any emc on vmware experts?
yes. don't do that
(unless you have to run Windows as the host, and you want to use the emc2-sim package)
i'm just doing it for a test env now
under ESX not workstation/server
does ESX have better virtualization? (I know it can do neat stuff like live migration and whatnot)
i'm trying to find a CAD package that will work under ESX as well, using RDP or NX (something that doesn't require opengl)
it's more efficient, because there's no OS overhead
hmmm. I don't know of any MCAD packages that don't use GL
plus i can run it lights out
ok, so the ESX hypervisor is the host?
hrmph.. i'm trying to avoid buying a cad workstation for a proof of concept
yeah.. bare metal hypervisor
hmmm. I wonder if you can use SM 3.0 with XP64 under ESX
my only boxes with decent horsepower are ESX.. my workstation boxes are either only dual core laptops or single core desktops
nope.. only under "experimental" under workstation or fusion
and fusion is only hosted on the MAC, right?
and its cheaper than workstation for some irritating reason
I say again: bastids
what's a decent linux package for CAM that can read solidworks models?
if there is one.. (i'm new to CNC)
I don't know that there is one
SheetCAM is now available
damn.. ok.. cause there's a opengl emulater for linux that'll sort of work
there's a CAD program that could do also, called Synergy
but only under linux
MEsa should work fine
it's software only
as far as that goes, you should expect to use one PC as a machine controller and a second one as a CAD box, or expect to dual-boot between the RT kernel and the standard kernel (and X config)
i've got tons of spare machines in the workshop for running the machine controller
the reason is that vendor-supplied (and therefore accelerated) video drivers have not been good for RT performance so far
i just havn't configured them yet..
probably dl360 or something like that
unless emc will run on sparc.. in which case i could fire up an E4000
(I don't know that model number)
well, it used to run on Alpha and SPARC I think
older 1/2ghz xeon 1U pizza boxes
there's a non-RT compile for PPC as well
it doesn't need much horsepower right?
I have two systems here, but I'm not sure which one is "faster": Tyan S2723 with two XEON 2.4GHz CPU's, 2GB Ram, http://www.tyan.com/archive/products/html/tigeri7501.html
-OR- Tyan 2865 http://www.tyan.com/product_board_detail.aspx?pid=85
Dual Core AMD Opteron 170 @ 2GHz, 4GB ram
well, depends on what you expect to do with it
i've got a dual opteron i could use as well.. so either dual 1.2ghz p3 or dual 1.4ghz opteron.. both with 4G
JymmmEMC, other than the memory difference, I'd bet the dual XEON would be faster
SWPadnos: I was afraid that might be the case.
depends on how the memory is accessed
the NUMA on the opteron board could work wonders
the Opteron has better IPC though, so it could be close
it's only one chip though, so one memory controller
single chip, dual core
oh, good point
Instructions Per Clock
i'm thinking dual socket
I have NFC on AMD stuff
yeah, the dual-socket ones are screamers
IPC is a generic term
I dont have nfc on Xeon's either, this is circa 2003 as well
I don't know much about Xeons, but I can tell you that my dual Opteron 275 is a very fast machine
i don't figure speed into anything really.. you don't need speed for anything other than gaming and cad, which i havn't needed to this point
unless your doing flow modeling or something like that
devi0us, what kind of machine(s) do you expect to control with EMC?
heh. i just batch all my transcoding and come back later
JymmmEMC, then the Xeon I think
bridgeport series 1
not sure which controller is on it right now
for encoding/decoding (and therefore presumably transcoding), the P4 does better than the Athlon
devi0us: I dont have that many decades left on the planet
may end up swaping it out
well, if you don't get rid of the controller that's there, you don't have much use for EMC2, since EMC2 is a machine controller
it is not a drip-feed / DNC program, it fills the same role as a Fanuc or Centroid (or whatever)
JymmmEMC: i batch it out across 16 CPUs so it really doesn't take that long.. never had a desire to benchmark it
devi0us: 16 cpus?! 16 boxes?
16 cores then?
2 dual quad core
devi0us: what are you using (sw wise) to transcode?
the quad or octal quad-cores are expensive as hell
heh.. i got the full dual quad dells for $629 each w/2G
then dumped in more memory and nics
JymmmEMC: i don't even remember.. i havn't done it in a while
i use them mostly for compiling, and testing stuff. i'm running 32 VMs on those 16 cores at the moment
dual quads are affordable, quad quads or octo-quads aren't so much :)
Well, both of these are in 1U right now, if I toss them in a mid-tower, I need to buy a PSU
15K RPM fans are killing my hearing
FUCK! I can hear agian
i have all my arrays out in the garage with the servers
so i don't have to listen to them
ears are still ringing though
i'd move them up to the workshop, but i havn't dug a trench for the fiber run from the house yet
hrmph. that bridgeport has whatever cnc controller bridgeport sold in 1981. i'll probably need to replace it with emc anyway
heidenhain tnc 145. whatever that is
devi0us, the question of performance needed for EMC is a hard one. EMC doesn't need a lot of compute power, but it does need good (low) latency
and there's no good rule to figure out what machine might work well, you just have to test and find out
which won't work in a VM
so if you want to play with the user interface and tools provided, then you can just install the sim version (on Ubuntu is easiest, since we have .debs for that)
thanks.. i've got the livecd running under ESX now
let us know the latency test rsults :)
i'm looking at the cheap harbor freight conversion now.. might be better than diving into a bridgeport with no idea what i'm doing
is there a buit in benchmark?
from a terminal, run "latency-test"
lemme nx into it and i'll let you know
actually. i'll just ssh. no need having the gui overhead
it's a graphical app
ok .. nx it is
ssh -X would do
nx is faster than ssh piped X
altho i don't think it matters. i've got negligble latency between here and the esx server anyway
yeah, over a LAN it shouldn't matter much
hmm. can you do a desktop-style lgin that way (like XDMCP)?
yeah.. that's what NX is
it's a high performance remote desktop thing
faster than vnc or X
is there a package for Ubuntu?
I only see the library when I search for "nx"
yes.. let me find the install guide
you have to add a repository
[02:35:18] <devi0us> https://help.ubuntu.com/community/FreeNX
does multimedia too. so you get audio
latency test is running now
only using about 50% of one cpu..
normally I'd suggest that you stress the machine
by running glxgears, doing ls -lR /, opening openoffice, dragging windows around ..
don't worry.. it's being stressed
just on other VMs
but I'm not sure how much that helps :)
there's 11 other VMs on this particular machine
once you see the numbers with the other Vms running, can you try it again with the others paused?
I'm curious to see whether a VM could be used for a real machine (depending on the hypervisor), or if it's just plain impossible
i can migrate all but one off.. my firewall isn't set up for HA yet
that would be more likely to be done on workstation.. normally you don't let VMs touch the parallel port or USB on ESX
ok. actually nevermind. presumably if an interrupt happens in our VM, it won't be serviced until the hypervisor decides to let our VM run again
USB we don't really care about (except for the mouse), but other I/O or PCI memory space is pretty important
on the question of doing a Bridgeport vs an HF minimill - I'd just go straight for the Bridgeport
the minimill will be mostly useless after it's done, especially when there's a BP sitting beside it :)
i haven't bought it yet
i can do a mini for $600 or so complete
the BP also already has motors, power supplies, some UI controls (buttons and stuff), limit switches, and ballscrews
in some ways, the BP conversion is easier
the BP is around $1k
(though I shouldn't talk, I still haven't done mine and it's been 3 or 4 years now)
does the latency test run forever, or will it end the job at some point?
it runs until you stop it
what kind of numbers are you getting?
ak.. it doesn't report to standard out when you kill it?
which numbers am i looking at? the max interval or jitter?
heh - no, it's graphical, silly :)
the jitter numbers
servo (1.0ms): 133062806
base thread (25.0): 131804942
configured for the mini mill sim i believe
if that matters
those are the latency numbers?
max jitter numbers.. it looks like there's ramp up
because the interval numbers on the right are MUCH lower than the max on the left
yeah, they'll spike every once in a while, and that's what you want to capture
those are last interval, so they're constantly updated
the max numbers are the important ones
i'm guessing those are bad
just to be sure, these are the numbers in the center column, "Max Jitter" ?
(you do have 3 columns, right)
what the numbers mean is that every once in a while, when you're trying to run a fast thread every 25 microseconds and a slow thread every millisecond, it will take an extra 132-133 ms to respond to the timer interrupt
which would be a very bad thing if you have stepper motors which you're driving with software generated step pulses on the parallel port
or if you had a hardware servo interface card with any sort of reasonable watchdog timer
that does substantiate the idea that VMs and RT don't mix - thanks :)
heh. yeah. that's well known
that's why you don't see VOIP systems running virtualized
time sharing and precision timing don't mix well
SWPadnos: thanks for you help.. bbl
sure. see you
man, it's cold tonight!
feels like it, low in the 3-'s they say
ok, then go complain somewhere else :)
i've been waking up to single-digit temps
above or below?
i wish my workshop was heated...
SWPadnos: No insulation
well, that is a problem
it was ~10 below yesterday morning (unlike the balmy morning today)
SWPadnos: Yeah, we stuck a fleece curtain between the kitchen and bird room - big difference
that sliding door probably doesn't help much
SWPadnos: Yeah, not heating the LR or Kitchen at night now.
SWPadnos: You mean the whole wall of glass =)
SWPadnos: the blinds help - tested that the other night actually.
[03:09:24] <skunkworks> http://cnczone.com/forums/showpost.php?p=536579&postcount=543
there are relatively inexpensive cblinds you can get that actually are made to help
SWPadnos: I might get some 6mil clear vinyl but not sure if it'll help or not
it helps, but it's not so great if you still want to be able to use the door
not the door, the wall of glass
glass - 3" gap - clear vinyl
is what I was thinking
not sure if it'll do any good or not.
wll, plastic helps a bit for general heat loss, but I think it's most efefctive when it stops air leaks
No leaks there
just single pane
well, yeah, it may help - I don't know how much
I'd get some of that 1/4" mylar bubble wrap, but need light in that room too
you'd probably be better off with two layers of the clear plastic
well, instead of 2 layers as bubble wrap
you want to let as much light in as possible - both for illumination and for IR heating
but air gaps are what insulate things
Well, the bubble wrap has sealed air pockets though, that's the difference
so in that regard the bubble wrap is better (lots of small air spaces)
but since bubble wrap will probably let in less heat, I think it's a net loss compared to the insulating plastic
yeah, no clue.
they make window (and patio door) insulating kits - double sided tape and plastic that shrinks when heated with a hair dryer
cheap, easy, looks good, seals well
I just did that on a sliding glass door and several windows
the door kit was under $10 IIRC
that stuff is much more available around us than it is in California :)
maybe not so bad in the bay area though
why bother now? isn't winter about over?
(I just paid for a propane fill today...)
that must have been nice
it's not bad right now
a lot of people here (including my mother unfortunately) pre-paid for fuel at a fixed and suppsedly reduced cost
of about $5/gallonfor heating oil
which is now $1.50 or something
previous tank was nearly $800 (!!)
yeah the folks who locked in a price for this winter are going to lose big.
("it's as high as it's ever been! I better lock in this price!")
too bad about how predicting the future is hard...
yeah. for my mother, the problem was that she's on fixed income, so any more increases would have meant turning the heat off
buy high, burn low
$500 for the cnc brain doesn't sound like a ripoff to me
unless it doesn't work
considering that it's about the cost of a low end (new) PC and a Mesa card, I don't know
it might not be a viable product, but certainly not an unreasonable price
no, cenceptually it's not bad
more than competitive products, and from what I can tell, it doesn't exist yet
and if someone was selling a low end pc with a mesa card set up to use, things would be quite different
well, Dell and others sell low end PCs, and Mesa sells Mesa cards, so it's not all that difficult to put together a more capable system yourself
the CNC brain still needs a PC and PC software to work, AFAIK
so it's also not an all-in-one solution
if it's only 6 axis, I don't really understand how it's going to drive a lathe, a mill, and a usable robot
one at a time
the software may support multiple "brains"
did you see the post on the emc forum about how when the guy boots his system it starts moving?
the one where the parallel port has noise unless EMC is running?
i think it's kudzu
but it could be noise
well, noise as far as the motor drives are concerned :)
all I'm saying is that the better than average system builder has trouble integrating a system such that things like that don't happen
do you consider people asking questions on CNCZone to be "better than average system builders"?
I'd say leaving out the potentially deadly robot might be a good idea in a single person shop
I think anyone that builds a system that moves is better than average
unless it moves when they don't want it to ;)
it sounded like that machine works when emc is running, which is fairly advanced
it must be different trouble reports
there were two similar one IIRC
the one I'm thinking of had some random motion whenever EMC is not running
not just at startup
maybe the cron job that he was supposed to get rid of runs faster on his system
the cnc brain seems a little atypical in that the first post says it's working and available for sale
on page 46, the guy has a job with the government and he can't be faulted for not making it work
if that is the Brain thing I've read about before, it sounds like a salesman talking
yeah - "double closed loop with dynamic soemthing-or-other compensation and motor auto-tuning"
SWPadnos: that's what I want!
does your something-or-other get out of whack with EMC2?
oh, phew! :)
I just want the auto everything
plug - n - play
stepper motor tuning you mean, based on the load and all
I know nothing about stepper motors
then you don't need that brain ;)
it makes step/dir signals
I must have entered the conversation a little late
it's a bit like a USB-connected, Windows-only, 6 axis PPMC :)
err - USC I mean
not PPMC, USC
Is Jon working on that?
I sure hope not
I think he may jump straight to ethernet
no, it is something that some dude on CNCzone is pushing
and has been for a while
peter wallace is funny: FlipFlop ahostmot2/makewatchdog.wdogabittus/wdbite has been replicated 1 time(s)
any get done quick product on cnczone gets true believers so fast it will make your head spin
the first post was "CNC BRain is ready", and as eric_u1 pointed out, on page 46 of the thread (!), a few people apologize for the guy not having them ready yet
"he took a government contract" ...
If I sent him $500 and got nothing, I'd be calling it a ripoff too
how do you "take" a government contract? grabnrun?
something like that is done every day I'm afraid
I want a little of the 'bailout' money.
1 or 2 billion would do me just fine.
turns out that doesn't work for everybody
if you are a banker or stockbroker, step right up
GMAC just found that out
they may go under
if you are a car maker, beg, and maybe
if you make anything else, forget it
oh come on. you could do with just a fwe hundred million, I'm sure
actually, you have to be a banker now
if you're in the finance system, you don't even have to beg
why ask for peanuts?
probably will not get peanuts
they'll tell you to go to one of the big boys
[05:12:04] <SWPadnos> http://www.theonion.com/content/video/in_the_know_should_the_government?utm_source=videomrss_89550
but the big boys are parking the money in 3 month treasuries
$700B for bankers, approved almost instantly, then $25B for companies that employ normal people who actually produce something of value, and they have to beg for it
the 3 month treasury rate was zero the other day
jmkasunich: is it possible to tell where the congressmen's money comes from?
yes. one sec
comes form us, of course
to some degree
unless you are the Governor of Illinois? :)
[05:13:55] <SWPadnos> http://www.opensecrets.org/
open secrets has the donations above $250
but you can search by zip code and see which of your neighbors are idiots
must go back to working on my 'magic' kinematics - the all geometry compensation model :)
or in the case of one of my neighbors, funneling politicians money for their brother
stustev1, did you go back to the cinci kins, or did you get it working already?
one of the people in my zip code gave BHO $1700 too much and was refunded
I am almost done the the 'magic' kins. It will be the cinci kins.
I like the guy on cnczone that wanted emc to compensate for slipping paper
I just responded to that
I saw that
it is a funny problem though
our previous dept. head worked on problems like that, it's not easy
the post I was talking about earlier involved a mesa board http://cnczone.com/forums/showthread.php?t=69280
moves on startup
I saw that one
that is strange, since the Mesa I/Os are tristated until the FPGA is programmed
the 7i37 may not be though. hmmm
it may default to some state that makes his system do funny things
you would think a transistor would be tristated fairly naturually
but he may be doing something stupid
the outputs are push/pull and isolated, so they could default to one or the other pole
err, I mean it could turn on if powered, even if the FPGA is tristated
if he is driving a high impedance input, it might be enough to look like a ground potential
it's a tough problem, and one that probably can't be figured out by someone who's never seen the machine, the components, or a wiring diagram for it
that's why I'm using mechanical relays, it's a lot harder to get them to actuate if you don't want them to actuate
yeah, there's a few millisecond filter on those
and a substantial current is required
I find it hard to believe that the output transistors are going to an on state
what gets me is that the input on his amp dies - that seems very strange
I'm ignoring that problem
he probably has a ground loop
I'm goin' in, cover my back
power up the PC, then the controls, then the motors, then the "router" (not sure if he means a router used as a spindle, or if "router" means the axis motors)
that's what I'd do
the PC needs to be able to run with the controls on or off, since e-stop should likely kill power to most if not all of the external electrics
ball is in his court now :)
I just reminded myself that I need to get the watchdog running :)
random question from cnczone: "Does G98 G81 work on EMC2 Software? G98 does the rapid move to R"
it's there, so I assume it works. I haven't tested it myself though
what the heck does that mean anyway?
it's in the handbook too, I'm giving the guy the link to that so he can figure it out for me
G98 and I think G99 specify whether a canned cycle will return to the retract plane or the initial Z height
I think in G98 mode, it actually goes to the higher of the two
g98 -- initial level return
g99 R value return
OK, so I have them backwards
I do that
can be a real problem
from a conversation with toastydeath, it seems that G99 isn't really useful these days
I think we decided that it must have been a holdover from the days of limited tape length or something, because all it really does is save you one or maybe two lines of G-code
I remember a weekly quiz in an early circuit analysis class where I had something exactly opposite in my brain (compared to how transistors actually work ...)
so I got 0 points on the quiz questions, but the professor did give me 1 point for the date and one point for my name :)
(out of 10, so it wasn't bad :) )
my advisor asks true/false questions with triple negatives in them
and then you have to explain why
so I would explain why correctly, but he would get me confused with the true/false part
argh. bed time. good night
encoder held together with duct tape, hmm
[06:34:07] <eric_u1> http://cgi.ebay.com/BRIDGEPORT-Series-1-CNC-Vertical-Mill-R2E4-Boss-9-N-R_W0QQitemZ130272884039QQcmdZViewItemQQptZBI_Mills?hash=item130272884039&_trksid=p3286.c0.m14&_trkparms=72%3A1205%7C66%3A2%7C65%3A12%7C39%3A1%7C240%3A1318%7C301%3A0%7C293%3A1%7C294%3A50
it's LIE-NUCKS, not LIN-NICKS, ffs
I used to have a recording of torvalds saying linux
I think it was a rt-linux demo
sorry, I had to vent
I understand completely
these brainless hacks on the table next to me are trying to develop a "completely secure test-proctoring" program that runs on hardware "so old that it can run Windows 2000"
could someone check this link: http://www.bikely.com/
ooh bike routes
it works for you?
I think they just blocked me because their software stinks, and I posted an example in their forum that demonstrated that
test the site with tor
or any public proxy
too much work, I'm moving to mapmyride
[07:36:54] <maddash> http://anonymouse.org
I was just being paranoid as usual at this time of night
Just this once .. here's the log: http://www.linuxcnc.org/irc/irc.freenode.net:6667/emc/2008-12-11.txt
[15:36:18] <skunkworks_> http://www.cnczone.com/forums/showthread.php?t=69653
ew strange macro language, but it has questions and answers
yes - I was just thinking how much easier it would be with emc (and cheaper) ;)
would it be politically correct to post a reply to that effect :)
heh - no thank you ;)
aw I wanted to watch the fireworks
if he already has camsoft, how does that make emc cheaper?
from the begining.
I don't understand how any cam macros can make a control do rigid tapping if it can't
sure, but he has entered the gates of hell never to return
that's a good point unless the macro language works in realtime
I guess the bottom line is that he spent $5k on a galil motion controller, and the machine control software is just a user interface to that
Art Fennerty stopped developing Mach, correct?
eric_u1: yes, some younger guy is running the show now
seen new owner at a talk in the UK
is there any indication that he is as good as Art?
new owner seems to be from a CNC using background
from what I can make of "drill macros.txt", the rigid tapping cycle sets an electronic gearing between spindle and axis, then polls every 25msec (according to comment: the code says 'sleep 0.0025', so it's hard to imagine how that's 25ms) until the bottom is reached, then commands the spindle to reverse.
if it can follow through the reversal, that seems like it would be just fine
he certainly knows his machining and the problems
is there mention of a spindle encoder?
COMMAND MG_TPE 'get spindle encoder count
IF \866>\863 THEN [STOPSPINDLE]:GOTO :TAPBOTTOM 'run spindle till total encoder counts
so Galil is taking care of it
last I checked, 20 year old ISA galil motion controller cards were going for $2k on ebay, but that's been a while
and people complain about the .hal and .ini files in emc ;)
some "volunteer" could write halconf
which would read the output so it could be re edited without breakage.. hint
* jymm volunteers archivist - Especially since he likes to always mention "some voluteer" alot =)
* archivist sends a press gang around to jymm's house, time he did something
archivist: Hey, we're waiting for you to start writing the stuff you keep asking for
Oh, btw... did you know you can "scroll wheel" thru tabs in FF when the pointer is hovering over them
micges, can you tell me " xxx Piastów, ul. C. Norwida 49, Poland" which part is town and post code etc and best carrier from abroad in Poland
except when ubuntu doesnt see one direction of the mouse wheel
jymm, you can also "scroll wheel" through applications, if the pointer is over the bottom panel (where the application "icons" are)
and if you have desktop effects on, you can "scroll wheel" through desktops
which is a bit disconcerting if you're trying to scroll through a web page, and the mouse moves over the desktop
jepler, that was a 2.5 ms polling cycle "sleep 0.0025"
I think that address means "your mail is lost, thanks for playing"
again - am I the only on that thinks this is sexy? http://imagebin.ca/img/CTkv0T.png
I have to send a parcel, I have the money :)
no, it looks great, can I download the schematic?
uh, yeah. that's what I meant too. stupid typos
how do you want it?
I don't know if all the bugs are out yet. Sort of a work in proccess.
question: where is the best place to look for a machine definition for a SYIL X5 (aka. BF20 Vario CNC)
what is a machine definition?
configuration for motors and switches
I don't know that you'll find one that's ready-made
(sorry, coming from Mach3, they have different lingo there
refer to the documentation from the manufacturer, then create the assignments in the stepconf gui
[16:30:48] <jepler> http://linuxcnc.org/docs/html/config_stepconf.html
zapfding, stepconf is a utility that makes it very easy to create your own configurations
jepler: ok. and is there a place where I can post once I found a working config?
there should be a place on wiki.linuxcnc.org
if there isn't, please create one :)
yes, add a page on wiki.linuxcnc.org
If you want to add information to this page or add a new page, follow some BasicSteps. http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?BasicSteps
I found it fascinating that everyone is apparently repeating the same process for the same machines
I don't think there's a page yet for users to contribute machine configurations. you should feel free to create one
I actually don't know anyone using EMC on a Syil X25, though that doesn't mean nobody is
zapfding: what brings you to emc from mach?
additionally, each user configures their inputs (limit switches, home switches, probes) differently
I think the machine is fairly new in that config. they use the same motors and controler as for their X3 etc.
there are a number of resellers not that sell the Syil stuff pre-configured bundled with Mach3 demo
do they provide a specific motor drive box with well-defined pinouts?
(Syil I mean, not the other vendors)
yes. the controler is made by Eastern Crystal
and gives you a decent parallel port interface
if there's a specific configuration of pinouts, and axis scales (meaning that the motor microstepping, gearing, and screw pitch are the same), then a stock config is a good idea
if I run into problems, is it ok to ask for help here or are you already worn out by newbies?
we're worn out, but you can still ask ;)
heh - that is what this channel is for.
it depends how many times you ask the same question
isn't that the truth
then the answer is cut'n'paste
questions are always welcome, though the channel isn't always full of intelligent, helpful people like now
seems that could be read two ways.
of course now, it's also full of dummies like myself
don't you love the english language?
archivist: Piastow is town, post is xxx
just move a comma and the whole meaning changes :)
thus, flame wars
thus flame, wars
while I am at asking questions: is there a usable gcode generator for linux already? I am kind of fed up with my current toolchain (QCad --> SheetCAM --> Mach3)
well, strangely SheetCAM is available for Linux now :)
as is QCad
(well, there's a program names QCad, it may not be what you're talking about)
hm, if EMC2 reads the output of SheetCAM better...
thanks micges trying to find the best/cheapest carrier to send 35kg over there maybe UPS or FedEx or ordinary post
I know, QCad is crossplattform
I believe there's an EMC post for SheetCAM
that's good, it's no longer windows-> Qcad -> sheetcam -> linux -> emc
it's not hard to write one
TurboCAD/CAM also works OK, with minor tweaks to the post
(in Windows, of course)
people are slowly working on various tools for Linux
oridinary cheapest but in poland you didn't know WHEN it will be on place
heh typical post
zapfding: what was the sucky part of your qcad-sheetcam-mach3 toolchain that caused you to be fed up?
archivist: best is UPS
seems to be £125 by UPS
lately, we've seen the sheetcam guy is interested in sheetcam-emc2 working well
cradek: I get strange circles in my gcode (which I suspect to be a SheetCAM-problem) and the import into Mach3 is different before and after restart of the programm
before and after restart of which?
it is not behaving deterministic
are the strange circles in the SheetCAM preview, or is that after loading into Mach?
after loading in Mach3
you should be able to verify the circles with a calculator and decide which program is bogus
from what you say here, sounds like it's likely to be mach
if you ever see strangeness in EMC its normally sorted very quickly
ok, try loading one of those programs into EMC (fromthe LiveCD if you like), and see if the circles are there
good point. will try
that's another way
another problem was that sheetcam does not generate radius corrections for outer pathes
but that is minor
you mean it can't generate gcode that moves the tool around the outside of a path?
the real problem is that the Mach3 author does not do bugfixes anymore
well, welcome to the brave new world
you can use cutter compensation in EMC2
yes. the path generated and shown in preview is not corrected for tool diameter
so you want to generate code for no comp, then do it in the control (EMC2) - that way you can actually change the tool diameter (based on wear, for example) and still have the parts come out right
though emc2's cutter compensation requires special considerations that may or may not be met by your generator
that's true. concave corners can be problematic
(because EMC2 may refuse to execute them, not because you'll destroy a part)
concave in 2D or in 3D?
an inside corner is a problem for cutter comp, even in 2D
is it difficult to compute or just a "somebody should fix it" problem?
if the corner is a smaller radius than the cutter
the way to make it work is to put fillets at least as large a radius as the cutters in all the corners
ie its physical problem
ah, ok. but thats the case with all other tools
so not a limitation per se
it's a bit difficult to fix. there's some work that's been done (a lot actually), but there are a couple of things left that are keeping it out of the main code base
I don't remember what they are, maybe cradek does
zapfding: emc2.2 rejects concave corners for two reasons: first, because the specified contour can't actually be cut exactly; second, because this limitation allows the algorithm to be much simpler
cradek: what is the machinist bbs you read?
as long as I get a defined error that is just fine
skunkworks_, the machinist bbs is #emc :)
zapfding: one way emc excels is giving errors for error conditions, or conditions it doesn't handle
skunkworks_: I sometimes read http://bbs.homeshopmachinist.net/
I thought this was mach - but it was turbocnc http://bbs.homeshopmachinist.net/showthread.php?t=27151&page=4
and most of the pictures are gone. :)
2.2.5 on the live cd is the latest stable version?
that was a case where the gcode was perfectly fine and the control was perfectly bogus
zapfding: you should run the updater and you'll get 2.2.7
2.2.5 is the latest on a LiveCD
2.2.7 is the latest, and there should be a 2.2.8 soon (in the next week or so probably)
zapfding: concave cutter comp eventually turns into something like this, which is a relatively hard problem (and is impossible in realtime): emergent.unpy.net/index.cgi-files/sandbox/offs-one-more-bug-fixed.png
fenn: I get the point...
zapfding, where in Germany are you?
The circuit looks 'solid' to me ;)
I just visited there this summer
I'll probably be back since I think my sister is about to move there
SWPadnos: how is your sister in texas doing?
better. she actually had electricity in the upstairs as of a couple of weeks ago
the regulations and various other shenanigans are astounding
.oO( there is electricity in texas already? ) ;-)
some parts, yes :)
interesting. that galil macro code has at least one error in it
unless \849 and VARB849 are different things
interesting. the university here switched to proengineer + emc2 because they were so fed up with the controller software they got for their machine (dont know yet what type)
yay University! :)
[17:52:16] <jymm> http://www.cyberdyne.org/~icebrkr/RvB_switch.mov
hmm doesnt play on my box
they switched to emc2 for schools in india :)
I'll publish a report soon (I hope, still waiting for some info ..)
hmmm. so that should mean that there will be more EMC2 users than all other CNC controllers combined soon, huh? :)
Sweet - when are is emc going to be gobbled up by redhat? ;)
when are is?
are is never going probable
unlikely not beeing done
Yes - I have problems.
speak like this only master Yoda can, hmmm?
so emc2 doesn't work in redhat ?
don't understand your question.(if that is for me)...just asking if it does or not
there's no reason why it wouldn't work, but you'd have to build your own RT kernel and also EMC2
it seemed that you were trying to confirm that EMC2 doesn't work in RedHat, which isn't true (and therefore can't be confirmed)
I have run EMC2 on redhat 9. Like SWPadnos said, I had to build a kernel etc.
there were some emc2 packages for RH
[19:59:11] <alex_joni> http://linuxcnc.org/rpm/
just wondering....thought as much ...since redhat is still linux
seems my memory is slowly drifting apart
there are packages for a RT kernel for FC3
aren't they up to FC5 or 6 now?
those rpm's are quite dated
emc2 has significant prerequisites that are not packaged by any major linux distribution (adeos-patched kernel and rtai realtime system). Most of us prefer to develop and use emc2 -- which we can do by building and packaging emc2's prerequisites for a single system every 18 months or so.
we frequently say we're happy to host emc2 and prerequisite packages for whatever linux distribution on linuxcnc.org, but rarely does anybody actually do that (2006 may be the most recent time)
har har http://www.twincities.com/allheadlines/ci_11195899
hm, what bogus reasons to stop them, though
where are all the Einsteins' now-a-days ? :)
chrome is out of beta.. yay
chromium is also quite close to run on linux natively, maybe a couple months to go
g'day all and i hope ya'll all have a great weekend ! (we had snow today:) (U.S.A. La.)
cradek, my sister was actually arrested for walking in the road, in her home town
and there was an impassable sidewalk, which technically makes it legal to walk in the road, even though there's an ordinance against it when there is a usable sidewalk
of course, we did think it was stupid
SWPadnos: that is very hard to believe
yeah, it's very stupid
she was actually stuck in jail, with no phone call, for the whole day
oh hasn't it fallen into the ocean yet?
mostly, from Ike :)
at least she has power in part of the house now
another reason not to ever go to texas
hey, my email won the UK lottery!
I am using EMC2 on my home converted 9x20 lathe, and am looking for some more info about tool wear offsets. I made a small batch of parts with a nominal diameter of 8.0mm, and they came out at 7.98mm. This is of no consequence for this particular batch, but I was curious if there is an easy way to correct for this after the 1st part is found to be no good
there is only one set of tool offsets - you would just need to change them by .02mm.
The method I used was to turn a test pass on the 1st part, measure it, and set the "touch off" button to the actual measurement
[21:48:31] <skunkworks_> http://www.cnczone.com/forums/showthread.php?t=62482
then for each subsequent part just ran the same program on it without doing anything differently
dareposte: it's best to leave G54/G92 X=0 and use the tool table for offsets...
cradek: I guess that's what I was thinking... i was reading through the manual here and I admit I'm a bit confused by how to use the tool table
you are not using the tool table currently?
is that the G43 function?
no I did not use the tool table, as I only had the one tool for this batch, and I don't really know how to use it yet
do you use CSS?
well not for that one, but I have the ability to use it
then you should probably not offset G54 (touch off) in X because it will mess up CSS
the tool table is the only good way
so for these tool tables to work, do I need to have working homing switches?
if you move the work coordinate system, the center of rotation (X=0) is "lost" so CSS can't do its calculation
I had been homing manually with a steel ruler at the moment, obviously not very accurate
(touch off Z is fine though)
can you put even just one switch on X? it'll make your life so much nicer
so it sounds like I need to get at least my X homing switch installed and calibrated to the true machine centerline, then I can use the tool tables
yeah I have the switches, just had been too lazy to install them yet
yes you can use tool offsets and CSS then
is there a good way to figure out the actual limit switch position once its installed? Or do I just measure and adjust the setup files accordingly
right now whenever I home with the ruler, the machine assumes that it is x-70 when I click the "home x" button
well you need to declare some position as X=0. This is when your "reference tool" is at the center of rotation
in fact its not that accurate, so I usually take a test cut and just offset the G54 (i think that's what I'm doing... it seems to work okay)
tool offsets are then the difference between the "reference tool" and the new tool
OOOH that's a good idea
so my limit switch could be declared zero, then have a 70mm tool offset or so
or whatever it happens to work out to be
then just measure the first tool by turning a test bar, and set the rest relative to it
if you have a QCTP, just pick a tool you won't mess with, call it the reference tool, and set it up so after you home, X=0 puts it at the center line
i'm using a qctp
once the homing switches are installed I can just adjust the tool wear offsets (in the tool table right?)
yes you'll cut one part, measure it, tweak the tool table, and then run the rest
is the tool wear offset (G43) equivalent to the offset value in the tool table, or is it a different value all together
G43 is tool length offset, EMC has no separate wear offsets
oh okay, so all adjustments are made in the tool table. i was thinking there was an additional set of wear offsets, but i can see how that might get confusing
yeah, there currently isn't, but I know some controls do have that.
that was my next question, I remember reading something about it being possible
okay well thanks for clearing that up for me
I will be installing a limit switch asap :)
my lathe had some mechanical problems (self inflicted), so I haven't made as much progress as I'd hoped. It turns out not pre-loading the spindle bearings is a pretty bad idea
you'll love the tool table once you get all your tools set up
It certainly will be nice to not have to manually home it each time I turn it on
or if it slips...
yeah no kidding, what a pain that can be
but I am a bit worried about the limit switch repeatability, I have omron switches but they're not super precision
it seems like the repeatability of the limit switches would limit the accuracy of the machine if you rely on homing to set X=0
but I guess measuring the first part, and changing tool length offsets would be able to overcome that without too much work
my lathe homes to index, so I get exactly the same encoder count every time (one count = .000005 inch)
is your lathe a servo then? mine is a stepper, with no encoders
yeah that's one big disadvantage to steppers - no index pulse
the steppers have worked better than I expected thus far, but I do see the limitations in their accuracy
you can gauge the short-term repeatability of your homing method by looking at how motor-position-cmd varies after repeated homings during the same session
dareposte: you might be surprised at how well even a cheap simple microswitch works
(since at the end of homing it'll move to the identical commanded position every time)
how do I view motor-position-cmd, is that in the HAL?
I will look into it
on my little stepper machines, .001 inch repeatability was very easy to get
at the terminal, something like: halcmd show pin axis.0.motor-pos-cmd
cradek: right now using the touch-off and measure method, my steppers are holding 0.02mm on 2mm cuts, and 0.01mm on finishing passes. It is beyond my expectations and I am very pleased. 0.01mm is the value for a full-step
actually I think its 0.0127 technically, as my ballscrews are inch based
I'm surprised by your .003/inch linear screw error
(reading the cnczone thread)
yeah it was no good, but EMC corrected it out, and now it seems to work great
they are cheap rolled ballscrwes
linear error is sure easy to fix
it would be interesting to see if it had periodic error too... you could use screw comp for that.
(but you would need good homing of course!)
jepler: I will try that out next time I power it up, I have turned it off for the day already
what is the pitch of the screw? .2 in?
the homing switches I got are "snap action" basic switches, omron #V-10G2-1C25-K
that is correct, 0.200 inches
nominal that is
i have it geared down on the X, direct drive on the Z
I wonder if the screws are 5mm instead of .2 inch
according to Roton they are 0.200" pitch, which is where I got them from. Cheapest screws on the market aside from E-bay
okay, thanks for the help
cradek: when you get back I might have another question about G code G43.1, the manual says it is a "dynamic" tool offset, and implies it can be set from the program. What's it useful for?
good night all
dareposte: probe your tool, use the offset
I agree it sounds like the tool table is the way to go, I'm just curious why G43.1 is available as well. If someone took the time to program it in then it must have some useful function :)
there is an example ngc file using it.. http://cvs.linuxcnc.org/cvs/emc2/nc_files/tool-length-probe.ngc
I use it for making circuit boards. I set the first tool - then the rest get automagically set by a microswitch.
* JPM slaps dmess around a bit with a large trout
my wife has a junker van, I have a fairly new car. So what does she chose to drive in an ice storm? my car
i don't mind trout....
i thought so
been fishing them for yrs
I prefer a large carp or a catfish
for smacking people upside the head with
me too hows things i bought my breakout boards today
Mesa 7I43 and 7I33
do you have the 5i20?
should be pretty decent
i hope so i decided to convet my mill instead of making a whole new one
oh really..... whats on it??
it would be better to covert it
nothing manual machine right now
easier to hide it from the wife
proof of concept sorta thing is good
JPM1: what kind of parport are you planning to plug the 7i43 into?
i'll need some motor mounts soon.. can you pound them out for me
straight from the PP
there's been a lot of trouble with the NetMos 9805 parport controller - it violates the EPP spec so it doesnt work well with the 7i43
just something to be aware of
Thanks for the heads up
most anything on the motherboard should be OK (AFAIK)
at least, it won't be bad for that reason :)
the netmos is a pci adder i take it
PCI or PCIe generallt
you found the right place JPM1
ive been doing my homework i cant even remember why i wanted to use mach now
because it runs on Windows, and is therefor "easier"?
haha that's a candidate for quote of the week: <JPM1> ive been doing my homework i cant even remember why i wanted to use mach now
jepler: new topic for the channel maybe? ;-)
well if you know anything about machine controls it's simple EMC is and mach is not
yeah. different things drive differnt people
at first glance i thought they were comparable products
an engineer I used to work with bought Mach because the screenshots on the EMC website weren't flashy enough
are you on a live CD yet JP
that's it, we need to add some flash to our website
I think the screens are a big part of Mach's appeal
not as many colorful things = worse, and there are more shots on the Mach site (complete with a flash-based viewer!)
yeah i downloaded it on sunday
reconfigurable screens are a big reason several people cite
you mean like pyvcp?
and the relatively easy scripting of machine functions
so show a whack of options
screen shots are nice but having ladder logic is nicer
as long as you arent scared to DO it
pyvcp can only interact with HAL, whereas there are "widgets" in the Mach screen designer that you can use for e.g. pause, load, set offsets ...
emc2 definately requires some "doing it"
we have a lot of the functionality in an easy to use way, and of course all of it if you can read and write a bit of code
what do you think of the feel??
Dmess -yeah i got stepper drives for know some old tolomatic axidynes but plan to use some a-b servos
but I wouldn't say it's as easy to customize in the ways that Mach is easy to customize
i like the feel
SWPadnos: and the fact that you can build a new GUI from scratch in c++, tcl/tk, or python just doesn't satisfy people :(
the idea of maping in signals from hardware to soft via hal is cool as well
right on.. glad your onboard... ;)
I am guessing that most people use the same mach screen
no the is SWEET
hal is a bit of genius
HAL is a dislexik DOG
eric_u1: i agree :-)
oh went over my head
ok out of friday nite routine it goes
coding is going to take some time but im sure i will get used to it over time
emc2 is kind of a coder's machine controller i think
same as me... were in it at the same point
yeah it is but based of of common principles
dam fat fingers
you should see the APT 360 with visual we have in linux
dmess: i've heard that name... what is it?
google... where else
i will check
sorry its been working on thi box for over 2 yrs and ive NOT wiped it out bcz of it...
seb... have you worked with apt??
just the other apt
what other apt??
debian & ubuntu's "advanced packaging tool", apt-get & friends
apt 360 is a cad/cam framework or soemthing?
i see... yes.. drives machine tools..