looking for ddc... brb
yeah - it's annoying that there are no templates for xorg.conf stuff any more
and some of those options are different for each driver
i couldn't find ddc in the xorg.conf but your statement inspired me to copy the xorg.conf from dapper and now latency is 4 digits. only a minute so far but very nice.
KimK, are you around?
or not so. :( base thread is 5 digits and the first looks like a 9.
bbiab, going to play with it.
SirXanth, you can change the font
this is a link to the cvs diff: http://cvs.linuxcnc.org/cvs/emc2/scripts/latency-test.diff?r1=1.7;r2=1.8
you basically take out the word "italic" from all the font definitions
check out my new monitor/keyboard/touchpad mount and power supply/stepper driver cabinet
[00:14:30] <Danimal> http://img.photobucket.com/albums/v20/danielwilcox/DSCN0298.jpg
[00:14:43] <Danimal> http://img.photobucket.com/albums/v20/danielwilcox/DSCN0299.jpg
[00:14:56] <Danimal> http://img.photobucket.com/albums/v20/danielwilcox/DSCN0296.jpg
[00:15:05] <Danimal> http://img.photobucket.com/albums/v20/danielwilcox/DSCN0300.jpg
that cabinet is mounted almost exactly the same way mine is
it's the old relay cabinet from the back of my shizuoka
didnt need it for that anymore, so i put the power supply and stepper drivers in it
mine is an ebay special that might be big enough to hold all the stuff that needs to go into it
i need to get another one for toolchanger stuff, but maybe hal it's size
how big is the one in the photos?
I think mine is 30x24x10"
had to measure it
oh - tiny :)
i have a bigger one, about the size of yours, but it wasnt as deep
and it was kinda ugly
yeah - the depth is a real killer. my power toroid is about 10" diameter and 4" thick, and the caps are about 8" tall
i got all new sheilded cables and 25pin connectors
yea... mine's huge as well.... it BARELY cleared the door
capacitor is about 1/2" away
we'll see how the touchpad holds up
I have a stainless-steel "Kiosk" keyboard from that ebay. I just brush off the swarf..
i didnt want some mouse hanging around getting in the way
My keyboard has an integral trackball
nice andy.. i'd like something like that, but i figure i'll just make a cover for it
oh i hate you... i was looking for one of those
I almost bid on a couple of those, but didn't. now I wish I had
how much were they?
nice stainless, backlit keyboards, with trackpad or trackball to the side
some were $50 or so, others closer to $100
i got 30 into mine
yeah. 10% of the retail cost
it's super compact though... barely takes up any room, and still has a 15" screen
But, no Ctrl key, no F5...
oh. very nice
i cant use my vfd.. i'm dissapointed
not enough torque
stalled it taking a cut with a 3/8" endmill
using the lowest speed belt setting?
variable speed pulleys
no, that would have been too slow
well, a VFD gives constant torque below the normal speed of the motor
so the power goes down linearly to zero
so use the lowest belt ratio, and run the motor as close to or above its speed rating
Hmm, I had that problem with my VFD until I reconfigured the V/f parameters. It was (be default) at 12V below 30Hz (which considering it is 240V at 50Hz was silly)
i'll have to look into that
it's a vfd from an a/c unit
might have some weird defaults in it
Yeah, definitely check that you have the correct max current limit (or perhaps higher if you have a thermistor) and that the V/f curve makes sense. Mine was very disappointing at first, but now chews through anything.
what hp is your motor?
Set to 4.2A max, and never less than 40V
yea i really gotta look into that
i had no idea there were settings for that
There might not be, but they are fairly common (mine is far from stat-of-the-art).
i have a pretty decent one... it has ALOT of parameters
I am no expert, but I would expect that to have those settings.
yea, there's so many parameters, i havent gone through them all yet... just enough to get her going
If it was set up to run a fan it will probably be wrong, as you probably want _more_ torque at low speed
makes total sense
Well, put it this way, I now have more torque than I have bed and slide rigidity.
and i bet i was still spinning 1500rpm's, it shouldnt have lost THAT much power unless a setting was off
what kind of machine?
This is a cheap chinese lathe/mll
my shizuoka is pretty beefy, so i was very dissapointed when it stalled
that thing can normally hog some material
40 taper spindle
but i'll check that out
last weekend I swapped the saddle keeper plates for tapered brass gibs (web page planned) and that made it clear that, actually, the two saddle slides were not actually parallel, so I spent monday scraping those in.
man that's a bit of work
andypugh: did you figure out how to get smp working?
well i gotta run and try to get the garage back in order
Actually, I wish I had documented the scraping process, as there are web pages about lapping them with emery paper, but actually scraping is much quicker and more controlled. Only took 40 minutes.
wow, i've never tried it, but my manual mill needs it
No, I lost interest in SMP months ago.
thanks for the tip on the vfd andypugh
is it an intel?
have a good one, adios
You need the right scraper. When you buy them they have a straight end, that is no use, you need to grind it curved
Sharpen with an oilstone
And use engineers blue
Blue up the bed, slide the mating part from clean bed to blue bed, lift it off, scrape off the blue bits, repeat
tlab Aye, Intel chips, intel board, dual (err, pentiumy things, err, early Xeons
do you know where it errors out when you try and compile the kernel?
Danimal: If you do need to scrape in your bed, email me at email@example.com for a more considered description of the process.
I am a geek, but my dad is an apprentice-served fitter with a bad second-hand-machine-tool problem, so I have reconditioned more tools than most.
I can compile a functional kernel, I can even get RTAI working.
I have had emc running SMP, but I could never get the smi.ko module compiled and running.
you can get smi to work with one core?
I think (and I am remembering back some months now) that something in the code references the .ko explicitly by path, and I could never get it to compile into that path. And if I moved it here, it couldnt find itself.
Yes, the live-CD install Just works
what chipset is the board?
Sorry, copy-paste went wrong..
Dual Intel Xeon processors using the 603-pin INT mPGA package and 604-pin FCPGA
hmm say anything like 845 or 945
Its a SE7501CW2
To be honest though, I can live with wasting a CPU. It works, it cuts metal, latency is 13000
server chipset, I don't know much about them
I like it because it was cheap and rackmoiunt
[01:02:46] <andypugh> http://www.bodgesoc.org/CNC_Controller.jpg
using the parallel port?
thats what I plan on doing
[01:05:56] <tlab> http://www.intel.com/products/desktop/motherboards/D945GCLF2-D945GCLF2D/D945GCLF2-D945GCLF2D-overview.htm
I have a dual PCI parallel port in a drawer if I need more channels
[01:06:10] <skunkworks> http://www.cnczone.com/forums/showthread.php?t=82364
I just notice a new bios update for my MB
skunkworks: Was that a comment on an existing discussion, or something new>
just ran across it. (I try not to be but I seem to be a mach biggot.)
actually - I really don't try
brb updating bios
I wonder if he will ever be back?
I think I only have made one brick in my life time.
Anyway, Mach seems a triumph, in that they manage to get Windows to drive stuff RT. I have tried that, and failed horribly.
practice makes perfect
But running an actual RT OS has to be a much better choice.
I would really like to test the 100khz pulse rate in mach. I don't believe that it acutally makes for a usable computer.
this was a triumph
But I have not had the drive to do it.
we had a meeting on lean manufacturing today
dudes came out and stuff
everyone seems to always come back with that 'mach can run 6 axis at 100khz'
i hope we get new toys to play with as a result
we are going thru lean now.
how is it?
How many axes can use 100kHz anyway>
what should the setting for the parallel port mode be? ECP?
I have never set that settigng, but probably ECP. I suspect that the stepper config wizard did it for me.
well it's a setting in the bios
I think ECP sounds right. You want bidirectionality
We have a few more people on so I will repeat a question: Should G43 change the machine coordinates as well as the relative coordinates?
the convention is no, but I recall hearing someone emc treating "machine coordinates" differently
[01:33:01] <tlab> http://www.linuxcnc.org/handbook/gcode/g-code.html#anchor259197
G43 tool length offset (plus)
yes, tlab, we know
andypugh: if the machine coordinates haven't changed, that's fine
oh I didn't read the whole question, sorry
on my machines, nothing changes the machine coordinate system, but I don't use emc so I don't know if there's a reason it would
toastydeath: not bad. Lots of closed minded people.
hopefully swpadnos sees this, he's probably the best one to answer
Emc _does_ change the machine coordinates, and I am not sure it gets it right either.
what leads you to that "not right" statement
Certainly when I put in my long boring bar the machine refused to move at all under G0 or G1 control, but would joug to the soft limits perfectly happily
i don't know how to help you there, most lathes don't have g43/44
I was too annoyed and puzzled to do a proper analysis, but I think that the G43 offset moved the machine origin such that the tool tip was outside the machine envelope, and it sulked.
is your tool offset data positive or negative
It is whatever "Touch off" calculates for each tool
(and yes, I need to do some tests to check it isn't getting the sign wrong)
I haven't tried mach at 100khz other than to run the driver test. I can try it next week though.
andypugh: the really fast way to check signage is to use g44
at 60khz it runs fine
and see if it goes anywhere.
Does EMC support G44, the documentationsuggests otherwise
oh, that would be a problem
andypugh: the mach driver actually runs windows from what I read, not windows running mach.
Yes, that counts as a triumph of determination over logic :-)
:-) I guess so!
Although doesn't rtai do something similar?
Or maybe I'm thinking of RTLinux...
FWIW I once did some development on a dedicated Real Time OS, and to this day I don't know why it mattered. It was a special labView box, running LabView RT and you had to log into it to download the code to be run, and could only see the results vicariously. Very odd for what is basically a totally graphical proogramming language
RTAI pathches the kernel to add real-time threads. I suppose that Mach might do the same thing, and a salesman decided to spin it as Mach running Windows, as Mach takes overall precedence
Not sure about that. Art explains it in the MachMysteries.pdf I'll see if I can find it if you're interested.
He does say this: "It took months to develop the theory and to force Windows to do something
it truly despises, to give reliable step pulses at high speeds."
Ae, that is true enough.
A little while ago I decided I wanted to emulate the cam and crank pulses from an engine from my laptop to persuade an engine to fire the injectors when it wasn't actually rotating.
did it work?
So, I tried Matlab, and that could manage a paltry 6 pulses a second from the parallel port. Not good enough.
Then, on a whim, I wrote an Excel macro to do the same thing.
Colour me surprised, but that would run at 100kHz
Which was plenty, except that every few hundred milliseconds it would just stop while Windows did "stuff" like checkin email, or whatever.
And those glitches were enough for the engine to set a fault and fail to synch.
I'm going to try and install a custom kernel. bbl
This is going to sound wierd, but right now Excel VBA is my development envirnment of choice, I have not yet found anything I can't do with it, and you can send your finished program to anyone with Excel, on any platform, through email dodging security traps, and it will Just Work.
why is the .deb of the kernel image 220+ MB btw? I used the config from my ubuntu build and removed some stuff.
that's weird alright!
So a macro beginning excel.visible = false with a custom Userform is effectively a freeform application.
And as I showed earlier, performance isn't at all bad either, as whilst it has the flexibility of an interpreted language, it selectively compiles when it can.
andypugh: I still do a lot in access vba
I don't know why I am drawn to it. (something to do with reminding me of basic or something.)
Well, it is BASIC, but with a full Object Oriented thing to it too.
The development enviroment is good, especially how the methods and properties drop-down for you as you type (even for your own classes).
In fact, if I am ever asked the best way to learn OO programming I say "Excel" as you have a _huge_ preexisting object structure that you have no choice but to work with, If you try to get into the concept in C++ from scratch you will be coding for weeks and wondering why
Excel has a problem though, in that it's relatively expensive
if you actually buy it, that is
Indeed, and I haven't bought it form home use, and the Open Office Macro language is basically undocumented.
ayeah, it's not all that easy to use
I wrote an openoffice macro a few years ago. I found a couple of documents on it
I liked it better than vba after I learned it some
So I only use VBA at work. (where I have written a Macro that replaced a commercial system that was costing us $40,000 per dyno per year and not actually doing all that we wanted, (whilst doing a brazillian things we didn't need)
but I forgot it all now.
At home I tend to code in Octave, as I am a great fan of Matlab
how well do you think Octave holds up v. Matlab?
It's pants, but free so good value
what kinds of things are missing?
I think my mother will want to use it, since she can't afford MatLab
does that present an "
"open file" dialog to the user?
I compiled linux 220.127.116.11 with the rtai patch from cvs. When trying to boot it said it needed acpi for smp systems?
ok, yep. that's what it does
Actually, the language is fine, but the lack of a development environment is a pain
some SMP systems need ACPI enabled, but you can disable almost all the sub-options
I think there's one that has a description that includes the words "power" and "multi-core"
that one should remain enabled
I'm running an athlon 64 x2 dual core and selected the *k8
mozmck: Might be a compile option set wrong, might be true
Ok, i'll look
if you're on 8.04, there is an experimental SMP kernel at http://www.linuxcnc.org/experimental
there are several actually
9.04, but using a vanilla kernel
I have an RTAI kernel on 9.04, but I don't know that it's stable yet
and there's a compile error that I haven't bothered to fix on TRUNK (joyfile needs the pow() function, which for some reason isn't getting linked in)
I suspect that, like me, you will give up on SMP. The Live CD works and lets you cut metal. I guess it comes down to why you want EMC
andypugh, the experimental kernel works quite well for me
on a dual-core core2
with smi.ko? That's the killer for me
I think cradek wrote it for his dual PIII box
I'm not sure if that's included
I want to learn emc2 better and contribute at some point. I want to be able to run emc on my main machine here in realtime.
To be honest, I don't now see why I even cared.
My shop machine runs 8.04
I've tried rtai and -rt kernels on 9.04, and haven't had good stability
it could be something else though, since this RTAI kernel has been running for a couple of days now (under more or less no load), and hasn't crashed yet
oh, joyhandle, not joyfile
It isn't clear that SMP will be more stable or have better latency, so why worry?
I noticed in the rtai docs something describing how to tell the kernel to not use certain processors at all, but then make the rtai run on those cores.
has anyone played with that?
looks like it would have the potential to greatly reduce latency from the docs
latency on a core2, and I think on some other dual-core Intel CPUs, can be improved significantly by running a CPU hog on the non-RT core
this does not work on AMD chips, AFAICT
once I got rid of graphics mode, and some other useless junk, I had latencies under 2 microseconds
Well, that was part of why I bothered, in case that was built-in. It is probably simplistic to imagine that you could have a quad-core running the OS in 2 cores, servo-thread in 3 and (err, the other one) in 4
and usually under 200 ns
you can do that, if you can get a reliable SMP RTAI kernel built
it's not too hard to get it working on one machine, but getting a package that more or less anyone can install on their machine is much harder
ok, I'm not sure I disabled cpu frequency scaling...
And that is where I failed, not really knowing what all the compile options meant, even in gui-menu mode.
that shouldn't cause a panic, but it is definitely a good idea to disable it
yeah, me either :)
I think that's why it wanted the acpi stuff...
unfortunately, there are times when the RTAI patches or some kernel version just don't work, and if you try one of those, you''re in for a lot of pain
<rant! fume!> "me either" is _so_ wrong! "Me neither" is what you mean. </rant> </fume>
neither me nor you would be correct
I'm not sure the colloquialism "me [n]either" matters so much
though "neither you nor I" would be even better
Aye, but to me "me either" means "me too" not "not me too"
[02:50:40] <SWPadnos> http://answers.yahoo.com/question/index?qid=20070527225055AA9nHSb
your form is the recommended form, but the last paragraph does point out that the form I used leaves no ambiguity as to my intent - I'm clearly agreeing with your negative statement
but nonetheless, I will now go to bed :)
Actually, 4am and no answer to my G43 question, I think I will depart too.
greetings, I have 3 gecko G320s, and I have the step and dir pins all wired correctly. do I connect the "common" on the geckos to the chrage "pump pin"?
no see gecko docs
charge pump is for an enable
thanks - I will look at them. I had the pins hooked up right, and I tried using "amifier enable" pin on the common and the motors went crazy until it they blew fuses. the motors did nothing without the amp enable, so I thought I might need "charge pump"
hmmmm. The gecko manual just says: "The +5VDC is the opto-isolator common anode line and must be returned to the pulse source +5VDC supply"
I am wondering why the person that wired it up has the common connected to a pin, and if it is connected to a pin, seems like it expects a signal
with opto inputs you can treat either as common if you know what you are doing and they are truly isolated from each other
if you have 1 common and a step and dir then they are not fully isolated
sorry - I am a total noob, you can treat either what as a common?
they are servos with optical encoders
I just downloaded the manual, one common must be +5 v
so what pin would I give it? outgoing pins are +5v, right?
I am just confused because it did nothing with just dir and step, but almost went through the roof when I used "amp enable" on the common
think vary carefully about pin functions
the EMC manual says:
"21.1.2 No Steppers Move
Many drives have an enable pin or need a charge pump to enable the output."
that is the prolem I guess, I know almost nothing about the pin functions
and would the charge pump pin be assigned to the common? thanks so much for your help!
no not to the common
so I just need to enable it, and that is it?
and the pin that is connected to the common, I just set to "unused"?
when the charge pump starts it enables the servos
wow - sounds like magic!
exactly which common pin are you talking about
what interface are you connecting too
+5v on the right most connection, page 6 of the manual
that goes to a source of 5v to provide current for the opto
he has them wired from +5v to each of the 3 geckos and the end one is connected to a pin, and the other end is connected to the emergency stop
i just double checked - the pin number 5 connects to the first +5v
which connects to the next and the next gecko
and the last gecko +5v splits to pin 10 and the emergency breaker
[09:49:14] <anonimasn> http://i46.photobucket.com/albums/f104/kendo679/090326104701_3.jpg
since is kind of hard to explain, I made a diagram: http://openosx.com/geckos.png
just what my CNC machine needs - a chainsaw! hahaha - better get it working right first....
jeshua, what is pin 10 or 5
brb, im gonna go out and turn some stuff
that is pin 5 and 10 of the parellel port the controller is wired to
i bought this controller all wired up, so that is why I don't have a clue! ;)
not all parallel ports can sink or supply all the current needed, but should just be a question of configuring
so the bottom line, is I just need to supply it with a good steady 5v power source?
and the guy that I bought this from might have been using 5v off his parallel port?
pin 5 on a parellel port cannot supply 5v for all the optos
but it just needs a contstant 5v, not a signal, correct?
maybe he had a breakout board or adapter board
yes 5 constant
cool - that would explain things
like why they went crazy when the got a signal!
just lucky I don't have it hooked up to a chainsaw yet :D
why does the gecko manual say "must be returned to the pulse source +5VDC supply" - that confuses me a great deal!
it just doesn't say anywhere in the manual how much current it wants (milliamps?)
anyways - thanks for your help archivist - I would have never figured this out!
i think I have a spare 5v DC adapter laying around
it was really kind of you to help such a complete noob! thanks again!
he states a ttl drive to the step and dir pins, (not true for opto) open collector is ok as well
Im at work so cannot concentrate
i think you lost me there
uh-oh - another question - if I hook up a 5v power supply to the +5v, where do I connect the -5v?
gnd on your driving parport
so I connect the -5 from the power supply to any of the available grounded pins?
i see says the blind man! THANKS!
hmmm. this wiring diagram shows pin 1 providing the +5v: http://www.cnczone.com/forums/attachment.php?attachmentid=12085&d=1132310389
no it says from pc power supply
and shows a resistor to pin 1 on the par port
which would tell emc that the unit is powered
you are right
i thought the PC power supply is 12v though?
red wire in a PC psu is 5v
yellow is 12v
nice - I learn something every day!
i was just thinking that I was getting power to the geckos before with the default stepper config, my motors would spin. Seems like X spun correct (in one direction) but Y an Z hardly spun at all, that is when I use the stepper wizard.
do you think it would matter what resistor to use? it is not specified
depends where you need a resistor, calculate for needs
on pin 1?
1k ish or not connected or dunno
and then do I need to give pin one +5v? or do I just set pin 1 to be "unused" with the wizard?
or it might not even need to be connected at all to pin 1 it looks like
File "/home/colin/emc2-trunk/lib/python/vcpparse.py", line 24, in <module>
ImportError: No module named ext
hmm am i missing something from my system ?
i guess I will just try +5v from the PC to the +5 common on the geckos, and -5v from the PC to pin 25 first
ah no worries fixed it
thanks for your help again archivist - I must have asked you a zillion questions! You are too kind.
i would try this now, but I have wait until the store opens to buy some fuses!
jeshua, not -5 thats the purple wire dont use it
black is 0v
LawrenceG! how is the weather?
hi skunkworks ... it has been pretty warm up here, but I have been saving the heat pump electricity for running the cnc!
nice! we have not run the air yet this year. (the plan is to try not to)
it issupposed to get about 10 degrees cooler for the next week or so... the forest around here is at extreme fire hazard... starting to get some wild fires
it's cold here
raining all the time
it has been a very dry spring.... we havent been living up to our usual soggy reputation
eric_unterhausen, it was a cold spring until about 2 weeks ago, but very little rain of snow
PEOPLE OF EARTH...
brain board for the power output stage
skunkworks, I need to see if I can mill it today
LawrenceG: are you following the emc-devel mailing list?
alex_joni, I think so... are you refering to ("hi Paul") git wars?
the point was to get your oppinion as a devel ;)
if you're in favour or against
* BJT-Work git :)
* archivist has a slight bzr bias
yes.... I dont do a lot devel on emc.... basically I got involved to add the step type 2 interface in the 90's because my stepper driver board needed it.
access to the source made emc viable for what I wanted to do and I have learned a great deal from emc because the source is open
LawrenceG: so you have nothing against it?
I am happy with whatever the board decides.... I pretty much depend on the active developers to discuss/review any ideas or code additions I may dream up
ok, so I can take that as "in favour"
everyone gets to be cranky..... but it is nice when there is some constructive component to a rant.... the strength in a software team is flexibility and not getting bent out of shape all the time
too bad some decide to spend time on bashing others instead of contributing
he made one valid point... no general email till after the decision
* archivist hates cvs anyway
the email was supposed to be a fully researched proposal
anything less and the board thought it was not doing its job
the email that was sent was actually asking for comments on a proposal, we're looking for constructive dissent
(there was a LOT of discussion at fest among most of the most active and interested current contributors)
and on irc for the last ... year
I saw the general email as a request for comment for something the guys at the fest spent some time on
so the email sent to the list, though it might have looked like a fait accompli, was actually the very email that he (and you) request
yep, that's exactly it
yep quite a bit of discussion as I recall
LawrenceG has it exactly right
at fest, several active users were involved too - it was a good opportunity to get a cross-section opinion
that is why we have a board to make decisions
* roh is in favour of git.. lots of other projects also chose to migrate from svn or similar old stuff to git. so when it comes to 'does somebody need to learn it some day?' _yes_, anyways. so chosing something less popular should have a good reason, since it will in the most cases result in: somebody needs to learn some scm for this project alone.
see gnome, openembedded, and some more
oe even changed from monotone to git, so they had a distributed scm already (which was slow)
not a proposal if the first para has "the board of directors has voted unanimously to transition to the git revision control system."
archivist, "We welcome discussion about this, because there may be dealbreakers in git that we're not aware of yet, or a killer feature in another system that we haven't considered." ...
SWPadnos for a 'simple' and small project like emc its not that hard i would say. even svn would be a step forward. cvs just plain sucks and is bad
e.g. you get different checkouts depending on your timezone.. which makes it unusable for rapid decentral working
skunkworks: I've scanned the drawings and schematics of my PUMA robot. intrested in a copy?
JanVanGilsen: can you put them widely available?
of course if they free ;)
currently i don't have webspace =)
that we can solve
heh - I have some space also.. Although I would assume it is still copywrited material..
question of storage and intention.
usually one doesnt built robots from it. and if you want to the documents are the easiest part to get. but for repairing and service such documents need to be avail, to make the value of a machine worth it
so its in the interrest of the producer to have such documentations avail. in the end.
JanVanGilsen: if you get the time - samco (at) gmail dot com
you've got mail =)
JanVanGilsen: sorry - it is actually samcoinc (at) gmail dot com
(it has been a long day)
imagine that.. long enough to change email addies
hehe somebody else got it then ...
Thanks for doing that..
JanVanGilsen: bet somebody else will report you as spam :P
Have you looked any at the index homing you guys where talking about?
I'm still looking at it =)
did have some problems with the component that determines the offset from the potentiometer :)
what are you using to read the pot?
12 bit sadc and a 4051 multiplexer, i read somewhere that the original controller used a 12 bit adc
nice - also - nice scan job!
I also included bookmarks in the pdf :)
And let the OCR scan the document so it's semi searchable, but it di have some trouble with the hand written drawings =)
Ours is a puma mark II 500 series.
the manual is dated 1984
It's a 560 (the 6 axis version of the 500 series) the 550 is the 5 axis version
we have the 560
nice little arm - seems to have a ton of backlash adjustments.
the manual is the same for the both the 550 and 560 version :)
is your arm newer or older?
someboby might want to reuse the power amp schematics, but i doubt they can compete with the modern controllers
I don't know other than the date on the manuals (1984)
are they pwm or linear?
they are linear (current mode)
does the puma have AC servos ?
my lil kawasaki js10 is AC
trying to find something to run them
thinking the mesa AC drive boards will do
pico systems also has a ac drive now.
will take a look
wow - pico systems has a webstore now
im gonna need 6 of there resolver to encoder converters
colin_, did the bot not come with any motor drives?
geo01005, it does but i cant find any information on them at all
I'm just wondering, why not use the original control?
the original robot arm control ?
its doesnt accept gcode for a start
just some weird robot language
Ahh, I see.
if i can get it running with emc then i can feed it gcodes straight from my cad/cam
Sure, I understand that.
you can get gcode to robot arm language converters
but they turn a 3mb nc file into a 500mb file
and try drip feeding that over rs232 :(
Hmm, it uses resolvers not absolute encoders huh?
resolvers are pretty common on bots these days
they are absolute for a rotation
all the ABB arms have resolvers
probably others too
No guarantees but we expect to have a 6 axis resolver input daughtercard (7I49) before the end of the year
it will only work with Spartan3 based FPGA cards (4I68,5I22,5I23)
PCW, What is the expected resolution? (counts/rev)
wow - you guys are just hammering things out
(maybe more if our firmware is any good)
colin__, yeah, mesa.
(I really like the absolute nature of resolvers, makes interfacing 3 pahse AC servos easier also, no silly hall signals)
yup, that's one of the main reasons for resolvers
factory mounted ;)
[19:43:43] <geo01005> http://www.austriamicrosystems.com/eng/Products/Magnetic-Encoders/Rotary-Encoders/AS5045
Also their very tough
geo01005: still need to align one of those if you want to use it for motor phase control
PCW: right, especially compared to glass in encoders ;)
Sure, I understand.
Wide temperature range also
and no real moving parts
except the magnet on the rotor
Those magnetic encoders can be programmed to where zero is though...
I've never used anything with resolver though, so I suppose I'm clueless.
lots of other reasons why I'm clueless too :)
Basically they are a variable transformer
my project for next week http://www.io23.net/ul/files/20090605164030545.pdf
figure out if there's a way to shim/load the bearings a bit more..
I always though a robot would have much higher resolution joints... more like 16-24 bits?
I would have though there would be a gearbox for each motor and then a high resolution absolute encoder right on the joint.
geo01005: there's a resolver on the joint
usually converted to 2048-4096 imp/rotation
you get about 3-4 degrees of joint movement / rotation
maybe a bit more, I'd have to look it up
There are also multipole resolvers, but they are not absolute
I'm just saying only 4096 counts on a rotary joint will results in large errors at the end effector.
geo01005: you still only rotate a couple of degrees for 4096 counts
that means that in the worst possible combo you move less than 0.01 mm for a count on any joint
Ahh I see, there is a gearbox on the resolver?
no, the resolver is connected directly to the motor shaft
the gear is between the motor and the joint
so two resolvers, one on the joint, and one on the motor?
sorry I'm wasting your time.
nope, just one on the motor
but while you move the motor one full rotation the joint only moves a couple degrees
archivist: you there?
Sure, I was just confused becasue I read somewhere that most robot arms had absolute encoders on the joints so they didn't run into kinematic problems when homming.
[20:01:22] <anonimasu> http://www.cnc-step.de/10mm_Aluminium_CNC_zirkularfrasen_frasen_1.jpg
<- about that is how my surface finish on aluminium looks
geo01005: that used to be true about 5-10 years ago ;)
nowadays you don't home them anymore
they have resolvers for absolute inside one rotation
on shutdown they store the current joint positions inside the servo controller, including the current shaft rotation angle
then on startup they check if the angle hasn't moved
if it hasn't all is well, and the stored position is assumed ok, and it gets loaded
hmm, well, I suppose that would work.
alex_joni, you work with bots for work right?
the pumas have a pot that rotates the same as the joint - and the encoder is mounted on the servo.
so the pot is sort of the absolute position of the joint
So the whole process is similar to using a limit switch to get the approximate home position and then using the closest index pulse to get the absolute home.
(for the pumas)
anonimasu: what machine is this?
a abene vhf3
a pretty heavy mill
skunkworks: any idea?
anonimasu: that finish looks very normal to me for circular milling with an end mill - are you thinking it's bad?
looks normal to me :-)
when Icut straight lines I get | | | | | | | marks on the surface
and I dont know why
steel ends up all shiny and smooth
are you climb milling it?
dosent make any difference
mekin summat with someone here
that's why im suspecting spindle bearings or something
climb should be better
hmm maybe I haven't seen a good machine then - I expect this kind of finish for any side milling aluminum
if I need something like that round (for pressing a bearing?) I use a boring head to finish it
hm, I just helically interpolate it..
I await the advice of the real experts here :-)
this has been showing up lately
and spindle is the only thing I can imagine
try smaller final cuts
0.05mm isnt small enough?
good evening.i have a problem with compile emc2.3.0 on new computer dell.i have install live cd ununtu 8.04 with emc 2.2.5.the install ok end emc 2.2.5 run ok.i have start synamtec and reload the package and install 354 new package.I have rebbot the system and run ok.But when compile emc2.3.0 and write ./configure --ebable-run-in-place the process stop with error: glib..configure required until somebody makes modbus optional.I don't undertsand because hav
e the problem.On another system i compile ok.please hel me.
yeah for finishing, use high speed, low feed, climb
motioncontrol: so install glib
[20:32:19] <anonimasu> http://www.boelensmachines.nl/UserFiles/Image/_machines_foto/groot/Frezen/IMG_7007.JPG
<- that kind of machine
anonimasu: gibs adjusted right? end mill large enough to be stiff? what surface speed?
cradek: surface speed I dont know
let me check the sfm
that's probably fine for finishing, but you could try faster
I cant spin faster
do you have a bigger endmill to try?
well, I have a 50mm tool
with inserts(lots of them)
using any oil or coolant?
inserts mostly just suck for aluminum
except there are some special ones for Al
ok i have resolve the glib error with install, but when compile search another package..what is the procedure for dowload all package for compile emc2.3.0?
you could try higher coolant concentration, or spray WD40 on it
* anonimasu nods
I dont think that will help
I don't either
im more thinking that it's a mechanical problem
and not the table/ways/screws
more like spindle bearings needing more preload
or maybe spindle tramming
tram won't do that
well, it's worse in one direction.. also
is it an end mill made for aluminum? and is it sharp?
no, it's not a special endmill
it's a dormer hsco-xp coated endmill
though I have some carbide made for AL
it gives me better finish
but still not good
I get |||||||||
over the surface..
what's holding the end mill?
a endmill holder
iso40 er 32
it should be
(I didnt measure it for a long time)
also spindle/motor could be out of balance and causing sideways shake (happens here)
it's a belt drive
very much decoupled from the spindle
anonimasn: you get vertical lines?
in your part?
and sometimes S lines
independant of feed rate?
dont say servo loops and stuff because that's one thing Im 100% sure it's not
I suspected that first, but both axes cant be doing the same
and my machine isnt so worn, at all
i wasn't going to say a servo loop deal
one of our machines has that problem and it IS a servo problem, we just can't fix it
both axes, it's the control itself
hm, that's odd.
this problem was fine a long time before
different aluminum or end mill?
different aluminium same kind of endmill
have any of the old stuff?
feedrate and 2000rpm
dosent mention anything about Ra though
toastydeath: sorry about pre-guesstimating what you were going to say :)
good night all
I should check my runout tomorrow..
then tram it properly
then try your old aluminum
* anonimasu nods
maybe I'll just do that now when I pick up my wallet at work
I don't see much information on the Mesa 5i20-5i22 setup with emc2. WHAT!?
Results 1 - 10 of about 988 for mesa 5i20 emc2. (0.26 seconds)
well ok, that's a little less than I expected, but still...
sharp learning curve ahead.
cradek: have you fondled jeplers servos yet?
cradek: you servoperv you
* Jymmm brings cradek up on servo harrasment charges
* Jymmm makes sure that cradek gets placed on the national servo offenders list
Jymmm: there is probably one living in your neighborhood.
cradek: btw, tir is 0.01mm..
maybe 0.008 or something
not a full 0.01mm
I think it's my cheap holders that does it..
im gonna try to preload the spindle a bit more...
and the surface finish issue was becasue of the material
well, I dont get perfect finishes..
but I think with a better holder for endmills it'll get better..
should 0.01mm cause bad surfaces?
err noticeable marks..
are you taking a full side cut on the wall?
or taking step downs
looks like full height which is surely what you want for a finish pass
how much stock was left ?
with this material and a new cutter..
I get _good_ finishes..
better then the picture I gave you
when taking 10mm deep and like 2mm stepover
are you giving it a second pass ?
and feeding faster then recomended feedrate -_- apparently..
but it's still fine..
ie run your finishing cut
then run it again
* anonimasu nods
I dont need finishing cuts if it works like it does now always
climb cutting ?
dosent make a difference
2mm is way way too much for a finish pass
other then that I have problems to keep chips away from the edge..
does anyone run any of the leadshine drives or clones using emc?
.2 mm maybe
cradek, yeah i thought that
I were doing test cuts
lumberjack_jeff: what do they take as input? if they are analog +/-10 volt then LOTS of people are using similar stuff
and I got good finish even taking 2mm cuts..
cradek; disclaimer: I'm a noob. the docs say ttl input
lumberjack_jeff: do you have a link to the docs we can see?
I have it built and working, but the basic machine information asks for a bunch of values that aren't in the docs.
[22:44:14] <lumberjack_jeff> http://www.driver-motor.com/ProductContent.aspx?ID=59
that's the drive I have. I'm told it's a leadshine m880 clone
oh it's a stepper drive. I assumed it was a servo amp for some reason.
step time, step space, direction hold, direction setup...? *clueless*
step time is listed
dir signal says 2.5us pulse width, which is nonsense
sorry, the docs are not very good
bbl, the dinner bell rings
excellent, why did I miss that. Sorry, I really did try to rtfm. :)
thanks cradek. I have to go to town too. bbl.
cradek: I didnt mess around more to get better finishes once I saw that at thoose cuts ended up good enough