Jymmm is now known as Jymmmm
steves_logging is now known as steve_stallings
my emc machine is still on.
I should probably turn it off.
steve_stallings is now known as steves_logging
I long home to play with the mill today
I'm logging. I don't understand '', alex_joni. Try /msg logger_aj help
no more logger_aj from now on :)
logger_emc is the new one
I'm feeling lazy .. but here's the log anyways: http://www.linuxcnc.org/irc/irc.freenode.net:6667/emc/2006-11-01#T09-49-32
is ti any better?
same stupid logger
but moved to www.linuxcnc.org
lerman: hi ken
* alex_joni needs to run
you back yet?
alex_joni, give them a MIME-HEADER
so I dont need to download it...
I am alive... just ignored... on irc as in life.
I have a question for you.
lerman he ain't here atm.
So, I gather.
do you miss him like I do?
he will be missed...
Actually, it is a general question for the qorld.
he lived so well... and helped out the community so much.
May the Lord let him rest in peace.
I'm working on a conversational mode gcode editor/generator. What resolution screen should I assume people have?
I'd make no assumptions
lerman, eeeeeeeeeew. 800x600
get a new monitor... from a dumpster.
Well, you can scale graphics, but there is a minimum at which things look good.
lerman, 1024x768 [for real]
but it should work well on a larger screen too (resize intelligently)
yeah - what cradek said
onResize() = call intelligent_scaling();
I have an 800x600 panel PC for my mill, and that's probably a reasonable minimum to expect
lots of the world is still 800x600.
touchscreen as well, so think about left-click only mice :) )
A-L-P-H-A, yes it is. :)
HOW Come lerman gets people talking, while you mofos ignore me. [albeit mine wasn't important]
I'm working on a mailing list thing, so I have no time for small talk
A-L-P-H-A: because what you typed was content-free
I've just barely started, but 800x600 seems reasonable. I'm building it using Qt.
there are a couple of other things to think about WRT touchscreens, like not having controls at the very edge of the screen (they're harder to hit)
lerman: can you give an overview of what it will do?
Qt has the advantage that I can do a lot of the work on my windoze laptop and than bring it out to the shop.
Take a look at the mach3 wizards.
is Qt Free?
it is for free projects
I don't have mach3 or windows to run it on
free / oss
Qt is f... what swp said.
so not DFSG-free?
the free version may be, but I'm not sure
they do distribute KDE, which is Qt-based
Qt is free in the Free Software Foundation Sense.
Uses gnu license.
right, but they also will sell you alicense if you want to write non-free software
[15:17:48] <cradek> http://lists.debian.org/debian-devel/1999/04/msg00808.html
It will have graphic screens for things like milling circular pockets, drilling hole patterns, milling surfaces, etc. It will need a keyboard to set diameters, etc (touch screens are lousy for that).
This license issue caused a fuss, and the result was (a) GNOME, and (b) TrollTech re-released the Qt library under a QPL/GPL dual license, which resolved the issue.
lerman, have you looked at cp1?
I suppose I should look at it before suggesting that it may be relevant :)
The basic idea is that there will be a library of templates -- each will have a directory containing:
A graphic showing the operation
A config file defining the locations of text, list, etc boxes on graphic.
The gcode of a subroutine that implements the operation
Haven't looket at cp1 (or heard of it, before). But I've used Qt.
sure - that'll look nicer for sure
is it possible to use gtk instead? (there is a Windows version, crappy as it is)
the main reason being that gtk/gnome is standard on Ubuntu
also, gtk is already a requirement of emc2
(well, if you want to use halscope, halmeter, or classicladder GUI)
cp1 is a tcl front end to some geometries.
it writes the gcode for them.
It was based around some early work at NIST that embedded geometry parameters in gcode comments.
A file generated by this thing will contain the subroutines used in the beginning (unless I can get lerman to implement an import facility). Lines that care generated by the conversational mode will have imbedded comments that tell the system how to edit them in the tool. So, you can use this to edit regular gcode and imbed the conversational stuff.
Also, a user will be able to build his own 'screens'. That means we will have the facility for our community to grow this thing.
I also hope to (later) add a tool that facilitates the construction of new screens.
Qt has a nice facility for designing GUIs (called 'designer') that is pretty easy to use.
I never quite got that to generate code (that could talk to the outside world)
Tools are always a pita because you have to spend a lot of time learning them.
I was trying to do a slightly more complext hing though - make controls that could be used in designer, and had EMC-related functionality
gtk has a designer called (mumble) that I used once and it worked
but yeah, pita.
all gui builders (and maybe all gui toolkits) are a pain.
03jepler 07v2_0_branch * 10emc2/src/hal/drivers/m5i20/ (10 files): fix references to "version 2.1 of the GNU GPL"; should have read "version 2"
03jepler 07v2_0_branch * 10emc2/src/hal/user_comps/devices/hal_joystick.c: fix references to "version 2.1 of the GNU GPL"; should have read "version 2"
03jepler 07v2_0_branch * 10emc2/src/hal/classicladder/module_rt.c: fix references to "version 2.1 of the GNU GPL"; should have read "version 2"
03jepler 07v2_0_branch * 10emc2/src/hal/user_comps/vcp/ (tokenizer.c tokenizer.h vcp.h vcp_main.c vcp_widgets.c): fix references to "version 2.1 of the GNU GPL"; should have read "version 2"
03jepler 07v2_0_branch * 10emc2/src/hal/hal_priv.h: fix references to "version 2.1 of the GNU GPL"; should have read "version 2"
qt designer is much more functional than glade
03jepler 07v2_0_branch * 10emc2/src/hal/drivers/ (15 files): fix references to "version 2.1 of the GNU GPL"; should have read "version 2"
03jepler 07v2_0_branch * 10emc2/src/hal/components/ (9 files): fix references to "version 2.1 of the GNU GPL"; should have read "version 2"
03jepler 07v2_0_branch * 10emc2/src/hal/utils/ (15 files): fix references to "version 2.1 of the GNU GPL"; should have read "version 2"
yep. they all do 1000 things, 998 of which each developer doesn't care abou
but it's a different 998 for everyone
My current thoughts are that AXIS could invoke this thing. The user would select the current line and hit edit (or some such). That would call my stuff with a file name and current line number. When my stuff finished editing, it would exit and AXIS would reload the file.
AXIS can already invoke an external program that produces g-code
A difference between the good gui's and the bad ones are the degree to which you can ignore the things that don't matter (to you).
but it (currently) does this for the whole program, not to assemble a program "piecemeal"
That's what I assumed. My thing would do the piecemeal assembly.
Have you guys seen this definition: Abstraction is the principle of ignoring the differences so as to treat things as if they are the same.
you can invoke your program for files with a particular extension by adding two lines to the inifile: http://linuxcnc.org/docs/html/gui/axis/#SECTION001111000000000000000
I'd have to search for the reference, but that's the best definition of seen.
that sounds like a good definition of Abstraction to me
I looked at the hole circle program. That's sort of the idea of what one of my screens would do, but it would try to show things more graphically. A problem with hole circle is that, for instance, it might not be clear what the definition of 'start angle' is.
Also, this program would be generic, in the sense that users could build new screens very simply and they would automatically be incorporated into the system.
that sounds really neat.. I have never played with conversational programing
or been exposed to it.
Neither have I... :-)
The emc subroutine facility is what will make it doable for me. Subroutines provide an abstraction that will let other people define the steps necessary to do an operation. I can provide the framework in which they can embed the operations.
that was also my idea when I made useful-subroutines.ngc
are there either parameters or local variables for subroutines?
Both parameters AND local variables.
look at useful-subroutines in cvs nc_files
ok - #1-#n
actually I'm not really sure how variables work (which ones are local)
(a pointer to the docs would be appreciated)
The first 30 (I think) are local.
In the Wiki, someplace.
and #1-#n are populated with the parameters?
lerman: did you see sample 5.3? :) really worthless but fun none the less. http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Oword
[16:14:12] <SWPadnos> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Oword
that was funny
I saw it, but didn't read it.
Optional arg1, arg2,... (up to arg30) will be passed in parameters 1 to 30 in the context of the called subroutine. On return from the subroutine the original values of those parameters will be restored.
this is the important part: http://www.electronicsam.com/images/KandT/Spiro-0codes.png
I can't quite tell if that means all 30 are local, or only the ones used as arguments
All 30 are local. They are pushed onto the stack.
they're all local, and any arguments populate the first N (as I see it)
ok, and nesting to 10 levels
I'm going to clarify that
one would hope that would be enough ;)
30 was an arbitrary number that seems sufficient. Also the 10. I could have writtent he code using malloc to allow any level of nesting, but I didn't feel like adding more room for errors.
BTW: you could write recursive code if you like. Anybody want to implement a flowsnake?
I suspect it's more than sufficient
only to 10 levels though
I suspect that recursion would blow that limit
err - blow by that limit
lerman: please check my clarification for accuracy: http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Oword
It's the Mandelbrot gcode generator?
I'm not following this discussion, but there is documentation for O-words in the .lyx documentation and I believe that should be the "canonical" documentation -- not the wiki
cradek -- correct.
even better if they match
No... Thank you!
SWPadnos: I'd rather that the wiki refers to the docs, rather than duplicating it
it just makes more work when someone revises one
Any idiot -- even me -- can update the wiki. I barely know what lyx is. And I probably don't want to learn how to use it. Hell, I'm trying to learn Qt.
jepler is right.
a simple case study shows that I'm willing to edit a wiki page but get huffy when I have to wallow in lyx
if something in the docs is wrong or not clear, and for whatever reason you can't update the docs, then put it in the wiki
I'm sure I'm going to have more problems when I get around to shipping this new stuff. Making it build and install on a variety of platforms will be a pita.
but that's so easy with a corss-platform toolkit
lerman: I think that's another good reason to use gtk instead, if your target is machines already running emc2 (or a superset of that)
If I call an O-sub with 1 parameter, does #2 have a defined value?
But I know more Qt than I do gtk.
The qt runtime library is trivial to add. It's already in there if you've added lyx-qt
jepler -- yes it has the value of #2 in the calling context.
So... you actually have two ways in which you can pass info. As an argument, or as a parameter.
I'm putting that in the docs too.
cradek: still on the wiki?
yes I'm still updating/building cvs before I can edit anything there
(I wish I could make it as precise using fewer words, please fix it if you can)
Some of my code uses the convention that parameter #100 is used for a return value.
maybe all that wordiness about context/modified/restored should be replaced with '#1..#30 are local, the rest are global'
When does parameter assignment happen with respect to the call? Consider this line: O1 call #1=3
Will the O1 sub see #1=3 or the prior value?
01 call [#1] #1=3
Assignment occurs before the call.
cradek: I believe you will see the old value.
cradek: that's different than what I have in mind. It should give the prior value of #1 to the O1 sub.
BTW: I've written code that looks like:
O#1 call  
* cradek shivers
Is this legal, or can O- not go together with other words? O1 call .... G1 X1
* alex_joni just got back
Suppose you have a hole pattern you would like to drill then ream. Call the code that traverses the pattern with an argument of the tool subroutine to use.
jepler: I meant to say that your call syntax was wrong (brackets required)
lerman: what you're trying to implement sounds a lot like cp1
cradek: yes, I said "angle" -- I saw that a bit ago too
jepler: I thought I typed the syntactically-correct equivalent question, but maybe I didn't
jepler: possibly legal. I haven't defined it.
alex: I wouldn't know. -- I'm not familiar with cp1.
lerman: maybe you should take a look at it
it's a collection of tcl files
tell me where to go... :-)
very easy to run (any normal tcl installation will run them)
hmm.. I forgot to put them back on linuxcnc.org
but I mailed a copy to SWPadnos
* alex_joni prods SWPadnos
it's on the wiki now
Cp1 on the wiki
SWPadnos: thanks :)
I just downloaded it from there ;)
[16:41:09] <cradek> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Cp1
lerman: I ran it on doze, and it worked pretty ok
it runs funny on Windows, but I'm not sure if that's tcl on windows or cp1
there are a bunch of .tcl's inside
there's a cp1gui.tcl iirc
I'll take a look later.
I think mshaver was saying that there were more operations added at some point (up to cp8, I think he said ;) )
it has a collection of MOP's
a MOP is like a screen you are talking about
So... why don't people use it? or do they?
lerman: poor advertising probably :)
it's menu driven, not graphical..
I'm sure it doesn't use subroutines... Since they are new. I think that can make a big difference.
also, it has limitations (the bolt circle routine only allows the first circle on the X axis, for example)
no, it doesn't
it does use comments to identify MOPs though, so you can edit them later
But tht should be a matter of changing a routine. So, it sounds like it is not a framework in the sense that I'm talking about.
it is a framework, in that the main GUI loads MOP functions from separate files
I'm not sure if the function list is in the code or dynamic
What are the MOPs written in?
well, tcl for the calcs, G-code for the results
ok - the MOP list is hard coded at this point
this is a nice conversation *lurks*
talking about cleaning floors
lerman: I just took some snapshots, I'll put them on the wiki
actually, there aren't many functions in the version I have - only 7. it may not be as much of a help as we're thinking :)
alex_joni: hurry up this is making me curious ;)
just kidding, but it's interesting
[16:51:48] <alex_joni> http://wiki.linuxcnc.org/uploads/cp1-circle.png
[16:52:07] <alex_joni> http://wiki.linuxcnc.org/uploads/cp1-textengrave.png
I wonder what the code looks like in AXIS?
SWPadnos: pretty ok
except you can't put 2 MOPS in one file
they end in M2
Sorry... I must leave you now. I'm taking my wife to lunch (salad at McD's). I'll be back in an hour or so. Ken
SWPadnos: disregard that
they don't end.. so you need an M2 at the end of the file
see you later
[16:56:51] <alex_joni> http://wiki.linuxcnc.org/uploads/cp1-axis.png
2 MOP's : one for text engrave and one for pattern drilling
big grill array :)
well, small holes :)
about right for an outdoor speaker grille
I think that's exactly what this MOP was for
yep, or ventilation grilles
03jepler 07HEAD * 10emc2/docs/src/gcode/main.lyx: improve o-code documentation based on wiki and irc conversations
I never thought I would see improve and irc conversations in the same sentence...
Lerneaen_Hydra: most we do is improved based on irc conversation
by IRC I didn't mean #emc as much as say efnet ;)
this is freenode.. not some pesky efnet or dalnet
I'm in some efnet channels too, that's what the humor in the first sentence is based on
RUN EVERYBODY! RUN!
You can run, but you can't hide. :-)
* Lerneaen_Hydra runs away, albeit belatedly
tag.. you're it
Nah, just getting all these folks off their asses and getting them some exercise
* Lerneaen_Hydra send his crab at alex
* alex_joni eats it
(if you don't get that reference you need to brush up your greek mythology)
We did sushi instead of McD's.
alex, you should know you just can't eat the crab of lernea
Lerneaen_Hydra: even better
Lerneaen_Hydra: sure I can
Had some kani.
lerman: even better
Ok, so when should we expect emc to be running on openMosix? a week?
[17:59:35] <Lerneaen_Hydra> http://en.wikipedia.org/wiki/Lernean_hydra
Jymmm: openMosix ?
see, I've even got an article :D
Lerneaen_Hydra: you just wrote that to seem important
Lerneaen_Hydra: I just assumed it was you real name ;)
Lerneaen_Hydra and your photo too... fbi 10 most wanted
Jymmm: are you submitting a patch for it, then?
jepler Wait for yours
alex_joni: yeah, all 70-ish edits ;)
Lerneaen_Hydra: it might bear some truth there...
"poisonous breath" :)
Jymmm: ssh! don't tell them that
* Lerneaen_Hydra readies a mouthful for alex
jepler did you even look at it, or was that your standard "feature request, are you gonna do it" response?
Jymmm: I have no idea why you would want to "port" emc to clustering software
* alex_joni asked what it was.. but got no respones
one axis per cpu
alex_joni google it
Jymmm: I'm not gonna google for not well defined user requests
I'm way too lazy for that
alex_joni #1 hit you lazy bastard! LOL
I suspect that one day, when someone will _really_ want this.. they will bother to make a better description of what they want/need
whats the limitation of PERIOD ?
the slowness of ISA
it doesn't matter if you have 1 or 8 outputs on the parport
ok, so if 3 axis are running off of a single cpu, what happens to PERIOD if there's only one axis?
at least nothing you would notice
that almost seems like going backwards
the calcs would be a tiny bit faster
so running one or 2 axis off a single cpu is the same performance?
I'd say about max 5-10% difference
but you have lots of headaches
because I bet you want them to run synchronized
e.g. cut a circle not a rectangle :D
ok, and if you distribute the work off to 3 cpus instead, the load per cpu , per bus has to be almost 1/3 less, no?
that's what I'm saying... out of a normal step output period calcs are not more than 10%
the rest is waiting for the ISA to turn on/off the pins
put if theres only 2 pins per cpu, that has to be allow for a smaller PERIOD?
depends on the pins
if you use processor pins (programmable I/O) then you can drive them really fast
ok, what about on a 6 axis machine?
but I haven't seen any i386 that has programmable I/O pins
Jymmm: same thing
calculations are 10% of the period
6 axis @ 10% == 60% =)
you can improve those 10%
no.. 10% for all of them
No, I reject your reality and substitute my own!
be my guest
hey.. my screen is bending
your reality is twisted
Jymmm: I agree the solution is interesant.. if you would have the workload to split
What if you bypassed paraport, and drove directly off of the pci bus? (single axis)
Really. And those kernel modes would interact with the real time kernel modes exactly how?
Jymmm: that would rock :)
but you need custom HW
O.o alex, what does interesant mean? it's one letter away from intresant, which is interesting in swedish
lerman would have to be multi threaded
aka lots of money
Remember thatmaximum throughput and minimal latency are not necessarily compatible.
sorry.. wasn't thinking :D
alex_joni Well I/O pins off the pci buss
Jymmm: haven't seen any yet
pci buss is address based
you need decoder hardware and interface
only afterwards there can be a chip that has the 2 pins
alex_joni you mean like PCi0, PCI1, etc ?
there are solutions with an FPGA that emulate all that internally
Jymmm: no, I mean like 0x54736589
outb (0x54736589, 3)
that's how you set 2 pins to true
compared yo outb(0x375) ?
well ox375 has GOT to be slower
not because it's a lower value
but because the ISA bus runs at a much lower frequency than PCI
and it takes a while for the command to actually cause the change
usually it's at around 1 usec
Does PCI really support single word input and output at memory speeds?
Well we're talking different addr space too (I was just saying the pci addr space compared to I/O addr space)
yes, but it still requires a bus transfer setup, I think
writing one word (which can be 32 bits, that being the major speedup for PCI) takes a while
Does it require a setup per transfer?
writing 2 words takes one extra 33MHz cycle
lerman: unless it's dma
I'm not sure what the setup time is though
you could theoretically set up a sequence of steps in memory, and DMA that
SWPadnos: when in doubt.. buy the specs :D
lerman, as you said, throughput and latency aren't necessarily compatible
SWPadnos: I recently learned that the full spec for PCI costs a lot
only a small amount is publicly available
didn't Pete W write on this topic on the user or dev list?
yeah - that's why I said "wcrew that" ;)
* alex_joni can't remember
I guess a better way to ask it is:
Suppose jon elson built his board to use usb (which runs over pci) instead of a parallel port. How would that change the performance.
[18:20:46] <Jymmm> http://www.ics.uci.edu/~harris/ics216/pci/PCI_22.pdf
Jymmm: read my message a few lines up
[18:21:09] <alex_joni> http://lists.linuxcnc.org/listinfo.cgi/emc2mail-linuxcnc.org
only a small amount is publicly available
lerman, it would make it totally useless for EMC2 with feedback
it can be enough for most trivial cards
for step generation, it would be fine
no rtusb? ;)
Jymmm: there are lots more in the official spec.. had a project partner buy one recently
skunkworks, not now, not ever :)
Thsi is from Intel
skunkworks: rt it what you want.. you only get packets once every 1 msec
unless they change the spec to use a much faster packet rate
Jymmm: there's a PCI consortium that sells the spec
SWPadnos: that's unlikely to happen
Well, how about the mesa card? That uses pci. What does it's performance (PERIOD) look like?
alrught, how bout we rape ancient SCSI cards and give em new life.... 7 addressable points =)
dunno, but I can say that people who have experimentes with the STG (?) cards were unable to get beyond a couple of KHz update rate with the ISA version, and 8 KHz or faster was easy with the PCI version
It sounds like a possible approach would be to move all of the kernel stuff (HAL and drivers) into a reasonable RT processor.
that's been done - look at the Gecko G-Rex
Ok back to basics.... how hard would it be to blink an led off the pci buss?
Well, I'm thinking more literally. Compile and build the HAL stuff. Get a board that can interface with a linux system and move the kernel stuff into it.
you need a PCI interface chip to decode addresses and provide setup / configuration information to the BIOS
Well, buy a mesa board and I think you can do that pretty simply. (blink an LED).
why are we worrying about this? Are we just trying to get by with slower computers?
lerman no purely plugged into a pci slot (w/ assoc components)
skunkworks: wasting time
you need to design a board that will operate at 33 MHz, with ~100 pins on the bus, and conforming to many timing specs
Jymmm: can't do that
alex_joni why not (seriously)?
No. Even the fastest x86 has relatively long IO latency.
what SWPadnos said (seriously)
the mesa card, minus the FPGA, is about as simple as you can get actually
Jymmm: the PCI is a bus.. that means all the wires go to all the slots
Jymmm: each board decodes the address and decides if the data is for itself or not
Why? we are only outputting data, not rx it
because peopl;e with a PCI bus probably have other cards, and your card has to play nice with those
Jymmm: do you know what an address/data bus is?
Well, the FPGA probably does everything on the chip. Including the entire pci interface.
no, the mesa card has a separate PCI intetrface chip
makes it easier to update the config if the FPGA isn't needed for the bus
OK. Is the interface essentially a single chip.
[18:31:13] <Jymmm> http://www.mev.co.uk/pci_design.htm
Jymmm: it's a bunch of lines called A0..A31 and D0..D31
you need to have a decoder chip to test for a certain address
has device drivers for linux too
of course it's a bit more complicated, because PCI addresses aren't fixed like early ISA
alex_joni I said YES
the PLX 9030 is what's on the mesa card
ok, NOW you said something... not fix addr space
that I can see as an issue
SWPadnos your facorite too... labview
the chip needs to talk to the board / BIOS to get an address assigned
and from there one it only get more complicated
argh.. can't spell today
and from there on it only gets more complicated
But, natively (in some twisted sense) the slots are hard numbered too
yes and no
that doesn't matter to software
it generally only affects the interrupt that gets assigned
they usually have some fixish irq number assigned to them
there ya go
and has nothing whatsoever to do with the data / address bus
some share an IRQ number though
but that means if you put the card in another slot it should know that happened :D
per pci spec yes.
well, in theory there are 4 PCI IRQs, and they're rotated one position from one slot to the next
or what to do if it shares IRQ with another one
most cards only connect to IRQA, so they did that to distribute the IRQs
alex_joni hard code IRQ in bios?
SWPadnos: seen boards with 3 slots, but using 2 IRQ's
many motherboards don't implement this properly anyway
Jymmm: except that linux won't give a crap about the BIOS in that case
sometimes it realizes something is not according to spec, and decides to do the right thing
but that's magic to me :D
and the onboard functions (IDE, network, sound, ...) are on the PCI bus as well, so they're sharing also
so, we disable those too
even ISA is hooked up to the PCI bus, through a bridge
no need for em!
likewise mouse/keyboard I think
at least used to in early days
anyways.. in 2-3 years there won't be any PCI around either
Te 66 MHz pci spec says that the target inital latency is 16 clocks. That's around 250nsec. That doesn't seem too bad.
[18:38:25] <SWPadnos> http://www.pci-tools.com/english/prototyp/index.html
only PCI-X and PCI-E and whatever sick names they come up with by then
bet that costs more than an integrated parport
bet that costs more than a computer
99 EUR for a bare board
Price/Preis: 279,50 EUR for the complete board
yeah - depends on the baord though - the PLX-M board is 539 EUR
you can get a MESA for a bit more :)
I can get 3 for a bit less ;)
$199 in singles
* alex_joni is not building an emc2 compatible hardware museum
ok, what about raping a add-in USB or Paraport PCI card?
parport PCI is nice
amazingly, they're $159 in qty 5 o_O
but it's still ISA inside :)
so it won't go faster than onboard
PCI parport cards - they're only $20 or so ;)
there's a big chip on the board
that's why I said rape the card for the "big chip on board"
you can get PCIe serial and parallel cards
talk about a waste :)
it contains pci-interface, pci-isa bridge, and generic parport
SWPadnos: the pci parports I got (2 parports/board, netmos) are about 20-30$
[18:43:23] <SWPadnos> http://www.byterunner.com/byterunner/category=Parallel+PCI-bus+cards/exact_match=exact
but they were too lazy to implement a new parport interface
and just put the isa one in there
well, that wold be more expensive, and useless for the normal parport applications
right.. and possibly cause timing issues :)
printers printing every 10th line :))
actually, I have the PCI-1284-P2, and I don't think I've gotten it to work with EMC
mine's a KWE 220N-2
with Netmos Nm9715CV on it
[18:47:01] <Jymmm> http://www.plxtech.com/products/sdk/
Jymmm: you're not gonna convince me to buy their boards :D
chips, not boards =)
even if they have a great sdk and open source drivers
* alex_joni stays away from PCI board design
me too, and I've got the software / test equipment to do it
<rant> why not design our own CISC processor with emc specific pins </rant>
that's what the MESA card is for ;)
SWPadnos: nah.. one that runs at 2GHz+
Because RISC is faster than CISC. (troll mode ON).
lerman: lol.. ok, we do not wanna go there
* alex_joni throws an microprogrammed processor at lerman
what processor do you want to emulate today?
an emc2-specific VLIW CPU, 2GHz+, and costing <$20, with integrated H-bridges
An Atmel, I hope. Does the first one to port the RT stuff to a single chip machine get a prize?
yeah 2 of them :D
Two atmels or two prizes?
lerman: does eternal gratitude and respect count?
well, you need 80 H-bridges - what if someone wants 8 axes of 5-phase steppers?
That's two prizes. Sure. (respect more than gratitude)
SWPadnos: pluggable h-bridges
lerman: you'll get both from me
I'll add that to my list... After the gui stuff.
SWPadnos: but it might be interesting to look at Axis (the processor)
well, for $20, I'd wat to be able to power it with up to 13200V, and program an integrated step-down power supply so that I could use different voltages on each output
I'm not sure I want to take apart my camera right now :)
yeah, and it would be low power consumption
at 13200V , it would only take a few microamps :D
of course - no more than 10 mW over the motor power, with full regenerative braking
And you could steal those with an antenna.
...the microamps, that is.
the app for the G-rex actually uses a cut-down version of EMC
[18:55:55] <Jymmm> http://www.openwatcom.org/index.php/PCI_Device_access_under_32-Bit_PM_DOS
that's on a Rabbit controller
Jymmm, the software to use PCI is trivial
it's the hardware and all the setup crap that's a PITA
[18:56:25] <alex_joni> http://188.8.131.52/search?q=cache:uc77T4wVdEUJ:users.student.lth.se/d98mad/etrax_and_rtai.html+ETRAX+100LX+rtai&hl=ro&gl=ro&ct=clnk&cd=1&client=firefox
cool - thanks
I may need that for the camera
SWPadnos: I sstill say that we could rape an existing card (somehow)
SWPadnos some cheap tW import that costs $5
I doubt it
SWPadnos: I am in touch with the people at the Lund university in sweden
they did that a couple of years ago
I can put you in touch with the right people if needed :)
the problem is that the bus state is transient, and you want to end up with a steady state output from that
alex_joni, cool. I'll let you know
I don't think I want to run linux on the micro. I want to run the RT application parts of EMC2 (HAL stuff).
I've been considering making an ARM-based controller that would work with EMC2/HAL
communications are an issue though - ISA/serial/USB have problems
I mean .. interface?
ethernet is good from many aspects, but is a bit harder to use
it's UDP :)
so no error checking ..
but it might be ok
and paraport has err checking?
kinda hard to setup
but it also requires an IP stack on the target, which forces you into an "OS"
Jymmm: nah.. compared to TCP
UDP is easy to use. You have to build your own error check and retransmit.
You could use raw ethernet packets.
sure, but ethernet drivers and DHCP and stuff get a little more complex
lerman: in motion control it might be too late :)
SWPadnos: tftpd and bootp
SWPadnos since it's only one-way communications.... why not just stay at layer 2?
wait - wait - did someone say h-bridge?
sure, but both require at least an IP stack (with UDP or TCP on top of it)
SWPadnos layer 2, not layer 3
it can't be one-way
It's two way. You need to be able to read switches, positions, etc.
SWPadnos what is a driver gonna say back?
err - the motor moved this far?
a limit switch was hit
like a paraport one does?
a button was pressed
Although the pid loop would be local to the processor. You do want to be able to run a probe, don't you?
SWPadnos ah, were talking totally two different things, nm.
actually, I'd want the embedded stuff to be as dumb as possible
SWPadnos as dumb as alex_joni?
no, even dumber
ah, very cool!
But it should work as hard.
whoa.. that's a really dumb board :D
oh should I say dumb?
* Lerneaen_Hydra pops in
ok returning back to my one-cpu-per-axis reading
For me, a key requirement is that we could update and rebuild HAL and then download it to the chip.
* Lerneaen_Hydra gets hit by an extra strong insta-dumb ray
Classic ladder would be built in.
* Lerneaen_Hydra whoa dude
a ladder evaluator would be a good thing, I think
since it's supposed to be a black box anyway
The idea is to move all of the RT stuff into a processor suited for it.
the motion control stuff is probably better left to the PC processor though
I disagree, but that's what Steve Hardy basically did with the early G-Rex code
What does motion control mean here?
but why again... Computers are cheap - even doing things the stupid way I want to - they still seem fast enough.
well, there are levels, of course
but all trajectory planning should be on the PC, I think
and path error checking
Is TP in real time, now?
things like step generation should certainly be done on a dedicated chip of some sort
but it only needs to run at ~1 KHz to be effective
Then it should be on the chip.
maybe up to 10 KHz for HSM applications
that's something that takes a lot of FP calculation, so it's better suited to a PC with high processing throughput
The problem is that x86 is not designed for RT/minimal latency.
so you take the stuff that needs floating point on the PC, and leaev all the low-level stuff to the RT controller
OK. So, where is the break between TP and HAL? Does TP have to be RT? Could it be pre-computed?
unless you stick an X86 on the RT board (which kind of defeats the purpose), FP calcs will bog it down
can't precompute, due to feed override, etc.
Well PID is currently done with FP, but it probably doesn't have to be.
I think you'd need pretty big numbers to use fixed point
big = many bits
I don't think so. Remember that PID used to be done (maybe still is, sometimes -- e.g. gecko) in analog.
Q. how big is EMC2? just the controller part?
analog is FP ;)
none of the GUI part.
would there be a huge problem porting it to say a cheap uController (uC)? Say an atmel?
Well, it is, but not very high resolution. So it only takes (say) 16 bits. But the code gets trickier because you have to scale the values.
yes - scaling is a problem when you're dealing with stuff at a 0.001-second period, and with coefficients that may be 0.1
I would go with an atmel. There is FP software available, I believe -- but it would be slow that way.
A-L-P-H-A, not sure - I only have a 64-bit machine with sim ATM
lerman: the new geckos are fpga
no more analog
it had to be cut down significantly to fit on a Rabbit, which has 256k or 512k, I think
I was thinking the old ones. Don't they have pots to do adjustments?
there's still analog for current sense - it gets sent to the FPGA as a bit
lerman: the old ones are analog
new geckos also have nano fuses, which Mariss says it's near impossible to blow up a gecko.
there are no FOGA based servo drives yet
SWPadnos: it's in the works
at least according to some mail by mariss I read a while back :D
Damn, SWPadnos. I don't mind if you can't type; your corrections aren't necessary.
sometimes they bug me :)
Besides, they make me look bad since I don't correct my stuff.
Now you're with the program.
actually, I'm trying to make a lens motor controller as we speak, so the tools are getting in the way of my typing
two servo motors and one position servo
pan/tilt zoom ?
all I'm doing mow is makinga switch panel to use the integrated controller
or is that focus?
no, focus / zoom / iris
I have a lens with a broken aperture motor
it's a cool lens, but may not have good enough resolution for us to use
that would be great in CS-mount (other than the broken part :) )
like a lot of security cameras
or the Elphel 333 ;)
this is Canon EF
and it's rather large for a security camera :D
anyone with a clue about electronics have a comment on this circuit idea? It's for transforming emc's "pdm" output into something more like a high-resolution pwm signal. http://emergent.unpy.net/files/sandbox/pwm-idea.png
* alex_joni looks anyways
hmmm - what does the PDM output look like?
I think you need a feedback on the op-amp
and all for naught
otherwise it will go berzerk
the op-amp is supposed to act like a comparator
no it just turns into a comparator
feedback and maybe a resistor across the cap so the input voltage doesn't float around everywhere when disconnected
it could get nasty otherwise
is PDM from stepgen or freqgen (or both?)
jepler: without the feedback it saturates I think (either - or + max) ..
but take that with a pinch of mistrust
it's been long since I studied those :D
alex_joni: yes, I want the output to essentially be a logic signal
alex_joni: I think that's why the output is marked "pwm" :-)
it will saturate at voltages > logic level though (assuming it's a standard op amp)
the filtered input voltage is compared to the sawtooth voltage and gives a logic output
Lerneaen_Hydra: unless you connect it to 0..5V
jepler: I finally get the idea of it
there are some opamps that work that way too
ok, you're extending the PWM/PDM period from EMC, but still outputting at 40 KHz or so because of the comparator
So, you take an analog voltage and compare it with a sawtooth to generate a PWM signal.
oh, a 0/5V op amp?
yeah that would work
yeah - it's basically a frequency step-up for the final output
jepler: ask jmk :)
that's my only sane advice I can deliver right now
I think it'll work
SWPadnos: in this case it would probably be generated by pwmgen, or by freqgen if you're stuck on emc2.0
yeah .. me too
but what does that say anyways?
it lets emc control the PWM resolution as BASE_PERIOD/(500 uS or so)
but still outputs pulses at 40 KHz or so
that's still piss-poor resolution though, I think
a 20 uS BASE_PERIOD only gives you ~25 settings
that's why you use PDM
how many levels do you get with pdm?
you still get the same number of settings
depends what you call a viable level
26 actually, 0-25 periods on
no no no
you can be on for 999 periods and off for 1
if that's a "level" then you can have 1000 levels
sure, but with a 500 uS time constant on the RC filter, it'll "catch up" after 25 cycles
given an infinite period, you have infinite "settings"
ok I see what you mean, it depends on that time constant
you can tweak RC to give finer or faster control
with a finite period, you still have period/BASE_PERIOD levels
but at least the real PWM output is still at a fized 40 KHz or so rate
where does period come from?
the smoothing time constant
it's from the comment "rc ~ 500uS"
or you can have several paralell outputs and make a quick & dirty DAC with an opamp
hm, howcome you arent generating your period with a microcontroller?
just being curious :)
you still need to be able to get the command to the uC
I chose 500uS because that gives two time constants per servo period -- you can get pretty much to the new commanded output by then
that's the hard part
that's not a bad point -- you could clock out a 24-bit value to a uC in that same 500uS
you can do a quick&dirty DAC with a parport and some resistors
as long as you use DDR clocking ;)
er, no -- a 12-bit value
jepler: have you tuned your etch-0-sketch any more?
jepler: doesn't that all depend on how many outputs you're willing to sacrifice?
skunkworks: nope, I did finally fix the wiring on the one encoder, and ran the motors in the air for awhile, but I didn't actually run the etch-a-sketch
Lerneaen_Hydra, you can output to 7 DAC channels easily with a single parport write per cycle (one clock + 8 7 data bits, 1 for each controller)
So for my app - what would be the best? pdm, pwm, pwm+dithering
that 8 was spurious
hm, you can use a shared clockline..
and clock all dac's at the same time..
SWPadnos: what do you need the clock pn for?
SWPadnos: you are faster then I am
one sec - for a parallel DAC, you don't use the clock line
or use a synchronus protocol -- configure all the micros for a baud rate equal to the BASE_PERIOD
but you need N pins per DAC, where N is the DAC resolution
hm, or to the micro directly..
sure - that could work
that would probably be easier..
theres lots of micros with PWM functionality now :)
for a serial DAC (or a microcontroller acting as one), you need a clock or implied clocking using a synchronous approach as jepler mentioned
you could get nearly two 8-bit DACs on one parport
but I havent used any
you can get several if you use some simple external logic
latch and address decode chips
with a serial DAC you get into the too-long-between-updates-becuase-of-base-period things though
yes, even if you use the serial port for clocking ;)
an 8-bit wide sets data 8 times as fast
see pico-systems.com ;)
it all depends on how fast it needs to be though
* anonimasu hugs the usc
it's awfully neat :)
Jymmm is now known as Jymmmmmm
* Lerneaen_Hydra hugs his 72 steps/mm unipolar stepper from 1983
though I only really used it when doing driver work
1 master serial clock + 7 ouput serial streams + 5 input serial streams would make for a lot of interesting possibilities
* SWPadnos hugs his crimp tool
or whatever the parport would permit
SWPadnos: is that one of thoose 600$ things?
I think it is, but I got it on eBay
it's a Lucent crimper with interchangeable dies
you can just hang a bunch of micros off it, make them input-only, output-only or input-output, and have them each "do their thing"
I got an awesome Weidmuller ferrule crimper as well
SWPadnos: the crimpers I think off have 8 pins *
like that :)
those are only 6
* alex_joni counted twice
anonimasu: * has only 6 not 8
the crimpers does..
some have 5
I think the ones I saw in elfa has 8
either way insanely overpriced
* anonimasu loves mil connectors
* alex_joni loves harting
I spent more on MIL connectors for my motors than I did for my motors
* Lerneaen_Hydra loves free connectors
I'd hate to buy them but they are really nice
ooooh - that's an expensive crimper
nothing like a 20x10cm connector
[19:54:24] <anonimasu> http://www.elfa.se/elfa-bin/dyndok.pl?dok=4953.htm
[19:54:34] <SWPadnos> http://www.luconda.com/artikeldetails/21/04/03/bilder/4120674-1-Weidmueller-PZ-3.jpg
340eur :) only
not quite the one I have
I'd love to have mil connectors on my box to the mill but I cant afford them
yeah - I should have thought of that as well
but I needed them forthe motor end anyway
elfa is always overpriced though
* Lerneaen_Hydra also loves surplus military grade hardware
I dont think it matters with the MIL connectors.. :)
they probably are as expensive wherever you buy them
actually it does - I got a few for my smaller motors at CNC workshop for $10 or so (total, with back shells)
for several connectors, not $10 each
50 eur for 5pole ones
oh that's just the back
jepler, I think the overall precision isn't affected by the circuit you made, but the PWM period is decoupled from the BASE_PERIOD, which is a good thing
SWPadnos: that seems true to me to
I'll be back in a min :)
what could also be interesting would be a set of 8 or 12 -bit R-2R ladders, with a driver that clocks out parts of the setpoint every cycle
basically, a serial DAC
with a full parport, you could do 12 bits on 4 axes (using a simple scheme) in 8 periods
but still hooked to the "PWM restorer"?
I ws thinking of using "multi-bit serial"
so clock out 8 bits (the data port) per cycle, with the address chosen by the control port, which also has the clock
external latches / address decoding would of course be needed
which means you might as well use a microcontroller
some of these are pretty funny. http://uncyclopedia.org/wiki/You_have_two_cows/26
oh, if it isn't brainfuck :D
There exist x and y such that x falls under the concept is your cow and y falls under the concept is your cow and x is not equal to y; and it is not the case that there exist x, y and z such that x falls under the concept is your cow and y falls under the concept is your cow and z falls under the concept is your cow and x is not equal to y and x is not equal to z and y is not equal to z
* Lerneaen_Hydra pokes cradek with the RPN reference ;)
I don't see it, but I can imageine "Cow" DUP
You cows 2 × have.
more yodaspeak than RPN
I like "Cow" DUP better
Intel Pentium 60 - A80501-60: You have 2.0000000056987983 cows.
You have two cows. One of them is an anticow. They touch. Now you have no cows. Or planet.
Potscript cows: (You) p (have) p (2) p (cows.) p showpage
In Soviet Russia, two cows have YOU!!!
You have one cow and one anti-cow. They collide. You now have no cows and a very large crater.
er, I guess "p" isn't standard; it's "show"
wasn't there a share ware site called 2cows at one point?
72 72 moveto (You have 2 cows.) show showpage
[20:12:20] <alex_joni> http://www.tucows.com/
wow - still around
32-Bit Wrong Endianness
You have 536870912 cows.
bad C programmer cows: a.out: cows.c:3: main: Assertion `cows == 2' failed.
[20:15:12] <alex_joni> http://www.bigzaphod.org/cow/
MODE #farm +c 2
yes I acually just did that.
You have two genetically engineered cows, given to you freely. In fact, the cows' milk cannot be used for any proprietary purpose, but feel free to sell it as long as the buyer can do whatever he or she wants with it. No, wait. We can't add a nice strawberry taste to the milk because then proprietary farmers might use it, not to mention that strawberry milk would taste terrible. In fact, we can't add any new gene features to the cows, because
a monasato cow?
hmm, Oh, right
GPL: You have two cows with evil tentacles on a deserted island but you're OK anyway
how do cows manage to have evil tentacles O.o
tentacles sure, but *evil* tentacles
Differential Calculus: You have 2.
the simple ones are the best
You have two cows. They are beautiful and work perfectly, but no one will buy them, probably because their milk isn't compatible with the majority of cereal. They will buy your chicken, however, which helps your stock.
Vous avez deux vaches qui sont cools.
quantum cows - If the 2 cows where created at the same time - the spin on each will be the same even over long distances.
You have no bulls
holy crap. I didn't realize they have *27 pages* of this stuff O_O
err - holy cow! ;)
2 holy cows
[20:28:18] <alex_joni> http://www.theregister.co.uk/2006/10/29/microsoft_vista_eula_analysis/
<- that's an interesting lecture .. if you care
<snip>4. USE WITH VIRTUALIZATION TECHNOLOGIES. You may not use the software installed on the licensed device within a virtual (or otherwise emulated) hardware system.</snip>
I'll throw up a little bit in my mouth when I buy my next laptop with vista preinstalled, but besides that I don't see how any of that bullshit affects me
or, in the best case, I'll look a little harder for al laptop without an OS preinstalled
if you're lucky to find one :)
you mean other than having to pay for that crap?
but I've seen more and more laptops with freedos around here :D
and pc's with ubuntu preinstalled
alex_joni: yeah, but you're practically from a foreign planet.
cradek: I'll buy without an OS if I can get the machine I want without paying a premium
this is extreme "The first user of the software may reassign the license to another device one time. If you reassign the license, that other device becomes the "licensed device.""
that's it.. no more hardware upgrades without buying new OS'es :D
wonder if it'll be enough to get people thinking about different solutions?
alex_joni: nah -- the geeks who have enough money to buy a new machine every 6 months, and $600 video cards in pairs, will just go ask for a raise so they can afford $400 in vista for each new machine.
and note the key wording, one time
* dastumster is gradually taking my organization off windows
dastumster: good for you
need some help with a couple of questions
Nick001: ask 'em!
I finally compiled EMC2 in my home directory and starting up it says "can't locate axis"
Nick001: dispite the fact that you heard all kinds of crap in the last hours this is the place to ask about emc :)
Nick001: how did you get emc2?
I don't have a directory either in USR, local/bin/access which should be in my EMC folder in the home directory
Nick001: emc 2.0.x, or the CVS version?
first I got it with the ubumtu package and then I got the EMC2-2.0.3 package and compiled that
after I got all the resources, of course
and I had to learn how to compile after I got the compiler to work
for emc 2.0.x, axis is a separate download. If you're on an ubuntu system, you can get the source with "apt-get source emc2-axis" and the build requirements with "apt-get build-dep emc2-axis"
will it direct it to the proper EMC2 directory in my home dir
it is my understanding that the one with the package is a demo that I can't access
I apparently do not have permission even as a root
The only thing about the packaged version is that you can't modify files in the /etc/emc2/sample-configs directory.
is that why I have to compile a new copy into my home directory
let me find some documetation pages that might clear that up...
Nick001: you don't have to compile
If you're not sure how to copy configuration files so you can modify them, read this: http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?CustomizingConfigsOnUbuntu
jepler: wasnt being bragging or boasting, its hard to do, but linux has evolved now into an OS that is in actuality simpler for a user imho
Nick001: unless you plan on modifying the software (EMC)
If you're having trouble getting your device connected to the parallel port to work, read this, particularly sections 2 and 3: http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?TroubleShooting
dastumster: no, I wasn't mocking you. I think it's a good idea, and I applaud those who can and do do it in their organization
dastumster: I can see/understand that ... I still feel OpenOffice has some more work to do before that's fully an option :)
the files I would be modifying would be to hook up motion controller cards and stuff like that to my machine
the machine I am trying to retrofit is a Hardinge HCNC with an AB control
Nick001: yes, after following the directions shown on "CustomizingConfigsOnUbuntu", you'll be able to use a text editor to modify .ini and .hal files.
has someone done this before so I don't have to reinvent the wheel
what's an AB control?
Allen Bradley 7300 CNC control
and to what level are you trying to retrofit it?
keep motors? keep drives?
keep parts of the control?
steppers? servos? etc..
getting rid of the AB control and driving the slides, spindle, and turret from EMC2 on a PC
Hi Nick001 I've got several of these lathes laying around.
I have stuff hooked up to an Ajax control system which, of course, does not work
what condition are these machines in
I think we talked.
Have you got the EMC running motion?
I'm still trying to get EMC fully loaded onto the computer
Okay. What's happening with it?
need to install axis first, apparently, and learn more about the system
What version of Ubuntu are you using?
6.06 I believe
Nick001: axis is included on the livecd
Okay. Did you load it from the EMC-live download or from a real ubuntu disk?
got the ubuntu disk that I burned
You got it from the ubuntu site?
I have some info on APT-get source which I need to try
Are you at the computer with it now?
no, from the linux CNC site-no I am at a different computer right now so I can use IRC
okay. can you work both computers so I can ask you to run a couple of things on the ubuntu?
need to walk 30 feet between them. can I get IRC on the linux box? I finally have a modem on there
yes but I can get a dsl online here pretty quick-10 minutes
NP I'm dialup also.
I use xchat for IRC.
On both 6.06 and 5.10
jmk_away is now known as jmkasunich
is it already installed with the ubuntu package?
Nick001, xchat, should be.
Nick001, internet -> xchat
or in a shell, type "xchat"
No I added it by connecting the box to the web and click system->Administration->Synaptic Package Manager
or in a shell, "sudo apt-get install xchat" [quicker]
rayh, I'm pretty sure xchat is installed by default.
A-L-P-H-A: I don't see it on my dapper
okay, so I'm out voted.
I personally perfer xchat-gnome
That works fine also.
lately it's been gaim
one less app to have open.
gaim is installed by default I think
anyone seen dmess lately?
I have not been able to figure gaim out.
Yes it is.
so is ROBOTS!
apparently I need to use the package manager because it's not listed in applications window
damn... it's such a stupid game, but I played it for hours.
what is not listed? emc or gaim (or xchat ...)
I'll go do that and I'll be back as soon as I can
SWPadnos: bet he meant xchat is not installed (aka doesn't appear on the application menu)
gaim is easy... like your sister... just add a new connection, specify it's an irc, and point it to irc.freenode.org, and then add a chat room, and specify irc.freenode.org as the network, and the room #emc
since firefox is there, and chatzilla is an extension for it, that may be the easiest thing to recommend to people
seems to me xchat is on 5.0whatever - and gaim is on 6.06
no root privileges needed
so how does one get chatzilla
go to firefox tools->extensions, select "get more extensions", search for chatzilla, and click "install now"
not significantly different from the synaptic approach, but doesn't need to be root
And setup to get this
ah - about the same. when you run chatzilla, the default message pops up with a link to freenode, you can click that and connect
you do need to know how to /join though
you'll confuse him for good :)
well, you have to /join in xchat as well, don't you?
I suppose the web client from linuxcnc.org is as good as any also - certainly esier to install ;)
yeah, seems like a lot of work to get onto irc :D
perhaps the default package list for the emc2-liveCD can be tweaked to include gaim or some such?
gaim is there by default
preferably not gaim, since it's unusable, but still ...
right - see above ;)
err - above this, below the original request
xchat allows for auto connect and join to a list
chatzilla does that as well, but you need to do it manually the first time (just like xchat)
you get tabs for the different channels, and just right-click the tab to choose to open at startup or not (and logging, etc)
oh wait - logging isn't on the tab
test please ignore
Took me a bit to discover that you need to click on the freenode link on the bottom of the welcome page.
ah - ok
you can also /attach chat.freenode.net
or click on an irc:// URL on a web page
I see there is a menu IRC->Open this channel at startup
yep, or right-click the tab for this channel
one thing I couldn't stand about gaim was the lack of tab-completion (at least by default - I don't know if it's available)
SWPadnos: seems to work now - when I use it on dapper.
okay so how do I include a password in chatzilla?
to automatically identify to nick services?
go to chatzilla -> preferences
click on the freenode.net item in the tree
check "enable identification server during connection process"
err - hold on - that's something different
ah - click on the "lists" tab
click "add" next to the auto-perform list
type in "/ns identify yournick yourpassword"
will try it.
see you le
rayh, is this chatzilla or $other
but the identify didn't work.
will try again.
I do that manually anyway
problems like the one they had before, with someone spoofing nickserv, made me not bother to set up automatic ident
* alex_joni had no idea he's so close to Isengard
[21:40:27] <alex_joni> http://uncyclopedia.org/wiki/Image:Europe_middle_earth.JPG
ah ha - it seems to have worked
your cloak is on
Yes it did.
Jepler: This problem is caused by migration to new version of
mailing list archiver. It should be fixed completely in 2
weeks. Now we uploaded emc-*'s messages in archive.
skunkworks: so the squeaky wheel does get the grease
sounds like it ;)
man I hate sourceforge
jepler: sounds like it gets a new wheel
probably the axle will still squeak
SWPadnos: only difference is /ns identify <your password here>
oops - right
/ns or /msg NIckServ ...
/ns is not accepted by all clients
it is by chatzilla, which is the configuration I was desctibing :)
how does one use the G81? say I want to drill a hole at 0.75,0.75... would that be G81 X0.75 Y0.75 R?? Z1
where Z1 = 1" deep.
what's the R retract for?
and what kind R is needed?
like how deep before we rapid retract?
[21:48:26] <A-L-P-H-A> http://www.linuxcnc.org/handbook/RS274NGC_3/RS274NGC_33a.html#1003380
I am... but there's no nice examples. :P :)
[21:48:48] <alex_joni> http://www.linuxcnc.org/docs/html/gcode/mill_canned/index.html
ddin't notice this entry. 184.108.40.206 G81 Cycle
I wrote up several examples of these canned cycles in the old user handbook.
A-L-P-H-A: check the link I sent you, it's from the EMC2_User_Manual
rayh: that's what I directed him to
I don't think they made it into the new.
[21:50:51] <alex_joni> http://www.linuxcnc.org/docs/html/gcode/mill_canned/index.html
<- not those?
I'm back on the linex box with dsl
Canned Cycles G81 through G89 have been implemented for milling. This section describes how each cycle has been implemented. In addition G80 and G98/G99 are considered here because their primary use is related to canned cycles....
Look under the Applications menu and see if there is a CNC category.
I put in apt-get source emc2-axis and it worked o.k. but when I put in build-dep it stopped and asked me if I was root
so how do I tell it I'm root?
in the console
yes, EMC2 and EMC2 user manual which is what the package put in
Nick001: you don't need the sources unless you want to modify the code
and I somehow doubt you want to do that
ahh... G83 is what I was looking for... but I'm only drilling wood, so it doesn't really matter.
G81 would be fine
Nick001: if you're not very advanced with these things, it can lead to a lot of confusion
I will need to modify the hal and ini for different parellel port devices to drive my machine
that is _very_ common
everybody does that
but you don't modify the default sample configurations that are shipped with emc2
there is only one advance with these things and that's DO IT
instead you make a local copy (in your home folder), and edit that
yea, what alex_joni said.
I know that is why I compiled EMC2 in my home directory and I'm trying to load access into it so it will open up
Nick001: if you want to get it to work rather fast, I suggest you delete the emc2 stuff you downloaded from yout home directory
and then do what?
then follow this: http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?CustomizingConfigsOnUbuntu
o.k. I'll do that and I will be back
that page will instruct you how to copy a set of configs so you can edit them lateron
Nick001: I won't be around by then (going to bed now), but others here surely will help
When you get a chance Nick001, start emc from the menu and click on the sim -- axis config.
that way you can verify that the axis interface is working properly.
good night all
rayh: I trust you'll get Nick001 to see the light :D
oh EMC2 doesn't default to axis does it?
I forgot about that.
depends on what config you use
I usually change that right away
if i move stg config to my home dir all i get is the piguin screen and then nothing - if i use sim then the program comes up
Jymmmmmm is now known as Jymmm
Do you have an stg card installed?
Hi Nick001 Do you have an stg card in the PC you are using with ubuntu?
I considering running this program for my cnc router, how much better is it than mach 2?
I don't know much about mach but EMC can certainly handle a router.
I've been running the demo of mach 2, but it only lets me do 1000 lines of code
BrMiHi08, uh. EMC doesn't have that many macro functions that mach does... but does handle everything else better.
and it's FREE.
A-L-P-H-A - I have bobcad cam so I can make all my gcode in it so I really do not need the macro functions
BrMiHi08, then welcome to the light side. :)
heh.... I've only been running linux for 6 months or so
BrMiHi08: I am a linux newbie also. I love emc2.
With ubuntu and EMC2, you hardly need to know anything about computers.
That should be a good enough intro to the os you run with emc
as long as you know how to right a compiler from scratch via assembler, and about to count to 32 in binary with your fingers, you'll be okay.
BrMiHi08: what kind of motors and drives are you using with the router?
I got a k2 cnc router
jymmm also has a K2 router.
isn't that what jymmmm has
mine is only 25x14 I think
does that have the xyrex or what ever drives on it?
have you checked the squareness, straightness, and level of the ways?
Jymmm. what where your problems... maybe you can help BrMiHi08 to check his...
mine has nema 23's
with the K2 you had.
I thought they fixed them.
share... maybe you should make a permanent pastebin, to direct people to it.
BrMiHi08 that's the problem?
I don't think he has any, or knows of any.
I am going to give EMC a try on it
what am I missing here (blames A-L-P-H-A)
I am the blame machine... I take the blame... and redirect it at americants.
blame the americants...
thanks for the warning
Jymm- was it had to configure yours?
bare, I installed Xylotex steppers and driver on it (wish I had bought Geckos instead )
ooooh....are the stock nema's not very good?
I gotta get ready for work, if you'll be around after 1900 PST (-0800) I can chat more then
Yes, I have horror storys (plural)
A-L-P-H-A goony goo goo
Jymmm the americant. :)
go make money on the street corner... wear the nice pink tube top... makes you look hawt!
BrMiHi08: emc2 is pretty easy to set up for step and direcion drives. This is the place to ask questions when you get stumped.
CANadians versus AmerCANTS. :)
I am taking a class on c++ at school right now.... but i might be a bit dumb when it comes to configuring it
and they are constantly improving it - in the year that I have been using it - they have added tons of goodies.
c++... the choice of language where people are sadists.
CANDadians vs AmeriCANS
I am downloading the live CD
I have a computer with debian on it right now.... but I think I'll put the live cd on it
Most all of the configuration needed is plain english but I'd read through the user manual as you begin.
EMC-Live is a real good system to stat with.
take a look at axis as the gui. pretty damn cool http://www.electronicsam.com/images/KandT/Dapper.png
Lets put things into perspective... Lets compare a 3rd world country where wages are low, but still have a decent level of living. As opposed to midwest, where people think they deserve jobs, and this and that. Yet they won't offer a competing service, just complain. Why do you deserve a $25/hr job putting nuts on a machine? How skilled is that?
Why not pay the mexican, say $2/hr or whatever... that can do the same thing.
hey now.... I live in the midwest
you should be paid what your skill is worth. Market economy will dictate the price. That's what a free economy does.
BrMiHi08: where abouts?
thats still in the midwest? :)
I am in WI
farther north than me
if I offer the job for $1/hr... no one will take it. If I offer the job for $100/hr everyone and their dog would be lining up. Somewhere in the in between is where the value of the skill and dollar amount can be meet. Americants just don't want to accept that.
if u offer the job, at 1 dollar a day, to raise exponetially, I will take it
Hence where there are migrant farmers, and illegals able and willing to work for those low wages...
1^1.000000001 daily, is still exponential increse. does that count?