hi Ray, how are things?
Good. Just started using 8.04 and like it quite a bit.
yeah, it looks quite nice
A fellow asked me about the number of tools that can be defined in EMC2. Is there a limit?
hmmm - good question
if anything, it's probably a number format limitation, like 127/255 ...
I thought that some of the original interpreter limited to 32
I don't know what it currently is
hmmm. is there some define MAX_TOOL or something?
I'm pretty sure it's between 32 and 255
Okay. I suppose the code for the interp would tell me.
if it's too small, charge him a couple hundred bucks, change the number for him, and recompile :-P
hi ray. you always know where to come for the helpful suggestions.
rayh: here's what I've been playing with. http://timeguy.com/cradek-files/emc/concave.png
I'm not promising anything though. It's pretty experimental.
There you go. That'll be 6K to do that.
Hi Jon, Steven, Jeff, Chris
#define CANON_TOOL_MAX 56// max size of carousel handled
heh, 56, a nice round number
err - yeah. it's - um - 2^6 -2^3
and yes, it did take me all that time to find it
is it actually used?
oh. bummer :)
it's that low because the whole tool table has to fit in an nml buffer ... or something
oh. that would be a real problem
I think we enlarged that buffer when we did some 64 bit stuff, so it might be fine to increase it again
yep ToolSts or some wuch
56 okay. I will join the rest of you in not asking why.
I had to lower it when I did lathe tools (since the entries got bigger)
rayh: I'll undercut you and do it for $2k
$1999 on sale today only
I remember a memory addition for a mazak -- 11K
I'm surprised nobody has noticed this limit, actually
how many feet of memory was that?
rayh: are you going to NAMES?
(my BP has 100 feet of memory!)
100 feet would be a lot of memory these days
even if you lay the DIMMs end to end ;)
I don't remember but remember that it was not a lot of megs by pc standards
jmkasunich: No. You?
hmmm. I don't think I've seen a computer that you could install 100' of memory into these days
yeah, just for one day - as a normal spectator
I had a Ustore on a ge550 that was something like 25 feet.
The first 5 feet were bad so I had to keep a program in there that I never used.
I wonder if someone ought to ask the NAMES folk if they are interested in supporting cnc on the floor.
I saw a little cnc when I was there in [mumble]year
a sort of open-project thing like we did for a few years.
two systems running emc, one was sherline
Right they do it there every year.
cradek: Did you go to NAMES when it was in taylor?
it was detroit I think?
somewhere up around there
maybe it was '03?
I don't remember meeting you there.
I did not go to the emc stuff afterward.
cradek didn't get sucked in until later ;-)
I was pretty new around emc at that time
I recall noticing that sherline was running their 4 axis mill by unplugging Y and driving A with it. I asked if that was because the rotary support in emc was so buggy, and mike (was his name mike?) said yes
I recall it gave crazy wrong speeds - I had a 4th axis at the time too
we still recommend inverse time for mixed angular and linear - thats simply not an easy thing to do right
jmkasunich: on top of that (hard for the programmer), it used to not work right
I've finally managed to figure out that my new server at work has 25,000 miles of RAM
that discussion started me thinking about the "bandwidth does a shipping container full of media of a freighter" discussion we had a while back
Always some &^%$ with one that's bigger than mine!:)
"never underestimate the bandwidth of a station wagon full of tapes"
shipping container full of 8G SD cards = 280,000 terabytes
SD are small. they still amaze me.
with or without the big blister packs?
take a look at micro SD
I have a 2G one, I think they may be available as high as 16GB
those are the tiny ones they stick in phones
about the size of a pinky fingernail, but around 1.5x as thick
hmmm. maybe a ring finger nail
the freight container, on a 20 knot freighter, going anywhere in the world (accessible by sea of course) has a bandwidth of 124GB/sec
I can't believe you've calculated all that...
latency suck tho
faster than the 6bone
there also the minor matter of getting the data in and out at both ends of the link
cradek: I've spent more time calculateing dumber things
That concave looks pretty impressive.
yeah, it's a great thing IMO
it will deprive a whole generation of EMC users of seeing the great "convex corner" message
well, maybe that can be an INI option
DONT_DO_WHAT_I_WANT = 1
rayh: thanks. it's pretty complex - I'm unsure how to find all the cases to test.
Some of the corners do cause the machinist in to pause.
sorry say again?
me to pause
I wonder if we would do well to get the user list to comment on it?
(that pic shows nominal path with right and left comp. entry moves are near the center)
it is on a branch. people could play with it. I am sure there are still bugs.
I'm really impressed with it's ability to find it's way round without gouing.
Did you try a diameter larger than some of those openings into near circles?
that will gouge - it doesn't check the full path ahead
it does one segment lookahead, so I'm sure you could break it if you tried
yes it's easy to gouge things that aren't "nearby"
rayh: did you ever seem my weird thread video?
I saw the mazak threading but give me the link?
rayh can watch videos!?
[02:48:15] <jmkasunich> http://jmkasunich.com/cgi-bin/blosxom/index.html
ray has broadband!
Ray is listed as a microwave tower with the FCC
oooh - so you can cook hot dogs too! ;)
i'm looking at concave.png but I dont know what i'm looking at (a toolpath in axis with cutter diameter compensation?)
one that doesn't barf at concave sharp corners
Darn I'm using 8.4 and have to add stuff to see videos. will be just a bit.
stupid flash player
just double-middle-right click or something
what does it mean for a corner to be concave?
you-tube has made it a defacto standard, for better or worse
I think you can just poke the "find a plugin for this" thing
seb_kuzminsky: angle < 180 degrees
hm... so what does it mean for a corner to be sharp?
seb_kuzminsky: a round tool can not exactly cut the concave corner
It means that a finite diameter tool can not cut clear into the corner.
if you tell emc2 to mill out a hollow square using tool diameter comp, it will say "I can't cut square corners with a round endmill"
rayh have you met seb? He's doing drivers for new Mesa stuff like their parport device (similar to the Pluto)
No I have not. I've seen the name several times.
ok, not sure if you guys had been around at the same time before.
seb_kuzminsky: emc has always made that an error because you don't know how far to cut until the next segment comes along. the old algorithm didn't handle any of that.
i see how the round cutter can't make a square "outside" corner when milling a pocket say
the inside corner can be sharp of course
I guess we have opposite definitions of inside and outside
I think you are using inside/outside opposite the usual convention
outside of the hollowed-out volume
it's called an inside corner when the cutter is inside the hollow
machining noob here
not when the remaining material is "inside" the path ;)
I'm inside my house
so the rest of world is outside then?
did whoever machined the rooms make square corners? or are they rounded?
that would sure be a lot of chips
milling out a house
anyway, emc's cutter comp was pretty much impossible to use unless you wrote your gcode by hand with its limitations in mind; I'm trying to fix that
"well Istarted with this 50 foot cube of aluminum
is the path in concave.png a slot in a workpiece? or is it milling the outline of a workpiece?
its just a test
neither. it's a test case that represents what would be an extremely strange workpiece
i understand that, but i'm still grappling with inside vs outside :-|
it tests all of arc/line, line/arc, line/line, arc/arc "corners"
just consider two rectangles, one inside the other
if the cutter is on the inner path, it's doing inside corners
seb_kuzminsky: suppose you want to mill a square hole in a plate
that would be a pocket
you can't make the corners square of course
but EMC wouldn't even let you try
with cutter comp on
if you programmed the square and then used cutter comp to say "mill inside the programmed line" it would throw an error
youd have to program in arcs in the corners or emc wouldnt run it?
cradek's changes will insert the arcs automatically
I'm impressed jmkasunich. When you going to post the nut that threads onto that?
rayh: its a special flexible nut - see the pic above the video ;-)
getting close to cam isn't it? where the line there?
more than one segment lookahead is probably the line between control and CAM
seb_kuzminsky: there's no consensus about where the line goes
(but I'm pretty sure this is not cam.)
a cam program could sure start with this
Ah. That. It reminds me of winding a string onto tops when I was a kid.
if you scroll down a few posts you'll see the whole vehicle, and a wider pic of the drivetrain
there are also some pics from our trip to Stuart's shop, dunno if you've seen those already
I got to the snow and stopped. We had a heavy 8 inch last night.
wow that's a lot for april
we'll probably get snow at some point this month, but hopefully not much
we just got rain - happy about that.
Lot of mud underneath. Snowblower didn't like the combination very much.
that's definitely the worst kind for snowblowing
I did see the trip to stuart's
where do you live ray?
What about you?
heh - apparently a few inches fell this past weekend (while I was in San Jose)
i'm in colorado
we've been getting just dusted the past few nights, no serious snow for months
I've got kids living just north of Denver.
you're on grizhfminimill too right?
hey, i'm just north of denver, in boulder!
daddy is that you?
I'll never admit it.
My son works in bolder for a web writing company.
i work at the university (CU), what company is your son in?
Um. I got a good memory but it's short.
second story office right across the street from the war surplus store if that helps.
sure, i go to the brew pub next door that that place for lunch sometimes, the java porter is excellent
Right you got it.
I'm really rural up here. About 140 miles north and a bit west of Green Bay WI
that sounds nice... i grew up rural and i miss it
i gotta go, bbl
The guys use to feel sorry for me on my 1.2 k dialup.
It was about 26 miles of copper.
I should goi too... gotta walk the dog and get some sleep
I knew I was missing out on a lot of internet. Just now realizing how much.
Yea good talking with you guys. Catch you all later.
good night all :}
cradek_ is now known as cradek
jepler: when you do get a chance to look at the 8255 - see if your relay works.
skunkworks: yeah I have been putting that off
No rush. You know how fast I work.
cradek: are you going to post a reply to dave on the list? your work will make synergy happy won't it?
yes but it's nowhere near done
(but it is in the right directions...)
more correct? ;)
if I pull it off it will be exactly what he needs for synergy's scheme (as I understand it) to work
is there any doubt? You are the man. (well one of them anyways..)
Is the path test you show + and - tool diameters - or is it 41,42.. Or are they one and the same now?
skunkworks: at the lower levels they are essentially the same, except for a sign difference.
skunkworks: I don't *think* that cradek has changed anything about the gcode..
yes 41+ is the same as 42- and same the other way
EMC: 03cradek 07concave_comp * 10emc2/src/emc/rs274ngc/interp_convert.cc: ok, arc-arc wasn't as easy as I thought. I think this fixes all the cases.
[18:43:14] <cradek> http://timeguy.com/cradek-files/emc/arcarc.png
I moved my two CDROMs around the graph paper for a while, so now I must have it right
try a DVD :)
I have round things of 7 different diameters here... funny
yes four sizes of coin plus three larger things
i'm still trying to understand what you guys are talking about
is it fair to say that with tool diameter compensation, the meaning of the path described by the gcode changes?
well, sort of :)
with out compensation, the gcode path describes the path that the controlled point takes
without compensation, the G-code path is the path of the cutter
with compensation, it describes the material left behind?
or cut out, yes
it describes the edge of the cutter.
it describes the edge between the removed material and the left material
so you don't have to offset the tool path in the cad software
for your tool diameter
in cradek's pics like the arcarc above, the center line is the controlled point, and the two lines on the sides are the boundaries of the area the tool has occupied?
here's an example: you awnt to mill a CD - both the inner and outer circles
say you're milling in the counterclockwise direction for both cuts
to be conventional milling instead of climb milling? ;-)
(another way cutter comp is used is to have CAM generate the tool center path for a nominal tool size (e.g., 1/8"), then use cutter comp to move the tool the difference between the nominal and measured tool diameter (e.g., 1/8 - .121 = cutter comp of .004 inch))
jepler: that's gonna take me a minute to wrap my head around...
for the outside path, you want the cutter to be on the "right side" of the CD edge
seb_kuzminsky: in that case, don't sweat it..
SWPadnos: the outside
this will be outside comp
for the inner hole, you need left comp - so it's inside the remaining material
how does emc know which side of the line is inside and which is outside?
in both cases, you'd program the arc as the actual edge of the material, and let EMC figure out where the cutter needs to be to put the edge in the right place
G41 and G42 are left/right (possibly not in that order)
left/right... ...when moving in the positive direction or something?
g41/g42 mean "tool follows the left/right side of the programmed path"
as if you were riding onthe tool
right, when walking along the path
now consider a square CD :)
its all coming clear
you might notice this makes g41 always climb mill and g42 always conventional
[18:59:44] <skunkworks> http://www.isd.mel.nist.gov/personnel/kramer/pubs/RS274NGC_3.web/RS274NGC_38a.html
the outside comp is easy - you can either do a bigger square or add some arcs to the corners of the outside comp path
although chris is working on getting rid of section B.5.1
so it 'just works'
yes - the problem occurs when you're on the inside of a corner - the inner square of our square CD
right, the naive approach can take divots out of the material at the corners
i think i see
if you move to the end of the programmed path, but offset a little to the left, you will have cut beyond the next line
right - you need to shorten the move by the cutter radius
(for the simple 90 degree case)
cradek: will your code refuse to enter "alleys" that are too narrow?
and you'll be left with a rounded inside corner in the part
seb_kuzminsky: no, it only considers the corners
not at the moment - it only looks at the next segment
one movement lookahead
if the "alley" is formed by something 500 lines later, it gets to be a harder problem
yeah, that gets hairy
I think throwing away part of the program (the part on the other side of the alley) is not the job of the machine control
and that's why cradek was saying that's more in the domain of the cam program than the machine controller
what he said ;-)
it's up to the machine programmer to choose a tool that will fit the part ;)
if the sw is going to warn him he's messed up, it needs to do deeper analysis
and that's something that's expected in CAM software
makes sense, sort of
it would be shiny frosting in a machine controller such as EMC2 :)
i guess i had imagined that cutter diameter compensation was handled by the cam too, and the mc didnt have to think about it
gcode gives me the willies
it's like some ancient broken assembly
seb_kuzminsky: often it is, but if you put in a resharpened tool, the control can compensate
seb_kuzminsky: man, don't get fenn started
that's what jepler was talking about...
the gcode describes the material to be left, the mc knows how big its tool is, and does the right thing
for hand programming it's nice too. you can just program the part outline you want.
if i break my 1/4" cutter and have to finish a part with my 1/8" cutter, a simple change in the tool table will fix it
and if you are cutting inside corners, it will cut away a little mroe than the 1/4" tool could
i would change the active tool diameter from 1/4 to 1/8 and the mc will automatically do the right thing (because of your new fixes)
yes if the program is written using diameter comp
i thought of another reason i got confused last night about inside vs outside corners
in g41 & g42 you're riding on the tool and the material is on the left or right of your path
when you make a turn, the outside of the turn in the path is the "inside corner"
of the material that is
but i'll be happy to bow to convention :-)
[Global Notice] Hi all! I'm about to do a teeny bit of re-routing, it shouldn't be too messy -- affected users should be around 7K. Have a great day and thank you for using freenode!
The wiki tells me to ask on IRC about testing an 8.04 package?
I'm also wondering if the Lyx documentation will be updated to run with the version (1.5.3) of lyx installed by default on that release?
dang sf with anonymous bugreports. it should at least ask for a name or email by default.
IMO we can change the "standard" lyx version whenever the folks who write most of the docs want to change it. (I think that's pretty much just jepler and tissf lately.)
I wish lyx was better about its cross-version problems
cradek: it's not so easy as that
cradek: we have to have lyx data that can be read by lyx on the oldest machines we want to support, on the machine that generates the online version (which isn't presently an ubuntu machine at all), and can be read by the lyx-to-xml converter.
for any of those three reasons, we have to use the "dapper" version of lyx in the CVS repository
and because it requires good hygene and because it ruins diffs when different authors use different versions of lyx to edit, using lyx2lyx is not a good solution.
"good hygene" = "run lyx2lyx before checking in, even though nothing will remind you to do it"
some of the things worth having newer lyx for (e.g., better support for non-latin1 character sets) get busted when going backwards
G0 X0 Y0 Z0
G1 X0.82322 Y1.00000
G2 Xnan Ynan Inan Jnan
hm this is not a very good gcode file
trying to visualize the arc....
yipee... got a new computer today.... old one died of capacitor rot
poor ubuntu site is slow.... trying to download 8.04-beta