03fjungclaus * 10emc2/src/po/de_rs274_err.po: Added to cvs ...
* robin_sz gets slowly drunk
03Zathras 07BDI build system * 10Babylon Cluster/comps.xml: File changed. New revision:picax.xml
wb is now known as fenn
West coast ?
discussion of CAD/CAM packages for linux came up again on the gingery_machines list
There are a few around now
everyone seems to think they exist but nobody knows any names
Qcad, Synergy, etc...
There's a bunch of links on www.linuxcnc.org
heheh i meant one that didn't cost thousands of dollars
i will check out those links
* paul_c finds a new(ish) toy has duff memory...
Now I know the west coast is stirring.
Middle of the afternoon...
at least it is light outside
sun's going down.
yup... we are at Z-8
I don't have jury duty...so am off to the treadmill...be back in an hour or so.
at least we have the winter solstice to look forward to.
* fenn wonders what duff memory is
faulty, knackered, US.
the thing that bugs me about all these cad programs is they are based on manual drafting, whereas what I want will go straight to a toolpath generator
what's the point of a 2d cad package when you've got a 6D mill?
you mean a six axis mill surely...
same thing if you think about it
er wellnot exactly
b/c a bridgeport has a knee and quill in the same dimension but diff axis
One would be Z, the other, W or Z'
03Zathras 07BDI build system * 10Babylon Cluster/RELEASE-NOTES: File changed. New revision:1.32
Begining to think you'd been hit by a major power cut.
Hi Paul. Got your note re the dl.
Got called away to repair a sawmill this morning early.
Much to much money in service to refuse.
well.. 4.02 is (slowly) installing on a 433MHz board with a paltry 64Meg of memory.
Wish that I'd borrowed a few mil 2 years ago and put it into euros.
Sounds like a good compromise with the proc speed and memory requirements. The 512 was just to big.
We'll see how it runs.... Might want to install icewm for speed.
Is ice on the disk?
nope - That is on the "Look at, and maybe Do" list.
It would work well in dedicated environments.
For embedded systems, fluxbox might be a better choice.
Not seen that?
[18:02:21] <paul_c> http://fluxbox.sourceforge.net/
- But their web site appears to be having problems.
I'm looking at 150,000+ blocks of gcode for my turkey array
god knows how big that file will be
[18:04:22] <paul_c> http://www.linux.org/lessons/short/fluxbox/
has some screenshots
so you were saying you did a script to automatically run multiple programs in sequence?
The final test for the Sherline was 4+meg and it ran without a flaw.
Yes. What version of emc are you running.
well at about 2k bytes per page....
I may need to do a bit of tuning cause I think something got broke.
thats 247 meg!!!
june 2004 emc build
The big issue with that sized file would be load time. Until the EMC intrpreter gets finished with read ahead, the tickle graphical interfaces don't respond.
right..i know that well
les. I believe that I'd burn that on a cd and try running it. You might want to use xemc rather than tkemc.
let me double check my numbers here...
Oh Pooh.... anaconda just crashed on me..
Ran out of memory and died.
les: if you could get it to run through, it would be something to crow about on the lists.
RAM or HD?
Darn. I was wishing...
still have text mode.
150K blocks*(20 bytes agv/block) ~ 3M
You mean the great text guru was running a graphical install? Say it was only a test...
Hey... Gotta test these things before the end user breaks them.
Hi dave. Didn't even look at the list here. How you doing today.
I knew he wouldn't admit to graphical anything.
yes my assumptions were wrong about file size...it will be only a meg or so
never mind haha
yesterday after i got kicked off the jury I came back and fired up vsitest...and left it on the encoder for 4 hours without a single pulse
no I need to try the same thing with the servo motor running...
checking for that drift you had?
just looking for emi
So the encoder -> board is good.
oh ok great
I really think that the problem is in the loop
lots more to do...
And the motor is rock solid on a battery box or a jumper.
on the battery box
I need to rebuild the driver...it won't read the encoder anymore
rayh: Did we get an ftp account in SD to use ?
How rigid can you get the motor shaft with a jumper across input?
We did get an ftp site there.
Probably be tomorrow morning/afternoon before i get it in there for them to try.
So tach is correct direction. What about tach gain?
don't really know...I probably should play with that
If I started an upload from here, it would be done by abot 06:00GMT
(going for Tea)
So we are not getting it at my guy's place?
Or will both run?
les ... I got a couple of mt30's from Doug....he did up the price...$220
no suprise....I need to change the web page then
still a bargain
ray...I have x and y running on the Mazak...tuned to +/- 0.0004 with just a couple of tries
yes it is
have not talked to him...how many does he have left?
Fantastic. Time to start making chips.
les...he is down to about 6 and is trying to save some for himself
oh....better remove them from the forsale page
we had 105 to start
ray...I really need the Z to start making chips. ...have 10 pin idc connectors ordered from Allied to do the encoders off the stg breakout board
You are getting close then.
well..I have to give up the board for repair or do the 4th axis workaround
also have to fuss with coolant
With that 8 axis stg you could shift the base address in software and use the last four channels.
so how much do I have to shift
guess i could try it and see what channels work
Not a clue. We need Paul or Les for that.
I'm headed for the shop .....will checkin later
It might barf at the high end.
* rayh may go away while trying fluxbox.
* paul_c is back for a few mins..
is STIX/STEP-NC/AP238 worth worrying about?
fenn: I've never heard of either of those, but after I searched for step-nc on google, I got a webpage that says I had better upgrade to netscape 4 to enjoy their site
it's hard to take a website like that very seriously
StepNC is touted as the next big thing in CNC
the www.step-nc.org web page looks pretty stale, and the features they describe are irrelevant to me as a hobbyist
i'm currently worrying about writing some sort of NC output from Ayam and am considering my options
since there is a lack of toolpath generators out there
Gcode is rather old and clunky
agreed... Nor is it particularly "standard" across the board.
since my interest is more in terms of hexapods (dont roll yer eyes) i'm looking for something that specifies surface normals. that way i can use a cylindrical endmilll and create a smooth (convex) surface without too many finish passes
i got the idea that gcode has problems with arcs in more than 2 dimensions
Arcs plus rotational movement seem to be ill-defined in the rs274ngc document I've read
There is no simple way to define an arc in 3D
Although... There are codes to incline a plane and use that as a reference.
you mean, in rs274 there is no simple way?
but not in EMC (yet)
I wouldn't mind getting axis 4 and 5 on my machine
that's why I made the z so big
not enough io on the stg though (i think)
hmmm.. G75 is listed as a multiquadrat circular interpolation move...
* paul_c is heading out for the evening...
back in four hours.
* paul_c is away: Out on the tiles.
* fenn wishes he knew what multiquadrant meant
According to the web, some rs274 interpreters think a G2/G3 arc can never be in more than one quadrant
(eg the quadrant where x, y are both greater than the center of the arc)
g75 would then be the default emc behavior, with arcs specified with I- J- to be up to 360 degrees with no regard to quadrants
[19:40:12] <jepler> http://www.advancedmsinc.com/glossary/glossary_G.htm
"Another important case ..."
* fenn blinks
so if you do g2 0 0 0 0 it will do a point but g75 0 0 0 0 will make a circle?
g0x0y0 / g2x0y0i.5j0 will make a circle
but if emc had g74/g75 (it doesn't), then g0x0y0 / g2x0y0i.5j0 would do something else, possibly drawing only a point
er, g74 / g0x0y0 / g2...
sorry if i'm a pain but i've never had a chance to actually use gcode
my friend was cutting a piece that had rounded edges and the CAM software would only generate discrete steps along the rounded edge
that's pretty typical
the result was that the machine had to start and stop fifty times each pass over each edge
that's a deficiency in the motion planner
emc smooths out the segments so the machine doesn't have to stop at each one
that was turboCNC i think
depending on your setup, it can run pretty much full speed over those kinds of programs
I don't know anything about turbocnc, but if it stops at the end of every segment no matter what, it sucks.
yeah but i cant convince him to use EMC because he "likes dos"
he does have a point though - it starts up in three seconds
well, there's nothing wrong with using dos for some things
uh, if you spend an extra two hours machining your part, who cares about startup time?
it took 24 hours to make the part :)
in all seriousness, emc gets this right and you would have much better luck with it
if the segments make a smooth path (no axis is reversing direction) emc will run right over them at nearly full speed
so emc will stop at the top of a bump?
or only the z axis will stop?
the Z has to stop and reverse direction, so its acceleration maximum comes into play. The other axes will do what's necessary to remain on-path
if your acceleration maximum is high, you won't see any discontinuity.
but someone recently had trouble with a program that reversed the Z axis every few thousandths of movement in X or Y
as you might guess, that can't be run through quickly.
depends how big your machine is
sure. I'm no expert on that.
I don't know who here is an expert on emc's motion planner.
all I can report is my experience using it
okay new topic
if you have a part geometry (not toolpath) and no time constraints, can you just trace over the part surface and steadily decrease the tool offset?
instead of defining each pocket separately
you will get an error if you try to fit a large tool into a small hole
it won't "ride over the top" like you want
it's not nearly that simple
pocketing routines can be pretty complicated
at the most basic level the toolpath generator has to know the shape of the tool
emc can't know this
i'm trying to figure out what exactly tool offset does
lets you use the same program after you sharpen your tool
even if i.e. you grind on the side flutes of an endmill?
I think that's a main reason for it
I don't use it since I don't handwrite any gcode.
even a ball end endmil?
maybe someone who uses it will speak up.
how do you generate gcode?
in general, I program in another language and have that generate it
in my opinion, gcode isn't a reasonable language to program in directly.
yeah it all looks like mush
I am having to do a bunch of manual cut and paste on gcode now
well, looks aside, there are no subroutines, loops, etc.
so what do you use, C? python?
A lot of the gcode I generate is from the Eagle circuit-board design program
it has its own "user language"
a text editor...for gcode
I would also recommend Python for that kind of task
.. as long as you are comfortable writing Python code to read whatever your input file is (such as dxf)
I just downloaded Autoeditnc from the sherline site...got tired of using word
heh word to edit NC files..
anyone know of a better freeware nc editor?
yeah...I use that in the office
fenn: I've also used LISP
I have one nc file with 4 tool changes that makes one part
fenn: use whatever sane language you're comfortable with
you're just outputting some strings, so who cares what language you use
I have to now make an array of 54 parts
i'm just wondering what languages i will have to learn
and have to do that manually ugh
I've also used m4 to get looping, but you might not want to use that as a first choice.
so you were going to paste part code, change coordinate system, paste part code.. etc 54 times?
les: use g54 / g10 l2 p1 x... y... and loop
how large is the program for one part?
I am using all fixture offsets to make a row of 9
then variables and another row of 9
why not use a language to generate the offsets for all 54 parts (one tool change)...
the change tools and duplicate that code, etc
well I could write a little code that does that sort of thing
it really isn't too bad and is often reusable
I'd write a lot of code to keep from changing tools 212 times
I have to lump together all 54 with tool 1, then all 54 with tool 2, etc
doing that now
wish this had subroutines haha
keep to tool change down to every 20 min or so?
dave-e: thelonger between changes the better...
les: want my m4 looping code?
I just tried to compile emc...not rcslib and it crashed as usual...on extbridgeportio
sure...but I read m4 as spindle ccw....??
the rcslib/etc has not been changed since the last build
dave-e: what error?
oh, the usual....it looks at -loutb.c ...and says it is not a directory...which is true
you mean -Ioutb.c?
the rcslib/etc has the tck...sh files in it...which usually cure the problem
dave-e: mkdir src/emcio/AAA
dave-e: then try again
well enough of this text editor stufff....I need to go out and actually run one of the custom bits I designed and see how fast I can feed it
usually run out of power at about 180 ipm
cradek...seems like voodoo but it works..clean compile but still no reading of the encoder..
dave-e: it is voodoo
trying encoder on vsitest shows counts
dave-e: there's a makefile bug somewhere but I couldn't find it when I ran into that problem a while back
dave-e: so I figured out how to work around it
dave-e: sorry, don't know anything about encoders
well I think I do but that doesn't help
so what does m4 do for you that C won't
smaller gcode file :P
I'm trying to find that out...on the gnu M4 page now
dave-e: it's a preprocessor, not a programming language
dave-e: they're really apples and oranges. you don't have to compile m4, for instance
I'm not sure I recommend using it, actually, but if you want a flexible preprocessor, you already have one on any unix
les: if you just want looping, I figured it out already and you're welcome to a copy of my file.
looping of a gcode file?
yeah, looping and calling "subroutines" in gcode with m4
yes thanks I would like to see that. Was M4 on rh6 distros?
-> rpm -qf /usr/bin/m4
yes, rh6.2 had m4
(any system that has sendmail has m4)
aw stupid xp wouldn't allow the transfer
what is the file name?
yes email would be quickest
if it's not obvious, run it with `m4 gcode-loop.m4 > outputfile.ngc'
cradek...I'd appreciate a copy of that file also.. email@example.com
having a look at your code
*cough* perhaps that could end up on a bdi distro since it appears to be useful
what is that gcode in the M4 program?
the define `program' makes a circle at 0,0 after setting the coordinate system offset to ii,jj
the part between the percents is the actual g-code that calls the "subroutine" `program' in a double loop to make a grid of them
oh...so I put my gcode to be duplicated between the define and )dnl?
and ii,jj are "passed in" to the define
ii,jj are integers so divide or multiply accordingly to get the coord offsets where you want them
and I guess preamble and ending are added to the whole thing after?
ok got it
just look at the output, it's obvious what's going on with those comments there
of course you can make many defines, name them whatever you want (`program' was a stupid name) and call them however you want, in loops or not
hmm g10 rather than g5x offsets...
it is using g54
see the beginning
yes...to set default
I was thinking of writing a little c thing to do this....but this looks fine
about 10 years ago I wrote a little c program to do arrays of micro optics with gcode
machine ran for about a week
oh...where do I set the array dimensions...in forloop(?
forloop(`ii', 0, 3, `forloop(`jj', 0, 3, program)')dnl
what is the 0 and 3? one must be the array size
start and end
ok got it thanks