go from .2 to .8
then another 20ms to level off at 1.0
so 40ms is exactly the rate you'd expect for .6ips
15 * ((inches / second) / second) * (40 milliseconds) = 0.6 inches per second
yeah, the linear ramp is right, but why the rounding?
maybe - I'm just throwing out (up) suggestions
and the ramp rates for the first 20 milliseconds and last 20 milliseconds is exactly half
i don't think the interpolation should cause that
the average rates
so it sounds like there may be some magic jerk limiting in there somewhere
so it's always 20 ms? - try going to 10 IPS
gonna try an accel to 0.2 ips
need to edit ini and restart for anything over 1.2 ips
i'd expect that to take 40 ms and be perfectly rounded
also, change the max_accel to some ungodly high number
can you take a second derivative too, jmka?
it would help to validate your methodology
if you see acceleration ramp up, level, and ramp down linearly
accel from 0 to 0.2 took 30ms
note that you'll have artifacting due to the differentiator, though it should be only one scope cycle per level
yah, i know.. this is still at 1KHz or something right?
(you don't have smoothing on the scope, do you?)
e.g. nyquist is >500Hz
it would show up as a delayed spike
second derivative should show you pretty nicely for sure what's happening
I'm going to run.... maybe go back and try g0 and auto mode just for grins
if -it- is rounded significantly, then there's probably a methodology problem
if it is mostly trapezoidal, then what jmka is seeing is probably what's really happening
the scope is sampling at 1KHz, the differentiator also runs at 1KHz
I think I have an ini for simulation, I'm gonna restart with that, after adding two differentiators for vel and acc
you'll just have one or two invalid samples at the beginning and end