iany one who the HARDINGE brand of CNC machines>> chnc II in particular
Just this once .. here's the log: http://www.linuxcnc.org/irc/irc.freenode.net:6667/emc/2007-07-05.txt
hello, boys and girls
renesis: chillin with my homies.
uh oh - its pepsi__
whats he doing in here
* skunkworks knows him from ##electronics.
im a spy
your not supposed to admit to it..
now that we have the evidence....
archivist: how is the whole terrorist thing going? All we get is americanized news.
I think they caught most of the idiots
sounded like they where idiots
which was good I guess
yes failures except to themselves
well one burnt
I hope the doctors dont give him too many pain killers
uk useless terrorists
sorry - they where 80 amps peak
[15:33:21] <skunkworks> http://www.electronicsam.com/images/KandT/servotag.JPG
dont you want/need current feedback for thoose?
or will you just fuse them?
fuse for now.
I guess I could buy pico-system amplifiers...
though I need a -10+10v output..
though thoose amps want tacho feedback.
hm.. thoose have a relay.. that trips on overcurrent.
anonimasu: catch The_Ball_MC -- he was working on a PWM to +-10V analog conversion board that would work with pluto (and presumably other sources of PWM)
jepler: though I still need the amplifiers :)
jepler: but if a simple fuse/relay will do for overcurrent, there's nothing that keeps me from using thoose amps skunkworks have/working on
cradek: he's a funny funny man
"Mach3 knocks the socks of emc2." like how>
jepler: the big thing is how it would work with small motors
well, Mach runs on Windows, of course
maddash: he's also going to send the guy an emc2 release that's over a year old ("if he gets a chance")
however I have little sympathy for the original poster who couldn't find our website with google (our "how to get it" page says if you have dialup, you can get an ubuntu disk shipped to you for free)
i wonder when tuxcnc will ship
oh, i just realized that this was yesterday
cradek: emc's still not a tiny download even if you have an ubuntu install disc
jepler: very true, but it's certainly possible
SWPadnos: that's how mach knocks emc2's socks?
maddash: haven't you said to yourself while running emc2 "if only I could open up IE7 and install some spyware while waiting for my motors to stall"?
jepler: it should work with small servos/lighter machine right?
If you prefer to start with the distributed Ubuntu CD, which the Ubuntu team will ship to you free of charge (for instance if you cannot download the above CD image), you can install EMC2 yourself with these instructions:
anonimasu: which "it" -- fuse as overcurrent protection?
jepler: oh. so it's not as though mach had some army of pixies doing the TP calcs?
jepler: oh the amps..
anonimasu: for small servos, skunkworks's amps will be overkill but I would expect them to work.
I need 72V and 22A
and 72V and 3.9 continous
anonimasu: just keep those ratings in mind when choosing the FETs
you can get something cheaper / lower Rds(on) / etc than if you're trying to do 200V
jepler: im very inclined to go with a full servo system.. with a pluto or with a mesa..
:( cubic splines aren't fpga friendly
and skip the usc..
it works well, but as I have serovs it'd be _nice_ to have true closed loop
anonimasu: the other day you were giving advice to buy a gecko stepper driver instead of screwing around with "less expensive" or DIY drivers, right?
anonimasu: if so, you should probably lean more towards mesa and a commercial servo amplifier than towards pluto and homebrew servo amps
(or do I have you confused with someone else?)
maddash: for a cubic spline, starting with f(x), f'(x), f''(x), and f'''(0) (f'''(x) is constant and all subsequent derivatives are zero) it is possible to compuate f(x+dt), f'(x+dt), and f''(x+dt) efficiently -- for instance, f''(x+dt) = f''(x) + dt * f'''(0). If dt is fixed then appropriate prescaling can be done so that these operations are all shifts and additions, which is efficient on fpga
but for finding these coefficients .. you want a CPU with floating point
yeah, floating point's the problem. i'm trying to convert to fixed point
OK so maybe I didn't choose my variable names most carefully
[16:19:08] <maddash> http://www.box.net/shared/gtt8aiaeve
the authors manage to get out of using polynomials
I bet it's hard to get coordinated motion out of that scheme
why not? it basically replaces the stuff inside cubic.c
so it works per-axis
OK, I don't know the specifics of your design
i wonder how the folks at fpga4fun managed it
personally I'd just ditch cubic interpolation (in emc2 on a PC anyway) -- it is based on the assumption that the tp and kinematics can't run at the full servo rate, which is simply not true on a modern GHz machine
I dunno, have they put the complete source code up yet, or just the starting pages?
I have a sneaking suspicion that it's not used in emc2. have you confirmed that it is?
[16:31:57] <maddash> http://www.fpga4fun.com/CNC5.html
("The complete code is too long to be documented here (about 180 lines of Verilog) but it is available from KNJN.com") -- bullshit.
maddash: yeah one of my peeves with fpga4fun/knjn is that they don't give out a wealth of information
in fact it seems like they give out as little as possible
the forums have nothing useful either
and if the proposed integrator (http://www.fpga4fun.com/CNC4.html)
is used, then there's no way he could've cut this (http://www.fpga4fun.com/images/CNC_fpga4fun_logo.jpg)
yep. cubicinterpolate is called.
am I the only one who thinks that cubic interp is a bit too much trouble?
<jepler> personally I'd just ditch cubic interpolation (in emc2 on a PC anyway) -- it is based on the assumption that the tp and kinematics can't run at the full servo rate, which is simply not true on a modern GHz machine
hm, didn't see that
who's/where's Fred Proctor? i'd love to ask him about cubic.c
anyone know the name of that algorithm used to draw lines on a raster?
and Fred is at NIST
I don't think he's too involved with emc at the moment, if you're thinking of contacting him
[17:36:03] <maddash> http://www.tellingthestory.com/
<--- the whole reason I asked in th first place
SWPadnos: clearly not.
that's not him ;)
are there any sample programs that uses the EMC digitizing G code ?
oops, breaktime's over
ds2: this? http://cvs.linuxcnc.org/cgi-bin/cvsweb.cgi/emc2/nc_files/gridprobe.ngc
that's what the comments say; thought I don't recogonize it as a typical G code program
? it uses loops and such (o-words)
I've never seen a loop in G code so that's probally why
looks easy enough
still trying to decide what to use to test my digitizing probe - turbo CNC would require the least work (boots from floppy, etc) but EMC is friendlier
eta is now known as lern_hydra
I was wondering, do you recall the current state of backlash comp? I can't seem to recall if it was reliable or not
I think it works fine -- we used it on the mazak to correct for a small amount of backlash on "Y"
or maybe it was "X"
ok, so for small amounts
I use it on my lathe X axis
would 0.07" be a "safe" amount?
wasn't there something about double the max_vel value?
you really mean inch?
(it's a manual mill with acme screws)
servos or steppers?
it's mostly hard backlash though
let us know how it works :-)
can't you improve it any in the hardware?
that seems awfully loose
not very easily, it's not loose as much as rather tight but weak contruction
what tolerance parts do you hope to cut with it?
oh crap the RT test seems to have left junk behind
backlash compensation (.01mm) on a circular move: http://emergent.unpy.net/files/sandbox/backlash-comp.png
that's what your parts will look like too :-)
try to cut only squareish things
jepler: looks like it was able to apply all that comp in one cycle - wonder how it works with a crazy large amount of comp
hmm how do you reinstall the packages that EMC depends on?
[18:27:08] <jepler> http://emergent.unpy.net/files/sandbox/backlash-comp-1.7mm.png
lerneaen_hydra: how did you install it in the first place? if you installed from a .deb, apt should not let you remove a package that emc2 depends on
did it respect the set maxvel?
jepler, I think I broke a depended file
oh -- so you didn't uninstall the package with dpkg or apt, you damaged or removed one of the files in the package?
RT test left ome files in use, and I thought that they were autogenerated, apparently not :)
that backlash looks pretty decent to me
I guess I'm not sure what to do in this case
the spikes in the acc are strange, but they don't violate any constraints
might be a minor bug though
wasn't there some way to see whcih files depended on which packages?
that file will be in rtai-modules-whatever
lerneaen_hydra: this can't be the "right" way, but: sudo dpkg --remove --force-depends rtai-modules-2.6.15-magma ; sudo apt-get -f upgrade
you can find the package a file comes from with: dpkg-query -S /path/to/file
(files don't depend on packages)
something like apt-get install --reinstall rtai-modules-2.6.15-magma
cradek: ding ding that's the right solution
ah, yes thats a bit ugly, but it should work
apparently a reboot was needed too
doing the circle at higher speed: http://emergent.unpy.net/files/sandbox/backlash-comp-1.7mm-2.png
that's going to be one distorted circle
looks like about the best you could hope for, though
it's a small circle -- radius 5mm, backlash 1.7mm
ah I think those spikes on the decel are because the commanded velocity is going down
what is backlash measured as? if I move X+, then reverse and have to "move" 1mm internally, and then X starts moveing -, is that 1mm backlash or 0.5mm?
conceptually: put an indicator on the axis. push it side to side. the distance it moves is the backlash.
in your example that's 1mm of backlash
0.07" seems a lot. even for a manual lathe conversion
.010" is more normal ...
hmm does backlash run at stepgen maxvel or maxvel?
0.5 * maxvel
so stepgen maxvel should be > 1.5 * maxvel
so pick all the numbers accordingly
is 1.5 hardcoded?
(I'm no expert, I only looked at the code just now)
anonimasu: the plan really is to use emc as the limit for most things. accel/max velocity and folloing error.
IIRC, was there not some problem with backlash not being jerk limited by the motion>
emc's regular motion planner isn't jerk limited either
backlash comp works great :)
lern__hydra: any pictures?
skunkworks, coming up, running a new part atm
skunkworks, I'll upload them when I get home, it's easier there, 30-40 mintes or so
jepler, I've got backlash down to 0.02 or so mm now :)
Is the paraport pinout listed here still valid (under the ' standard_pinout.hal' section)? http://www.linuxcnc.org/docs/devel/html/config/stepper/stepper.html
if in doubt, look at the hal file
one of us would have to do it - might as well be you
cradek: I'm mostly speaking electrically
skunkworks: yeah but that's not related to the MAX CURRENT..
anonimasu: agreed.. It will not be perfect.
jepler: Well, geckos _work_ and homebrew stepper drivers dosent..
jepler: there's a big difference between a 120$ stepper driver vs a 500$ servodriver..
JymmmmEMC: why arent your machine running with homebrew allegro based drivers?
actually why arent your machine running at all?
Cause the power switch is turned off
cradek's mill works just fine 'home brew' depends on the performance you want. works great for him.
This is just tooo good. http://arstechnica.com/news.ars/post/20070704-riaa-sued-for-using-illegal-investigatory-practices.html
anonimasu: or are you talking about the new controller?
anonimasu: I was kidding! Eeeesh =)
anonimasu, hey, stop that
anonimasu, dont wind up JymmmmEMC ... he's way, WAY mor advanced than you imagine
skunkworks: I think a reasonable demand is that it should work, always.
skunkworks: that means no lost steps.. no fucking around with resonance frequencies..
anonimasu, well, in the "fitting a power supply and three drives in the longest time ever" competition, hes way, way ahead of the rest of the filed
2 years and counting :)
robin_sz: drop it..
on his foot?
anonimasu: Don't sweat it, You know robin_sz is being an asshole, I know that robin_sz is being an asshole, and even robin_sz know he's being an asshole. It's all good
I ordered my X3 on black friday - its only 3/4 out of the crate...
these things take time when they are not your "Day job"
day after thanks-giving
the prayed for mass attack of christmas shoppers day.
JymmmmEMC, simple question
JymmmmEMC, fitted that PSU and drivers yet?
Skullworks-PGAB: By the NEXT black-friday you'll have it 1/2 way out
that would require putting parts BACK IN the crate
I pulled the table and saddle - as one moveable chunk
skunkworks: fixing images now :)
* robin_sz decides the answer must be "not yet"
ok - I will be back later to check
pulled the electronics cabnet and spindle gear head casting
still have the base and column bolted to the crate
(got distracted when the big machine showed up...)
what are there called in english? http://ironbill.se/product_thumb.php?img=images/6003.jpg&w=76&h=57
* jepler looks
that picture is so small I can't tell what it is
it's a drill used for making a hole when you want to use a tailstock
I probably don't know the term either
they've got a distinctive shap
small (2mm) "lead" (maybe 3mm long) which is just a normal drill, and then a 45* taper outwards
the result mates nicely with a live tailstock
wow .. just looking at embedded processors ...
# Includes 80/40 MHz
# 5 RS232
# 3 timers
# 256 KW Flash
# 64 KW SRAM
# watchdog timer
# 512 byte EE
# 3 timers
# 20 TTL I/Os
# 35 solenoid drivers
# 20 opto-couplers.
lernean_hydra: you'll find a lot of lathe words in here -- http://www.littlemachineshop.com/Info/print_chapter.php?FileName=GS_MiniLathe_01.htm
lerneaen_hydra: maybe one of them is the one you're looking for
jepler: slightly larger: http://www.ironbill.se/bilder/6003.jpg
thats with 100 baseT and a compact flash card ...
hmm, center drill?
sounds like it
I use them for the first opearion of any drilling on a lathe
stiff enough not to get pushed off by the pip
[22:02:59] <robin_sz> http://www.tern.com/portal/content.asp?contentid=701
look at that
[22:03:07] <lerneaen_hydra> http://www.lerneaenhydra.net/images/rsgallery/original/IMG_1285.JPG
<-- you wouldn't think that there is 0.2mm backlash in X now would you ;)
I got backlash down to 0.02 via software
lerneaen_hydra: I have 0.02-0.04 of backslash :/
* anonimasu is not really happy about it
at the price of the screws..
I've got cheap acme screws
next thing I do will be ballscrews with zero backslash nuts..
oh, you don't have that now?
just the default ones(standard)
robin_sz: nice boards
robin_sz: I wonder if you could run a TP on them..
lerneaen_hydra: looks really nice
the entire setup cost 30k SEK new :)
all new parts
I've seen thoose :)
that's a good price
especially for a completed machine..
well, it was manual to begin with, though it was easy to refit
how well does the software work?
like a charm
* anonimasu has no idea what they run with
the opti stuff..
or did you run emc?
I'm running EMC
I just bought their steppers
bought geckos to drive them with
it looked like windows :p
and EMC to control
they sell drivers and stuff..
not that expensive I thought :p
[22:33:09] <anonimasu> http://www.industritorget.se/default.asp?visa_sida=kop/kop_annons.asp?MaskID=10283
1200 sek for a driver
no midban resonance compensation?
[22:34:02] <anonimasu> http://www.industritorget.se/pics/kop/upload/mvnnuyx8_b.jpg