Has NE1 successfully set up EMC in unbuntu? Or do I have to go back to dapper drake? My Bootmenu is filled with linux versions
cradek: what are you making?
jmkasunich: a vacuum table for max
OzarkMark: emc2 packages are prebuilt for Ubuntu 6.06 "Dapper Drake" and Ubuntu 5.10 "Breezy Badger". The next release of emc2 will be for 6.06 only.
I got to drill a lot of 5" deep holes
the next pre-built release that is
OzarkMark, um - most people here have either (a) been able to install and set up Ubuntu/EMC2 successfully or (b) are the people who created the packages and liveCD :)
Thats what I was afraid if. I installed Feisty before I new about the 6.06 ISO
you can steill get it to work on Feisty, but there aren't any precompiled packages
I tried B=Live CD. I have a screen Res problem that I can see the OK button to even start EMC
OzarkMark: the precompiled packages won't work on Edgy or Feisty. It's a lot of work (you have to do things like patch and compile kernels) and if you're more interested in just using emc it's a better idea to use Dapper.
It defaults to 640x480
Can I wipe my linux partition and start over?
OzarkMark: are you using a KVM?
<--newb to the penguin OS
Cradek ..you have a peck drilling macro
dmess: of course :-)
OzarkMark: yes, carefully note which partition(s) you used and tell the installer to use those when you reinstall. You'll loose any files you put on there (including your home directory), though
deep drillng macro??
OzarkMark: this page may help you fix the screen resolution after you finish the install: https://help.ubuntu.com/community/FixVideoResolutionHowto
I dont care about losing anything.I only have linux for a week now. But tried doing unbunto kernal builds and have 8 different startup options then some1 told me about 6.06
dmess: I used G83 for everything
ubintu has worked on most machines i tried it on...
ATI RAdeon 200 series
Unbuntu resolution is great
no that was meant for OzarkMark
the dapper res defaults to 640x480
5 inch deep holes? for connecting all the vertial vacuum holes togather?
skunkworks: then one 6.5" deep the other way to hook them together
nice. your going to wonder why you didn't make one sooner.
getting frustrated. was gonna go back to mach3(cringe)
no.. whack the box again..
cradek: what are you going to put on the top for actual drilling thru the copper clad?
OzarkMark, if you don't mind losing everything, then a fresh install with 6.06 will be your best bet. Just boot with the 6.06 CD or the EMC2 LiveCD, and while it's at the CD boot ment, press F4 (I think) to change video modes. You should be able to choose 1024x768 or thereabouts
err - menu, not ment
ok thx. for the 411. I've been on CNCzone but the turnaround time on info is way slow
when you install, just select "use entire disk" if you really don't care about what's on it :)
will wipe the linux partition and install 6.06 exclusively
yeah - especially for EMC questions. I think only one of us here is actually on cnczone
Cant use whole disk, have to maintain a small windoze part.
ok, then you need to be more careful. you may want to delete the Linux partitions and then select "use free space"
can I do that with GREP?
no, I don't think grub lets you delete partitions
(which is probably a good thing)
do you have one Linux partition (plus a swap partition)?
the boot dis has the capability ti sesize partitions..
Ive got Windows and Linux. But the linux has about 8 or 9 start options
ok, start options and partitions aren't the same thing
it would be a very good tihng for you to understand the difference before doing the install :)
remember, I've still got water behind my ears
SWP- ya, in a perfect world, I would take the time to learn everything, but I just want to get my comp set up for the cnc machine. There were several people within 100 miles that run linux and emc so I figured that would be the way to go if I needed support
way to go :)
So if worse comes to worse, Ican take a road trip
G2G will be back now that I found this place. Thanks again, The pest will return!
I dont think grep allows that either ;)
swpadnos: i ordered the usc today
the emc2 disk does a very painless dual boot setup if you wanna have windows also
use the create partition in free space option,use the slider to select the size,it does the rest
did it on 2 different computers this week
Anyone in here have renter's insurance?
I don't rent
when I did rent.. I did not have the insurance
gEDA components for m5i20 7i33 7i37 on wiki http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?M5i20_Halvcp_Test_Panel
SWPadnos_ is now known as SWPadnos
[03:51:49] <Skullworks-PGAB> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?CoordinateSystems
- last line - "I'd also enable G43 while doing the Z Touch Off."
um - why?
Skullworks-PGAB: if your tool is not your "reference" tool length you'll get the wrong number otherwise
"the acceleration and deceleration functions on most controls are much different for positioning than for feeding. G1 generally has a much shorter, sharper speed ramp than G0. If you routinely move at very high speeds with G01, on a machine that's not specifically designed for very high-speed machining, you'll beat the screws and thrust bearings to death."
there's more detailed explanation here http://www.desktopcnc.com/swarf/dog_leg.htm
it probably doesnt apply to milling machines though
I've seen this in some youtube videos
it seems scary and emc works better
I think the ngc spec doesn't specify that the axes will be coordinated
the dog leg thing is stupid
just that they'll get there rapidly
i think i understand what this is about now - the guy specifies the amount of time a move will take, and the control determines the accel from that
no, it's max speed on all axes, with each axis stopping once it gets to the correct position
SWPadnos: ngc specifies that g0 moves are coordinated
at least, that they'll all end at the same time
'set the target and let pid take care of it' is an easy way to do rapids on old machines.
'tell me when you get there'
i dont really see the point since you have to be able to do G1 anyway
hey - it does say coordinated now :)
I'm pretty sure it didn't say that a year ago
I bet you're mistaken, I doubt anyone changed that
i remember it saying coordinated since always
perhaps you fell through a time portal without realizing it
that could be
I'm obviously wrong about that, I've just proven it to myself
"but I could swear ..."
what said coordinated?
the RS274NGC spec, revision 3 (or later)
and maybe earlier as well
here's one: http://www.isd.mel.nist.gov/personnel/kramer/pubs/RS274NGC_3.web/RS274NGC_3TOC.html
I have a version from 2000 on my local disk, which says the same thing
SWPadnos: well 2000 aint a year ago buddy =)
it's at least a year ago
I'm quite sure of that
a year ago is did NOT say coordianted?
if I had read the 2000 version a year ago, and it didn't say that, then it would be true that a year ago, it didn't say that
(of course, I was wrong about it anyway)
ok, pick a year from 2001 to 2006
fine you search then... http://www.archive.org/web/web.php
the 24 1/2th Century!!!
ok Duck Dodgers!
SWPadnos: (I got all those on DVD too)
I saw one on New Years Eve
a movie theater had looney toons playing from noon to midnight as part of the First Night festivities
just walk into the theater (free), watch some cartoons, leave when you want ...
it was cool to see them on the big screen
it seems endmill holders should have a slot cut out opposite the set screw, thus giving you two lines of contact and one point contact, instead of one line and one point
2007.09.15 gcam released
maybe axis could have a 'halmeter' to show values in the 3d window like with x y z vel
thinking about the guy with the 800x600 display
axis is slow.
isn't anyone coming to axis' rescue?
SWPadnos: That would have been very cool.
SWPadnos: A friend of ours has a place in the river in AZ, in the summrertime, him and allthe neighbors hit th bakcyard, and they bbq whatever and play movies on the side of the 2 story wall outdoors when the breeze picks up.
everyone just brings their food, lawn chairs etc
who brings the CNC? har har.
I want to install emc2 in debian etch (stable - 4.0). I just read http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Debian_Etch_Compile_RTAI
, and I have to say I was surprised by the complexity of the installation. Is there any other way to install emc2 on debian? Anyone here running emc2 on debian?
just pretend you're on ubuntu and install normally
it should work? I'll try, thanks.
well, officially it's 'not supported' but i've done it before and it worked. i'm having trouble getting apt to work with linuxcnc.org lately though
do you happen to know what's the equivalent Ubuntu version of Debian Sarge? (I guess the stable version of Ubuntu)
sarge is pretty old, i think etch is stable now
i'm also making cheese
too many carbs
cheese is ok, as long as you aint cutting it
carbs are good for you
cheese is evil nasty stuff
fuck no thay aint
i stirred the curds as they were forming so i got this grainy stuff that isnt really cheese at all
That reminds me, I gotta find a cheese tasting place
curds?! are you seriously making cheese?
we get raw milk from a local farmer; had some extra this week
making butter too? lol
CNC butter churn
nah not enough milk to make it worthwhile, but i just didnt want to dump all that pretty milk down the drain
* JymmmEMC hates when an auction will end when he'll be aslepp
* fenn hates auctions
Eh, a 7" LCD TV I wanted to suprise my gf with
tomp: ever heard of a vernier optical encoder?
never, the reader is usually pretty simple. vernier implies the reader is a staggered array . is that right?
idea i just came up with - you have a mask grating that has 1 more line than the encoder disc. as the disc goes through one phase cycle, a dark spot appears to circle around the encoder 360 degrees
so you have two (low resolution) sensors 90 degrees apart on the disc, that gives you sinusoidal quadrature output
problem is that the two gratings have to be close together, if they're touching that means wear
top & bot of 2 glass plates?
i was thinking printing on transparency from a laser printer
do you think glass would work better?
etch top of one and bot of other? how close you gotta be?
glass is whats used in encoders by people better at it than me ;)
well i suppose the light could diffuse/diffract in the space between the two gratings
uh, short answer is "i dont know"
oops, google it 'vernier encoder' (1st hit is 'priorart' )
you could do two dimensional gratings too
pattern of holes and dots
or, more likely, hexagons
anyway i thought you could use that to get more resolution out of your encoders somehow
btw i have been playing with postscript for generating encoder discs, i can probably make one pretty easy
i read some discussionon that, like alex's cd printer. i remember old 8" floppys had a series ( many ) of holes used to make precise timing
alex_joni:have you ever fooled with a usc board?
around here only SWPadnos, anonimasu, jmkasunich and elson have
and rayh I think
k,thought some said you had
got 1 coming,was wondering how hard its gonna be to do the pid loop
on the emc end
alex_joni:do you know anything about the adaptive feed in emc
i noticed its in my hal
jlmjvm: I know a couple things
1. you can switch it on/off with g-code
2. it basicly works as a realtime float input (from 0 to 1)
3. the value will be used to scale the speed (0 - 0%, 0.25 - 25%, etc)
so if your adaptive-speed input is 1, emc2 will go at the normal speed (taking into account programmed feedrate, feed override, etc)
if the adaptive-speed input is .5, then it will be half of the above
so if its set at .5,would the mill go down to 50% feed under a load?
jlmjvm: if you set it at .5 then it will always go at 50%
ugh writing the code for these encoder things makes you dizzy
what you want to do is take some value (maybe force measured on the spindle, or the following error), and scale/offset that to result in a 0..1 value
link adaptive feed and following error
adaptive feed,m53 is feed override
say you have following error from 0 to 0.05 "
at 0 ferror you want 1 adaptive feed, at 0.05 you want 0 adaptive feed
connect ferror to an add2
connect -0.05 to the other input
and scale the output with 0.05/1
tomp: it works, try yourself if you have a laser printer http://fenn.dyndns.org/pub/irc/encoder-vernier.ps
edit the .ps if you want to change any parameters, but its sorta hacky right now
fenn: can you generate a pic?
[12:25:48] <fenn> http://fenn.dyndns.org/pub/irc/encoder-vernier.pdf
you can print on plain laser paper and hold up to a light, put a pin through the crosshairs
thanks alex,im gonna try this later today,gotta go work on some new mazaks for a few hours
jlmjvm: ok, cool
let me know how it works
i should put a giant one of these on my bicycle
will do,i think it will be like you said,an unstallable stepper
jlmjvm: we'll see :)
see you guys later today,gotta run
although I wonder..
lets assume the stepper does lose some steps
that will be noticed by emc though ferror
but the commanded position will not change
and stepgen (taking a commanded position) won't compensate
fenn: care to back me up on this?
thats why you use adaptive feed to slow the commanded feed down
i guess you need PID too
if you need to recover a lost full-step
that would assume some velocity command, isn't that true?
or PID on position?
i think it would recover a lost half-step or less by itself
pid on position
I'm thinking about a couple lost steps
till adaptive starts to kick in
or a stepgen with an external feedback :)
it already has a PID inside
i think your P gain would have to be pretty low or you'd overshoot
because of the delay in step/dir signalling
how much do steppers normally lag behind the magnetic field phase?
cant be more than a fullstep
I have no idea, but I'm thinking about a really bad case
where for example the encoder has less resolution then full steps/rev.
i'm trying to figure out what the stepper's "internal PID loop" looks like
not inside stepgen, inside the motor itself i mean
there's P due to the magnetic field, and D due to friction (?)
eh, nevermind that train of thought
I think the motor lag gets to 1 full step (360 degrees)
and then it starts losing steps
ISTR Mariss mentioningthat regarding his new step/servo drive design
I guess simply linking ferror to adaptive feed, will reduce the speed of the machine in case of error, and never get back to full speed
you probably want to use freqgen instead of stepgen for a feedback system
you could add ferror to the position command
fenn: that's an interesting idea
you still need the adaptive feed stuff, but you'd want to have that decay toward full speed (so it's always trying to get to full speed, and only slows down when "actively" pulled away from full speed by a following error)
I think adding ferror to the position command is unstable
its like P in PID
with a gain of 1
(max_ferror-curr_ferror)*1/max_ferror -> adaptive_feed
so you'd add it in HAL
you can scale the value to suit your taste
SWPadnos: only as input to the stepgen
that's OK for servos (right - it's like P), but steppers need you to slow down
so you'd have to slowly add it in, after the adaptive feed has slowed things down a bit
no, you want to speed the motor falling behind up, relative to the commanded feed which is slowing down in response to ferror
so multiply it by some value less than 1
so the motor that's falling behind is slowing down in reality a bit
its slowing down by K
no - I think there's a problem there
there's too many constants :)
motion outputs points along a path, sampled in time. at some time, a motor can't keep up
so the whole system is slowed down using FO
hopefully, the motor can catch up , but the question is, what if it has lost steps
(ie, the impending stall wasn't detected and corrected for before a step was lost)
but motion is still happily tossing out uncorrected positions - though at a slower rate of motion
SWPadnos: that's why we want to add ferror to the input of stepgen
so it travels towards the real point (where it should be)
sure - it's just that you'll need to add that error in until you restart EMC
but if you add ferror*1 to the input of stepgen, the motor that's struggling doesn't slow down at all, and presumably continues losing steps
motion says "go to 1.000", but the motor only makes it to 0.999
do it in hal
so the AFO code automatically slows everything down
there's another problem
but stepgen still thinks it's at 1.000, not 0.999
so you get past the roadblock, but there's still this following error
once the motor catches up (ferror =0) stepgen will be at 1.001 not 1.000
so you lie to stepgen "go to 1.001"
and it will want to go back to 1.000
the difference between stepgen's position and emc's commanded position is the number of steps you've lost so far
SWPadnos: emc.pos-cmd + ferror
I think we need another component here
you have to keep that ferror correction until you turn off the machine (F2 = machine off, where I think stepgen will reset its position (?) )
emc -> component -> stepgen
naw its simple
a scale block and an add block
fenn: actually a bit more complicated
lets take an example
emc-cmd is 1.000
stepgen only goes to 0.999
actually, it's an integrator block - you want to develop a positive error in the stepgen command to counteract the negative error caused by the missed steps
ferror is 0.001
why does stepgen only go to 0.999?
lost step (it thinks it's at 1.000)
stepgen thinks it's at 1.000, but the motor missed step
ok, let's talk about what the motor's doing then :)
ok, so motor is at 0.999
ferror is 0.001
stepgen cmd will be 1.001 (emc-cmd + ferror) right?
motor goes to 1.000
stepgen thinks it's at 1.001
ferror is 0
stepgen input will be 1.000
stepgen goes back one step
and it starts to wobble
I'd say you need some smartish (not sure how smart) component
p=1, i > 0 ?
as ferror increases, it slowly increases the error output
no, that would be equivalent to ferror
you need to keep track of ferrors that have developed over time and feed those back into the stepgen command
PID would have the same problem as just adding ferror
pid that takes emc command, and encoder feedback
fenn: no, as it doesn't know/care where stepgen thinks it is
but PID only works on the current state, I think some history is needed
you need to count the number of lost steps and always add them to stepgen-in
I really think we need a velocity-loop
then this adding in goes away
emc-cmd > ddt > pid > freqgen
bet it will be a pain to tune :D
does stepgen really have an internal PID loop or is that just a myth?
fenn: it does have something
it does have limits on accel & vel
it's not PID, it's an accel/vel limited ramp
so it needs to have something..
SWPadnos: I think jmk said it's some crude pi
not sure about d :)
obviously P and I are there to assure it reaches where you want to go
could be. I don't remember (and I think I'll make coffee before looking :) )
even if it's not explicitely stated
you guys are going about this "unstallable stepper" thing all wrong ;-)
you _cannot_ deal with a lost step
the idea is to prevent it from losing a step
I'll just design a step drive that does that. I hear it's easy
(once you know how to do it)
jmkasunich: why can't you deal with a lost step?
its only if you're a motor drive designer that you can't deal with it :)
has to be done in the control
fenn: I know that..
if you lose "a" step, you will actually lose 4 full steps
one electrical cycle
um, isnt an electrical cycle 1 full step?
and when it happens, the motor will make a jump
you think it bounces back?
it starts accelerating backwards
you can get the motor to run backwards at 1/4 (?) the speed
winding currents during a full cycle (4 steps) = 0+ +0 0- -0
in fact, full step winding currents look a lot like quadrature encoder signals
on a side note, I had a little fun connecting two stepper motors together. turn one shaft fast enough, and the other motor stays in step
lets talk about the "bounces back" thing
then simply put another stepper and use a hall probe to make an electrical encoder
assume you energise one winding in one direction - that is a particular full step location
alex_joni: i believe that's called a magnetic encoder :P
fenn: ok then
there is a magnetic field in the motor from the windings
and if the torque is zero, the magentic field from the rotor lines up with it
now apply some torque
the rotor gets displaced a bit from alignment
and the further you displace it the more it resists
up to a point
the resisting torque is a sine wave
increases from zero displacement to 90 degrees
above 90 degrees, it begins to decrease
at 180 degrees it goes negative, and begins driving the rotor away from the initial stable position
to a new stable position at 360 degrees
and then depends on whether you keep applying torque
experiement: clamp a 200 step/rev stepper to your workbench, energise it with your favorite drive (no steps, just zero speed)
put a wrench on the shaft, and apply torque
what is actually pushing on the motor to make it go backwards, in a milling machine?
count the number of cogs per revolution
fenn: load torque
my experiement is on a stationary motor, but it also applies while spinning
inertia for forward, or backward during decel and accel phase
relative to the (rotating) magnetic field
if the observer rotates with the magnetic field, they will see load torque trying to turn the rotor backwards
so, if we slow down that magnetic field, and the motor stays in the same place, the motor is "accelerating" forward in the electrical cycle
and if the torque is high enough, it will slip backwards just like in the zero speed experiment
you lost me
"the motor stays in the same place" relative to what?
the rotating field, or the world?
relative to its own inertia
sorry.. there's no change in the motor's inertia but it appears to be accelerating from your observer's point of view
inertia isn't a reference frame
it's accelerating because we're decelerating the magnetic field (from emc)
as soon as you start talking about acceleration and deceleration things get a lot more complicated
i dont think it should be complicated though, let's talk about something else :)
first you have to have a mental model of the nmotor behavior
which for me is a rotating magentic field (rotated by the drive), with a spring connecting it to the rotor
a floppy compression spring maybe
a spring that can easily break :)
if the spring gets stretched too far (90 electrical degrees) it "slips" one "tooth" (4 steps)
not floppy at all - in the range from -45 to +45 degrees, it is quite linear
ball detent is a good analogy
from 45 to 90 it rolls off instead of being linear
above 90 the force decreases and eventually goes negative
the ball detent is a _fairly_ good analogy
but balls often lock up completely until you exceed some force - steppers will deflect from perfect magentic alignment for _any_ force - hence the spring
this doesnt explain why we can never make up a lost step
the spring, combined with the rotor inertia, is a spring-mass system, with a resonant frequency, and that is what mid-band resonance is all about
its not that you can never make it up
its that when you lose the step, you have a completely uncontrolled movement of 1/50 of a turn
once you go over the peak and the "spring" breaks
IMO, the goal of an unstallable stepper is not to recover from that, it is to prevent it
right, but that all has to happen inside the stepper drive
not with microstepping
you _do_ need a microstepping drive, so that you can control where the current vector is
the encoder tells you where the rotor is
given enough accuracy
you watch the difference, and as it approaches 90 degrees you back off to allow the rotor to catch up
alex_joni: exactly - you don't really know where the current vector is
because the actual current lags behind the command, due to inductance etc
except if you do it in the driver.. then you might know
seen mariss's latest info on the geckodrive list?
I read it on-and-off
I'm aware of what he's been doing with the unstallable stepper, but haven't looked recently
he is basically building a vector drive
[13:56:06] <alex_joni> http://finance.groups.yahoo.com/group/geckodrive/files/Stepper%20Servo/
there are some results there
i think we were going somewhere with the 'integrator' component
this component should add ferror to an accumulator, then set ferror to zero (?)
why shoudl it set ferror to zero?
emc needs to know it's off-course
so you dont add it to the accumulator twice
and it needs to slow things down
that's what a pid does usually
how does pid know it already added the error?
it doesn't add anything
it increases output untill the error goes to zero
it compares prescription and feedback, and adjust the output
gotta go - places to be, things to do
ok, but, that oscillates because the commanded position is not the same thing as what PID outputs
fenn: except if it's velocity
does velocity -> freqgen actually get you to the right place eventually?
why shouldn't it?
velocity -> DC servo does too
there's a problem with PID though
it usually increases output as ferror builds up
and that's bad for steppers
ok, so i guess the only thing left is to try it and see
* fenn volunteers jlmjvm to solve this pressing scientific issue
well.. I don't have encoders on my steppers.. so it's up to jlmjvm
but with high enough resolution, slowing things down (fast enough), might actually make it work
and if it really can't keep up, it will stop
F2 to reset and you can continue without issues
why wouldnt it be able to keep up?
I'm talking about the simple setting
ferror connected only to adaptive feed
as jlmjvm is going to try first
and assuming the system speeds down too slow.. the motor keeps losing steps
eventually it will decelerate to a halt
'error 54321 Operator needed to turn handle'
error 54322 would someone please help me?
I seem to be stuck on one axis
i think the complicated solution sounds more useful
would be nice to have a pin to axis you could hook up any hal signal to, and it would color the backplot segment that color
error x: error reporting error code
then you could see where feeds are too high etc
error code missing while trying to report the error reporting error code
(that's it, just "eror:")
programmer stuck in a loop
why not grub like?
is it all figured out now?
[14:12:39] <alex_joni> http://youtube.com/watch?v=WefWH8OwtDU
skunkworks: not really
but some problems are clear now :D
we finally have all of our bits time to put them to use
gotta run for a couple hours
* jepler installs dapper yet again
is this a german or english chan?
I don't know if anyone here now speaks german
my english is not perfect, but good enough to annoy others ;)
looks good so far
if you have any questions, just ask them
yeah, i want to use emc2, but i dont know which hardware i can use for a good performance
how much money do you have to spend?
millie: By "hardware" do you mean the PC, or other parts of the cnc system?
right now i don't have a machine. but i try to start learning how to use emc and i have install it on a duron 1300 wit 512Meg Ram and an oooold Riva TNT2
512 megs RAM is plenty
the computer isnt so important really
it runs, but a "good performance" is soemthing different
I found a 1GHz Duron was a bit sluggish when running emc, but it was usable.
(for instance, when switching windows or running image-to-gcode it would take a long time...)
* fenn uses a 1GHz duron daily
mhm, ok. i will try it :)
it should work, but it might feel sluggish
millie: more important than the computer is what kind of electronics and motors the CNC machine will use
Any of you guys metal casters?
Ziegle1: i've done some aluminum
fenn: yes, thats my next question
what can i use?
Well if you are interested... we have a "just for fun" contest starting up:
[15:17:28] <Ziegle1> http://www.backyardmetalcasting.com/forums/viewtopic.php?p=22612
millie: how much money is available for this project?
we are planning to build a 3 axis machine and we want to modify it that i can use it as an SMD placer
hrm. what counts as 'sand cast'
sodium silicate wouldn't fit my definition
basically non-die cast or investment cast
millie: interesting project!
money is a good question.
i don't know
lol... yeah resin bound, or SS bound are probably a no go
if i know what kind of electronic i need i can build it myself
millie, there's a new Yahoo group for people who want to make pick-n-place machines:
[15:19:30] <SWPadnos> http://groups.yahoo.com/group/picknplace
but... i dunno... might be worth it
also, if you look over the recent messages on the Yahoo gecko group, there is a long discussion on the subject http://groups.yahoo.com/group/geckodrive/
millie: inexpensive CNC machines are often based around NEMA-23 stepper motors, driver boards based around chips like ST L297/298, Allegro 3997, all controlled by "step and direction" signals from the PC.
my friend who builds the machine said that up to 3 printerports are enough, but i think its not
for pick/place, it may not be enough
I don't know what kind of I/O a pick and place uses .. seems like there will be a lot of them, though.
I'd suggest using something more, maybe the Mesa 5i20 or 5i22 ($199 to $429)
there will be a lot of io
you need I/Os for the feeders, multiple vacuum lines, the axes (at least 4) etc
count up the I/O you anticipate needing, and then figure out what I/O card to buy
I'd say at least one I/O per feeder, or a really nasty mux setup :)
i think i dont need to buy a i/o card. i just need to know which kind of hard- and software interfaces i need
or you could use a serial protocol and talk to some microcontrollers :)
fenn: thats what i meen
most of the i/o is not critical timing after all
i'm working with AVR Controller all the day
millie, you will almost definitely need an I/O card of some sort, unless you're interested in writing EMC2 drivers for serial ports and the like
SWPadnos: that's what m-codes are for
an AVR can certainly do some of the PLC-like control (advance a feeder, monitor vacuum, etc)
but you'll still want some real-time feedback to EMC, at least to tell the PC when it needs to stop motion
thought there are different io mappers availibel
there's a simple serial protocol that's meant for non-realtime reading of some I/O bits
there are drivers for various servo controller cards and dumb I/O cards
yeah - dpp, I don't think its in CVS
SWPadnos seems to be under the impression that I ever actually wrote or tested dpp
I only talked about it on IRC a few times
I thought you had some code or something
You thought about it, therefore it is :)
if I have the code I don't know where it is
ok, I'm very capable of being mistaken :)
millie: emc2 lets you write new drivers and components, but there's a learning curve
hm, it would be easyer for my to put the whole machine control into external µC than writing an io mapper für emc2
how does a pick-and-place advance the component spools?
fenn, depends on the pick-n-place
lets say you were building one :)
some use pneumatics, others mechanical ratchets ...
well, JonE mentioned an interesting thing about the one he bought
I think he said that as the placement head nears the feeder, a blade slides back (pushed mechanically I think) to slice the cover off the next part. when the head moves away with the part, the feeder ratchets forward, and the blade remains over the part, to keep it from being bounced out from the motion
the part to be picked up would have been uncovered by the first inbound motion that pushed back the blade
that sounds smarter than putting a motor or solenoid on every spool
yep, but mechanically more challenging
what are we on about controlling push off??
especially because the spools have different-sized parts
dmessier: pick & place machine
linear drives are the way to go if $$$ is no object
yeah, except that you may need 100 feeders
deltabots work well
(that's what we were talking about, not the head motion stages)
hmm, it should be possible to control 3-4 steppers with the printer port
at least 4, probably 5 or 6
what are you feeding?
you know - PC board assembly
Surface Mount Technology
the little square things on circuit boards
it is also possible to control som io extensions via rs232. i've done this bevore. its fast enough
i see... speed requirements??
as fast as your wallet allows
well, the fast machines place 18000 parts/hour
no requirements so far ^^
it should just work
and different end of arm stuff required for every component shape
and it should work faster than i can do with my hands ;)
i saw a video of a machine, and there was a red light that blinked every once in a while, looked like it was an infrared lamp heating something
any idea what that was?
each placement is a sequence of move to feeder, pick up part (move head down, engage vacuum, wait for sensor, raise head), possibly center part, rotate part, move to place location, push part down, release vacuum, wait for vacuum sensor, raise head
you only need somewhat different vacuum nozzles, the head is mostly the same
(and some machines have an auto-nozzle-changer, so it can place a bunnch of tiny parts, swap nozzles as it goes for a bigger part, and go on from there)
do they ever use hot glue?
i think it sould be possible to controll the steppers with a printer port and extend the io via rs232 with a self made io extension
I think the glue they use is activated by pushing the part onto it (pressure-sensitive), but I'm not sure
actually, they can't use hot-glue - it would release in the oven
well its only for holding it down until it gets to the oven
SWPadnos: yes, afair it is preasure sensitive
usually glue is used for bottom-side components, so ...
the solder paste is tacky enough that most parts don't need to be glued down
unless you have very uncoordinated people moving the boards from pick&place to oven
SWPadnos: that's depending on the handling
if you move the assembled boards very soft the solder paste is maybe enough
does the pick-n-place machine dispense solder goo too, or is that done with a stencil?
I think early machines did dispense solder, and they can still dispense glue
you can also use a dispenser
i guess you can only stencil one type of goo
stencils are much more common, at least for production quantities
but dispensers are limited.
prototypes are a different story
smaller grids than 0.7mm are very difficult
0.7mm is pretty small
i'm working with 0.5 and less ;)
heh - 7 of the chips on this board I just did are 0.5mm pitch :)
you dont have to put a dot of solder on each pin
most terrible housings are MLF and BGA
and I've done hand soldering of 8x0402 resistor packs before
MLF aren't so bad if there's enough rooom around them, but BGA are inmpossible for hand-work
SWPadnos: me too. i know the techniques
smd hand soldering is one of my jobs :)
I usually do it myself too (though this board has ~275 parts, so I had it done for me)
thats why i want to use a machine. but for the solder paste i will use ... hm, damn, i don't kow the english word
(I guess mask works too though)
mhm, stencil? ok
[15:49:12] <SWPadnos> http://en.wikipedia.org/wiki/Stencil
but do you think that it could work? steppers controld by the printer port and an io extension via rs232 for any general io?
it could work, but you'll have to write the serial driver
I'm working on a modbus driver, so the low level serial port handler should be done for you, but the protocol will need to be defined and implemented
any other low cost alternitives? especially for self made
err - lots of PCI parallel port cards?
it also depends on your definition of "low cost"
hint: you dont need it :)
hm, i don't know why, but i'm not a friend of such pci cards
fenn: what i dont need?
the card swp is going to recommend
actually, I wasn't going to recommend anything :P
no, really, i believe you
even a 5i22 only has 96 I/Os, so it could possibly be enough
but it may not be
this futurlec 8255 is more expensive than i remember
$80 72 io
it also depends greatly on what the intent is. a pick&place that only has 5 feeders and is used to place all the filter and bypass caps would be very useful, and doesn't need anywhere near the I/O of a full production machine
it should be modular and extensible if anything
this board of mine has 34 line items, and 275 placements - there's one part value that's used 54 times. that kind of thing is perfect for a 5-feeder machine
feeders are a bit complex, possibly enough that an AVR board is a good idea for each one
you need "end of reel" detection, missing part detection, actuator control and feedback, and probably more
hmm, ok guys. thx for your help. i will think about a solution. if i found one or if i have any questions i'll be back ;)
ok. good luck
now its food hunting time
and definitely come back if you feel like working on a serial protocol ;)
swp you mean i cant just stick a webcam on my robot arm i bought from lynxmotion???
sure you can
it son't be a pick-place, but you can do it if you want ;)
ohu, one last question: is it possible to run emc on an "normal" pc without rt kernel for sumilations?
yes, as long as the normal PC runs Linux
you can run it in vmware or whatever too
any requirements or settings?
I think there's a precompiled sim package for Ubuntu, but I'm not sure
SWP, fenn I'm going to buy a torroid xfmr this weekend for my bp clone conversion. 80v secondary. I've seen a few 800Va, 1000Va and 1500. Whats substantial and whats overkill?
otherwise, just install Linux of some sort into a VM and compile the CVS version
substantial is enough to run the machine, overkill is way more than you need
overkill trips the breaker when you turn it on
on my normal pc i'm running ubuntu gutsy
mhm, i will try it. try and error ;)
should be fun :)
skinnypuppy1334: what's your biggest stepper's current rating?
if i could'nt find some binarys i compile it myself
er, actually need current for all of them
stupid stepper motors
ok, my food's escaping ;)
have a nice weekend
Fenn, I was looking at a limit of 5A for x&y prob 3A on quill
so you need (3+5+5)A*80V plus fudge
for steppers, it's power, not current (depending on the drives)
skinnypuppy1334: you can get away with less if your drives cut the current when idle
it's the chopping that does it, not idle current reduction
the step drive is a switching power supply - power out = power in
(less a little loss)
if the motor's consuming 5A then how much current is the drive taking in?
well, let's do an example
Yeah I got the 203v, I was figuring on the microstepping and current off feature to help out there
you have a 3V / 5A motor that's stopped but providing full torque (5A)
3V * 5A is 15W
the drive is taking in 15W plus a little to cover its efficiency loss
where's 3V come from and why does it mean anything?
that's the stepper rating
but you're driving it with 80V
it's the voltage at DC
no you're not
you're feeding the drive with 80V, not the motor
(at least, not always)
you're driving it with 80V but the reactance limits the voltage across the motor coils to .. ~3V
the drive is a current-limited supply, so it flips the switch to the 80V supply, and as soon as the current reaches the trip point, it shuts off the supply for the rest of the PWM cycle
no, inductors don't limit voltage change, they limit current change
what's all this back-emf stuff about then :)
I'm talking about a stopped motor for the moment ;)
so there's no back EMF
i always figured the inductor generated a reverse emf, otherwise current would flow in the wire
at low speed it's the resistance that matters
at higher speed the inductive reactance takes over
guh. i'm just talking about a coil
I think the power supply rating guidelines are int eh step motor white paper
(and I sure as hell can't explain it as well as Mariss)
I didn't see any swp
power supply ratings that is
of this http://www.geckodrive.com/photos/Step_motor_basics.pdf
he just posted a different formula for choosing a voltage though
btw skinnypuppy1334 you know that DC voltage is 1.44 times the rms AC voltage right?
Fenn I needed that reminder
V=1000*sqrt(L) (V in volts, L in henries), or V=32*sqrt(L) with L in millihenries
if you have access to the Yahoo gecko group archive, look at Mariss' post from 9/13/2007 (4:10 PM eastern time), subject "Re: Step motor supply voltage"
pretty busy list
lots of pick&place discussion lately ;)
I think I subscribed to gecko drive list months ago, but didn't subscribe to the e-mail digest
whee, new system up and running
7) unless you really really need the speed, in which case you can try to go up to 110V and see if the geckos explode :)
much more responsive than the old one (2.1GHz Athlon XP vs 1GHz Duron)
then back off a bit until they stop exploding
Thanks for the help again guys... I'm kind of in and out here.
Been priming painting the back half of the house, can't wait to be through
Hi, I'm looking for a simple program to create tool paths on Linux (without switching to windows). I was thinking of using Skencil with the special exporter, but I can't get it to run on Dapper. Any ideas?
Thanks, I'll have a look.
Thanks dmessier, apt is very interesting, too. And there's a tutorial on the EMC wiki
inkscape is a much better version of skencil, btw
Another question, this time about Axis: when using a rotary table (as axis A), is there a way to tell Axis the rotary table position and rotation axis. For the moment, A only rotates the tool itself around the X axis.
chk out #cam also
A is defined in the G-code standard as having its axis rotation parallel to the X axis
fenn, I don't know if inkscape will work with the script that jepler wrote to export GCODE
you'd have to call it B or C to change that
(to Y and Z respectively)
Thanks SWPadnos. So there is no way I can get "WYSIWYG" when using the rotary table?
heh - I don't think so at the moment
you'd have to define the kinematics of the table
OK, that's good to know.
I guess technically you could make a model of the machine with the 3D previewing code that's somewhere
then you'd get correct toolpaths and also you'd have a graphical simulation of the machines movement
First we make it work good - then we add extra bells and whistles
In any case, I'm already impressed with what I can do with EMC and AXIS. They sure aren't the limiting factors for me yet.
throw HAL into the mix, and they're unlikely to be the limiting factor for quite a while
I'm teaching myself Python for the moment, to use as a gcode generator (like holecircle.py). Haven't touched HAL yet.
When trying to compile gcam, I get the error: 'GtkAssistant' undeclared. Any idea which package to install to fix this? I can't find any instruction from the src or the website.
you need a recent version of gtk, 2.10 i think
i wish he wouldnt do that
thanks fenn. I have found a post about this problem. Looks like I won't be able to compile it. I'm starting to feel that Ubuntu 6.06 is a bit dated.
maybe.. iirc ubuntu didnt have the right gtk libraries when gcam development started
[18:15:21] <fenn> https://lists.ubuntu.com/archives/ubuntu-devel/2006-December/022989.html
Thanks fenn. I have to think about updating to the latest Ubuntu and building EMC and RTAI.
there is a live cd that has what you need..
dmessier, I only know about a Ubuntu 6.06 live CD. Is there another one?
will it not work??
there is a live - Breezy edition
I think I need a newer Ubuntu, so I can get GCAM and Skencil to work.
sounds like you know more than me
print multiply add 4 add 6 5 divide 1 square-root 3
% stack underflow
SWPadnos: got the tamron 18-250
SWPadnos: it's quite nice
alex_joni: how much?
JymmmEMC: a bit pricier than us
alex_joni: dont like changing lenses?
JymmmEMC: depends on the situation
but when going on a trip I somtimes wished I had only one
on trips where I leave even my battery grip at home
alex_joni: Maybe it's time for you to look at two cameras... an all-in-one and the one you have now.
I have about 5 cameras already
the rest are film though
I carry my lil 2MP everywhere/everyday, and pull out my D70 as needed.
If I can find a phone I like, maybe I won't have to carry it anymore.
I do have a 2mp on my phone
but it's not really nice
There's one phone with a 5MP camera
But, the phone I want either a) doens't exist yet, or b) I haven't found it yet
The biggest issue seems to be a qwerty kybd
I really like my phone
I want: kybd, A-GPS, miniUSB, WiFi, Stereo Bluetooth, Quad Band, EDGE/G3, CF/SD slots, 2.5mm headset jack, 3+MP camera
sounds like a laptop to me
alex_joni: Nope, it's a phone. Many out there, one big drawback is they have everything but a 1.2MP camera
I was joking
but CF slot on a phone is f-big
iPaq has both CF and SD slots
I dont care a whole lot about CF, other than being able to transfer photos from my Nikon across the network
hmm.. mine isn't in there
[20:14:53] <alex_joni> http://www.sonyericsson.com/spg.jsp?cc=GB&lc=en&ver=4000&template=pip1&zone=pp&pid=10653
Yeah, like I mentioned... haven't foudn it yet =)
Hello Alex; I am trying to download the CVS for the simulator using the wiki command, but I get: 'lin-dsl-static-206-222-212-221.inetnebr.com: No route to host'. Any idea?
can you try 'dig cvs.linuxcnc.org' ?
it works, and pinging the host works also.
My command is: 'CVSROOT=:ext:email@example.com/cvs cvs co -rv2_1_branch -demc2.1 emc2'
do you have ssh set as the preferred cvs medium?
try 'export CVS_RSH=ssh'
I don't think so. I CVS on a local system using pserver.
then the command above
Bingo, it works!
no problem ;)
Bye. I will now try co compile it on a redhat 8 box.
have fun ;)
is rtai working?
we'll find out in about 30 minutes if it is
* JymmmEMC hates when you're trying to learn a new lang, they don't make the examples clear, and/or especially when they toss in things they haven't/don't bother to explain.
03alex_joni 07TRUNK * 10emc2/src/emc/task/emctaskmain.cc: fix stepping issues (step after resume wasn't working). Both normal stepping and run/pause/step should work now
03alex_joni 07TRUNK * 10emc2/src/emc/motion/command.c: fix stepping issues (step after resume wasn't working). Both normal stepping and run/pause/step should work now
03jepler 07TRUNK * 10emc2/src/hal/components/stepgen.c: remove useless statement
220 nano kilo-whats?
no idea; on a component on a sw ps
only 2 leads
sq lookin g thung
could be a crystal, but I don't see what the markings would mean
c1 is the designator?
Yeah... 120V --> F1 --> C1--> L1
ok, then X2 may be the precision / temperature rating (like X7R), 275V, dunno why there's a K there
this being C1
or th nK
or the nK
I should have said I don't know why there's an n there
caps are usually rated in mucrofarads of nanofarads, but not milli or nano farads
now to find a 3.3V@4A supply
so a 220 nanofarad cap is labeled as a 220k picofarad cap
maybe they're covering both bases - 220n and 220K ;)
then again, they also use the exponential notation, so that may be 22 (whatevers) instead of 220 - 22 with no zeroes after it
I found a 24p netgear sw. I think just the PS is bad
single voltage supply 3.3Vdc @ 4A
tiny lil thing too
made by http://www.dveusa.com/index.html
Martini_ is now known as jlmjvm
SWPadnos: think i'll fry it using a 5V supply?
put a resistor in series
or a couple of diodes to drop the voltage
3.3 sounds like core voltage, so 5v will definately fry it
oh gawd, brain in gear... ohms laws... that slike murphys law, right?
you can use a computer PSU
those have 3.3 iirc
and plenty amps too
if you confuse ohms law and murphys law, then murphys law is sure to dominate
orange is 3.3V on ATX
[23:12:49] <alex_joni> http://pinouts.ru/Power/atxpower_pinout.shtml
alex_joni: I was JUSt gonna ask that
now to find a paperclip
JymmmEMC: don't forget the fire estinguisher
alex_joni: There's one very nice feature in ATX supplys.... short it out, it shuts down =)
alex joni:you gonna be on tomorrow
JymmmEMC: yeah, but it surely can provide more than 4A
jlmjvm: I think so :)
k,wanted to get with ya and try that adaptive feed
alex_joni: "Amps are nuttin, it's resistance that ya gotta worry about" ;)
gonna try to put the stuff you told me earlier in the ini file
but im sure it wont work the first time around
SWPadnos:have you ever set the pid loop in emc for the usc?
I think the only time I had a motor spinning with the USC was during testing of the encoder driver code
im gonna have to do it next week
damn, my extractor tool is too big =(
I've been wishing I could find my tweezers lately (while doing rework on baords with 0603 parts)
what have you been using?
I just push the parts into place and hold them down
* JymmmEMC shakes his head
I have a nice pair of tweezers for that though, so it's annoying that I can't find them
they're so damned thin, they may never show up
yeah, no doubt... I keep a seperate lil contianer for all the tiny stuff
so do I, but every so often something doesn't make it back inside ...
just due to their nature... I ALWAYS make sure they're returned.
everythign else, eh.
all the tools when I used to do microelectroincs
ok i guess thats a lie... just saw the on pair of tweezers onm my desk
well it's dead fred
all LED's stay lit upon ppower on