sweet, its a bonafied dual boot box, emc with a gig of ram to piddle with
i guess i should rebuild my motor controler and hook it up and see how it does
maybe in a few hours ill do that
Has there been any discussion on changing the interpreter to "round" concave corners when tool comp is on?
And what exactly is "Dynamic Cutter Radius Compensation" as compared to "Cutter Radius Compensation"
whats with the two cradeks?
If one is good, two must be better.
I wouldn;t go that far
Does 'redd' mean anything to anyone here?
(RTLinux Ethernet Device Drivers)
since RTLinux went commercial, we've focused more on RTAI
Damn, I never remember which it is that we use (and the wiki didn't help).
rtapi actually has compatibility layers for both
but the RTLinux is a very old version
Anyway, REDD lets you use ethernet from within RTLinux. IMHO, that would be a neat way to support external devices.
RTNET does the same for RTAI
USB is too slow, but a dedicated ethernet card is about the same price as an LPT card and is a lot faster.
Cradek's arduino efforts made me think of this (again).
What exactly is "Dynamic Cutter Radius Compensation" as compared to "Cutter Radius Compensation"?
[00:44:43] <Ziegler> http://www.linuxcnc.org/docs/2.2/html/gcode_main.html#sub:G40,-G41,-G42:
so dynamic uses a g-code specified diameter, and "regular" uses a diameter from the tool table?
I wonder how image-to-gcode would do on this: http://jmkasunich.com/pics/ddg72coa.jpg
Ziegler: thats how I read that page
probably not bad
theres an awful lot of detail, and not that many pixels
convert it to gray scale to see what will be raised and what will be embossed
I might edit it with gimp a bit to darken certain areas
make the rope braid the highest point
and the lettering
gray out the white insides a bit
do you guys know if a module such as nvidia_agp can affect RTAI?
it wouldn't surprize me
nvidia's proprietary driver and realtime don't usually get along
even if I'm using vesa drivers in Xorg?
you asked if the nvidia module could hurt, I said yes
now you say you aren't using it
so why ask?
There's a distinction between Xorg driver and a kernel module isn't it?
a driver might be a kernel module (might not, I'm not sure)
but a kernel module can be many other things as well
I though nvidia_agp originated from the kernel developpers, not Nvidia
I have no clue
basically, _any_ kernel module CAN affect RTAI
most don't, but your guess is as good as mine as to which ones will
if you have bad latency when its loaded, and good latency when its not, then you know
yeah that's an easy test
but my issue here is different
I have no clue what that module even does
I see Axis stalling for 1 second. I think the parallel port pins are freezing too at the same time
think based on what?
4 channels scope
real scope, not halscope?
I will output a siggen on one of the output pin of hal_parport
its a lot easier to stall a user space program like axis than to stall realtime stuff
overtime tonite and i got work at 6 am tomorow in queens!
don't bother with siggen
loadrt the parport driver and a not block
connect the not's output to its input, and to one or more parport output pins
put the not and the parport write function in a thread, and start it running
jmkasunich: ok that sounds good
it will toggle on every thread execution
actually that was my original idea, but I wasn't sure on the implementation
I'll do that
and have my scope triggered on an external trigger
I should be able to catch the glitches then
you really want to trigger on absence of transitions - thats hard to do unless the scope has fancy trigger modes
can't do that on mine... having it triggered on a non PC source would let me see the glitches tho
such as the line 60Hz or something like that
but you gotta constantly be watching it
of course of course
and it would still be easy to miss a glitch
the bug is frequent enough and long enough
when I'll know what the problem looks like, I'll start ripping off modules
do you see anything when you run the latency test?
probably some ACPI or CPU fan control or whatever
this isn't a laptop is it?
nothing outstanding with latencies
non ASUS mobo with AMD processor
A7N8X I think
lemme setup the HAL and I'll get a good confirmation
must be raining in nebraska - timeguy.com is very slow
ok I have a period time close to 200us
this is without "machining"
Axis stops on "joint 0 following error"
why is irc blinking at me
[Harlock]: your period is too long wrt your max velocity and step size
either reduce max velocity or reduce the base period
200us is really long, normal is like 20us
thanks fenn, i've reduced the feed rate for now...
if I have 200us, it means the thread is running at 100, isn't it?
uh, its 200us between consecutive executions of the thread
yeah but I have a not block inverting it's own signal
so one cycle the pin will be high, second cycle the pin is low
so the thread was running at 100
I've put 50us now
200us = 5kHz, so your signal coming out should be 10kHz square wave
40kHz for 50us
er, 40kHz square wve
I was measuring 200us period on the scope
did that backwards
yarr. is there a 10kHz wave on the scope?
or close to that
now that I've changed the base period to 50us, yes the freq on the scope is 10kHz
so now you just wait for it to stop working
you could set up a charge pump and trigger on that
there's a feature on the scope to trigger when the period is higher than a level
what if I want to send more than one NML msg in tandem?
e.g., I write a module that handles an external jog device, so I'd have to read the inputs and send EMC-AXIS-JOG. But, I wouldn't know which axis mode (coord v. free) I'm in, so to be safe, I first send EMC-TASK-SET-STATE.
wouldnt that screw up any motion currently happening?
thus, I'd send TASK-SET-STATE, then send AXIS-JOG --- In this case, then NML is inefficient, because by sending two NML packets, I'd also have to wait twice the amount of time for a response.
Skullworks_PGA1 is now known as Skullworks-PGAB
anyway, what happens if you just send the jog message?
fenn: it just bounces off emctaskmain.cc
excellent -- i see that the usr_intf/axis section has been cleaned up and organized
it seems the hiccups I've seen were only graphic
my scope triggered only once on one cycle of 57us instead of the expected 50us
[Harlock]: could you check to see that the scope and latency-test match up? i'm just curious
I'm not sure I'm following you. My thread is setup to run at 50us, scope says it's about right, except in some occasions where i'
I've seen the thread last 57.1 us
basically, i'm skeptical of the concept of using a computer's internal clock to test its own jitter, and since you have a scope all set up to test whether it's an accurate test, it would answer that question for me
is there a g-code that pauses until the user does something?
G4 pauses for a fixed period of time
jmk i think (MSG, does that
look at nc_files/skeleton.ngc
I got 58.80 us
so 8.80us over 50us
not what I would call fantastic
if you just look at the signal, what is the period? and what is the period reported in halcmd show thread
and presumably your scope can measure jitter in the signal
I can see the jitter
I can tell you with the cursors
I set persistence to infinite
what are the two values (Time, Max-Time) in halcmd show thread?
i think Time is the amount of cpu time used by the thread
the difference between when it started and stopped
i dont know what max-time is
ok so these are the actual run time of the ask
time is the most recent time, max is the largest value that time has had
and max-time is the worst since the thead started
max is writable, so you can reset it to zero with a setp if you want
what do you setp? which param?
ok so here are the results: halcmd show thread tells us that base-thread programmed to run at 49448us took at worst 38647us to execute. Meaning no overruns
now the scope tells us the the smallest time the task took to run was 50us, largest was 60.80us
i can set time for each funct, but not the thread.
thread times aren't params, but they are visible if you do show thread
it would be nice of thread times are params, but params like pins need to "belong" to something
and threads don't belong to any one component
I would have handled them like objects
there are complications when doing things in kernel space (threads must be created in kernel space)
ah I see
so fenn, anything else you want me to look into?
does the jitter in latency-test match up with what you see on the scope
the average value, not ovl-max
I would guess the scope average is 54us
there's no average in the latency-test, is there?
uh, no i guess there's not
I would be more concerned about overall max
no, you have to do an eyeball average
the last-interval column
max-jitter should match up too
base thread (25.0us) is hovering at 24400us
max jitter is 24532us now
the period gets set to the nearest integer multiple of some internal clock, so your nominal 25uS might actually be different
I think show threads will show the actual programmed period
halcmd says 24305us
thats a bit closer to 24400
so my jitter is twice the thread period??
not sure I understand the question
my statement is wrong
you said "max jitter is 24532uS now"
I think you meant ns, but was that measured by the scope, or indicated by a latency test or something?
or 24.5 us
was that measured by the scope, or indicated by a latency test or something?
i'd expect it to say something like 132ns
jitter of 24us for a thread supposed to run at 25us, that's weird
max latency of 132nS would be incredibly good
I though jitter was the error between the actual period and the commanded period
if the jitter is actually 24us in a signal of 25us, the waveform would be so distorted you couldn't make out where it was on an infinite persistence scope capture
[Harlock]: jitter is not a precise term
latency is what the latency test measure, not jitter
so why does it say jitter?
the latency test program
I hate inprecise words
yeah there's no latency column in the latency-test
yes there is:
"RTH| lat min| ovl min| lat avg| lat max| ovl max| overruns"
[02:37:26] <jmkasunich> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?TroubleShooting
ooh not RTAI test...
there are two latency test scripts, one in /usr/realtime.. and one in scripts/latency-test
we're talking about EMC2 latency-test
oh, well I don't know anything about that one
its a wrapper on the real one as far as I know
i dont think it is
and it might be using that annoying inprecise language
the real one uses HAL?
the latency-test from EMC2 does
the real one uses RTAI
I don't know what the EMC one does, never used it
scripts/latency-test loads two threads and timedelta
then it must be using HAL
someday I'll have to figure out why Axis freezes
sounds like a pile of module to take out one by one
timedelta has this line: jitter = max(max_ - period, period - min_);
axis is just a user space program
I can understand that
I doubt modules have anything to do with it freezing
What do you think is the cause then, beside being a poor user space program?
I don't think its a poor program - it works fine
something else, higher priority, is prempting it
something that doesn't happen on most systems, because most people don't have this issue
you mean in between RTAI and Axis?
you mean between in priority? yes
he means something is preventing Axis from running
for a second
overall priorities in a running system are:
1) realtime threads
2) linux kernel drivers and such (interrupts)
3) linux kernel system calls and such
4) user space programs (further divided based on "nice" values
the parport and not block are in group 1
axis is in group 4
along with your web browser, IRC client, and just about everything else
is there a 'ktop' or something that measures kernel processor usage?
Axis has priority 16, which is higher than Xorg (pri 15)
fenn yes I use Ksysguard
right now I use top,but kSysguard is easier
sucks resources too :)
eh, actually that's "niceness level" which goes the other way, -20 is max priority
nice and pri are different
nice is how "nice" something is to another application
lower niceness = mean
let's call it "aggressive"
oh top shows both PR and NI
harlock I still dont understand what problem your having, are things still taking that moment of pause?
we found out that RTAI/HAL/EMC2 don't seem to be affected
just the display is freezing for a second
it might be Xorg as far as I know
sometimes that kind of thing is caused by things like journeled filesystem flushing things to disk
or a disk drive doing retries on a bad secotor
I have ext3 indeed
bad sectors would show up in dmesg
I'm just thinking out loud
that's good, I do that all the time.
pretty much everybody has ext3 these days, and most of us don't see those delays
As long as there's no fight...
how many process are you running harlock?
I think SWPadnos saw delays that he tracked to disk journeling
I got a primitive SATA driver
maybe it's not helping
Silicon Image 3112 I think
worst piece of !#$#%@%R
that's when SATA was the new thing around
it gave me problems with some kernels
why do you need a seperate kernel module for your drives?
gezr: that's the way things are done these days
gezr: I have 110 tasks total
so you can download updates without havin to recompile a kernel
fenn: not only that, you can unload them from memory and reconfigure them
kernel compiling was all the fun, still is
but oh well
gezr: you still compile, the kernel is just as big as it was, but now we get to compile tons of unused modules too
I dont compile anything unessary
that's what I liked to do in the past
but my new approach is to stick as close as possible to the distro
.deb packages, etc
usually when I have a single piece of source code, I debianize it
then compile and install
makes it easier to remove and clean
do you use checkinstall or something else?
fenn: I'm not sure what you mean
checkinstall is a script that sets up a fake root and does make install
then it turns the fake root into a .deb
no I use the "official" way... I think. With dh_make
or something like that
I do dh_make first, answer the questions
then fakeroot debian/rules binary
oh the joys of trying to play debian's rules
it's a real pain sometimes
I wish there was less dependencies... everything is depending on soo many packages
there should be a dependency tree limitation
some packages should be compiled static
ok I gotta go now, I feel I'm only whining
see you guys
Need help getting two DTE to talk
serial link stuff?
I think you are in the wrong channel
Yes, I'm trying to get data flowing between a old computer and my WinCE device.
I'm not sure best channel. Sorry
oh, you are definitely in the wrong channel
this channel supports EMC2, a Linux based machine tool controller
I suggest a web search rather than IRC
I saw machine and controller and I thought RS232 question might have a fair shot
well, I think im going to hook up my motor drives
what do you guys think?
sounds like a bad idea
whats the xwindows config tool on ubuntu called?
got it, restarting x is done ?
sweet, got it looking right finally
time to power up the drives
i have a idea is emc2 suitbale to install in a usb flash disk?
all flash disks are slow
maybe interesting ,and more convience
year,i do think so,i have the experience to port emc1 in COMPACT FLASH CARD,and things go well
hahahahahahah, it works
I think I have my drives set up wrong hehehe
na, they are okay, just cant do much with the little bittie motors I have connected, they act like speekers
find comm first
they seem to be doing okay with the basic axis stepper set up, but max out at 5% of feed overide, running the axis.ngc file
now just go ohm
what about it?
i think im feeding back on my Z axis
i just like saying ohm
one touch maxtor hard drives are annoying
they wont shut off
the hard drives are called "one touch" 500 gb
cheap now I know why
i need some putty
epoxy putty ?
playdoh will work, oh well
im going to have to get some real drives I guess
something about this tig welder looks unsafe :) http://www.cnczone.com/forums/attachment.php?attachmentid=19323&d=1151469872
ive been using unc8054b chips, they just arnt cut out for this
you guys pretty much all running geckos?
is cement an insulator or a conductor or wich of the best ?
I dont understand your question
don't think you can go wrong with geckos, unless you want to drive a really big motor
its all going to be a matter of what im going to want to do I guess, i have a huge registration bill for school due, and well, I dont know how much I can spend on emc stuff right now
I cant really develop without some things working I dont figure
just make a teeny tiny mill
I think that what I have is probably going to be fine
I like the idea of a teeny tiny mill :)
I have a bunch of steppers, and I dont know why the drive chips I have wont work
I could just mess with the settings till I get something I can use and then run with it
i wanted to make a palm-top hexapod, but never got around to making miniature timing belts
the pager motors arent strong enough to drive a leadscrew directly
you need what 6 motors for a hexapod?
I liked the 3 motor arangement for say a ploter, that may be the shiz
the tripod thing?
one of the videos I watched has that
one of the guys in here made it
that would be so easy for me to make
yeah, you can also do a "capstan" drive
like an inkjet printer
I want to have at least 2 linear axis to work with
that's what i mean
the tripod method is all funky kinematics
well, the inkjet using axis 2 as a rotatary
well anyway, there's lots of options
yeah, thats for sure
for mu-mill , http://www.hsi-inc.com/canstack_template.php?series=36000
these steppers have lead screw built in
hi,i have ibm t23 computer as my platform to run emc2.2.1,i found it is very slow
cpu cellon 1.1G HZ,and 512M ram
is it enough?
oh,but i alway found unexpected realtime delay,is there sth wrong?
you may want to read this page http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?TroubleShooting
maybe sth in INI file about period about task
and this http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?FixingDapperSMIIssues
and this http://rtai.dk/cgi-bin/gratiswiki.pl?Latency_Killer
that is enough,i will read it carefully
check dmesg, see if there are any warnings or error messages that might help narrow down the cause
oh it's a laptop
that's usually difficult or impossible
they have a lot of power management stuff in the hardware that bothers the cpu
bad for realtime
i will just have a try,but some guy around me using t41 laptop,things go well
anyone around familar with aptos / apt360?
try #cam, but i doubt anyone is awake
and, no, not really
are you asking about getting it running or using it?
I'm.. a little confused
it seems like there ought to be a handy macro for pocketing things without linear sides
out of curiosity, what do you use to generate nc code?
nothing - i'm working on a script to turn an outline into a spiral pocket
supposedly ace converter turns dxf into g-code, but i havent tried it
and there's gcam but it is rather limited and you have to start from scratch
yeah, it seems to not like debian systems too much
and crashed a time or two
yeah, stupid developer if you ask me
it seems like there isn't that much to to 3d work
well, he might be stupid, but he started a project and got something working, so i give him that much credit
I've seen a few projects based on opencascade
stupid is the wrong word.. close-minded i guess
are there at least decent (3D) cad programs out there?
opencascade is so huge and unmanageable it's hard to even get started on it
not really, most 3d stuff in linux is graphics oriented
brlcad is solid under the hood, but aging and has a bad interface
at the same time, i could actually make things in blender well enough
yes but it is very hard to do things precisely in blender
it is quite powerful for artistic use though
very true, but still possible
err, precise things are possible
just not easy
k3d is interesting and undergoing active development, but it is a graphics program in the end
as far as free-libre software, there's nothing i can point to and say 'this is it'
so, if you're just dealing with triangles, .stl seems to be the way to go
but is there a common format for, erm, maybe that's call NURBS-based models?
so instead of approximating a sphere using triangles, that format allows you to say "there's a sphere, r=suchnsuch, centered at XYZ" ?
there is a more complex extension of IGES in ISO 10303 AP203 and AP214
nurbs ,that is interesting....
the complex math formula prevent programmer to use it,
i'd settle for a format with triangles, cylinders, spheres, cones
well, yeah, but it is a cleaner solution
eh, i guess
i mean, there's no approxamating involved
and it seems few nurbs source lib can be found to aid our job
it's good to know that there's a standard though; is it somewhat followed in the real world?
part of the reason iso-10303 is so complex is that everyone got to have their say
i've never laid hands on, say, mastercam, so I don't know what formats are "normal"
any idea if there's anything free/open that can speak IGES?
there is a something called gCAD3D that can open IGES files, but i dont really understand what the author is trying to do or whether it's free or not
i mean, i can throw my brain at making iges->nc, but if I can't even edit an IGES file...
i see what you mean about not usderstanding
and I've even had a few years of german
it is "freeware" that runs on linux
but wtf is the point of that
i love the ToDo
its like teasing a dog
AP001: Dokumentation -> English
and it's actually got source
kostenlos doesn't translate well
it looks like he's cool with anything non-commercial
thanks fenn, you've given my brain a new place to start
perhaps a place to start tommorow
definitely look at brlcad
it can do iges, which is good
i'm there now
this from the current lead developer of brlcad
anyone know what kind of tool would one use to adjust a "castlated nut" that doesn't have wrench flats?
(this is apparently the 'lash adjustments for a ball screw)
a pin wrench?
[08:17:15] <fenn> http://images.google.com/images?svnum=10&um=1&hl=en&q=pin+wrench&btnG=Search+Images
is that the same as a face spanner?
let me see how the phone camera did with this
assuming I use a face spanner to try to turn it, how do I prevent the screw itself from rotating? Poor pictures of the assembly:
[08:23:52] <ds2> http://imagebin.ca/view/tbRfE64x.html
[08:23:56] <ds2> http://imagebin.ca/view/51GSu-G.html
does it matter if the screw rotates?
well, from what i can tell, that nut adjusts the preload on the bearing
if the screw rotates, I can't tighten the nut to reduce/eliminate the lash
this whole thing started with finding lash on the X axis on the order of 0.050"
you can wrap some aluminum shims around the bearing and use vise grips
and what I found was the entire screw shifting on the mount
er.. s/bearing/screw shaft/
that sounds doable
what keeps the nut from loosening itself?
apparently there is a star washer underneath it that has one of its points bent up
ah i see
i wish the picture is better but phone cams don't have a good macro mode
and that probably keys into the shaft somehow
eh? I didn't think of that
otherwise the star washer would spin
hey does that shaft have a flat on it?
not anywhere that is visible
ok, it sorta looked like it from the picture
there is normally a timing belt pulley held on with 2 90deg spaced set screws on it
and they dont dig into the shaft?
there is 2 predrilled dimples for the set screws
well, you can use that pulley to hold it steady i guess
you arent going to be putting too much tightening force on the adjustment nut
more worried about breaking that pulley
would it be a bad idea if I take it to someone with a conventional lathe and stick it in a collet with the lathe in low gear?
uh. why would you do that?
generally if something requires a lot of force, you're doing it wrong
u guys realize what time it is right? lol
go back to work, sucka
cuz I figure a collet is a more event grip then shims + vise grip
i got work in queens at 6 am until 6 pm
im makin 400 dollars in one day...
what are you doing?
i work at a power plant
fenn i think im buyin a rotary phase converter, i donty want to deal with messin the machine up at all ....
as in thermal or electrical?
im buyin a cheap one....
i didnt know you could buy such a thing
fenn are u serious?
it's just two motors
Gamma-X: how about buying a vfd instead
and just pretending it's a rotary converter
cause i dont wanna mess with the wiring in the machine.....
fenn the whole ting is that i cant use the vfd to control the reverse and all that cause other pumps run off the 3 phase power.
so? just use the existing contactors and stuff
i thought u said it would mess the vfd up?
then at least you'll have a vfd later on when you decide to do it right
i never said that.. but i'm no vfd expert either
shoulndt mess it up...
should i get the 300 dolalr one or 350?
300 has more power?
sounds like you've spent more time worrying about it than required to earn 50 bucks
i would buy the pretty one. then at least i wouldnt hate it
it uses up 10 amps the shitty one uses 11 amps
i wouldnt worry about 1 amp
just wanna me sure my machine will run lol
im a bitch lol
looks like the iges standard is on my reading list
de2 you figure out your spaner wrench issue?
if you didnt find a way to hold it the answer is simple
de2 : you still there even?
oh shoot, its ds2
all he needs to do is take a piece of angle iron or sheet steel, whatever he has acces too, find a drill bit that matches up with the holes x2, or about the same size with 2 similiar drills, grind out the relief area, then drill 2 hols in the steel, lay it flat on the part he wants to hold, then stick the drill bits in upside down and ito the holes, bingo, instant spaner wrench
that will keep the big piece from spinning
then for the smaller nut if he has a drill press, same principle but he will need to make it a pure spanner wrench, drill 1/4" or smaller hols along a curved path cept for the very end, leave it like a fish hook
then grind out whats left to form the wrench
vise grips are for the loss
gamma-x reguardless of what vfd you go with or phase converter, get an electrician to look at it
gezr: i think the nut is threaded onto the ballscrew shaft
fenn : you may be right, its hard to see in the photos hu?
unless thats some sort of bearing block, I really cant tell
no I see, in the first image, its the stuff on the right, thats the nut that sets the bearing preload on the bearing block
not ont he right, on the left my bad
just a bit above the bare foot
fenn : you see what im talking about?
yeah, Ive seen that sort of set up lots
ds2 you still around, ive got questions for you about your .050 play your seeing
no it's the stuff on the left
the lathe table is underneath the stuff on the left
yeah, I didnt mean to say right, im tired
but thats probably not where his loosness is, it could be in the gib
could be lotsa places
if he has .050 in that bearing pack, then the pack is shot
that sort of set up is funky too, the ball screw has a timming pulley, that goes down to the servo, and the servo goes back under the bed of the machine
he, anything a bit lose could really add up in terms of a play with that set up too
even the servo motor, if its going bad, can make a growling sound that makes it seem like the screw is shot when its moving
im going to go to bed, good night ya'll
yay only 800 pages for IGES
i like this IGES standard - it's written in plain english and gets the job done
(btw.. nice thread you posted earlier)
they sure dont tell you that stuff on the ISO website :)
fenn: before or after you pay the fee to download the stuff?
well, i never read the actual standard documents, only the drafts
but sean says they are essentially the same
all of a sudden, my 2.2.2 stops working! 'scripts/emc configs/stepper/stepper_inch.ini' gives:
insmod: error inserting '/home/bryant/emc2-2.2/rtlib/rtapi.ko': -1 Invalid module format
what's causing this?
gcc version mismatch?
alex_joni: but that can't be, b/c I've not touched the kernel at all
you need the same compiler for the kernel and for modules
oh geez, I 'aptitude upgrade'-ed
madtry uname -a
maddash: try uname -a
the version's 18.104.22.168, and some other misc infos
that's not a RT kernel
perhaps you booted with the wrong one from grub?
here we go:
Linux athome 22.214.171.124 #2 SMP Wed Jul 25 20:56:32 EDT 2007 i686 GNU/Linux
did you build that kernel?
i've only two kernels, 126.96.36.199 and 2.6.18 (the non-RT) -- I'm sure of it, because the last time I've even touched the kernel on this box was over 6 months ago, and 2.2.2 ran fine last night
yes, I built it following the Debian etch RTAI guide
in the wiki
check dmesg for more errors
I thnk the problems gcc
did you recompile emc2?
usr/realtime/... somwhere in there is a file that says GCC = and tells the version used to compile the kernel
rtai-config I think
and you can get the gcc version from /proc
yep, last night, I upgraded (among other things) gcc 4.1.2-15 -> 4.1.2-18
maddash: well.. might be the culprit
geez, that's just stupid
why is that?
now I have to rebuild my kernel and RTAI??
how about putting the old compiler back in place?
apt-get install gcc=4.1.2-15
but before doing any of this, I would still follow the advice further up
check dmesg for a more specific cause
then look for rtai-config in /usr/realtime/..
alex_joni: "rtapi: version magic '188.8.131.52 SMP [...] gcc-4.2' should be '184.108.40.206 SMP [...] gcc-4.1'
if module versioning support were permitted in the rtai kernel...
guess you compiled with gcc-4.2 then
it's disabled for a reason
fenn: that's probably the easier solution, but returning to 4.1.2-15 might bork the dependencies
you can install multiple versions of gcc btw
fenn: guess he did
4.1.2-15 and 4.1.2-18 should both be fine (compatible to each other)
ah, nvm. just relink `which gcc` to gcc-4.1
maddash: I really suggest you fix your dpi issue
alex_joni: what do you mean?
fenn: the other day he said he can't read stuff properly because of a misconfigured dpi issue on his X server
might be what's causing gcc-4.1.2-18 != gcc-4.2
massive amounts of "deprecated conversion ..." msgs
is freenode doing this to me intentionally?
maddash: can't you do something about that toggling?
no, I think it's on your end..
alex_joni: yoou mean, gcc-4.1?
13:08 -!- maddash [n=bryant@unaffiliated/maddash] has quit [Read error: 104
(Connection reset by peer)]
you might have a too small timeout set somewhere..
alex_joni: no, it's not, b/c yesterday, I connected from work, and right now, I'm connecting from home
is there some administration channel for irc.freenode.net?
try /stats p
that way you reach freenode staff
you can wail into the abyss a #freenode (if you get voiced)
maddash: how about trying another irc client?
hi ho, from inside gaim...
maybe it works better :)
* alex_joni is using irssi without issues for the last couple of years
* maddash thinks that cli is the wrong place for a chat client
you must have never used screen then
what emc_io_cycle value gives assured responsiveness to iocontrol.0.* input?
what does assured responsiveness mean?
what do you expect?
usually iocontrol stuff is far from realtime needs, so running it 10 times / second is more than enough I'd say
well, if I'm running some overweight gcode file, I'd like iocontrol to sense a .1s signal at >=98% reliabiility
if the signal is too small for iocontrol to notice it I suggest adding a debounce filter in HAL
wouldn't debounce just make things worse?
why hsould it?
debounce semi-anesthetizes the receiver of its outputs
then a latch
meaning, i'd have to hold the signal up for, say, four times the duration before iocontrol sees it
ooh, so there's a HAL latch component?
all I could find was estop_latch
look for flip-flop
ah, i see what you mean
I didn't think of flip-flop that way until now
time to go to work
hmm.. I really love my wife's new laptop..
covet not thy spouse's laptop
it's nice and small, and has great battery life..
[12:08:23] <alex_joni> https://mail.rtai.org/pipermail/rtai/2007-December/018367.html
what happens when a valid EMC_TASK_PLAN_OPEN is sent while emc is still running a g-code file (eg, in the middle of EMC_TASK_PLAN_RUN)
you'll probably get an error that you can't do that
let us experiment...
alex_joni: emc simply stops executing, and loads the file without an error. smart.
argh, whose brillliant idea was it to add all these underscores to the NML nomenclature?
under what criterion determines whether an NML type can be called by emctaskqueuecommand?
eg, what's to stop someone from stuffing EVERY nml type into the pre-condition check function inside emctaskmain?
if the answer to my question is 'nothing,' then the pre-condition check becomes pointless
and every command should be allowed into taskqueuecommand
wow, i think i scared off everyone
in HAL I see "net" used with "<=", "=>" and without what is the difference?
what does the "=>" do?
BigJohnT: they are only for reference
BigJohnT: e.g. to help the user reading the hal files
HAL itself doesn't regard them
alex_joni: ok, thanks
maddash: the precondition checks are done for certain messages
and the actual preconditions are coded there
I must say, I'm not sure it's the best place ..
why does emctaskmain.cc bother with EMC_AUX_ESTOP_OFF? why can't it simply set emcStatus->io.aux.estop=0?
Hey, anyone in here a materials engineer?
hi Mad. "whoa" what?
Not the "girl geek" thing again, eh?
sort of. and i don't think there are any materials engineers in here.
ok, so, I'm trying to locate an odd material, but I have no clue where to look. freenode doesn't have a "materials" room. I'm thinkin I might have to open a dialog with some materials prof at a university.
do you mean you don't know what to use for an application, or you know what to use but don't know where to find it?
yeah, mad... i get it here, and in electronics as well. that said, people are starting to get used to me.
cradek: I DARE you to emp my box
I'm looking for a fluid that expands and contracts based upon an external stimuli. maybe something that phase shifts when it gets zapped by 'lectricity or something.
sounds fun, but no clue
Jessica: those are experimental at best.
mad, emp'ing equipment is easy.
Jessica: actually, water works fine, too, assuming you've severl hundred thousand volts
yeah, or a change in temp...
no, I was being serious
yeah, so was i/
well, maybe not hundred thousand...
what about liquid crystals?
mad, i think they just change orientation, not volume
dunno for sure
yeah they work just by aligning the crystals
the fluid is just a carrier
ok, on a different note, anyone know of a source for dual shafted steppers in very small sizes. kinda on the order of 3.5" floppy steppers, but with more torque?
hmm, is it possible to queue nml messages onto emccommandbuffer? ie, emccommandbuffer->write((EMC_AXIS_HOME*)) followed immediately by, emccommandbuffer->write((EMC_TASK_SET_MODE_AUTO*))?
I think there's only room for one. you have to wait until it's acknowledged.
what if I write one, acknowledge one, and immediately write a second?
hm, let me experiment...
wow, it just bounced right off
damn, I forgot to whois her
what the heck is the point of such an elaborate IPC protocol if you can't stack messages
and how dare it be named, "emccommandBUFFER"
maddash: I am in another predominantly male chan that a "french lady" stops in every so often... never says anything and leaves. I swear its just a bot to get everyone to do a /whois
actually... I have noticed quite a few nicks pop in in the last couple weeks that have the similar domains... that never say anything but a quick "hello" and then leave about 10-15 min later.
HAHAHAHAHAHAHAHAH HOHOHOHOHOHOHOHOHOHO WHOWHOWHOWHOWHOWHWOH HAHAHAHA LOL
im tired of being rite
lets play a new game of choice
chess anyone ?
play code the emc better than Unit41
cause, it wont take much im a washout coder
does anyone know what EMC_AUX_HALT_TYPE means? it is not called by anything inside src/emc/
maddash: there are a couple NML messages which are not used anymore
jesus christ, that file is hideous
anyone here know mdadm any well? specifically how to manually force a resync?
when using an external estop switch (eg, net ext-estop parport.0.pin-01-in iocontrol.0.emc-enabled-in), how does emc know to transition from estop to estop-reset as soon as I release my estop switch?
the mini doc at the top of iocontrol.cc states that user-request-enable neeeds to be active, but I can't find any code in src/emc/ that comes in contact with that pin
external estop buttons being OK only mean that you can get into estop reset using the gui
really? that's not the case here...
ok, I bet there are many ways to set it up, that's the one I'm familiar with
maddash: it depends how you hook up emc-enabled-in and user-enable-out
when I fire up tkemc using configs/stepper/stepper_inch.ini, with the only modification to standard_out.hal being the ext-estop signal, pushing the switch causes ESTOP, while letting go causes RESET
I just want to know how this is happening
iocontrol + task
I like the setup in demo_sim_cl
I followed the dir at the bottom of http://linuxcnc.org/docs/2.1/html/config/stepper/index.html
this is super funky
maddash: that description is correct
although it's not the best way to implement estop, it's the easiest way for new users
that's not the point; how is it going back to RESET?!!
the only thing even remotely related to emc-enabled-in is aux.estop, and aux.estop is only read inside emctaskmain's while(!done) loop inside main()
if ( *(iocontrol_data->emc_enable_in)==0) //check for estop from HW
emcioStatus.aux.estop = 1;
yes, that's the first part of my statement
but even if aux.estop =0, there's nothing inside emctaskmain that checks for this condition and sets emcStatus->task.state to RESET
hold on,I hacked in a few extra debug msgs
maddash: When you said "pushing the switch causes ESTOP, while letting go causes RESET" is that kybd or bigredbutton ?
FAT red springy button
So, twisting the fat red button, "release" (resets) the Estop function, and the machine is functional again?
Oh, that's is just SO WRONG!
JymmmEMC: not functional
you need to push machine on
I thought I was clear the first time: PUSHING red button => ESTOP In the upper left corner of tkemc, RELEASING red button => ESTOP RESSET In the same corner
maddash: there's no reason to get angry
that is the intended behaviour
so it works right
not angry, frustrated
it works perfect, but I don't know how!
It's kinda confusing a bit.... softESTOP is a toggle on/off, HardESTOP is ON/OFF (In/out).
my big fat red button is a push button
maddash: look for determineState()
taskupdate is called each cycle of while (!done), which in turn sets task.state to determinestate(). god, so convoluted.
now I've forgotten why I even chased after iocontrol.0.emc-enable-in in the first place...
lol @ maddash
* robin_sz has been staring at the servos on his laser ...
150A peak, 30A cont ... quite scary really
drives a re a bit old though, bosch indramat
Mmmmmm instant fried eyeballs
hopefully the servos on the laser wouldn't fry your eyeballs
"do not stare into laser with remaining good eye"
you know its a big servo motor whenthey mount a fan on the back of it :)
that reminds me, must ebay those other two servos
size of a pallet, sodding things
i finally bought a vfd