videolan won't work :\
sxpert__ is now known as sxper
sxper is now known as sxpert
* anonimasu_ found inspiration for writing code again
hi anonimasu_ :)
i'm tryng to get videolan working
i'm asking on #videolan for help..
isnt there a deb
the apps are installed
just can't send the video to another pc on lan
get error : main private error: cannot add this stream
I tried with avi file, tv out /dev/video0 , same thing..
I'm missing something
seem work now
but the video quality is poor..
and i can't see the local video
hm, you can configure it lots..
kinda quiet today :(
probably my fault
I am in a bad mood & doing dishes
I cant find the mop
so I cant finish my cleaning
alex_joni: i was the one that was quiet, didn't you notice? ;)
what's up with you guys today
well damn, i love to be contrary.. i feel great this morning!
you will get over it, as will I...
a bit of coffee and I will be fine
yes just wait for those four waffles i just ate to kick in, and i'll be moaning like the rest of you
* fenn bellyaches
there is a bird that comes to my window every morning and peck peck PECKs every five minutes
i hate that bird
get a cat
the cat is so clumsy it falls off the kitchen table.. no way it'll get a quick little bird like that
i hate that cat
feed it to the bird
constructive crew aren't we?
[14:59:28] <fenn> http://www.somafm.com/groovesalad.pls
what is a .pls file?
a playlist file
internet radio station.. a .pls file is a playlist file
opens in doze players
i'm using xmms
sorry, I keep my doze box locked down too tight, no auto execute, play or whatever
fenn: ever tried musicpd?
ValarQ: nope, why, should I?
i seem to have misplaced all my mp3s
fenn: because it works better than xmms
fenn: it works much better for me anyway
but xmms is easier to type than musicpd :)
you never type musicpd
the programs name is mpd
well that's a little better
and that is a daemon which you only start once :)
then you can control it from a great range of different clients
* alex_joni goes away
back later guys
[15:05:37] <ValarQ> http://musicpd.org/
if i ever do more than just listen to mp3's, like if i set up a radio station for my house, i'll probably use something like that
but it's one more daemon i have to look at every time i type ps -ef
EMC gets a quick blurb in a trade rag... Cutting Tool Engineering. Pics at http://aquila.aldor.com/ebay/emc/
fenn: why look?
thanks, I can even read the text
hmmm i thought emc did more than 6 axes.. i wonder if they got it wrong or if I've got it wrong
thanks for sharing that thalx
current motion planner only handles 6 axes
i wonder why.. seems like an arbitrary thing
all the functions are coded like for( i=0l i<MAX_AXES
actually that's emc2, but i thought it re-used the code from emc1
the planner does not support that many axes I think..
* anonimasu_ is curious about rotary interpolated axis:es
rotary interpolated? like g2? or like a rotary table?
I am not sure why not more than 6. The move from 4 to 6 was to support hexapods. Fun starts with anything greater than 3 as feed rate becomes undefined.
like a rotary table
you have to specify which axis is being fed at that rate
that's one of the limitations of rs274
like g1 x10 f300 r40
(you cant specify which axis)
the rotary axis should be interpolated..
so when x is at 10 the rotary axis should be at 40
Limits of Cartiesan geometry. Beyond that you must know the kinematics of the machine.
nothing fancy about that really
anonimasu: that was just brought up on the developrs list yesterday.. it works just as you would expect
yeah I saw that..
although I didnt have time to read it too much
steve_stallings: could you explain further? cant understand what you mean by limits of cartesian geometry
like, cartesian has xyz, rpw
Rate of travel is the square-root of the sum of the squares of three Cartisian axes.
This equation does not work for more than three axes.
isnt the issue that the motion planner is only capable of moving one object?
the issue on the list?
no, why emc only supports up to 6 axes
you cant move one axis independently though
while moving the rest..
So you must now take into account how axis 4, 5, etc. alter the effect of axis 1, 2, and 3 on feed rate.
yeah, or write a function to disconnect your rotary axis..
or some axis, for continous motion
you should be able to move them independently
fenn: how do you do that in gcode?
like, axes 1,2,3 are synchronized, axes 4,5,6 are synchronized
anonimasu: you cant? i dont know.. i hate gcode
To do this, the planner must understand the specifics of the kinematics of the machine and thus cannot be generalized into the planner.
fenn: you cant..
emc shouldn't be designed around the limitations of gcode language
NO - all 6 axes are interpolated at the same time without problem. Feedrate is the issue.
not really the problem
or is it?
the trouble is more that there's no code to run a axis continously
small group thismorning
dave-e: lively discussion
anonimasu: i think if the motion planner had a separate thread for each separate object in 3d space, you could do so much more
The 6 axis limitation could probably be removed without too much difficulty, as long as you decide how to control feed rate.
ini file variables?
specify which axes are synchronized to which other axes
fenn: define what you mean...
anonimasu: one object is the table, another object is the spindle, another is the tool carousel
that's not parts of the tp.
EMC doesn't know or care about objects.
well it should
it should not..
well if you cant keep your spindle spinning unless the table is moving, you have a problem
maybe have the tp threaded and have the functionality to remove interpolation from one axis
all axes are part of the same "world view" as Ray likes to say
ah but you can
spindle is independent
dave-e: only because it's hard-coded
dave-e: look at the latest thread onthe developer list.. some guy has a rotary "spindle" axis on a cylindrical grinder machine
he wants to move it at a constant rate, independant from the other axes
S2000 M3 will get you a spindle until you do an m5
roughly constant rate could come from a reference supply and a switch
unless you put load on the motor
EMC is concerned with co-ordinated motion. If the motion is constant or not co-ordinated, then it is a PLC or speed controller function the EMC need only turn on and off.
or if you want to coordinate two separate sets of motion
* anonimasu_ nods
if all he needs it to turn ...
fenn: that means interpolatedmotion..
coordination is a different matter
like say i want to do threading at the same time as turning
Two sets are just more axes, EMC need not be concerned with sets.
fenn: multi tool machines?
yeah, but there are many more applications
well, that wouldnt still be a big deal..
how are you going to control it from one interpreter..
so many of these things are simply in concept but quite difficult to implement
oh, the trouble I see is how to control it..
i think it might be easier just because everything would be generalized
so there's less code overall
One classic example- my Hardinge Superslant has two turrets. They run two Fanuc 6T controllers. They use sync-events/handoff to stay co-ordinated.
generalization is difficult to implement and keep it simple
that's how I'd do it..
run 2 emc's with some form of coordination..
maybe n ot might be the easy way
fenn: running 2 controllers?
controller 1 says "go" controller 2 says "stop"
against days of coding
fenn: one master one slave..
well, i wouldn't want to use it
* anonimasu_ sighs
Controller two runs whatever until it needs to be in sync. Then it waits for an event from controller one.
fenn: so you would spawn up multiple tp threads? threads
I know a little about two turret lathes
The two controller concept does not require two computer, but it would take some planning of threads.
the trouble I can see is how you write your code.. for it
the computers for the two turrets hand control off to each other with M codes
dan_falck: how do you do stuff at the same time then? like prep the tool carousel and cut a part
you don't do stuff at exactly the same time
the tool carousel isnt inside the tp..
with these machines
it's aux stuff..
plc's/sub threads of the controller
but what if you had two sets of NC motion
they just need to avoid hitting turrets together
like a robot arm and a milling machine working together
that does not mean that a toolchanger is interpolated..
or creating vibration problems while doing finish cuts
The G code programmer must know if it is safe for the two controllers to run at the same time. They can and frequently do on a lathe, just stopping where inteference would occur.
anonimasu: ok a toolchanger was a bad example
fenn: as bad as anything..
so the guy doing the programming has to do the syncing
fenn: it's STOP/WAIT/GO signals
the robot waits for the mill to finish..
yea, it's hair raising to proof out a new program
what's so bad about the tp knowing the kinematics, anyway?
so it knows if the robot arm is gonna crash into the milling spindle
actually that's not what the tp does
the tp plans the motion..
it needs to know if the robot arm is within its acceleration limits
that's dependent on the robot arm controller..
nothing wrong with TP knowing kinematics, that is the basis of how a hexapod works, it is just more complicated than a plain 3 axis machine
fenn: it's also aux stuff..
acceleration in meters/s/s at the tip of the robot arm
anonimasu: i'm talking about using emc to control both a robot arm and a milling machine simultaneously
robot arm, even two axis, may need kinematics if the motion axes are not Cartesian
all machines need kinematics
yes, but the case of a three axis mill uses what EMC calls trivial kinematics
some machines, like most of them, use identity kinematics, where it's practically the same
it's not a big deal to go from cartesian to machine space.. just multiply it by a 6x6 matrix
at least its not a big deal in non-realtime
treat this as a cell...with multiple emc's?
re: feedrate, I see feedrate as similar to threaded cutting, but the axis that everything is synchronized to (the controller axis if you will) is no longer the spindle, but a linear axis
dave-e: you can't synchronize motion in realtime between two different emc's
but you can do the state machine thing
how's that work?
feedrate is not max velocity (as in axes limited) but rather based on process, vector sum may exceed recommeded cutting rate
pause, go, etc
more like a robot as a tool changer
or part mover
steve_stallings: yes, that's what i'm saying.. but "recommended cutting rate" is in feet/s instead of surface feet/s
for feedrate vs threading
dammit.. people throw extra axes on mills all the time and dont worry about it
i just wanna throw more axes on a hexapod and not worry about it
fenn: extra axis:es isnt the same as a separate machine
can you really consider a robot a separate axis?
feedrate control for 4 or more axis machines is usually handled by the CAM software, which is made aware of the machine and tool kinematics, then machine runs in inverse time mode (this move should take xx time to complete)
anonimasu: yeah if it's holding the workpiece!
steve_stallings: just wondering, does emc do inverse time mode?
i believe so
i dont see the need for the separation of CAM and CNC controllers any more, in the era of PC-based CNC controllers
that's what STEP is all about
the cam programs are highly specialized on what they do..
the problem is CAD programs can't seem to use a standard file format like the rest of the software world
so a CAM program is really just a specialized CAD file interpreter
the actual toolpath algorithms are sorta an afterthought
Yes, EMC supports G93 inverse time mode for feed rate, G94 is regular mode.
it's not all that simple
it really should be the cad program's job to output a standard format
STL was a good idea but not for CNC machines
only geometry, no process definition like "is this hole drilled? or bored"
not to mention feedrates and such
but it's stupid to have the cam program try to guess at what feedrates to use, since that depends on your machine tool
STEP still does not solve the multi-axis problems since it is intended to be generalized and does not know the machine kinematics.
I dont belive in step..
right, you should only have to program in the kinematics in one place
really, what happens when you want to do HSM with it..
and special cases..
i dont think step is the solution, but that's where things are going for some reason
hsm machines can have special scenarios for a piece of a part with delicate detail..
or well, any part for that matter
i'd rather there be a standard "cad" format that contains part geometry and design data like material type, tolerances etc
Funny that we are accepting of CAM needing to know machine/process limits in order to set feed rate, but find CAM handling kinematics strange.
yeah, that would be good
and the cnc controller uses that data to generate a toolpath, based on the kinematics and "machine tool info"
I think step does have to tolerance data
yeah CAM should be a part of the cnc controller
not part of the cad package
but there is a reason for having custom posts for cam programs
just to hande a controller/machine
fenn: have you heard the the saying "let the professionals take care of it"
as far as I can see this issue is the same
anonimasu: have you heard the saying "trust us, we're from the government"
fenn: well, I've yet to see any opensouce algorithms for making toolpaths
i know it sucks
i'm gonna write one for brlcad
fenn: there are people investing lots of money to make cam programs good at making toolpaths..
so they say
i've seen lots of university research too
fenn: why is there no opensource algorithms?
but for some reason the uni's never publish any fucking source code!!!
who's ass are they kissing?
no...it is very much who is going to fund my next grant?
this whole world is gonna blow up soon anyway
everyone's in bed with each other, and if you're not in the loop you've got nothing
If the universities are funded by US government grants, the code is public domain. You just have to fight to get them to obey the law.
fenn: the reason is that it's not that easy is because generating toolpaths isnt that simple..
well, easy in 2D but not in 3D
no it's just as hard in 2d or 3d
but we have really low expectations for 2d toolpaths
There are a couple of relatively cheap ($150) programs for CAM that have been developed for Windows and targeted at hobby/home shop use. Sheetcam is 2.5D and MeshCAM is 3D, but both drive only 3 axis machines.
how well do they work?
Both have active user communities on Yahoo and the comments are fairly positive. The programmers listen to bugs and update regularly for free.
'course they are Windoze
and I'm trying to go Windoze free
i'm not going to taint the purity of my system :)
so now i gotta write a cam program :(
* anonimasu_ nods
i gotta learn how to use brlcad first
its got a lot of quirks
openGL shading is still experimental!
designed by programmers for programmers
why, CAM program should be free standing and import common data formats
steve_stallings: CAM has to know machine parameters somehow
steve_stallings: where better to get that data than from the CNC controller?
BRLCAD supplies machine parameters, or am I crossing messages in my mind?
brlcad is mainly for simulating nuclear bomb effects on tanks, stuff like that
neutron penetration, ballistic simulations, radar reflection
ballistics research labs.
but it's got a good suite of modeling tools and imports a wide variety of formats
OK, tying to be clearer now, how is BRLCAD related to your CAM proposal
brlcad is a good open-source CAD platform that runs on linux (and windows too?)
there is no open source CAM program
so.. in order to avoid having to write code to understand/manipulate 3d geometry and lots of file formats, i'm going to use BRLCAD as the base
so you only have to learn the BRLCAD internal data format?
it's a short term solution.. not in any way related to how things ought to be
but right now its either do that or set up a new computer and waste hundreds of hours dealing with windows
i really hate windows
actually XP works pretty well if you keep it isolated f rom the net
well, w2k is good..
actually there are some open source CAM programs but none that do more advanced stuff like waterline or >3 axis
fair enough, I would like to love Linux, but WHICH Linux? (not seeking answer about favorites, just exasperated at variety)
I'm fighting Debian right now...
I like linux, but I cant say I love it..
steve_stallings: they are all the same once you get the hang of apt-get
it's a messy operating system..
downloaded the net install. 180 Mb and installed...
nice and fast but it doesn't come with X windows component
so I get to ftp X windows and try and install.
can't you apt-get install xorg ?
not fun ... 4.5.0 doesn't know my optical mouse
um, what kind of optical?
tried it both din and usb
ps/2 and usb
set the input to /dev/input/mouse
usb mouses should end up there :)
dave-e: dosent exist if you dont make it first :)
or if the dist comes with the symlink in place
that's one good thing about udev
I'll try dev/input/mouse
look in /dev/input first
I can have a look in a hour if you want what I am using
need to sleep a bit
only have to swap disks
course it on this machine
off debian.org...well the se mirror ... I did a desktop install...and thought that would be me an X windows...apparently not
did you try "startx" ?
and X86Free -autoconfig
and XF86Free -configure
if that's there it means you have X installed
look in /var/log/messages and see if there's any messages about X messing up
it tries...but stalls on the window with the X and one has to kill it and look at the log file
mouse is one problem...
but it also looks for dev/hb0 which doesn't exist
cannot find anything in the config file to change to fix that
sounds like someone did a poor job packaging X
I tried the straight bdi-4.20 install but the display was awful....jerky mouse... at 1280 x 1024 which my 64 Mb card should handle easily
I'll sort it out somehow...it just isn't easy
enough complaining ... back to work
thanks for the try :-)
some day, i'm going to fly a spaceship with a general-purpose piece of machine control software similar to emc
(but emc isn't all that general purpose enough yet)
* staggerlyTom sez Hiya Hiya Hiya
right now i'm writing to some uni professors whose research is developing CAM algorithms
asking them to "do the right thing" and release their source and pre-proofs of the papers under the gpl
it will be interesting to see what happens
has anybody seen alex_joni today?
2 hours ago.. said he'd be back
I have a set up question
Is there any way to set up EMC to skew an axis that isn't square to another axis?
you're looking for "trapezoidal" kins? :)
I have a mill with a big problem. Y is not square to X
I will ultimately get a better mill, but wonder if there's a way of faking it
I think the only answer is to write a kinematics module that corrects for the geometry errors
since you need X motion to get linear Y motion
it might be possible to do evil things with HAL to get a similar effect, but I don't want to think about them :)
i thought emc had some sort of axis compensation built in already
for calibrating leadscrews
maybe I can do some screwy stuff in CAM
that's non-uniformity along a single axis, not having one axis dependent on another
where is that done?
you could just use corel draw to shift the envelope of a DXF file, for 2-D stuff
there's a linearity file - I don't remember what it's called
I've got some CAD/CAM software that will probably help me a bit
or re scrape the ways :(
Option 2 will likely provide the best results
depends on how out of square it is
well - of course it was bound to happen. I start talking on IRC, and my wife decides it's time to do some activity together
It's a pos mill/drill and this morning I checked it with a square that had been surface ground true -.020" in 6"
at least it's not a "Honey-Do" thing
I had never checked it before...
you aren't gonna be able to scrap 0.020
SWPadnos Suggest that she get on IRC too! LOL
I need to scrap it
you could scrap it ;-)
use it as a drill press
hmmm - I think I'll pass on that - thanks anyway :D
I need a real knee mill anyway
you could give it to me
fenn: you got the lathe together and running yet?
I'll use it as a drill press
i'm slow :(
does Roland have any Boss mills left?
SWPadnos is now known as SWP_Away
SWPadnos is now known as SWP_Ball_and_Chain
fenn: not as slow as me.. I bought a South Bend 13" lathe about 2 yrs ago
still haven't run it
SWP is living dangerously, better hope SWMBO doesn't look at the screen
Anyone run steppers by chance?
steppers? what's that?
Jymmm: what kind of couplers does that router table have?
does anybody know what i can do when the emc.run script can't find the tcl module msgcat ??
when i search on the debian homepage for msgcat it says that its includet in tclx8.3 that is installed on my system
/usr/bin/msgcat on my page
dan_falck it's hard to explain... an aluminum tube with slots cut in it
ok. helical coupler
those are nice
If I spin the steppers sitting on the bench - no problem. Pick em up off the becnh - they stall. Tight grip in my hand - no problem, loosen my grip - they stall.
fenn: /usr/bin/msgcat is a file, right ?
resonance- your hand is dampening them
Jeff (from cylotex) says this is resonance, but this happens on MotorA when slow, and on MotorB when fast.
Imperator_: yes it's a binary16KB
both motors are the same?
hm, i have that file but it don't find it
I bought everything as a kit from Xylotex
power down, swap motors (where they connect to the driver board)
see if the problem remains with the motor, or with the driver channel
Well, no good now. I think the board got fried.
Imperator_: tcl is a GUI thing i think, try switching display to keystick in the ini file and see if it still does it
do i have to tel tcl which modules it has
bbl. going to look at POS mill again...
dan_falck: use a bigger hammer next time
What I need to know is this common with steppers? Would the RPM make a difference? How would you compare a good/bad stepper in respect to resonance?
fenn: keystick ?? don't know what this should be
I#m on emc2
oh, maybe it's not in emc2
is that a emc1 thing ?
it is definatly a tcl problem
then we will know if it is in the emc.run script or in the gui
i need to fix my own emc2 install before i try to do debugging
it stops when starting the display now !! DUMMY DISPLAY module, press enter to continue
right that's what dummy is supposed to do
so i guess that means the problem is in the gui.. were you using tkemc?
i have manualy installed tcl because after the installiation of BDI2.18 it was broken or not there or something
do i have to configure tcl ?
* fenn is getting out of his range of knowledge
ah! and another thing about feedrate and synchronized axes, is EDM
you synch the feedrate to the voltage across the spark gap, not to the position of any axis
I would use feedrate override, modify it based on voltage
that will slow down or speed up all axes together
what, by hand? a hundred times a second?
sometimes you have to back up, too
a) not by hand... feedrate override can be set by a NML msg, so write code that reads voltage and sends the NML
b) yeah, that;s a problem
on a EDM you have to go all the time backwards and forewards
can that be programmed in to the part program?
or does the machine have to sense when it needs to back up?
the machine have to sense it
it depends on the current, i think
we have a lot of wire EDMs and normal EDMs is the cellar at my new job, i can ask if you want to know something
i'll fall off that bridge when I come to it :)
How often could EMC2 via Ha
gawd.. How often could EMC2 via HAL , ipdate the analog Feedrate Override? 10ms? 5ms?
the HAL part could read an analog input (gap volts) and act on it in realtime, 1mS or better
think it was something like 1 ms
but if you go thru NML, that is user space and slower
wow, thats fast, roundtrip sound like possible 5mS
the problem with NML is that in user space there are no guarantees
could be mostly under 2mS, but sometimes 100mS
why would you do that in userspace though?
for the EDM thing I would try to keep it entirely in realtime
there's 3 basic forms of edm control .. for hole, sink, and wire, each needs more sophistacted control
and sink only needs to modify fwd velocity from programmed rate to 0 THEN able to runaway on interrupt
that would require modifying the motion controller to allow access to the feedrate override value from realtime, but the ability to back up would require significant changes to the motion controller anyway
so sink does not need to reverse ( just run away , sky blue safe haven not on path)
there's no easy solution to backing up with the current interpreter
is "run away" always just "pull up on Z"?
run away for sink is... return to top of pyramid, and path is base of pyramid
sink needs side clearance so the electrode is smaller than the cavity, right?
and it moves around in a little circle?
oh, you start at top (point) and move down in increasing orbits as you get deeper?
yep, pyramid.. you begin work to cut a cubical hole..
your position is like being at top of an egytian pyramid,
your approach is anyweay you like towards base of pyramid
your safe retreat is anyway back to peak, away from stock
this is ONE orbit only
I call the top of pyramid the 'roughing point', where the heavy power cut ends
from the 'roughin point' all excursions take place
what's it called when you do undercuts, like a t slot?
or is that "not done"
you don't do undercuts with a sink EDM
I do, the path is still triangular in section, with safe place at ctr of bore
I cut the top face, side wall, bottom face, running away to bore center
the tool is a flange smaller than the bore to be 'grooved'
where can pix be sent?
steve can post them for you.. i don't know how to use the dropbox
or i can put them on my page.. but it's not very permanent
i need to get that new wiki up so we can post pictures there
but nobody else seems very thrilled about it :)
I'd like it, I just glommed your whole site :-)
fenn.freeshell.org has a 50mb/day limit, which is pretty low when you think about it
fenn.dyndns.org is a dsl line, i'm typing on it right now
Didnt mean to steal all your wind, but I wanted the photos of the 'encoder'
the magnet thingy?
yeah, just ran to your new site, kewl! how can we dog ear dead old misleading docs?
I really get po'd when i find the stuff I'm doggedly following is outofdate
the wiki on fenn.dyndns.org is not "official" yet
it's mostly just a mirror of the other one, running different wiki software
oh, there really is a 'them' huh. who blesses it?
i tried to organize the docs into EMC1 and EMC2, but it's not working so well
yeak, symlinks with annotaion... neccesary
well, the "them" is people who have a stable server to put the wiki on
also its kinda dumb to have two wikis for one project
and... you could work on the original wiki??
no, it's a pain in the ass
i'm not going to work on it
ok, nuf sed
i need to ask a Q about HAL, but need a bit to compose it..
i've got an inadequate answer, after some hemming and hawing
Tom: ask when ready
i get: unresolved symbol adeos_get_sysinfo when I su, then cd /usr/local/src/emc2, then 'scripts/realtime start'
what distro are you running?
the realtime script is tested on BDI-4.20 and some other BDIs
John, where do I check distro version #?
did you intstall it?
now I have to guess cuz I cant just look it up. I Think I compiled it.
you didn't compile the distro?!
not the linux distro, the emc2
are you using a BDI, or Redhat, or vanilla Debian, or Suse, etc?
Debian with emc1, (emc2 in sep dir according to some notes i found)
Debian as in you got it from Debian and did the realtime OS install yourself?
or Debian as in Pauls BDI-4.20, which is Debian with a custom installer and RTAI-3.1 already installed?
uh, the old Live BDI cd ( prior to fest), cuz ther'es a 'install to hd' icon on deb desktop
how do I identify it?
ok, that is BDI-Live
thats the one
probably BDI-Live rc46
(dunno how to check that... stand by one sec)
i see user Morphix, so I guess Paul C's Deb based live bdi cd from Names 2004
do "uname -r"
BDI-Live rc46 will say "2.4.25-adeos"
2.4.25-adeos, thats it
John, dont fix old stuff, I got a new partition waiting
what should I put on it to test the gcode/mcode/wait/ gcode idea?
BDI-Live is only slighty old
I can reproduce the problem here, need to investigate
if you want to use a current setup (highly recommended) get BDI-4.20
that is a nice Debian with pre-installed RTAI-3.1
anything interesting today?
BDI4-20 it is , thanks John
hey John, around?
jmkasunich: hi there
seen you were asking...
yeah... still fighting with the 2.6.12 stuff
just now I was looking into a different prog
the realtime script doesn't work on BDI-Live
seems it can't find the adeos module, it's not in the same directory as all the other rtai modules
maybe it's compiled into the kernel?
I used to have that configuration
adeos compiled in, the rest as modules
if I manually load adeos, then there is another problem on unload... that damned rtai_ksched -> rtai_up symlink
ohh.. bugger that :(
not on Live rc46, adeos is a module
tried anything on 2.6.12 ?
the symlink prob... we have code in the script to find the true name of rtai_up, and load it using that... however for unload you need to use rtai_ksched cause that's how it's listed in lsmod
yeah, 2.6.12 stuff in a minute
anyway, I think on other systems rtai_up appears as rtai_up in lsmod, and that's how you delete it
hmm... I have a thought...
why not parse lsmod ?
I'm tempted to modify the unload list to contain both names
and remove the modules are there
check for hal_lib for example
only works if they are parsable (IOW, something that you can reliably parse on, like RTAI*)
might work tho
lemme check something on the farm 4.20 slot
(this 2.6.12 stuff has my main box irreversibly fscked up)
you can always go back
apt-get remove the newly installed modules
and apt-cdrom add
apt-get install the 2.6.10 stuff from the cd
John, I know how to make realtime run
I have to run EMC from start menu ( as user)
then stop it
then su and cd to /usr/local/src/emc2, then 'scripts/raltime start' is ok
yeah Tom, there are ways around it... but it's still busted as far as I'm concerned
alex: gotta find my cdrom
btw, Pauls repository at 188.8.131.52 seems to be down...
jmkasunich: the debs are on the net aswell
check the wiki
installing on debian
they are listed there
you can ping the box, but apt-get update hangs on that url
well the 2.6.10 are somewhere else, iirc
anyway... just tried realtime start on the farm 4.20 slot (expecting no probs, just wanted to see how rtai_ksched appeared in lsmod), and I got a kernel panic
things are just going to hell around here
probably smthg strange...
rebooted the slot, gonna try again ;-)
so I should put that url in my sources.list instead of the 81.100... one
yup.. if you want to go back to 2.6.10
well I'm gonna want to switch back and forth for a while
I guess since the 81.100 ones is down, if I switch back to 10, I won't be able to go to 12 again?
the weasiest might be swapping hdd's :)
not that bad
jmkasunich: check /var/cache/apt
can't remember the exact path
the modules and kernels coexist, and I can pick kernels at boot timne
but the debs should be somewhere around there
how do you specify which directory a kernel looks in for modules at boot time?
its only the rtai-dev package that overrites things
the kernel just knows
probably uses uname or something
rtai-3.2 (that was on 2.6.10) was in /usr/lib/realtime iirc
but my uname is always buggered
and the magma is in /usr/realtime
so you should be able to make them coexist...
uname -r gives 2.6.10 but it shuold be 2.6.10-adeos, right?
but most easy.. (what I would do).. boot 2.6.10
on bdi-4.20, it is 2.6.10-adeos
and get rtai-3.2-test2 (I tried that one, not sure about test3)
yeah.. i compiled 2.6.10 and added adeos patches and stuff.. i dont know why the uname didnt change
and compiles /install it to somewhere else than /usr/realtime
jmkasunich: follow me so far?
did you change EXTRAVERSION in the kernel makefile?
that's why uname is fscked
ok, back to alex
jmkasunich: basicly .. you have a moer or less working 2.6.12 - magma combo
more or less because of the shm bugs
right now I'm running 2.6.10, and trying to get a RTAI version of hello.world to work
then reboot to 2.6.12 and see the bug
still have the 2.6.10 rtai modules?
spent 5 fscking hours last night fighting with kbuild tho, for a dumb typo
in /lib/modules/2.6.10-adeos/rtai ?
then you should be set
just don't have a 2.6.10 version of rtai-config
ahh.. and ./configure fails for emc2 .. right?
(or if I do, it's hiding)
./configure finds the 2.6.12 one... so I can only compile for 2.6.12
you "may" tweak Makefile.inc
/usr/lib/realtime was deleted
well. then it seems that rtai-config (for 2.6.10) is gone
but you should be able to compile against magma includes
I was afraid of this
and use the 2.6.10 modules during runtime
* jmkasunich goes to retry scripts/realtime start on the farm 4.20 slot
should do a make clean and make next time, just in case something is borked
realtime start only installs rtapi and hal_lib
it loads all the rtai modules too
what I was getting at.. it's code that hasn't been changedd in a while
btw.. I got the STG today
you would think...
yeah. I woudl think..
isn't it true?
yeah... and on top of that, I use realtime on this 4.20 box all the time (at least until 2.6.12 fscked it)
yeah.. 4.20 worked ok for me too
oh, and I got a copy of "The RCS Handbook"
* alex_joni goes home...
will be online in half an hour
bleh.. multivariate, differential, parametric, n-manifold sweeps in n-space
you should wash your mouth out with soap after saying things like that
gonna go chug some glucose
charge up my brain
if I make drawings for EMC group, what is useful format?
depends on use
if folks need to edit them, probably somethign like dxf
but for veiwing, pdf or gif is much easier
when I've done docs and wanted drawings, I used easycad (makes dxf files) for the masters, then printed them to pdf files and included the pdf in the final document
for simple drawings, xfig works fine
ok, looking at easycad & tools for dxf. These dwgs must be correct, not just look good ( cad vs art ) no offense.
Tom: easycad is windows only (unfortunately) right now
linux version is coming one of these days, but even then it's not free
it's about $200, and I personally think it's money well spend
but for fast and free, try qcad
fast to get that is, I don't know how fast it runs
staggerlyTom: install vmware + vmtools and you'll find the piece..
well, are there useful viewers? I have Mech Desktop from AutoCad to create files (Win, but files can move)
I don't post dxf files for others to view... too many complications
I usually export as a pdf, anybody can view that
[20:28:48] <cradek> http://wildspark.com/dxfscope/
(I haven't tried it)
note the comments about not all entities work, fonts, etc
so when you give somebody a dxf and they view it, they may not be seeing what you saw in your cad pkg when you created it
cradek, just grabbed, it, will try & report ( once shortcomings are know, avoid 'em :-)
jmkasunich: but that's *always* the case
jmkasunich: even between versions of autocad
I can use what you saw is what you had... preview in viewer before uploading
hence my preference to send them a pdf ;-)
staggerlyTom: what do you want to draw and share?
yeah, pdf is good to look at but you cant get a dim of fit
I want to show parts for the MAZAK and how EDM motion works for Hole & Sink
Jymmm, so what's the plan now for you?
ah, I see.. you wnat them to be able to click and read stuff that wouldn't be available from a paper print
yeah, often I dont put in the dims others need, bnut the math model has it.
where I was aiming to give them a paper print that looks just like my copy
staggerlyTom: there's probably no good free solution.
differet problems, differnet tools
staggerlyTom: (I use autocad but understand most people don't have it)
what about the WEBER on the new BDI 4-20 cd?
fwiw, xfig does let you draw to scale
Ok, so all wont pop for Weber's Synergy , so not a common ground, ok
staggerlyTom: it works good/great :)
looks real nice, but the reason why ACAD isnt a common ground applies to WEBER.
All interested in a nice cad tho, should look at it. They were nice enuf to give us a 30 day trial.
Weber's Synergy system is on the BDI 4-20 cd.
any luck on your fights?
well I'm making progress on my little shm test prog
(basically building the rtapi shmem example, without rtapi, using raw rtai calls)
note that the probs with rtapi may be because it's own internal shmem is getting messed up
was wondering.. did you try different methods of calls to rt_shm_alloc ?
haven't got that far
it's own shm?
yeah, there is a shmem inside rtapi, used to hold data about tasks, allocated shmem blocks, sems, etc
its shared so ulapi (user mode) stuff can access it
will look at that
the info you see when you do cat /proc/rtapi/* comes from there
well.. that's ok
not from what I saw
don't recall the details now, and can't redo it right now
from looking at /proc/rtapi/shm I always saw all the modules there
but maybe the info should be more...
will parallel test today
should list the shmem blocks, and how many kernel and userspace "users" have each block open
I loaded the shmemtast, it showed 1 rt and 0 usr "users:
I loaded the shmemtast, it showed 1 rt and 0 usr "users"
then started shmusr, it showed 1/1
then started another shmusr, it showed 1/2 (and both shmusrs worked)
then I exited from one of the shmusr's
the other one kept working, but ISTR the 1/2 got messed up, it should have gone back to 1/1, but instead I think it looked as if the entire block was no longer in use
starting another shmusr just printed zeros instead of the incrementing count (even tho the original shmusr was still inc'ing)
so probably the rtapi_free call is borked
or whatever the mechanism at freeing the shm is involved
I'm gonna give it a try with raw RTAI
if that works, then serious debugging in rtai_rtapi.c
ok.. let me know
* alex_joni starts reading in rtai_rtapi a bit
hmm.. one thing jumps to my attention
jmkasunich: can you look briefly at smthg?
int rtapi_exit(int module_id)
at the end of the function
just before the next function (rtapi_snprintf)
who asked of x.org was better then x-free?
there's a rtai_free after the release of the mutex
ok alex, looking at it
the mutex is stored in the shmem block
oh.. silly me, you need to release the mutex too
(that is the rtapi internal block)
hmm.. is there a mutex example in rtapi?
maybe the mutex is failing.. and the operations get screwed
most likely mutex failure is lockup, somebody waiting for a mutex that never gets released
no.. what I mean.. maybe the mutex gets released to two modules
not for the kind of testing we're doing... we aren't attempting simultaneous access to anything
these tests would work even without mutexes
I know.. I was thinking at the whole problem
I know you're testing the shm now.
but the problem might not be shm, but mutex
ok, good toi keep a bigger perspective
but I seriously doubt the mutexes
I'm not saying you should abandon testing shm
rebooting to 2.6.12
well... it appears to be rtai_free
so rt_shm_free might work better?
doubt it, rtai_free is #defined to call rt_shm_free
here's what happened:
I have a user program that simply rtai_malloc()'s sizeof(int), then spends 30 seconds incrementing that value once per second
I don't even need to have a kernel side program at all
if I have two instances of the user program, each one increments, so the number goes up by twos (as seen by each program)
start one copy of prog, number starts incing
wait 10 seconds, start second, number increments faster
when first program ends, number slows down
start a third, it begins counting from zero, while the 2nd one is still counting by 1, they're not accessing the same memory...
and when the 2nd one ends, complete lockup
malloc(), malloc(), free(), malloc(), free() leads to the 3rd malloc not pointing to the same memory as the first two, and the 2nd free locking up
something tells me the first free is marking the block as completley unused, instead of remembering that there is another prog using it
I think I'll re-write it to use rt_shm_alloc and _free, see if that does the same
should probably also try the exact same sequence under 2.6.10 again, I'm not sure I did same thing
if you change to rt_shm_alloc try fiddling with different alloc's
instead of USE_VMALLOC?
and her persky lamb
you brits and your livestock... gotta wonder about that
so .. what goes on?
the TP fixed yet?
damn... the only shm functions that aren't labeled internal use only are the heap ones
no, your countryman keeps throwing spanners and disappearing
when you say "throwing spanners" ?
he built a new kernel with 2.6.12 and the bleeding edge of rtai, then when HAL didn't work he posted a bug report at me and disappeared
oh. helpful huh?
a week later I finally have a copy of the kernel to work with (no help from him there), and the problem looks to be internal to RTAI, not even remotely related to HAL
why, exactly, did we need a new kernel again?
to rock ;)
livin' on the edge man!
personally, id settle for a working tp, but hey, im weird
gonna reboot now, want to make sure it still works on the old kernel
today was a good day
I decided to move the chiller from on top of the offices
to near the back door
so all the hot air can blow straight out
works a treat :)
sucks cool air in the front door too
moving a 1t chiller on your own is fun
and constructed a big air duct to catch the hot air from the fans and send it sideways
i just installed BDI 4-20, uname -r sez 2.6.10-adeos
and now theres a different problem with scripts/realtime, it doesnt exist
no point asking me ... i dont get involved with the BDI stuff
staggerlyTom: there's no scripts/realtime on emc1
and on the BDI there's a emc1 (kinda=
it's a bit modified though
emc1? why is it emc1, ithought this'd be 2
I was just advised tro nstall this to solve the proble, I must've misunderstood. Do I need to load src & build?
staggerlyTom: in order to get emc2 follow, http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?BDI-4_Compile_EMC2
that should cover it
thanx alex, doin it now
* alex_joni prepares for bed
ciao Alex, see you in Stuttgart
take pictures for us! :)
i wanna see your 30 kw hexapod
* jmkasunich is baaaack
(several reboots later)
i'm having this issue with hal not loading into shm
i remember the fix has something to do with doing scripts/realtime start instead of just starting hal
it's fedora 3
yeah, scripts/realtime loads the RTAI modules (there are about 6 of them - without them you don't have a RTOS)
the chain of dependencies is: adeos which supports the RTOS stuff (rtai_xxx) which supports rtapi (our generic RTOS interface) which supports hal_lib
if you are running rtlinux or something the first couple steps are different
i have these loaded in lsmod: rtai_math, motmod, hal_lib, rtapi, rtai_sem, rtai_shm, rtai_fifos, rtai_up, rtai_hal
if rtapi and hal_lib are loaded, then you should be OK
but then i get RTAPI: ERROR: could not open shared memory
you're not using RTAI-magma are you?
it worked before, but now it doesn't
i dont know what happened
3.1 is known to work, that's on the BDI-4.20
magma is known to not work, I'm writing a bug report for the rtai list now (shared memory problem)
jopingo is having the same problem
3.2 is unknown
does realtime stop ; realtime start fix it?
emc2# scripts/realtime start
insmod: can't read 'adeos': No such file or directory
insmod: can't read 'adeos': No such file or directory
insmod: error inserting '/home/blipkowi/emc2/rtlib/rtapi.ko': -1 File exists
insmod: error inserting '/home/blipkowi/emc2/rtlib/hal_lib.ko': -1 File exists
RTAPI: ERROR: could not open shared memory
do you have adeos built into your kernel?
HAL: ERROR: rtapi init failed
well, i thought i did
i dont know how to check
ok, hang on
you are running realtime, then you are running emc.run?
emc.run calls realtime
that explains the "file exists" messages, realtime already loaded the modules, then emc.run called it again to reload them
run realtime stop to clean up whatever mess is laying around
then run realtime start, followed by lsmod
is adeos a module? or is it suposed to be a module and i compiled it in non-standardly?
if hal_lib and rtapi are loaded, then everything else (including adeos) must be there
it can be compiled as a module or as a built-in
the BDIs have it as a module
if you have it built in, you'll get a message when the script attempts to load it, but that's ok,
since it's already there, the other things that depend on it will still load
OTOH, if it's not built in, and you are getting the message because the script can't find the module (maybe it's in a strange place) then nothing else will be able to load
all the rtai modules are loaded
anyway, if everything loads, then just use halcmd and/or /proc/rtapi to take a look around
bin/halcmd show all
halcmd errors out, could not open shared memory
ok, something is fscked - it seems suspiciously similar to the problem I just found
i dont really know what to look for in /proc/rtapi
it just prints some statistics...
HAL_LIB is the only module loaded
there should be one shared memory block listed
shm size is 65500
that;s HALs block
halcmd should be able to connect to it, but it can't... something is busted
Timer status = Stopped
is that normal?
yes, you haven't attemted to start a real time thread yet
how do i turn rtapi debug messages on?
echo 4 >/proc/rtapi/debug
hey look at this
note that the messages mostly go to the kernel log
Jul 17 17:03:32 snacker kernel: ***** WARNING: GLOBAL HEAP NEITHER SHARABLE NOR
USABLE FROM USER SPACE (use the vmalloc option for RTAI malloc) *****
yeah, that happens normally
including on my RTAI 3.1 / 2.6.10 system that works fine
(the code does indeed use the vmalloc option that they are recommending)
btw, a few lines up in the log is probably a line like "RTAI 3.2 over adeos.stuff.more"
that tells you your rtai version
RTAI[hal]: 3.2 mounted over Adeos 2.6r9/x86
3.1 mounted over Adeos 2.6r9c5/x86. works here
magma mounted over Adeos 2.6r11c2/x86. fails
magma is 3.3 right?
Yours is closer to the working one than the busted one, so I don't know if theyre related or not
I think so (3.3/magma)
i swear i had it working before
i dont remember if it was working after i rebooted though
I wish I could help more
but I have enough trouble keeping up with BDI related problems
there's simply too many variables when you build your own RT kernel
I can send you the test program for the problem I've found
jmkasunich come on... get REAL =)
/msg me your email
* jmkasunich is not in the mood
this is the one with two rtai users, and it counts, removes one user, keeps counting
maybe i'll call it a day
heh, now heres a thing ...
my wifes sister just topped up the oil in her car ...
and it only needed a teeny weeny bit ...
seems she put it in the brake fluid resevoir :(
jmkasunich is now known as jmk_away
is any one here awake?
we are all sleeping
where is country code 0035?
i am new to the whole cnc thing. and was wondering if u could just help me out with 1 or 2 q's?
i am wanting to cut wood. but i am needing angles cut into it
robin_sz: what's up?
I'd use a chamfered mill..
need the wood cut something like that
that's the quick solution
how many axces would the cnc machine need to cut that?
with an ang;ed cutter, 2
with a manual z
but 3 is perferred for any serious work
I've had 2 and one manual on my machine a while back..
it was so painfully slow to make stuff
ok... now what exactly is a angled cutter?
a cutter with the edges \/
a V cutter
but what happens if it's a stupid angle like 22.5 degree's?
then you buy a stupid angled cutter :)
i make speaker boxes... and some of the angles are totaly wierd
you can always tilt the cutting area..
and use a straight cutter
or the spindle, but that's slower..
what about another axis on the machine?
or wouldnt that solve it?
it's troublesome to generate the toolpaths
my friend makes speakers
you are better off using space blocks to machine the angles..
its easiest to just hack out the panels on the CNC router
instead of doing multiple passes with a ball cutter
weird angles are done on a standard router table
you can make a fixture that allows you to adjust the height
i am guessing it just moves the z axis up while moving the y axis?
make a router table with a tilting router, or use an edge planer
you can cut angles like that but it's painstakingly time consuming..
tilting router <---- wouldnt that be a extra axis?
no .. not CNC
its not worth trying to CNC it
unless you go 5 axis ... and have say, 50K $ to spend
just make a standard router tabel witha tilting router
let the CNC hack the panels out
tilt and cut the angles
and then if you need on or two edges with crazy weird angles, pass them through the router table
ok... so u recon 2 machines would be best for this?
* robin_sz nods
nah, you can use one
well, one CNC, one manual router
robin_sz: if hes on a budget that's not really a solution
realy apresiate the input... so difficult to figure out what is needed when you have actualy never seen one :)
anonimasu: more useful and cheaper than trying tilting head on the cnc
a manual router table is always handy
robin_sz: yeah, but the more usage on the cnc the better :)
well, my mate gearoge does it with a second op on a manual router table
robin_sz: unless he's in production and wants to do it at the same time..
well was hoping to have the whole thing automated... but hey... i goto do what i goto do :)
he makes .. oh I duno, 10,000 cabs a year maybe
well, he probably would know >(
actually, hes moved the whole factory to Poland ...
I'd use a V type cutter..
and mill the edges in one setup.. directly..
robin_sz... live in africa... dont get cheaper labor than here :)
if it was in production..
if I had 2 or 3 known angles, and a toolchanger, then yes
anonimasu... where can i find some v shaped cutters? know a site off hand?
hm, not really I am not into wood routers
if I had many weird angles, then I'd get creative :)
and u robin_sz?
he's a laser freak ;)
i was also realy interested in that till some one told me i would need a 400watt laser to cut the wood :)
yeah, I have a ferranti laser that cuts 12mm plywood .. thats 450W
i think my gf is going to be getting herself a small desktop laser cutter
robin_sz... freek... so u are telling me i would need something over 450w to cut 18mm ply?
18mm .. hmmm ... 400, 500 would do it I guess... 600 would be a sure thing
* anonimasu_ wonders where les is when you need him
i am guessing every one's machines run off ballscrews?
almost some use acme screws
big machines always use racks
but dont u get alot of play using gears?
i always thought gear's where evil
robin_sz... so your's is using rack's?
helical racks and pinions .. almost zero backlash
my commercial machine does, yes
gets 0.1mm accuracy
.1mm arent that great when u consider a ballscrew only has .001mm
sure they do
oh, heh considering the size of the machine it's great
what machine u ppl using? would just like to take a look at what u's are running
ballscrews on big machines dont do 0.001mm .. thats tiny ballscrews in tiny actuators
and ballscrews wont move fast
u get big ballscrews... 50mm dia
you cant get rapid motion out of ballscrews .. especially long ones .. as they start to flap around
whip & snap ;)
and wood routers dont need 0.001mm accuracy anyway
guy's i am not trying to cause shit... i am just trying to find answers to my questions. hence the reason i am acguing abit
Jacky^: how are things going?
Yuga: oh, dont worry about it :9
Yuga: well, ballscrews cost about 5 times what rack and pinion costs
robin_sz... ya.. that's probibly y i thought it was better :)
oh it is "better"
but ... it might be more accurate than you need
and it might have drawbacks that you dont need either
like rapid motion
dust + ballscrews dosent work too well..
robin_sz... what is considerd rapid motion? 200ipm?
no .. thats SLOOOOOOOW
well les wants 1000 ipm
yeah, but he's not like everyone else ;)
ok... so what is the average?
that's pretty insane :D
anonimasu_: i'm fighting to compile ffmpeg with x264 support :\
600ipm on routers
Jacky^: good luck
600ipm.. shit that's pretty fast
what's the diff's between a Helical Rack and just a plain old rack? and when i say "rack" i arent talking about your gf's :)
helical gears are //// and normal ones are ||||
well, you know what helical gears are?
more contact area
robin_sz... not a clue :)
ok... so one is ground at a angle?