is there a way I can log into this box while emc is running, and query its positions?
I'm assuming shelling in
keystick is CLI interface, can run concurrent and even control it too.
scarry huh =)
where can i get that, and which machine does it run on?
[01:08:13] <JymmmEMC> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Keystick
gene__: it runs on the machien you connect to,
this one then.
does "THIS" machien have emc already on it?
yes, its running a program that won't be done till midnight or so, v2.1.7
I THINK it has to be configured in advance of starting a job, but best to ask the other guys to be sure
from the sounds of the link, yes,
I'd have to restart
that link also mentions emc1 too, so not sure.
you could use live cd I guess
I saw that also.
It worked great on BDI when I tried it a ways back...dual control
ok, back in 90
something to investigate later I guess
thanks for the link
I'm going too let this finish in the dark I guess, I'm bushed. I might check emcsh later also, thanks Fenn.
emc_rel_act_pos Returns double objs containing the XYZ-SXYZ actual pos in rel coords, including tool length offset
emc_rel_act_pos 0 returns X position
you have to run emcsh in the directory your .ini is in i think
you should be able to fire up keystick to talk to a running emc, but I wouldn't try it the first time when an important program is running
On axis.unpy.net, I saw an entry about making a custom usb HID device...
Here's what I'm getting and am going to play with: http://www.ultimarc.com/ipac1.html
that looks very interesting!
...and it's fully programmable by doing a pass-thru connection with a regular keyboard and a text editor :)
jmkasunich_ is now known as jmkasunich
jmkasunich: re your post on the mail list, did you think i was talking about stepping with the new doublestep driver? i thought it would have a 50% duty cycle normally
stepgen always generates pulses
normally they are one base period long, with doublestep they are whatever the programmed time is long
the only time you get 50% is when you are at the maximum possible speed (one step every other interrupt)
i guess i confused myself
oh man i hate opencascade.. sigh
#($*@% InstallShield wizard for linux..
his License does not grant any rights to use the trademarks, trade names and domain names "MATRA", "EADS Matra Datavision", "CAS.CADE", "Open CASCADE", "opencascade.com" and "opencascade.org" or any other trademarks, trade names or domain names used or owned by the Initial Developer.
i wish i knew what they were thinking
hey i got this neat new cad library, from SOMEWHERE
i'll give you a copy once sed finishes chugging through 70 THOUSAND FILES
With technology provided by <name withheld by request>
perhaps we could call it IceweaselADE
or just weaselADE
or <name withheld by request>
would somebody do me a favor? can you see if this page loads reasonably fast: http://jmkasunich.dyndns.org/cgi-bin/blosxom
(I had a problem with the thumbnails, which I think I've fixed, but it's always fast from here)
yep. the thumbnails are popping up pretty quickly
the two puma images are the slowest
they're pretty big, and png instead of jpg
I should figure out how to shrink them and do thumbnails like all the rest
or make one more post so they fall of the bottom ;-)
same way - using image magick convert
jmkasunich: slow man, slow
jmkasunich: If you like, I can "test" your connection for you =)
it loaded fine here
maybe a little slow
nothing compared to compiling opencascade though :)
fenn: Well, I have a tiny connection here, just ask SWPadnos
"In the absence of long setup or hold times, the absolute maximum step rate is one step per two BASE_PERIODs."
whats "long setup or hold time"
renesis: it is a characteristic of your stepper driver. However, I used different words there than in the rest of emc's documentation. The normal terms are "step length" and "step space".
ooo that makes alot more sense
microstepping is neat, very quiet
renesis: where did you read that text you quoted? I want to correct it..
[12:13:04] <renesis> http://linuxcnc.org/docs/devel/html/config/ini_config/index.html
k ty too
hm maybe somebody besides me wrote that :-P
03jepler 07TRUNK * 10emc2/docs/src/config/ini_config.lyx: improvement of wording about BASE_PERIOD suggested by renesis
jepler: I made some progress on the vacuum table yesterday
(between problems with the mill)
I also did a lot of troubleshooting and think I have a couple of its problems figured out
cradek: I barely touched my machine all weekend
I need a dial indicator or something to help me quantify the table slope
but after tightening the spindle holder the "dots appeared as small circles" problem seems to have gone away
oh, and the threaded inserts are 1/4-20 as you had thought
did you run that small vacuum pump yet?
[13:57:01] <cradek> http://www.cdcotools.com/item.php?itemid=96
nope not yet
this is the most obvious kind of indicator to use, but the mountings are both too big for your spindle
there's also the extremely limited height available on Z
see right under the dial there is a dovetail that you can use for mounting
oh, I see
so, it takes very little height, the dial sticks out horizontally
well then, just stick the 3/8" clamp in a lathe and turn it down to 1/8
I think these probably have screws through them, so it's not that easy
I'm not familiar with that style of clamp
mine has a nut on the outside
oh, I can kinda see that
(I'm going to order some stuff from these guys today)
if I could see how to use it on my machine I'd sure have you order it
hmm, I blew right past $70
is it possible to order from a machine/tool supplier and not spend $70?
need anything else from cdco?
SWPadnos: jepler and I joke that ordering one little thing from mouser etc. costs $70
heh - so true
SWPadnos: this company has a lot of the cheap-but-seems-good chinese stuff that 800watt sells
ah. I got a lot of that stuff from Machine Tools Direct (eBay)
and another company I don't remember the name of
CDCO also takes payment by paypal, which I like, because they don't get my cc#
jepler: regarding turning down the indicator mount - it's probably hardened
you dont need to mount it in the spindle to check for table flatness (parallelism)
spindle mount checks spindle squareness to table (can do if offset and rotated)
you dont even really need a DTI for checking spindle squareness
compiling opencascade only took 9 hours
you need to try something other than those 200 MHz panel PCs :)
and no, having 10 of them doesn't make it any faster
what if i make a cluster of 30 of them?
that could be faster, unless they only have 10Base-T
upgrading from an 8080 is an option
but I think they never released it
nah 6502 was faster
the original risc
closer to 0isc
JymmmEMC, a troll in ##php to kickban Farzad
i want to test how trajectory planner influence the positioning accuracy
where is it implemented the trajectory planner ?
says "The source is documented and meant to be easy to read/understand. The files are in emc2/src/emc/kinematics"
ok, thank you
i found some topics on trajectory planner from march, 2006 wrote by cradek
there have not been any huge changes since that time.
mostly addition of small features such as spindle synchronized motion, as well as bug fixes
i want to compare the old trajectory planner with this one
on which version is the old one ?
I'm just curious - are you having a problem that you think is a bug?
all released versions of emc2 incorporate cradek's trajectory planner.
i use emc2 on a research regarding the optimization of cnc milling machines
with your help i understand how to implement feed rate override and spindle speed override almost in real time ( 0.2 seconds )
now i want to modify the source code of the trajectory planner and see the modifications
cradek : i want to try with the same file - narrow.ngc from your example but the link don't work
what example refers to "narrow.ngc"?
it was an example used by cradek with axis in march, 2006
sorry I don't remember what I had for lunch yesterday
if you have a screenshot of a test case you should be able to approximately rewrite the gcode
the files were on timeguy.com/cradek-files/emc/nasty-blend.png
at that moment i want to try that example and see what hapens
those very old screenshots are from just before or after the "simple_tp" planner was merged into the TRUNK; they probably refer to bugs that existed for a short time and were fixed..
i don't know if they were bugs, for me is useful to understand the old trajectory planner and the new one
[18:48:45] <cradek> http://timeguy.com/cradek-files/emc/.old/nasty-blend.png
this is one of the wrong behaviors of the old planner
cradek: thank you,
you can see enough gcode in the image to retest this case if you want
of course it will work today
cradek : i want to test it on YOZ plane
is there a way to import the emc.ini from EMC2 into EMC2?
er EMC1 to EMC2 that is
rap2: the numbers to copy down are your max velocity, max acceleration, and axis scales
other than that you should not keep your old ini file
just set those values in the new file
I use a stg card. would that be the stg.ini?
yes, when you start emc2, pick the stg sample configuration
it will copy it to your home directory, and there you can modify it with the values you need
so I have to set forth axis, sever versus stepper and all that?
is the stg.ini file the only one I need to edit?
yes you have a little work in front of you. So much is different there's not a straightforward conversion.
rap2: I can't say. It depends what all is on your mill.
cool thank you for the help at least I know (sorta) what I am up agenst.
good luck rap2, come back with specific questions if you get stuck
one characteristic of the current trajectory planner that I don't think has been extensively studied: does the commanded path actually remain within the tolerance specified G64 P- compared to the ideal path specified in the gcode?
(actually I am quite certain you can find conditions where the difference between commanded and ideal is 2*P, because the interpreter combines multiple "straight feed" moves that are all within tolerance P of a single straight feed; and then the trajectory planner can stray an additional P from that modified path ...)
the constraints test have also not been run for more than 4 axes (XYZA); the development version of emc now permits up to 9 axes (XYZABCUVW)
i have only a 3 axis cnc milling machine, only for research
i will put some optic sensors in well determined positions and check if the real path coresponds with ideal path
depending what you want to measure, you may not need to use a real physical machine. Most of the testing I refer to was done on a simulated machine, with xvel, xacc, yvel, and so on computed by hal 'ddt' blocks, and exceeded constraints being flagged by a 'wcomp' block for each constraint.
I think optical sensors will test your machine more than they will test the software
it sounds fun
here in Romania we have a few word: theory like theory but the practice killing us
In theory, theory in practice are the same.
In theory, theory and practice are the same.
i want to see not only on the screen but also on the machine
the emc2 is a very good platform for research
if you think that my experiment is interesting i can described it
thank you for your help
if i move really really slow (say 0.01mm/minute) and am moving a vector of 3um X and 1um Y, (G91 G01 X.003 Y0.001 F0.01)... will i observe something like steps, even though it's a servo system?
will it smoothly move thru the 3um?
tomp: given that you have small enough encoders to handle it..
how can it resolve less than the 1um on my scales?
i got 1um
then I dont know
tat 1 um scale rez, and 1um smallest command unit, i think it will act like steps
tomp: yeah, you cant really move smaller then your motor can see..
tomp: dont they usually say that you need 10 times better resolution then the minimum detail you want to be able to cut?
than the control can see from the feedback. it's just that the step-like quality 'scared' me.
tomp: well, add a tach/encoder/anything that tells your machine your motor is moving
I dont know about your machine.. or how that works.. though your servo jitter should be around 1um too right?
yes, and i cant afford that, and if i use .001" scales, it'd be the same story for .001"x .003"y ( wether i interpret it correct or not, it'd be the same story, for me and all )
servo jitter is .001 ( dro and lag show 0 or .001 )
0.001 = what?
hm, I wonder if the motor will interpolate so you move 1um in 1 minute.. or whatever you specify..
dosent really seem plausible though as you have no feedback
i think i have to sub count like you said ( use 1/10 smallest command unit ) and display only 3 decimal places to >reduce< this steppiness
tomp: I think to do it you need to interpolate it somehow
im wondering what would happen as 1um is the space pid has to keep in position within
slow speed is a bitch, right, no feedback for long periods of time (many servo cycles)
tomp: I dont think that would work
tomp: though maybe you could write a tp that handles that
or modify it.
way ahead of me
well i begin by feeding analog voltages to an amp, and hope that the tp i use can make motion as nice as that. or consider that a goal to get closer to
tomp: well, you cant do pid that way..
no no no no not pid, just p
hm, well, that'd work
velocity is proportional to my control voltage
yeah, kinda off topic for here, thanks
I dont mind
it's just that i dont understand how to do it
monitoring lag is interesting, today ( non-emc) i used it to find a bad ballnut. once at speed, a good system has a constant lag (+/- 1 unit jitter), and this one had up to 18.
i removed/replaced the thrust and tension bearing, but still had 12 units... removed the hi-win ballscrtew to feel a real click every rev :(
i bet hal can be used to help debug such problems
you can plot acceleration vs current
no current difference seen, lag is more 'fine'
if there's a dip in acceleration and the current stays the same, it means something is binding or not working right there
the amp's current output looked pretty steady ( copley 423 )
500mV max output tho, so signal buried in noise
sorry current monitor, not output
well, just feed the amps a known signal
fenn: any ideas on very slow speed control? can you scroll back?
use sensitive tachos
fenn: yess, used a lab power supply eventually my brain kicked in.
tachos for slow speed or debugging stickies?? ( both i 'spose )
for slow speed
they're just motors with a high volts/rev ratio
if you spin them too fast i suppose the coil inductance kicks in to limit current
but i dont really know
hmm.. another option would be to use a sinusoidal encoder somehow
all encoders are sinusoidal at some point
yeh, but mine 'frets' ( jitters ) and the overall motion is slow velocity ( maybe several hundreds of +1/4um -1/8um a couple whole ums that cancel each other... meaning, well ... 0 )
all inside 1 servo cycle of 4mS
hm.. what's moving the motors if its within one servo cycle?
fenn: he's not running PID..
fenn: just P..
right, didnt point that out, its just P
tomp: resolvers might work for that
ok so why do you have a servo cycle at all?
just to check up on things ( out of bounds, off track, too big a signal swing since last )
and maybe... ooh we made progress, move over here now ( change the target )
so there is a PID with a wide deadband, and an analog P loop inside that?
or not analog?
no pid at all, no deadband desired ( theres always some ), only a p loop with a cyclical check
loop may be a bad word here, only a p control is better
i wonder if there is a such a thing as a resonant resolver
[22:16:52] <Ziegler> http://www.xkcd.com/309/
the loop that exists is rechecking and varying the P to get on track if necc, or to speed up if poss, or to abort becuz the position feedback says i fell off the edge of the map
i dont get how it can back up if P stays the same
P doesnt, p is created eaqch cycle from the process control
p doesnt care what it was last cycle
then how does it do "+1/4um -1/8um all inside 1 servo cycle of 4mS"
i'ts LIVE P, it's an initial value and live modified by the process. so it can change a bit
imagin connecting a pot to an analog amp, replace the pot by a process, limit the pot by a range
you got live dynamic fwd/bwd in a range, and you checdk it cyclicly
good q tho... i got n servo updates per evaluation in the system i'm modeling after...., i dont have this stuff in emc yet
tomp: I think you need to figure it out how to omit the normal servo driver(modify it) to just handle what you want
fenn: you right , what i observed can't be inside a servo cycle, it was inside many, i was reporting 'tracking'
doh, i cant see the dro that fast , and wasnt logging
aw crap. ran out of disk space while linking the opencascade files
2.9G and counting
fenn: do you have any idea what changing the paper "thickness" for a laser printer does?
uh, makes the paper thicker? :)
heh, does it print with more toner..
more heat less heat?
i think the printer usually keeps the fuser a constant temperature
changing thickness with a lever or in some config dialog?
in some config dialog
im doing toner transfer to a pcb.. and im wondering if changing it will give me more toner
I don't think so, but I'm not sure what they're doing, to be honest
try it and see
or well, same amount of toner and faster/less heated paper
yeah maybe its the speed the paper goes through
in which case you'd probably want it set to "thinnest"
that's what I'm thinking about
I want the toner to come off easily :9
what kind of paper are you using btw?
I found some glossy laser printer paper that works
[23:29:18] <anonimasu> http://imagebin.org/10243
anyone know why copying files takes up 100% cpu?
that's the first pcb I've ever made :)
it just took ~5 hours of messing about finding a paper that would transfer ok
heh 5 pins. i cant see much detail so i wont try to analyse anything
[23:31:04] <anonimasu> http://imagebin.org/10245
it's a smd opamp
yep looks good enough. try lowering temperature and increasing iron pressure (use the edge of the iron)
hm, use a higher dpi printer..
the rounded corners and wobbliness are from too high a temperature
but it works eh?
now I have a stupid 5x5mm chip with 8 leads..
next time make more than one
unless pcb stock is expensive (i got it practically free on ebay)
depends on where you buy it though :)
fenn: I think it's low dpi because everything came clean off the paper..
23points per square mm...
0.04.. that's big :p
ah good old linux. re-mounting part of one (mounted) filesystem over another one would have been impossible in most operating systems
fenn: does that sound sane?
fenn: at 1200 dpi the dotsize of 0.02mm :)
look at the printout and see if it's satisfactory
it looks like that on the printour..
if you have wobblies and rounded corners on the printout, no amount of ironing is going to fix it :)
1200 DPI is only 1/8 the resolution of a real photoplotter
laser beam on x/y table, draws on photographic film
what the professional PCB houses use to make gerber plots
at least, that's what the prototype house I just ordered from uses :)
Results 1 - 10 of about 2,610 for pentium "heat sync" </google>
SWPadnos: I'm happy if I can prototyep smd stuff in my kitchen :p
well, you need synchronized heating or SMP performance is unpredictable
anonimasu, mmmm - I love the smell of baking etchant
hmmm. I should go get some donuts with my cream. bbl