tho it is an integer number of 'smallest command units'
:) itof.c ?
I wonder when my sister will be able to return to her house (if it's still there)
SWPher name isn't dorthy is it?
no, and no red high heels either
that I know of
SWPadnos: hey know a way to make a cheap bulk tape eraser?
get an old woofer
no got one
SWPadnos: next suggestion
uh - what's the question again? you have a bulk tape eraser but you're asking how to make one cheaply?
oh - you don't have an old woofer
I have bulk tapes and want to erase them
no woofer, no tweeter, no sylvester either
well, big speakers are the easiest thing for me to think og
[00:14:35] <tomp> http://www.youtube.com/watch?v=1pOYY14zS8s
how to erase tapes with hf osc
rotating magnetic field
SWPadnos: I just have a lot of 8mm tapes to erase
I suppose you could put them next to an AC induction motor (inside it instead of the rotor would be best)
[00:15:36] <tomp> http://www.audio-restoration.com/erase.php
are you trying to prevent others from reading the tapes or prepare them for re-use
prevent and reuse
it's a one time thing
[00:29:53] <SWPadnos> http://erase-o-matic.com/
it looks like all it is is some of those high-power magnets and a flat surface to run the tape across
has anyone seen issy's gui code?
I'm looking for a backlash compenstion tutorial.
or at least the config values to add/change. Stepper based.
the backlash idea is simple,average the error,i think
Well, sure. The dial tell me .008 backlash on the x, for example. But what is the line/lines in the config file to update? That is the info I'm not finding.
The ini file sections of the Integration manual.
add BACKLASH = 0.008 to the AXIS_0 section for X (AXIS_1 for Y, AXIS_2 for Z)
that would be where to look for more info :)
oh,let me think my lab process.we got a laser to find the correct posion,and our range is from -100cm -100cn for example.we make the cnc go 10cm per step.
find errro postion and correct postion(from laser),and get these values into ini file,that's all.
if you want to find more,then you have to turn to the source code.
stustev: issy said that he's working on making it linux friendly
DanielFalck: It was only working in windows?
I think some of it was
it is C isn't it?
stustev: I'm not sure.
stustev: I sure liked what I saw
I liked it a lot also
SWPandos: found it. thanks again.
.008 is quite a bit to take up. You may need to lower the max vel a bit if you see following errors.
I have the basic cnc fusion kit for x3. Is this a normal amount of backlash?
and/or increase the STEPGEN_MAXVEL and STEPGEN_MAXACCEL (if applicable) inifile settings.
scutsxg, if you want to compensate for screw errors, in addition to backlash, then you should look up how to do a screw comp file
backlash is compensated by an extra move at the time of any axis reversal -- that extra move represents additional acceleration and velocity. if your "stepgen headroom" does not permit this, you get following errors.
basically, if your config was working before you added backlash compensation, double STEPGEN_MAXVEL and STEPGEN_MAXACCEL when you add backlash compensation
or reduce MAX_ACCELERATION and/or MAX_VELOCITY, if your BASE_PERIOD or hardware can't support double-speed motion/accel
as for the amount of backlash to expect, that depends a lot on the machine. I have a small router with plastic body & triangular leadscrew, and I have backlash of .0055in on one axis and .01something on another.
Yes. the machine works as it is, but it is time to work out the bugs. I'll give it a go.
(with emc's backlash compensation I get good results on this machine, enough to mill circuit boards with .012 trace and sepraation)
jepler ,nice to meet you again. could you give me some advice on how to compile emc2 from scrach to let it go in 586 machine.
and i searched some website,cncuser have done some work on puppy linux,but his website faild. I install emc2 based on ubuntu 7.10 with a CF card in 586 256M ram pc104
there's a puppy ISO on linuxcnc.org, if you want it
but it doesnt work ,maybe it need too much resource
or have you tried it and can't find/figure out how to get updates?
your biggest problem wil be building a good RT kernel, not compiling EMC2
there are several pages on the wiki written by acemi giving details of compiling rtai kernels on debian; it's also easier to get a light debian system than a light ubuntu system from all I've heard. maybe this is not a bad starting spot: http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Installing_EMC2#Debian_Lenny_compile_RTAI
swpadnos,no my problem is on comiling emc2,rt kernel is not a problem,and how to make it work on limited resource system
if you already have a working RTAI system on a 586, EMC2 should compile, and maybe even run, just fine :)
ok,jepler,thanks.It is my pleasure to talk with you.
that page jepler pointed out should have some pointers as to what options to use with configure
(but then again, maybe not - haven't looked at that in a while)
two much package needed to compile emc2.
i will try ,thanks all.
what are you trying to get in the end? a system that runs keystick, tkemc, AXIS, headless using remote X for the GUI ...
at the moment, you need python and gcc, plus some libraries like gtk-(something)
it's kinda assumed that emc2 will be compiled on a full-featured system, if not run there. patches to make configure able to exclude more parts as optional would sure be considered for inclusion, though.
but the bottom line is: it's work to do that. somebody has to do the work, then contribute it to the project.
(well, you can also optionally do the work and then keep it to yourself:-P)
swpadnos using emc2 to drive my motion board,and a little cnc.
if you plan to use a UI provided with EMC, then choosing wisely may reduce the requirements for both compilation and runtime resources
jeper,i should listen to your advice,do some contribute.I learned lot from emc since 2003. and i think when finished my phd paper,i will give some job for emc
scutsxg, You're located in CN?
but how do you know? my poor english? :(
IP lookups work too ;)
I did a whois
I see Guangzhou. I was there last year. Beautiful!
oh,some guys in china also works on emc,most in university and some research place.
thanks,next time you can call me,it is my pleasure to show you around this city
Really. That would be fantastic. I was just there for a day.
How close are you to finishing your degree?
and some engineers also like to using emc to control their home made cnc,but they dont konw too much about emc ,for example rcs,hal trajectory and so on.
may be one year.
rayh,my msn firstname.lastname@example.org,if you come to guangzhou again,do contact me.
lucky Ray - China would be very very cool :)
You bet I will. And I'll try to find a bottle of MauTai to celebrate your degree.
haha,you know MauTai,ok no problem.
Oh his city is very beautiful.
indeed. one of these days I'll get there
ok,i will come to deal with my emc lab.and we should give some room for guys who want to talk about technology question.
EMC Fest China! :)
thanks rayh,swpadnos,nice to talk with you ,and jepler for always answering question.
I'm packing a bag.
good to talk to you. see you around
Nice to meet with you.
wow. backlash comp works. off to a good start. i see what you meant about the accel values - I can hear the z axis spin up to take up the lash.
You got it. Congratulations.
why, when I load a mastercam made gcode file do I have to touch off .5"? Is this a setting in mastercam? Never had to with other files.
do you have some G92 offset or something?
it's purely up to the gcode where the "top of material" is. Is there a mastercam setting that places the top of the material at somewhere other than you expect?
or is this only with MasterCam-generated code, but not with hand-written code?
only with mastercam.
it's possible that there's a setting in Mastercam for "gage block height" or similar - so you'd touch off on something of that thickness rather than the part
I set the material to size, I looked for a cutter zero point setting and didn't see one.
mastercam will put 0 wherever you want it to be, and you can move it
i've only used mcam for simple stuff, once upon a time
so i can't tell you how, only that i've done it
It is set in the material setup area
i thought zero was defined in the part model
that's how I've always done it, anyway
and the in the material setup, you can move the model aroudn on the material
it doesn't actually change the zero
I discovered to my great surprise that I can drill through a .062 circuit board at F60 with my .80mm cutter. I know this because I decided to drill an additional hole in mdi mode, carefully considered what F-number to enter for the drilling move, but neglected to enter G1 or G81 before hitting enter on that MDI
Its been a few weeks, but as i recall I have to zero the axis, select z, touch off .5" and press play.
ended up with G0?
emc2newb: that's because you have the part zero in the mastercam drawing/model .5 from the bottom
heh - well, now you know ;)
.. which is the same as F60 on my machine
you could edit the model, shift everything up .5, and it would then want you to touch at z0
the table or whatever
So, .5" thick material should be set for 1.00"?
i'm very confused now, i've never had to worry about material size in mastercam
only for the backplot has that been an issue
... or is the z axis zero at the top of the material?
z axis zero is wherever you define _in the model_
before it's cut, the top of the material is the only thing you can point to...
if you set it like cradek is saying, Z0 is the top of your material, and the model in mastercam lies below the Z0 plane
Okay. So I've been setting parts to start at 0,0,0, which is on the table.
any number you pick can be made to be on the table
... instead of on the top face of the material
so all your Z moves are positive
like, go back and ignore the material setup thing
It must be a mcam setting. Everything else works fine. Will re-check the material setup
because the mill doesn't care about mastercam's backplot
what the mill cares about is where the material is in mastercam's drafting mode
however you drew it, whatever Z values you used
are what the mill's going to use, regardless of how you set the material up for mastercam
because that's just used to verify the program
so if you find you have Z values below Z0 in the drawing, that's why you are having issues
you have to shift everything up so the lowest point lies on Z0
(if you are touching off the table/vice/whatever)
Ok. I thought the material setup might tie into machine limits somehow.
nope, all that does is the graphics for you
so no negative z values? am i getting that correct?
if you want to touch your tools off the table, correct
but like, there's some caveat to that?
if this is the first operation on the block, you are going to have to bump Z up a bit in the control so that you have material on the other side left
second ops are fine to touch off the table.
i prefer to program Z0 at the top of my part
for first ops, so that I can just touch the top of the block, bump it down a wee bit, and know i won't be milling into the vice
yah, just contemplating that
emc2newb: are you completely new to machining?
next boot into windows I'll check it out
more or less. wanted a project to combine my hatred of electronics and computers. And to make parts for projectsa dn work
eventually tooling for work
yikes, buying mastercam is a very expensive way to start...
you can do a lot writing gcode by hand.
good evening cradek
i will second the g-code by hand thing
it's good we didn't take apart the mill. I needed it today.
if you have some time to work on getting comfortable by hand, do it
what did you make?
we could have had it done by now - I would have stayed another day
i don't ask where the s/w comes from
some parts for the table saw
did you break something?
adapt some neat anti-kickback clamp thingies
dude knows a dude that knows a dude that belongs in prison
nice to have that option
emc2newb: bad style to ask Free Software folks for help you use illegal software
damn. good point
bill was making some shelves and propelled two of them to the other side of the barn at high speed... so, time to mount the clamps
but it was free ;)
motivation to the max :)
cradek: no, the time to mount the clamps is _before_ you propell boards at high speeds
nah, that happens all the time, no big deal
you just learn to not put anything important in the way
(I hate table saws)
actually, not as impressed with it as i thought I would be.
emc2newb: when you have a legal copy, start talking to us about it again
or call them for support
cradek: if you've gotten to the point where you consider kickbacks routine, something is wrong
most of us don't have software like that
jmkasunich: maybe I'm exaggerating a little bit.
"it's never taken off more than one finger at a time"
emc2newb: it's impossible to pick up and learn
pretty much by design
you need to get your hands on a tutorial
or you will never pick it up
I think that's enough on the subject
tutorials tought me what (little) I know now. But not specific to my emc setup
cradek: I bet you have been told a 'million' times not to exaggerate
that's more than enough on the subject. as jmkasunich said, come back and talk to us about it when you have a legal copy. until then, the subject is closed in this channel, OK?
stustev: did you friend get your $12000 touch screen working?
that's something you don't want boards going through
at high speed
I don't know - I will know tomorrow
that's really amazing.
that's a scary number
how does one find a $12000 touch screen?
hmmm. maybe one that goes with a $17000 monitor
call Cincinnati for a replacement 12 inch touch screen monitor
oh, so it's a $120 touch screen sold to big uindustry
found one for free here in town - much better price
that's slightly better
you can certainly go out to dinner a couple of times for the difference
I'd like a free 12" touchscreen CRT... I bet they work great compared to the LCD I bought
there will be a cost - the gentleman wants to learn more about EMC2
I can ask him - he said he has a shelf full
the touch screen crts do not work as good as the touch screen lcds
maybe he doesn't know the curent price
oh really? I had guessed (no experience) the opposite
he purchased a big lot of cinci spares
LCDs are a bit easier to calibrate, because the image doesn't move around at all
probably more parallax with the CRT
I thought they put the magic goo right on the glass, so no or little parallax
also, the LCD is flat, which all CRTs but the Zenith FTM aren't
and flatness isn't needed then either
true, if they do that
the lcd image doesn't move
but of course you end up needing to replace the CRT instead of the touch overlay when some important spot wears out ;)
Thanks for the help. It _is greatly_ appreciated
the 'magic goo' is a glass plate glued to the front of the crt - if the shape is not exactly the same you have differing parallax
do they directly drive the CRT from some custom board, or is it a standard monitor of some sort (probably hercules or CGA if anything though)
afaik they are all standard monitors with the extra touch layer in front - I have limited experience though
I bet a retrofit board that takes direct CRT drive signals and makes a VGA output would be pretty popular (unless they're doing vector displays)
a modern multisync monitor (not LCD) will sync to damn near anything. it would probably not be hard.
Guest173: don't know the type of touchscreen makes much difference - maybe the different types have a different thickness of the touch layer
to pickup the touch on the crt only?
I've seen "bare CRTs" in some old stuff. I don't know what the underling signal was, or if there was a standard
or many to choose from
I'm thinking interface options to get the signals to emc
stustev, they differ in sensitivity, endurance, resolution, and resistance to contamination
I understand that difference - what is the installation difference between them?
I don't know of any systematic installation differences - I think they're all available as external overlays
maybe SAW isn't
if you can project a field across the crt you may be able to pick up position directly from a touch - some may do that
xerox macine i ran in college had ir beams running across the screen. Worked most of the time
there are some IR devices that use several LEDs on one side and several sensors on the other
if you left your finger on the screen too long - did it burn you?
they tend to be low res - like for arrays of buttons in known locations
only if you're susceptible to burning from milliwats of heat ;)
heh, the HP-150 had a touch screen like that
yes, one side send, other recv. Didn't hurt unless i punched the screen.
any screen will do that - that is why they are called 'touch' screens
[03:26:25] <cradek> http://www.hp.com/hpinfo/abouthp/histnfacts/museum/personalsystems/0031/0031threeqtr.html
it was maybe 10x10 sensors
Yes, some touch panel overlays have a limited resolution (a button grid), others use xy position
but the zones were small, since that's only a 6" monitor ;)
yes it's pretty small
like an imac!
or a paperweight
yup. xerox was ~14" monitor, prolly 10x10 grid as I recall
cincinnati quoted $12,000.00 for a replacement monitor - $6,000.00 and three weeks to repair it
for $6000, I'll put in a whole new computer with EMC2, with a touchscreen LCD!
a big one
two aftermarket places quoted approx $2000.00 to replace it with an lcd touch screen and adapter - would just plug in
cinci doesn't have the exec software for this machine anymore - it is a 1996 vintage - it is a candidate for EMC as the other components are wearing out also
machine needs two monitors. one to run the thing, one to watch the game.
cinci would have to rewrite the exec to fix or change anything - hows that for buying name brand
the cinci in Eminence MO is a horizontal 5 axis magnum with a pallet pool. I want to put EMC on it but I need to get better (and a lot faster)
what part do you think takes the most time?
cradek: can you fix monitors - I know have a broken touch screen crt from the cinci?
stustev: I have mixed success with that. do you know anything about what's wrong?
no - the manufacturer said it is usually a cheap component - they were going to fix it overnight until I found the free one - they would probably work with someone to help them fix it - they seemed very helpful
does it respond funny to touch, or is the image screwed up?
or just dead?
it takes a lot of time to make everything work once the wiring is done and the control is installed
the monitor is just dead - nothing wrong with the touch screen
stustev: if you don't need it, I would like to give fixing it a stab then
maybe next time I come down.
It may take me a while to get my hands on it. I will have them send it to me.
ok, but don't go to any big trouble.
Can emc accept input from 'grey code' output type rotary encoders?
Guest173: yes definitely
it will be no trouble - we send stuff back and forth all the time
two channel grey code is commonly called quadrature and is the basis of most encoders used in cnc
Guest173: "two bit gray code" is exactly the same as "quadrature"
no, it like 10 or 12 bit
oh - that is an absolute encoder
emc2 can read quadrature using a variety of dedicated hardware, or in software only
for machine position, not jog
at the moment EMC doesn't do >2 bit gray code
but a HAL module could be written to decode it without too much trouble
figuring out what to do with it is a bit more trouble
you need a lot of I/O pins to get the gray code into the PC tho
cradek: why do you say that? it is a position, just like any other encoder
if you want to use it as incremental, sure
i know. just thinking about a paper tube labeling machine at work that needs digital control
cradek: I don't follow you
no, you can decode all the bits into a reasonable number with an appropriate HAL module
read all 12 bits in, and encode them into a position
but absolute encoders are for homing, and that's a whole ball of wax
or am I in left field here?
yes, there are some potential issues there
well, homing might get interesting, but that depends on the application
paper tube labeling machine - might not involve g-code or homing at all
absolute in this instance is required - we can manually jog the machine
initialization of position without requiring a homing cycle is an interesting one
also maintaining offsets if machine home isn't identical to encoder zero
Well, it doesn't need g code, but it would make changovers faster
err - machine zero
Guest173: does the machine repeat every revolution? or do you only need absolute position over one turn of the encoder?
my viper has absolute encoders - the homing is done by changing a parameter bit to zero - moving the machine to the home position you want and changing the bit to one
if we had a nice parameter editor screen that could be good :)
the encoder tells the machine when to on/off pressure, vaccuum, etc. Cycle timing based off of position
one revolution = one cycle of the machine?
does it have a turns counter, or is it only the fraction of a revolution?
1 rev = 1 operation
4 turret machine
that would be fairly easy then
12 bit signals -> new HAL component -> floating point position, goes from 0.0 to 1.0 over one rev -> group of wcomp or other components
thats what I'm thinking. Its either emc or make something avr based with lcd and gui
sort of an electronic cam switch
I'd be willing to make a gray-code to floating point position component
huh - binary == gra_code ^ (gray_code >> 1)
I didn't know that
yes. electronic cam switch.
SWPadnos: that seems astonishing
I'm looking at the wikipedia page, running through the 4-bit gray code
(they tell you that formula, I'm just proving it to myself :) )
I'll poke around @ work tom. and see what the higher-ups had planned for the machine.
[03:53:11] <jmkasunich> http://www.dspguru.com/comp.dsp/tricks/alg/grayconv.htm
hmm. it doesn't seem to work for 0110 gray == 0100 bianry
I's a ca 1974 made german machine, btw.
this guy says you have to do it a couple bits at a time, so its not a single shift and xor
yeah, maybe it's pairs of bits
anyway, its not hard - the component could be written in an evening most likely
getting 12 or more bits into the PC is more of a trick, parports don't provide a ton of inputs
what kind of speed does this machine run at?
10 RPM, 100RPM, 1000RPM. 10000RPM?
60 rpm ish
~ 1.5-2 seconds/cycle
a 12 bit gray code will have 4096 distinct positions per cycle, so 2000-3000 per second
the boss wishes it ran that fast
EMC's main control thread normally runs 1000 times/second, will the various outputs need timing resolution better than 1mS?
actually, if its just an electronic cam, it would be HAL only, and could easily run at several KHz
nice thing about gray code is that you don't need to sample super fast
that is awesome
and we'd be able to change settings on the fly?
yes, you could have for examply pyvcp widgets that let you set the on and off points for each output
I need to find out what type of motor is in the thing. I do have a cam timing diaghram
I need to walk the dog and go to bed (midnight here)
this is an interesting application - feel free to contact me about it, either here or by email
jmkasunich at fastmail dot fm
11 here in chgo. thanks
SO, in theory, one could make a say 48in, 48 out ISA board and use HAL to control each i/o individually?
yes, ot you could spend $199 on a Mesa 5i20 and have a PCI board with 72 IOs and an FPGA on it
but i have a bundle of old 8255 pio chips
and isa proto card
well, you can do that too ;)
you'll have to write the driver, but there are two or three to base it on (for ISA and PCI 8255 cards)
there are several I/O cards that are probably less expensive than the components you'd need to build one. except that you already have the components so ...
they do need and init sequence to set the ports
Is a work proj and I bet I could get them to down $200
that's all in there. the 8255 card drivers take a text string that tells it which poins to make input and which to make output
$200 only gets you the card, you'd need some I/O breakouts also
[04:09:21] <SWPadnos> http://www.mesanet.com
72 io is much more than an avr would let me do
with a touchscreen...
and a lot faster, with a 200k-gate FPGA on board
well, 2-3 machine cycles/ second
ah - is a competing idea to just use a microcontroller for the whole thing?
don't need lightspeed. I could see a 100 lb mandel fly across the factory...
Well, so many i/o...
hal fits this application well
time for bed. eastern time here. see you
also may allow us to do more logging, costing, tracking and upload via wifi. gnite
thanks 4 chat
gnite all. When i return i will have handle 'topls64'
another day in work begins
awesome i just found out why my stepper just randomly stops. Didn't have a pull-up resistor at the reset pin of my PIC,... induction + TTL == ?????? :p
PIC is evil
so i've heard
but for me, to just randomly fiddle about, it works
..most of the time
[11:05:31] <fragalot> http://omploader.org/vcXk4/P1000953.JPG
good afternoon alex_joni
Guest259 is now known as skunkworks_
fragalot, even worse, those prototyping thingies have capacitance between the rows
archivist_ub: shouldn't matter too much at these low freqs.. should it?
beats re-soldering 100 times over while trying a circuit out :p
well if a line is open, no pullup then its easy to couple noise to it
true, true. I thought the internal pullup would 've done that for me... forgot that the reset pin (obviously) didn't have one, lol
i'd like to learn emc
so it's not in the repositories?
[13:43:15] <archivist_ub> http://www.linuxcnc.org
[13:43:42] <skunkworks_> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Installing_EMC2
there is also a live CD which has the realtime kernel needed
i'm running on the realtime kernel already fortunately
mr_boo: what kernel?
hmm, is there a shell command to determine?
i run 2.6.24-19-rt
ubuntu hardy heron
-rt is no good
not for emc2
what's wrong with that kernel?
the -rt kernel has some realtime improvements, but is not a deterministic real time kernel
what do you mean by that?
the -rt extension changes the scheduler somewhat
to give faster response time, etc
but it's not good enough for driving stepper motors out of the parport
so the hardy heron realtime kernel is a crappy one?
here is a more extensive description: http://axis.unpythonic.net/01190912545
crappy is a reletive term ;)
mr_boo: crappy is not a metric
[13:50:10] <skunkworks_> http://axis.unpythonic.net/01190912545
skunkworks_: I was faster :P
this was not music to my ears
mr_boo: the -rt kernel is pretty good for what it's targeted
and that is (as I heard servers, and sometimes audio/video apps, etc)
but not emc2 where you need a couple of usec latencies
amazing how a cnc controller software can't fall under the term "realtime application"
i think they're pointing out the degrees of "realtime application"
it is - just the realtime kernel in hardy isn't 'good enough'
mr_boo: CONFIG_PREEMT_RT is a fairly new feature of the linux kernel
what would happen if i attempted it anyway without installing a more suitable kernel?
and it's still under development
one day it might be good enough for emc2
mr_boo: nothing will happen, apt-get install will fail because of missing deps
if you want to play - You could build a sim package.. (it just will not have the realtime componants)
ok, it's written for the preemp kernel only?
no, it's written for the RTAI or RTLinux kernels
no.. emc2 currently works with 3 different kernel types
regular kernels (no patch) only in sim mode (no hardware control)
not PREEMPT, not Ingo's -rt
RTAI patched kernels (ww.rtai.org)
and RTLINUX patched kernels (although I don't think there are patches for 2.6 from RTLinux.. )
and that support may have bitrotted some
there is RTLinux for 2.6, but I don't know how recent it is
lets say the developers of the regular -rt kernel would sort out the pieces that are missing for emc2 suitability, would that mean sacrificing something else?
mr_boo: probably not
there's a layer used in EMC2 called RTAPI, which is the interface between EMC2 and the realtime OS underneath
but there are not missing pieces, it;s just missing performance
one would need to write an RTAPI layer that sits on top of something like POSIX.1b
SWPadnos: an wrapper for RTAPI is just a SMOP
I'm describing the problem ;)
for what i've understood so far its not a compability issue but a performance issue
mostly done by jepler already
the solution is left as an exercise for the reader
the compatibility can be overcome quite easily, however atm it's not really usefull
performance of the underlying RT system, particularly regarding timing jitter and predictability
when -rt will be "good enough" we'll surely jump ship
would it be tedious to complete the -rt kernel to suit emc2 demands?
it's god enough for servo systems already, or other systems where there's no need for a BASE_PERIOD
* alex_joni has no idea
mr_boo: I'm not familiar enough with -rt to appreciate that
does it work most of the time with pretty good latency and predictibility, but once in a while it can hog?
when the -rt kernel isn't good enough for emc2, how come it's sufficient for other realtime stuff?
something like that is not acceptable for emc2
mr_boo: it always depends on the realtime needs
so one could claim that emc2 has particularly critical realtime demands?
if you can give us some examples for "other realtime stuff" we might comment on your statement
more critical than a video player.. yes
the -rt kernel clearly exists for a reason
it has to do with needing a ms accuracy and a 30us accruacy...
software step generation requires an interrupt rate about 10x as fast and latenct about 1/10 as big as audio applications do.
* alex_joni runs home
imo -rt is targeted at audio apps where sub-ms latencies are very acceptable -- but 500us latencies are huge for software step generation
mr_boo: because a machine tool, moving a cutter accurately is not the same as making sure a server reponds with video very quickly
shouldn't these things be more connected to hardware performance mainly rather than kernel type?
mr_boo: no, because the kernel determines what the hardware is doing at any time
can i even use emc2 with hardy heron?
mr_boo: yes. there is both an installation shell script if you have an existing hardy install, or a live cd based on hardy for a new install. you can find both on the linuxcnc.org website
yes, but you need the rtai kernel
i don't want to modify anything in my system but feel relaxed to add a kernel that can be removed later
it's a package, just like the -rt kernel
you can remove it with apt
what am i waiting for
mr_boo: just out of curiosity, do you have an actual machine in mind to run, or are you just experimenting
yeah, emc installs and uninstalls like any other package on ubuntu. same with upgrades, they go through the ubuntu package manager.
i've got a wire winder
like, a real wire winder?
no, a crappy home made winder
let me show a photo
still, that's cool
here it is, http://www.carmi.se/misterstarshine/img/DSC00217.JPG
not sure it's gonna work well
but i've moved the wire guide closer to spindle
ah are you the fellow formerly known as bo-dick? the leading part of that url looks familiar.
uhm, i guess i'm revealed
a challenger approaches!
don't worry, i'm still a troll
i am about to put debian base back on my pc
upgrading to xubuntu 8.04 slooooowed my laptop down
you could also fix it by buying a new laptop :-P
i have been seriously considering that option
* mr_boo clicks the freshly downloaded emc2-install.sh script
a script that does everything automatically for me?
how lame is that
or shall i call it convenient
yeah, that's the right word
it's like magic.
*smart* is also appropriate
particularly when one needs a patched kernel and stuff
wonder if i can have irc running on that kernel
You bet you can. I am.
that's soo cool
well, lets reboot then
just make sure you have the right kernel selected ... oh well
and he's off
I presume he is making home made guitar pickups with his coil winder
see a lot of him in #electronics
im running rtai kernel now
it didnt recognize graphics card or usb mouse
so ive got keyboard only now
close enough ;)
the mouse thing is most annoying
lets try with the ps/s dongle then
that's quite surprising
anyone else running usb mouse on rtai?
so all hardware that works on the other kernels are supposed to work here as well?
I can see it not recognizing the video - that happens.. but not the mouse. what hardware is this?
"all" is a very strong word
mr_boo: not 'all'
this is a logitech mouseman mouse
but sure, usb mice should work fine
well, lets reboot and see
you shouldn't need to reboot to check the mouse
unplug it and plug it back in
rebooting is also unlikely to help
worth a try
you can try the PS/2 dongle
you don't need to reboot
what hardware is this?
ps/2 isnt hot pluggable
maybe ctrl-alt-backspace to restart the X server
it is on Linux
it won't automatically start programs like USB would, but you can certainly unplug/replug PS/2 devices with Linux
ps/2 dongle did the trick
maybe i should accept that the x server graphics driver doesnt work with rtai
the vendor drivers screw up realtime anyway, in many cases
mr_boo, re coil winder, half the battle is back tension on the wire feed, felt between washers with a light adjustable spring to squeeze together
often with lighter sizes of wire pull off the side to the feed real does not rotate (wire through a guide hole)
ive got a threaded guide
cant stand this graphics resolution
how can i change it to vesa for this kernel only?
that goes in the xorg.conf file, which is not specific to one kernel
i cant stand 800x600
"For the prototype I used my usual 1.5 sided board technique where the ground plain is brought to the top of the board and the alignment can be off by a couple of mm." --http://dorkbotpdx.org/blog/feurig/benito_my_first_at90usb162_project
Well this is obvious in retrospect: carefully lay out the board so that alignment after flip is less critical.
* jepler files away for future reference
jepler: if you make up your pinned alignment table - they will be right on.
did you ever get any spring dowels?
it's been so long I can't remember what we planned
cradek: no, I didn't order any
I acutally use brass pins. (I take them out after it gets vaccumed down
jepler: I think they have them at the hardware store
or if you want them semi-permanent, I could press in real dowel pins
mmm vacuum table
only way to go.. ;)
it felt like it took ages to get the tape and left-over adhesive off the last thing I cut from copper-clad http://axis.unpy.net/01221358219
how come the login screen in hardy is a different res then the desktop?
skunkworks_: it is?
on this computer it is. The monitor can't handle the resolution that the login screen is at.. (seems to be max res within hardy)
because the desktop resolution is a user preference
the login screen always uses the first mode specified
then it switches to the user preference
jepler: that's extremely cool (use lighter fluid)
(once you've logged in)
cradek: I used one of those "citrus"-smelling solvents .. undiluted it worked pretty well
first I wasted time using it diluted (which is what the label says to do) and then I was cautious because I didn't want to break the thin bits
so - if the xorg file actually had the resolutions in it - I could delete the highest one? ;)
or re-order them
just stick the resolution you want at the front of the line
hmph I can't immediately find an eagle library with at90usb82/vqfp32 in it
seems like the vesa driver is used when everything else fail\
so now i can toggle between 640x480 and 800x600
vesa should be able to go up to 1600x1200, or maybe even 2048x1536. the only thing it doesn't seem to do is support widescreen aspect ratios
everything above 800x600 seem off limits to me right now
im using the proprietary nvidia driver usually
but it doesnt appear to work with the rtai kernel specifically
no, and if it did, it would probably intorduce timing problems
for running emc, remove nvidia card, drop in circular file, get matrox pci or agp card, insert
whats wrong with pciexpress?
nothing much, except that most PCIe cards use proprietary drivers which screw up realtime performance
no idea, none of my computers have it
and if you don't use those drivers, you're stuck with problems like not being able to get above 800x600 sometimes
some Intel and SiS graphics apparently aren't always as bad
maybe i should consider not to use my personal computer that i use for everything else for cnc purposes as well
a more elegant solution would be to have the realtime part of the system implemented in hardware at machine side
pc computers arent known to be the best choice for realtime purposes
let us know when you get all the fetures of EMC2 on such a board
mr_boo: even high-end machine tools don't have that level of abstraction
that wouldnt be necessary
well, one or two do
just have the realtime part there
and let computer be the brain to determine what to do rather than exactly when
search the IRC logs and email archives for discussions of what is easily done externally and what doesn't work well
it's a good idea, but a difficult thing to engineer so it actually works
for what ive understood the step pulses are the critical part of the system
mr_boo: the steps in your case are a simpler thing, the steps for multiaxis machining are not so simple
only if step pulses are your chosen method of control, and you have no desire to control motion from external feedback
so this is why the -rt kernel is sufficient for servo systems
no, it's not
if the PC needs to generate step pulses or PWM waveforms directly, then it needs exceedingly tight RT performance
the control loop, which looks at inputs and makes decisions on what to output, only runs at 1000 Hz by default
so -rt is probably fine for that, but it isn't for the 30000 to 50000 Hz needed to generate steps/PWM
so emc2 features 50kHz step rates?
think about 9 axis all being controlled at the same time mr_boo all to a few tenths at speed
if you have something like the USC from Pico Systems, or one of the Mesa FPGA cards, those will generate steps instead of the PC CPU having to do it
step rates depend on the CPU
and other things
btw, i tried to start emc2
i get an error dialog
gonna post it in pastebin
here it is, http://pastebin.com/m5e44d7ee
wonder if ive screwed up somehow
im pretty sure im the cause for this confusion
Looks like your system is missing something to do with OpenGL
_tkinter.TclError: Togl: X server has no OpenGL GLX extension
rayh is right. your Xserver cannot do opengl.
this is surely because you messed with the nvidia binary driver
the server doesnt even use the nvidia driver on the rtai
yes, but the nvidia driver has a special way of crapping on the software opengl bits
sorry, I'm not even sure how to get all its bits uninstalled. maybe you can find something on ubuntuforums?
so you recommend me to remove the nvidia driver entirely and run vesa?
does it work using a livecd?
i dont really wanna remove any drivers from my system
mr_boo: does it work using a livecd?
Jymm: to run emc2?
what would be the difference?
you haven't touched it =)
Jymm: do you mean the emc2 live cd?
Jymm: what os is it based on?
Jymm: err, distro
does it matter?
there are 2 - hardy and dapper.
not really, just curoius
cant see how the live cd would sort out my hardware compability issues
did you install an nvidia driver?
It's a simple test, if it works then you know it's your install config.
if it doesn't, no loss
lets try it then
burn it at 4x speed, and be sure to MD5 the iso after you dl it
you might also try sudo dpkg-reconfigure -phigh xserver-xorg
ill go for the cd i think
lets uninstall everything on the hd then
if this is an install you want to keep, then the CD will only be a diagnostic tool
i dont think its gonna work with my pciexpress card
noone else uses it with such a card
PCI express works fine, you just can't use the proprietary drivers
shame that emc2 requires opengl
yesterday I had to drill a rectangular pattern rotated 2 degrees from horizontal on the mill. I had a brief moment of panic when I couldn't rotate the vise - I keyed it to the T slot.
mr_boo: only one GUI does
you can use EMC without X
then I remembered my brain and rotated the four points instead of the vise. jeez.
nto that anyone has tried in a while
you do have a CNC after al
my guess is that the simulation tool is the part that needs opengl
even al has a CNC, after ALL
the preview is what needs openGL
personally, I don't see why that's a shame
it's a shame when you've installed proprietary drivers that have borked your openGL setup
and they're hard to un-bork
I think it may be as simple as removing a couple of links and renaming a couple of files
I think nvidia links libgl.so.1 (or similar)to their own driver
or using apt ... somehow
yeah, I don't know. it may be possible to remove --purge nvidia and mesa drivers, then install only mesa
lets see if the livecd has any answers
how do i remove the stuff?
is it just to remove rtai kernel and emc2 in synaptic?
try booting the liveCD before messing with what's installed
im not prepared to do any major modifications to this machine after all
as to package removal, emc and its RT kernel are packages which can be removed the same way you'd remove any other package - using apt or synaptic
Lerman__ is now known as Lerman
cad apt retrofitting http://www.me.uvic.ca/~mly/bookpdf/book.htm
somebody here to help french
bah only the index
the apt book?
u can get the chapters, clik on the entries in the index, save as
nice apt chapter
just come indoors after fitting a new alternator to the chariot
experienced mechanic quoted 2 hrs, us two amateurs took 2 hrs!
2 hours, or 2 man-hours?
so it only took you twice as long - nice! :)
but if you can get paid for it, then you have experience and are pros
nah two mechanics at the auto electricians
bu we know where we wasted time, will be faster next time
I still hate working on cars
the belt adjuster was the hard part, you have to remove trim behind the front wheel to access to loosen as its behind the transverse engine below the alternator
I shall have to leave the cnc on all night as its mid job
I need help
I can't adjust the PIDs
My servo has a dead zone
what is your hardware?
continu current ?
uh, DC - Direct Current?
does that mean no tach feedback?
a user has contributed a component to compensate for this kind of deadband in servo amplifiers. It is called 'idb' (inverse deadband). For more information, see http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?ContributedComponents#idb_Inverse_Deadband
tuning will depend on the type of drive you have also
to add that component to your emc installation, there are instructions at the top of the wiki.linuxcnc.org page
that would be because of 'fast decay' drives?
sorry, i had trouble
thanks for this information, i will try later
fast decay drive ???
what are you using to run your servos?
motenc -> drive -> servo
Didier: it refers to name
"fast decay" refers to one of the ways the transistors in the H-bridge can be configured when it's in the "off" portion of the PWM cycle
motenc -> mdp -> dunkermotoren
i can't change anything like this in my servo...
it seems that the drive doesn't between +-300 mv....
work , sorry
it may have to overcome static friction when starting up
I think 'idb' is the thing to try -- if you have the PID output in units of 1 = 1V, then an idb "amount" of 0.3 would be the place to start.
it does not seem like that would make it totally untunable
neither should a little bit of deadband, I'd think
what does deadband ?
the dead zone you mentioned, where nothing happens until you get to 300mV
i move the motor by hand
there is deadband in the ini which is different than what we are talking about... are you adjusting that by chance wrongly?
i let deadband with 0
that's good. what do you mean "I move the motor by hand"?
no reaction (hand) between +-300mv and hard reaction after
ok, with the motor off the machine?
no with the machine on
i watch the pid output on halscope
in the past, i used brushless drive, and i never see it
inverse deadband seems to be good
i say later if it's ok, thanks for all
see you later Didier
good morning for you, and good night for me, bye
getting a servo to respond to the lsb of the dac is a holy grail, but the fewer bits it takes to respond is a good measure of your systems..... uh, responsiveness
300mv out of 10V is a bunch, i'd look for reasons why it's still in the deep freeze till i hit 50 to 100 mv. dont quickly put band aids that bang it out of the stall.
anyone ever install ldap for home use
for like, login and stuffs
not for home use, but i have, toastatwork
toastatwork: that's probably the most common use of LDAP
toastatwork: using clear text is NOT recommended
i don't plan on using cleartext
at one point i tried to do ldap + kerberos
kerbos is a module to ldap iirc
but i didn't understand how to manage ldap, and i couldn't get kerberos to grant a ticket for anything
toastatwork: learning LDAP is NOT going to be easy
maybe i'll get a book
toastatwork: I printed out the manual for OpenLDAP today, but been reading it for about a week
a little casual philosoohy eh
all 200+ pages
toastatwork: then, you'll need to read up on LDIF, openssl as well
s/,/ and /
I have about 5 RFC's printed out, still not done yet. LDAP has about 20 RFC's directly related to it
toastatwork: do you have a M$ box?
i do, but that's not the one i care about
if i do this i won't be using ldap for windows
softerra ldap administrator 0 comes witha 30day trial
softerra ldap administrator - comes witha 30 day trial
but win only, there is also phpLDAPAdmin
toastatwork: or, say screw it and just issue client certs =)
or look at gosa, which is simialr to phpldapadmin, but geared more towards user administration and has some kerberos support i believe
gosa is a ldap admin tool that's geared for ldap user administration, adding/removing/etc
it looks cool enough
steve_stallings is now known as steves_logging
hi all... another day of Catia training in... its almost harder than work... staying awake that is
catia is some fine software
i like it alot
u work in aerospace?
they seem to be the only people able to afford it :)
boeing, Bombardier,Boeing Philly...
its not that bad for a seat these days.. for what you get
and Dassault... go figure
I thought Bombardier made yellow things that fly over snow
last i checked it was around $45k for a seat, plus yearly maintenance
they bought out boeing after they bought out deHavilland
hmmm spilling a glass of wine on a keyboard doesn't seem to have any ncbpaqh/ fljaogoih?Ld
for what a global express??
a Global is worth about 30 million US dollars
i was trying to talk work into getting us a license and thats what the quote came back for
dare... how much do you need... i know fixture guys on 12000 dollar no maint. seats
it was for automotive with sheet metal, machining, cam, and fea
even with prismatic and surface macining maybe 25 tops
auto and sheet metal.. is NOT catias forte... they dislike class A surfaces
look into icem or Cadra.. they are true class A surface engines
well its just what our design dept uses now
i work in manufacturing so we get the 2d prints and limited iges
what function do you have in MFG??
used to use a proprietary system but now they've migrated all new models to catia
autmotive manufacturing engineering.... just about all the functions I think
mostly tooling specification and design
crap your 1 a me
i used to was a auto plant Engineer... for a caddy fuel pump the.. a whack of
synchro-sleeve assemblies... in the 1000's and 1000's per day
yeah we do some volume
i quit... the #'s were too much for my likin' always the same crap
my lines produce over 2000 vehicles for sale each day, and the 4-cylinder line is making almost 6000 motors a day now
went back to prototypinf for a few yrs
i enjoy it for the most part
they kept me in set up mode.. as soon as it was running i was pulled to some other disaster
they wer tier 1 to GM and Chrysler
we run pretty much 99% for weeks on end, not too many disasters really
Powder metal product and assemblies
that sounds cool
i've never seen a powder metal process before
not really..its miseable stuff to machine
after sintering and compacting.... there is no real grain flow direction... so the best analogy is imagine turning a grinding wheel ( face grooving in our case)
did you manufacture from the powdered metal?
or get blanks in to work on
we were the sister plant to the black shite so the sent it to us to finish
we did end up all under 1 roof by the time i left... juts alot of dept.s
anybody know what optoisolator output is most commonly used for level conversions?
i like the idea of ramping up to 60 shipsets a year for a few yrs... but some parts take the better part of a yr to make
i see a ton of choices here
i know very little of hardware on board level...
i know enough to be dangerous
but no more
me too.. so i dont pass on advice
i want to use a 5v / 20ma signal to push a 12v signal out the other end
so do i... LOL
i.e. parallel port signal convert to vfd signal :)
preferably without using mechanical relays
looks like we may have to go for 1 of those 17 dollar boards
i have relay board to turn it on and off... but i want variable speed
the problem is that there are just too many choices
on digikey there are over 5000 types of optoisolators
i'm sure probably 3000 of them would work fine for my application....
i'm trying to build a board that has 5 digital IO (optoisolated), and 1 pwm to analog signal
i've got the pwm to analog signal design done, but for the digitals i'd rather not use relays if i can use a $0.50 opto instead
phototransistor output, common collector, with or without base, logic output, so many choices