is 2.2.1 a recent release ?
fenn_ is now known as fenn
amazing. not only do I have a 2-56 tap like I need, I have dozens of them
they're so small, I can imagine overlooking them
a lot of them
buying boxes of unknown crap at an auction sometimes works out
yeah the box of a dozen is about the size of a box of matches
small matches ;)
sure they are not LH? ;)
ds2: I checked
I think it would funny to go swap out someone's tap set with a bunch of LH taps }:-)
funny but expensive
I think that would be very funny if you could do it without leaving fingerprints or DNA evidence
like DNA evidence as in the blood that will be flowing out of you when the guy finds out ;)
the evidence that would cause him to know who did it ;)
more like: thanks for the LH taps, I always wanted a set of those, now I'll go buy another RH set at harbor freight for $6
a cheap LH tap isn't that much more expensive then a decent RH tap
and if you buy taps at HF, you must be tapping wax
the ones in my set twist but don't break
heh, that reminded me of something
a couple years back I needed a LH metric die - talk about a rare animal
we once had a drill bit at work that untwisted without breaking when it caught in the hole
so you ended up with a reamer
of unknown diameter ;)
and only two flutes
and a random spiral - parts of it still spiraled forwards, parts backwards
I guess that should have been labeled a "twist or drill"
ISTR having a photo somehwere, but dunno where
I think I've seen that photo actually
it may be on the old metalweb site or something
How do you launch holecircle and involute ?
[02:17:25] <jmkasunich__> http://www.metalworking.com/dropbox/_2001_retired_files/untwisted.jpg
if you have .py files specified in the ini or axisrc file, then you just load them with axis
I don't remember exactly where to specify the file types / handlers though
heh. it would be cool to be able to write a C program, and have axis compile it for you, then run it, then use the result of that as the NC file :)
Do I need another python package to run from terminal?
that would be easy, your handler would be for .c files, and your handling program (script) would build and run the program, which would send the output to stdout
I think you should be able to use something like `python holecircle.py`
cradek, yep. seems like it should work
no, open holecircle in axis
your handler should be .py = python
can you not run it from the terminal?
SWPadnos: could be, but it will just spew the gcode to stdout
`python holecircle.py > my_nc_file.ngc`
I get an error in axis when I open a .py or .png
SWPadnos I'll try that didn't know python was part of the syntax
only at the terminal, if necessary
it should work as cradek pointed out, if the ini file is set up right
try sim/axis and if that works, you only need to fix your ini
Yep works in sim. I'll look at the sim ini thanks !
I'm happy to hear that...
crazy. each of these taps has "G#1" engraved on it by hand
wonder whose job that was, and what he did to deserve it
they're all a bit different
I think these are old. The labels on the boxes are about faded away, and the boxes have wood inside to protect the taps
(and they say USA on them, another sign they're old)
PeterDumbassW is now known as PeterW
I thought I recognized that name
so PeterW, any news on the 7i43?
Saw that discussion concerning watchdogs, Hostmot has a watchdog and I would expect most other most other controller card do too
yep. I think the WDT write is part of the misc update function
whatever it's called
7I43 cards came today, 2 minor patches and they are good to go, will probably ship some tommorow
still the $79 or $89 price I think I remember?
Yep, Ive updated the price list, there are three versions ,200K Parallel, 200K para+usb and 400K para+usb I think 79,89,99
nice. Hopefully i"ll have time to order some tomorrow :)
* jepler is tempted to order one too, despite how little he needs it
* SWPadnos is tempted to send you one anyway
(secret-- if you order parallel only 200K you will get USB for a while)
but it's so small ;)
Well its just a little smaller LUT wise that the FPGA on the 5I20
with the reduced number of I/O, it's probably fine
how many gates are sucked up by USB and/or EPP?
I am going to make a reduced version of HOSTMOT2 for it
The EPP stuff is pretty small, the 48bit GPIO config being about 4% of the 400K and 8% of the 200K
(the epp interface itself is probably about 1/3 of that including pretty good filtering so long cables should be ok
oh yeah - I was looking at the ethernet stuff earlier today, and noticed that the 4I71 isn't on the proce list
ok, not too bad
USB stuff is about twice as big sice we have a weensy 8 bit processor to do our little protocol
hmmm. which one?
or is it an embedded one>
Embedded Harvard Arch 100 Mips ~ 150 slices
small instruction set I imagine
is it the picoblaze or something else?
Simple accumulator oriented: lda,sta,add,addc,sub,subb etc etc
what, no FPU?
oh, is it similar to the softDMC processor?
No, its home grown, Picoblaze is not piplined so about 2 times slower
and microblaze is probably a bit too big for general use on these (now considered) small chips
Very similar to Sweet16 But Ive learned some trick to make it smaller
I also wanted something small and with no licensing restrictions
ah, ok. the prices are in the PDF, not on the HTML page yet
the picoblaze is relatively unrestricted, except that only people who buy the EDK can use it I think
Not sure its legal in non Xilinx FPGAs
right, and I don't think they give you source anyway, it's one of their IP blocks
luckily, Xilinx has very good support for encrypted Verilog/VHDL files
We originally used PicoBlaze for SoftDMC but wanted something faster
Who wants to mess with licensing unless its something you cant do yourself
indeed (unless you're in a big hurry, and price/philosophy don't matter) :)
Anyway, the source for the processor (Dumbass8) in the USB GPIO config is on the website
I suspect a lot of emc users will be chomping at the bit for these.
like the pluto, but better
they'll expect one of us to write a driver
We still need to get the protocol firmware working
jepler: yep surely so
Of course I do!
it's not just vendors ;)
jepler: actually I expect you to do it
that's why I'm considering sening jepler a board, of course
but I was too inept to get it right on the pluto
I don't want to have to do the dirty work ;)
I'll have to get sam interested, he'll send you one
well, with enough gates, I have confidence in you
* cradek ponders his plan
err - sounds great
jepler: my pluto works great. I don't know what you're talking about.
* jmkasunich__ ponders the summer that he didn't spend getting the 5i20 driver working
why dwell on the past
why not dwell on the future?
right, ponder the summer you WILL get it working, whichever one that might be
and in the future, you can dwell on how the future (now the present) came about
I certainly hope it isn't a summer
it should be this winter, if I can get my ass in gear
jmkasunich__ is now known as jmkasunich
hmmm. an ass-sized gear
maybe MSC has those
I have an ass sized gear in the basement
if I find my round tuit, I'll send it over
it's a worm gear, have no idea why they made it so big
I really don't want to know any more
too bad the 7i37 doubles the price of a 4i71
the 4i71 is an ethernet card
ok, so I"m not good with numbers
isolation ain't cheap
for dozens of pins
If I was buying it for work, $80 is very cheap
but my wife spent all our money
When I open a simple (or any) .ngc file in emc2 it just hangs.
what user interface are you using?
Whatever the default is.
It may be AXIS
I think some configs use other UIs
Should I try another ui?
does AXIS hang when loading the splash file (the one that says EMC2 AXIS)?
well... not at start anyway
i haven't tried loading it afterwards.
ok, try that
what's it called?
I think you can click some reload button
I don't recall where it actually comes from
what do you mean when you say hangs?
it's an old slow laptop
when I mouse over it i get a spinning wheel and nothing works.
AXIS becomes unresponsive, all apps are unresponsive
i have to kill it.
ok, wait a while
How long? I've tried waiting a few minutes.
if the laptop doesn't have much memory, then generating the preview could take a while
though it is strange that it loads fine at first
It's not swapping.
I think this may be a jepler or cradek question. I'm not very familiar with the inner workings of AXIS
well, that is a n openGL preview, so it could be crunching also :)
top says it's crunching...
but i think it's just looping doing nothing.
I wouldn't bet on it
It loads fast at at first though.
it's certainly possible, but this would be the first report of that kind that I've heard
also, if you've killed EMC / AXIS a number of times, it's possible that something isn't getting cleaned up correctly
I could try slowing down the real time stuff.
what do you have the period at?
I lowered it form default... looking.
raised I mean
went to a number less than 50000 or more than 50000 (which I think is the default)
ok, that's plenty slow - what's the CPU/RAM
ram is ~160MB
ok, that's skating a bit for Ubuntu
cpu is 233MHz amd
ok, that's quite slow
what about video?
At the moment I'm ssh'ng in... but was doing it locally earlier.
I dunno about the video. How can I tell?
lspci or check the X log
there's probably an even easier way too
non responsive i suppose is the correct term
is the neomagic on the host machine or the laptop?
heh - right :)
I'm sshing from a laptop to a laptop. The slow laptop running emc has the neomagic.
i think I should try a different gui
What do you use?
I suppose I could try slowing the CYCLE_TIME
I don't really use any GIU, since I have no machine at the moment, but I usually test with AXIS
err - GUI
I've loaded very large files with it - like 20MB, ~160000 lines
it takes a little while for that size file, even on the Opteron system I used for that test
I suspect my slow machine has found a race condition of some sort.
I don't think there's a race, and there's no RT usage generating the preview
there could be something that's somewhat inefficient, which combined with the slow CPU speed and potential lack of MMX-type extensions, gets exceedingly slow
well clearly something's wrong.
yep, sure is
I'd be curious to see what happens if you leave it overnight
is it eating up memory?
i just killed it.
i'll pay more attention next time.
I may leave it overnight.
was it eating up memory ;)
Not that I noticed.
I just got bored.
you can change the startup file to something more complex
I don't remember exactly how at the moment
You mean a more complex ngc file?
is axis the only one with the graphic preview?
so you ran emc from a terminal?
no messages or other status printed there?
I just cut and pasted from the icon it put on the desktop
oh. all you need to do is type "emc"
it should be on the path, and needs no arguments
doing that brings up pesky questions.
I guess that wasn't so bad. (just tried it)
whoops... still on tkemc
I think there's some option to just use the last config (and you can just hit enter to do that I believe)
you have to select the correct config
Just tried loading my test file again (which is just a couple lines... just goes to 0,0,10mm).
No messages on console
err... in terminal
well, it did the same thing reloading hte axis splash gcode, right?
memory usage isn't growing.
hmmm. you may want to post a message to the user list, or use the name jepler a lot on this channel so he can see the problem :)
I suppose I can use the more basic gui's in the mean time.
been ~15min... no messages, no increase in memory usage.
milltask is taking 10% processor and axis 76%.
ok. at this point I don't expect anything to change, but it shouldn't hurt anything to leave it up overnight
it's possible that there's a problem with the openGL implementation on your computer - that's not a very popular video chipset you have there
I'm just guessing here, if you couldn't tell ;)
well, right now it's running through ssh
Local machine is i915
hmmm. true, that should make a difference
I've tun it quite successfully on an i965 I think
It behaves the same local or through ssh
I'm going to fiddle with it... but may leave it overnight to humour you. ;)
the evidence point sagainst that helping, but what the heck, right?
Maybe the processor will burn out and I'll be forced to use a faster machine.
oh, that would be helpful
Increased BASE_PERIOD to 500000, no obvious improvement.
I just ordered a pair of these for experimentation: http://www.clubit.com/product_detail.cfm?itemno=A4842001
no, there wouldn't be. the RT system is running, but isn't really used when generating the preview
I figured that at $60 with PCI slots, they're probably a good option for EMC2
They have processors?
no ram tho?
it's an Eden I think, so it's soldered on
no, no RAM, but I just noticed the link on that page for 1GB DDR2 for $21.99 shipped
so still $81 total
err - $82
I should figure out if one of my old cases could power those.
get a picoPSU for $40-50 or so, and you have a very low power system that runs off 24VDC
it looks like a 20-pin ATC in the photo, but a newer version may be 24-pin
I have less than no room for new stuff.
if you toss the old laptop, there'll be plenty of room for it ;)
though you might be able to use the laptop LCD with it
I can't think of a good 24V source...
that lcd blows
oh, the PicoPSU also comes in a 12V version
I have a mostly unused good lcd monitor
I just like the one that you can use anything from ~14-~30V with
ok, haul that out, throw away the laptop, and you're golden
you could almost run the PC off the monitor power supply - I think it's ~40W total
probably less at idle
that's what I should be using for my server
Maybe I should turn one of those into a server and use the server for the mill.
oh - good point. good thing I got two
Seeing as it's on... well, never.
the S3 Unichrome graphics could be an issue - I don't know the state of that driver
For running emc?
I also do embedded consulting work, so I figured I could test these out for other purposes
would be a pain to move the server... ugh
Yeah... low power cheap PC's are great.
ok - time for bed for me. it's almost 2:00 AM
heh... west coast here
good luck with that crappy old slow laptop ;)
err - hi JymmmEMC
SWPadnos: I gotta
hit you up some time on led displays/drivers
ok. sometime later though
will email you instead
SWPadnos: atmel your thing?
cool beans, sleep well!
does the mesa card have a watchdog type circut to sence that the computer locked up?
* skunkworks_ could look
it doesn't have hardware to do that, but the hostmot config has a watchdog in the FPGA
so you have to program the FPGA to get it to work (which means the computer has to have booted and loaded the config, etc)
ok - that is what I ment.. I didn't mean extra harware - I ment programmed into the fpga.
also, it doesn't detect that the PC has locked up, it only knows that it hasn't been refreshed by the software, and sets its outputs to "default
[14:36:08] <jepler> http://linuxcnc.org/docs/2.2/html/hal_drivers.html#r1_5_2
the default value may be 0 (watchdog not enabled); you'd want to setp ...watchdog-control 3 to enable it and tie it to the HAL dac-write function.
yep, and it doesn't do things like microcontroller watchdogs do (reset the chip), all it does is set the outputs to 0 and turn off PWMs (or anything tied to the internal enable signal, I imagine)
actually, I haven't looked at the VHDL to see exactly what it does, so don't listen to me yet
I was just wondering in generalities.
don't worry, I don't plan to look at the VHDL for you (that's left as an exercise for the reader ;) )
[14:39:06] <skunkworks_> http://www.cnczone.com/forums/showpost.php?p=364052&postcount=20
sometimes I consider registering on that site to correct some silly comment, but then I think better of it
thanks for doing that for me (us) :)
no problem :)
looks like the hostmot watchdog may only stop the DACs; I can't see where it defaults or tristates the dumb I/O pins.
if ((MissedIRQ = '1') and (StopOnMissedIRQ = '1')) or
((WDTimeOut = '1') and (StopOnTimeout = '1')) then
StopPWM <= '1';
too bad peterw isn't around, he'd know for sure
oh, could be
after I said it I was trying to think of how he'd do that, which is why I put in the disclaimer about not having looked at the code
have you been reading back?
On HostMot, WD timout turns off channel enables, and clears PWMs (just to make recovery painful...)
I think on HostMot2 Ill just clear the DDRs so all I/O goes to startup mode
maybe apply the inverts first
I forgot - you only apply inversion masks to the alt functs
What Ive done for a custom design for someone is have a set of safe settings, to be swapped in on WD timout, but thats probably overkill
I think it may make sense to apply the inversions to all bits - you essentially set the polarity of a particular I/O, and it gets inverted or not regardless of whether it's simple I/O or alternate function
that would work
just a few registers with default values that get written on WD timeout
I'm probably not making as much sense as I think I
m thinking, due to lack of caffeine
just depends on space available
the 5i22 certainly has room, the 5i20 possibly not, but the hostmot config isn't all that space-hungry
Re WD: I guess I got pissed a little by a CNCzone posting suggesting something about non-step&dir servo systems crashing with +10V stuck
yeah, there's a lot of misinformation to be had out there
I wonder if there are any servo interfaces that _dont_ have a watchdog
it is possible, but since non-engineers can think of this problem, you'd think that the engineers who design the servo systems would also think of the problem and design in safeguards
5I20 configs with Hotsmot2 will just have smaller option sets, 12 axis servo or 4 Step+dir and 4 Servo should be easy
Very nice. I have the 5i20 for a large mill convertion.. have not played with i though yet.
The only thing that I'm having trouble fitting into 5I20 is my UART (8 channels takes 80%) so a serial amp + quadrature maybe limited to 6 channels
does it need to be a full UART?
I'm assuming that means you have a higher clock rate, you need edge detection and center-sampling, etc
Well this is my funny UART with 32 bit interface, Amp packet is just one write
is this a TX-only UART?
a UAT ;)
Yes, UART uses DDS as sampler, so no fixed /16
No this is for TX and RX
Is this a standard? (amps out there right now that would work with the uart?)
No, our wierdo stuff
ok - just wondering.
* skunkworks_ would not call it weird
it's weird from a UART purists perspective
It is just normal Async RS-422/RS485
with a 32-bit interface instead of an 8-bit one :)
We just wanted to minimise the wires going to our Amp, have smart bidirectional comms (forsetup out and bus voltage, temperature etc back)
Plus an interface the we can isolate with just 2 (well 3 because of XMIT ena) high speed optos
the AD11xx (??) series is apparently wuite good for that
We used optos on our first Serial Amp, because the AD11 series is not rated for 3750VRMS but they have a new series that is
how's this from a university professor: "I'm certainly not interested in actively supporting an open-source replacement of VRONI"
does "supporting" mean "encouraging development" or " answering stupid questions all the time"
ah, well too much of the stupid questions might make the guy a bit pessimistic I guess.
right - the context could change the meaning wuite a bit :)
but the thing is he is making money, significantly so according to his email, on this Voronoi-diagram code
oh, in that case he's being an ass
I would understand it completely if he were employed by a private company
unfortunately, universities are getting more and more like private companies these days
but simultaneously being faculty at 2 universities and 'information-hiding' doesn't really fit together
guess I will have to look at this Voronoi-business then from scratch when I have time... but it's supposedly not a trivial algorithm to get right (i.e. robust when the input and floating point calculations are inaccurate)
awallin: have you seen the method of using opengl to draw cones, and then looking at them from above?
cradek: I know the concept, but can opengl give you the data and not only the rendered view?
what's the relationship between voronoi diagrams and polygon offsetting?
jepler: the offset is fairly easy to construct once you have the voronoi diagram of the edges
[15:48:09] <awallin> http://www.cosy.sbg.ac.at/~held/projects/pocket_off/pocket.html
cradek: let me know if you have some links to the opengl based offsets
awallin: sorry it's only something I remember, no references
are you working on pocketing again?
not working, just thinking about it ;)
I have Held's book on the voronoi diagram approach, but it's pretty cryptic math/comp.geom. for a physicist to read
[15:52:28] <cradek> http://www.cosy.sbg.ac.at/~held/projects/vroni/img/arcs26_off.gif
yeah, stuff like that. http://www.cosy.sbg.ac.at/~held/projects/pocket_off/38_off.gif
the VD for that looks like this http://www.cosy.sbg.ac.at/~held/projects/voronoi_2d/38_vd.gif
the arcs look wrong in that one
you could look at how inscape does offsetting
or does it need to be 3d?
I think jepler examined that and found it did it badly
it just looked inaccurate i think
so a voronoi diagram is the same as the "skeleton" from cgal?
hmm no i guess it's different because voronoi has polygons in the skeleton, not just lines
anyway.. awallin should look at http://www.cgal.org/
iirc cgal has a license problem
CGAL is distributed under a dual-license scheme. CGAL can be used together with Open Source software free of charge. Using CGAL in other contexts can be done by obtaining a commercial license from GeometryFactory. For more details see the License page.
from that link
2D Straight Skeleton and Polygon Offsetting: QPL
QPL is basically a dual license
[16:07:34] <cradek> http://www.cgal.org/Manual/3.2/doc_html/cgal_manual/packages.html#Pkg:StraightSkeleton2
The Q Public License gives you the right to use the code under the condition that any program using it be released itself under an Open Source license.
if you are the copyright owner of the GPL code, you can amend the license by adding an exception allowing the use of CGAL with it
[16:08:51] <cradek> http://www.fsf.org/licensing/licenses/index_html/view?searchterm=qpl
Since the QPL is incompatible with the GNU GPL, you cannot take a GPL-covered program and QPL-covered program and link them together, no matter how.
they show right on that page how to fix it
right I see that now
if you're the copyright holder of all the gpl code
playin around with some of the sample configs and I try to start scara and tk window opens and no axis gui loads then I try to load halui_halvpc and I get emc still running do you want to restart I select yes and halui loads and then axis then tkemc loads too just letting you know.,if you didn't already
so as long as you contain the CGAL stuff to its own little area, it shouldnt affect a larger collaborative project too much right?
alSMT: thanks, I also noticed scara doesn't work. it will be fixed in the next release.
no thank you
what went wrong with scara?
alSMT: if you want, you can correct the problem yourself by modifying the scara.ini file as shown here: http://cvs.linuxcnc.org/cvs/emc2/configs/scara/scara.ini.diff?r1=1.8;r2=184.108.40.206;f=h
it was probably still busted from when you changed that stuff about active axes
is there a way unload emc from terminal?
that unloads all the realtime stuff
cradek: been on the indexpage touch probe, I was wondering if there any other sights on building a touch probe?
alSMT: sorry I don't know, I bought mine (long time ago)
it seems like it would not be a hard thing to make.
there was some discussion on emc-users, you might check the archives
I saw a design with ball bearing balls..
and a pcb
I have some ideas using and gate with contacts instead of the ball and rod type renshaw uses
I'm interested to hear what you come up with.
though I dont know if it's worthwhile when you can get probes fairly cheap..
atleast not for me :)
i want to talk with my nef. he's eng for borg see what he thinks
eng for borg?
ah, I cant remember right now
anonimasu: cheapest I saw was 100$+
yeah.. that's cheap
I think 200~$
well.. you can probably build one for ~20-30 if you want
I cant remember..
do you remember brand name
it was something like homeshopdepot or something
i'll give a shot
pulleys and stuff and things like that
lerman_ is now known as lerman
here's an interesting one http://www.aml.t.u-tokyo.ac.jp/research/probe/probe_e.html
[18:24:38] <SWPadnos> http://www.indoor.flyer.co.uk/probe.htm
that may be the one you were thinking f
ya I ckecked that one out
it was a porbe for sale.. complete..
oh. could have been IMService then: http://www.cadcamcadcam.com/index.asp?PageAction=VIEWCATS&Category=14
they don't list the price
they don't seem to have any in stock, but they're pretty inexpensive
i'll givem a ring
note that you're unlikely to get any support if you tell them you're using EMC
they sell their own controller, and I gather they're not too helpful for others
and don't ask Ray what he thinks of them :)
others like you if you're not using DeskCNC or their new ncPod
heh - hi Ray ;)
we were just talking about IMService
I remember him well.
EMC is top shelf
I'm not sure he's too eager to help out those who use EMC
Hey got a question re software encoder read.
but then again, he may be, so give it a shot if you like
9600 counts per second.
How fast does base period have to be?
oh, that's the question ;)
depends on latency
let's say 8
but you'd probably be safe with 25000-30000
or even slower
Okay. I'll get the stuff and try it out.
at 50000, you get 20000 interrupts/second, which is beyond Nyquist for that count rate
beyond = better than
but since you can't guarantee less than 42000 (with the hypothetical 8000 latency), that's the absolute minimum phase time for the encoder
which still gives you around 20k counts/sec
more than double what I need.
should be, if I haven't screwed up somewhere ;)
It should be fairly safe then.
You never screw up ... do you?
as one person I work with said "he's never been wrong yet, but when he is, I want to be around to see it, because it's gonna be a big one" :)
what's alSMT want Fred to do?
he was just looking for a probe, and IMServ has one
but they don't have a price - I guess they're out of stock
so he was going to call and ask about it
I mentioned that he might not get the best response if he told them he's using EMC
which I suppose isn't fair, because I have no direct experience with them, I've only heard about that from several others
That shouldn't be to bad. He might get a brief lecture about switching to his "better" systems.
heh - right :)
Just have alSMT tell Fred that Ray Henry sent him.
i'll do that
That'll get fred going big time.
all in good fun
I have a friend who uses deskCNC, but went out of his way to buy it from someone else because he didn't like the treatment he got at IMServ
Yep. Last time I saw Fred he told me is was obsolete.
i was obsolete.
oh, you. interesting comment, that :)
I knew it already but wasn't certain I wanted him to tell me that.
"you are obsolete" - what a statement
on second thought maybe i don't want there probe
Oh I think it's a good probe for what it is.
got ma new suse RT kernel
finally alsa will work properly and emc can run
unrelated to each other, but still good :)
I'm surprised that nobody else sells a probe
Yes rayh, you ARE obsolete and have been replaced by a 17yo McDonlads "Do you want frys with that?" punk ass kid.
Yup. You go the irony of that 'eh.
rayh: Eh, fuck em if they can't take a joke
cradek, I can't seem to figure out how your new probing algorithm works (http://timeguy.com/cradek-files/emc/smartprobe.png),
I'll assume the first row started from the left, how did it know how far to go up before going right? it seems to intelligently choose height based on how far it needs to go, and I don't see how it can know that with a Z-only probe
lerneaen_hydra: you forgot the 12 other axis
lerneaen_hydra: THEN it all makes sense =)
yes most true
especially with renishaws probe
There is the probe that someone made using a pcb and plastic and ball bearings.
lerneaen_hydra: It goes UP until the probe opens again, then LEFT or RIGHT until it contacts again or until it reaches the next desired X coordinate
the probe can be triggered by contact on any side, not just the bottom
I have a set of the papers from which the renishaw probe was designed. Speaking of obsolete!
jepler, it would seem to me that the amount of UP varies from probe to probe
oh, it triggers on the side too
jepler: only one axis can be in motion at a time during probing, right?
The old EMC routines allowed for three.
rayh: using what probe?
you can move any machine axis during a probing move
jepler: but only one axis at a time, or no?
you can move up to 9 axes
but at least one
so for instance you could improve the smartprobe algorithm by having a guess at the local slope of the item being probed, and instead of going left or right, go along the guessed slope in X and Z simultaneously
We were probing with a hexapod motors with 6 but using world coords and head tilt.
double flip tripple side retarded face plant
* Unit41 the crowd goes wiiiild
If I where to make one - it would be the 6 ball-bearing one. Looks cool.
oh, is 2.2.x out?
yep, 2-3 days ago
where have you been? don't you live and breath emc2?
come on now
please, go brush your teeth if you're breathing EMC2
Guest718 is now known as skunkworks___
ah, I fear I've fallen asleep at my post
the member is unmasked
please, leave your member masked
* skunkworks___ has internet issues.
has anyone had any problems with emc2.2 on late model amd systems?
I run it on AMD Athlon(tm) XP 2600+ without trouble, but that system is a few years old now
one is a socket 754 2.2gh,the other is socket am2 2.0gh,one bout 6 mos old,1 2 weeks old
what is the problem you're having?
both had to have the base task period changed from 10000 to 20000 to keep it from reporting a realtime error on startup
emc 2.2 reports realtime errors that emc 2.1 did not report.
problem is i dont get enough steps with the slower period on the new board
specifically, the rtai function rt_task_wait_period() provides a return value which says whether the requested realtime deadline was met. emc 2.1 did not report when this failed, but emc 2.2 does.
thats the error
please tell me it can be turned off
not without rebuilding from source
im finishing a mill monday,dont want to have to turn off the error everytime it starts up
will just use 2.1.7 for now
in src/rtapi/rtai_rtapi.c, change "if(result != 0)" to "if(0)" and the "Unexpected realtime delay on task #" message will not be shown.
2.2.x has the stepper step/dir speed enhancement right+
lerneaen_hydra: "doublestep"? yes.
i forgot about that
yes, that's what it was called
It thought it was called doublefreq
skunkworks___ is now known as skunkworks_
you can call it whatever you like
lemmee try that
jepler:i cant find a src folder,or were you talking about something else
I think you have to build a RIP
if the double freq works that would solve my problems on the new machine,can lower the base task period so the error wont show,but then i dont have quite enough step pulses
dont want to build a RIP
It worked for me. Takes a bit to setup - I think using the stepconfg would be the easiest. It is fixed now I think
you can do it from stepconf?
i got an update yesterday,and my stepconf will open now
havent used it yet
lemmee check it out
I have not used it either. but I think it is the easiest way to get doublestep setup. It will do it automatically for you.
work great had to try it
alSMT: how much faster was your top speed with doublstep?
maybe 38ipm to 57
i should say went from 38 to57
sounds like 50% to me
does stepconf have a way to use an output for a spindle brake?
stepconf generates a set of ini and hal files. if there's something you need that isn't supported, you can add it later with a text editor
indeed - that's why I said that ;)
* SWPadnos does simple math quickly ;)
and sometimes accurately too!
I ran a lot higher base before 2.2 realtime delay that i didn't get before the udated error mess.
right. I suspect that the system operates the same as before, except now it tells you there could be a problem
faster rapid and more responce from the gui
very happy should get rid of onboard graphic sone
yep - the longer base period should help with "responsiveness"
would this be correct ? net spindle-brake=> parport.0.pin-16-out
that is syntactically correct, but it doesn't connect the signal-brake pin to the motion controller brake output
I don't know the exact name, but it would be similar to this:
was gonna say its not working
net brake-out motion.spindle-brake => parport.0.pin-16-out
it's net <signal name> <pin name> <pin name> ...
you the man SWP
SWPadnos:can bad memory cause latency problems
or is that vidoe and cpu only?
bad memory usually cuases problems much worse than latency - like spontaneous reboots and crashes
skunkworks_: way to scare off the new guy
A long time ago (too long) I looked at the EMC project.
jepler: that happens a lot.
And wrote the beremiz project (www.beremiz.org)
Now, it is time to collaborate for industrial free software.
I'm also author of the canfestival project (www.canfestival.org)
are you related to Francois TISSERANT?
I dont think so.
All together we can do : G-Code, whole IEC-61131, CANOpen.
(he recently started helping the emc project by translating some of our user interfaces and documentation into french)
So, vous parlez Fraçais.
nous parlons english ici
J'avais understood. sorry.
heh - pas de probleme ;)
several of us can understand a bit of french, but not at a technical level
it's not "against the rules" to speak other languages, but most of the programmers are fluent in english only (I'm sad to say)
So, the beremiz project is a IEC-61131 IDE. With SFC (graphcet), LD (ladder diagram) FBD (Fonction block Diagram) ST (structured text) and IL (instruction list)
can it output to a classicladded CLP file?
or is it also an actual PLC controller?
No. It generate textual ST/IL/SFC and then compile it to ansi-C code
edouard_t: are you at all familiar with emc2's hal yet? It's the method we use to connect different components (real-time and non-real-time) together.
not at all.
I did re-invent the wheel.
the image links on the beremiz.org site seem to be broken
on the PLCOpen editor page
emc2 is very different from the original emc
yes, on page http://beremiz.org/The%20PLCOpen%20Editor/5.html
if I left-click the image below "- Configurations, Resources and Tasks" it is not found
All missing images are in the slides in the dowloadable PDF
ok - just thought I'd let you know
It will be fixed tomorrow.
At that time, just right click and 'show image'
OGG video on beremiz.org have to be seen with vlc or mplayer.
edouard_t: one way I see to work together is if your ansi C code can compile into a hal component, which communiates with other modules through pins of 4 types: bit, s32, u32, and float. To a C program, access to pins looks something like this: *(data->pinname) = newvalue; /* to assign a new value */ localvariable = *(data->pinname); /* to read the current value */
there are just a few API calls needed to set up "data" in a shared memory area
and give names to all the pins
another would be to replace the sometimes unusable CL editor ...
(at least unusable for me)
We have also made a kind if HAL.
it should be just another output target to make a clp file instead of a c file
my opinion is that the G-CODE related stuff of EMC could be easily be a kind of "plugin" of the PLC
it's not that simple, if you take into acocunt some of the premises of how EMC communicates between its different parts
There is no stcreenshot of the beremiz plugin manager at that time.
beremiz.pdf is giving my evince (ubuntu dapper) fits .. it's stuck at 100% CPU and won't display page 3 :(
Hard to explain the way component are seen in our IEC-61131 paradigm without some drawings.
Known bug. Use acroread. Sorry.
we'll each have to learn more about the other project, then find out how to work together.
you've made contact at a good time: we've just released a new stable version and are starting to work on new features again.
exactly. I have to give a try to the lifeCD.
Have you ever considered CANOpen as motion control bus for emc ?
I'm not sure that has been updated with the latest version
edouard_t: can is expensive..
that's the catch..
we've talked about a number of possible protocols, but none are implemented
CAN, modbus, ethercat, miscellaneous USB or other serial protocols, RTNet, etc
the live cd still has an older version (2.1.6, I think), but stuff like HAL is pretty much the same in 2.1.6 as in 2.2.1.
So, except parport, what is used to control operative part ?
a0hah anonimasu is calling CAN expensive
edouard_t: motion control cards and such things
but a $200 probe is cheap
[22:05:55] <SWPadnos> http://www.linuxcnc.org/component/option,com_weblinks/catid,11/Itemid,7/lang,en/
That doesn't look like a complete list to me
but it may be
edouard_t: from what i gather CAN may be too slow for motion control
it's being done all the time with commercial hardware..
down to 1ms sychronous cycle time for 10 axis.
CAN can go up into the negabits / second range, I think
edouard_t: besides software step generation on the PC parport (by far the most frequently used option), existing options include PCI and ISA cards as well as parport cards that use the EPP protocol. Depending on the card or board, they may ultimately output step&direction, pwm, or analog; in the case of servo motors, they typically take quadrature feedback inputs.
or well, faster too.. probably if your hardware supports it..
or slower, depending on how long your wires are ;)
though can isnt guaranteed to be realtime..
anything that fits the hal realtime model (accept a position command at the servo rate, typically 1ms; return a feedback position at the same rate) can work in principle
there's some attempts like rtcan..
with guaranteed timing..
but because all the included drivers are realtime-layer-agnostic (using our homebrew realtime API layer), it is hard to take advantage of advantaged things like rtnet, rtusb, and so forth -- for the most part, the existing drivers just bit-bang very simple hardware interfaces
*advanced* things, that is
the problem is tha error message may break CAN determinism
well, there's also a conceptual problem with using external smart controllers/drives - EMC likes to be "in control"
so, you depend on good wiring, with no transmission errors.
with feedback coming to the EMC trajectory planner on the PC, which then sends out new velocty / position commands
fenn: Any hardware with "can" interface costs a shitload of money..
so as jepler said, it should be possible to use just about anything that can send feedback and accept commands at the EMC servo rate
here's the mailing list post i was remembering http://article.gmane.org/gmane.linux.distributions.emc.user/4147/match=can+mbit
I worked for a company (where i developed CanFestival) hwo does RT motion control with CANOpen, without problems.
and of course you can always lie to emc, by looping back commanded to actual position, then doing whatever it is you want with the commanded position
(PPMC and UPC are motion control cards)
ooh the plc editor is in Python? count me in!
crap.... the wife just wiped out the Malibu into a poll..../...
edouard_t: the thing to remember is that most people that use emc are people with sane budgets ;)
dmess: I hope she's OK
she is ... the cars not
10 axes should be able to survive on 80 bytes per update - 4 in and 4 out. Add in some digital I/O and you might have 100 bytes. I think Jon's calculations may have been wrong
just more to add to my scrap pile..
but think of all the servos you'll be able to scrounge
(sorry :) )
pls.. be carefull with 7 + axes systems
I'm careful with 0-axis systems
the good.. your 1 of me
edouard_t: nice to meet you, and I'm excited to hear about your projects
dropping a brick on your toe still hurts, it doesn't have to be a meat grinder to do damage
have you beeen wrong more times than correct
I don't know
I'd be interested in running can to talk to io modules..
the GO with the flow bro... ;)
I decline to answer on the grounds that I may incriminate myself
(now if that isn't a confession, I don't know what is :) )
does emc bind with comedi (http://www.comedi.org/)
no, but it could
er, it could and it should, but it doesn't
We are working on such an absraction layer
it will let IEC-PLC exchangr at the same with all the different IO modules plugged in.
at that time, are planned CANOpen (done), comedi, ethercat (with EPL), powerlink, ... and any kind of modules (i.e. G-Code interpreter ?)
PLCOpen motion control blocks will also be implemented.
i work on problemed components.... so what do i know....
i upgraded the new computer to 2.2 again,did a new config with stepconf,added the brake,all looks good,no rt error on startup
very nice setup feature
is this still a valid command setp parport.0.pin-16-out-invert 1
what's the boards stance on gpl2/3?
the additions in gpl3 are mostly irrelevant to emc aren't they? but it certainly seems much more "legalese" than v2
it would be applicable to specialized IO hardware if I've understood it correctly, though the hardware would hardly use EMC code
it's too bad they arent compatible with each other
lots of things say "or later" though, fortunately
lerneaen_hydra: I don't know about the board, but I personally think we can't switch to gpl3 without a huge ordeal
I don't know enough about gpl3 yet to know whether it would be any benefit
lots of other applications that would also need to change?
a lot of files say gpl2
a lot of original authors are no longer involved
oh, and not gpl2 or later
EMC HAL seems to be in conflict with iEC-61131-3 configuration/ressources/tasks concepts. I have to go deeper into the manual to understand basic concepts.
hal is "parallel" and flow-charts are sequential?
in IEC-61131, tasks (what seems to ba called threads in HAL) already exist, and are defined in the PLC program.
in our project programs are compiled. There is no dynamic task/ressource configuration creation.
it helps to think of hal as an electronic breadboard simulator
edouard_t: is that a result of iec-61131 or just your implementation?
in some sense iec do not define anything on this.
edouard_t: hal isnt iec-61131-3..
although HAL can be restructured dynamically, in general a system is left alone once loaded and configured (and the HAL configuration can be locked if that's desired)
I know that. But it would be very interesting to make it collaborate seamlesly with IEC based PLC.
your PLC can simply export hal pins and functions...
edouard_t: i think the best thing would be to run a softplc and export hal pins to it..
In some sense, it would also be nice to see all HAL pins inside the PLCOpenEditor don't you think ?
yes, but it might get confusing if you used both systems at once
unification is the only good solution.
you will be assimilated. resistance is futile.
edouard_t: unification needs a solid standard..
also, it's nearly impossible to see the exact HAL pins you'll have unless you ar running on the target system, with the target hardware installed
try asking about "fddi" on #osadl.
Open Source Automation Development Lab
some drivers detect the installed hardware and modify the HAL pins they export based on what's found
unfortunately, there's no way to simulate that at the moment (but that is on my list of things to think about)
edouard_t: I still think hal should be hal and the plc should be the plc..
and keep them well separated..
IEC-61131-3 is about PLC programmation languages. IEC-61131 is more general.
I know ^_^
err it's too bad emc dosent have a plc that does that kind of stuff..
other than CL
for sur it will have it soon :-)
I dont like cl.. :p
me either, but it is there ;)
*orders a codesys runtime for linux*
I know Marc le Douarin personnaly. I'm telling.
To know more about OSADL Feild Bus Framwork : http://www.osadl.org/Fieldbus-framework.fieldbus-framework.0.html
oh I can live with the classic part..
just not with ladder.
FBD looks a lot like HAL to me
i think the schematics would get large very fast though, if you drew everything out like that
well, you have workspaces.. usually
so you dont stick everything at one place..
and you can nest fbd's..
anonimasu: is there a concept of namespaces?
er, hang on, let me figure out what i'm asking
ok, say you have 2 parports
one is at 0x378 the other is at 0x478
hal labels the first one parport.0 and the second parport.1
then you link pins to 0 and to 1
if you chop those up into different workspaces, is there any way to specify which parport you're linking to?
you connect blocks with lines(data paths) I guess you could call them
fenn: the thing is that there are 20 ways to do it..
that doesn't sound good
it's the beauty of it..
you can limit what can access things if you want to..
or export the pins as globals..
IEC defines directly represented variables.
%IB0.1.2.3 for a input byte
%QX2.1 for output bit
who was looking for a probe earlier?
%ML220.127.116.11.5 for memory long
alSMT was looking for a probe :)
fenn: yeah, I got that..
he's not around anymore though
found this: http://www.cnczone.com/forums/showthread.php?t=28308
fenn: maybe you can pass it on, if he shows up again
edouard_t: yeah, but it depends on what tools you use.. if you take sigmatek as a example their hardware are very modularized
edouard_t: they dont export any globals, because it wouldnt work out, if you have 40 modules in a system..
That is what we solve with our plugin framework.
structured text is ugly :\
hehe, you get used to it..
I know, if wrote parts of the compiler. But it is standard.
there has been a lot of talk of integrating all of emc's configuration (including .hal files) into a single .xml file
perhaps it would be more useful to translate .xml into structured text than to write a special hal interpreter
we map directly represented as tree leaves, and all tree nodes are plugins or sub plugin instances.
hu... sure that none did understood that.
yeah it seems a bit out of context :)
%Q... is an end node?
I think the explanation of this concept is out of the scope of this channel.
fenn: he's talking about how the compiler handles memory..
I'll publish a paper soon about all this. If you are interested into this please register to the beremis project annouce mailing list.
fenn: And how parts of code get's mapped..
got a homepage for it?
ok, i've never written a compiler but i can imagine
[23:28:44] <edouard_t> https://sourceforge.net/mail/?group_id=201091
Sorry. I have to sleep. Midnight and a half... Too tired to keep concentrate. By
I feel like im going to dream in st.. tonight :p