cradek: jepler: never mind. A simple yahoo search found it. duh
so i just burnt a huge pile of old bills/paperwork out back.. after that the idea of my own small forge out back sounds like fun!
...and the livd cd so is done.. burning it now
there was a DRO written for Goslowjimbo http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?PyVCP.
it uses 2 hal components, a mux2 and a sum2.
it seems to perform a 'zero' by subtracting the current count form the current count.
couldnt it just toggle the encoder's 'reset' pin?
form / from
cradek: The eagle script is just awsome.
03petev 07TRUNK * 10emc2/configs/boss/boss.hal:
-Made axis assignment match machine wiring.
-Added A axis to jog wheel.
so what kinda cool stuff you got hidden onthat server electronicsam?
nothing really hidden
[00:58:12] <skunkworks> http://www.electronicsam.com/images/
everything under that is open
stuff you did with the cnc?
a little bit of everthing
i had all kinds of nifty hidden gems on my webserver...
but.. that box is running emc2 now and my page is on someone else's server :(
this is hosted by hasweb - a division of hostdime
cradek: i'm going to boot the emc2 box off the live cd now.. and try the touch off thingy
ohiopctechDOTcom: how are you going to do that? you need to run trunk. (don't know how that would work booting off the live cd)
well uhh.. i need to try something different that what i was working on anyways
and i wasn't running on the live cd
it's the live cd installed to my box wiht all the updates
...the cd booted and tese\ted it's self as good
cradek: it works fine with a live cd
so it's possibly something i mucked with that is causing it
but i don't know what it would be
wow, I don't know either
BTW: axis is the shit...
that's a compliment right?
aka... da bomb
ya get it? lol
yes I think I get it
I'm definitely confused by the problem you're having though
it's a really really useful feature
cradek: your eagle script is the bomb ;)
you younguns and your fancy talk!
skunkworks: where did this kid learn to code like that?
why, in my day, we didn't talk in code!
it's all about throwin off the 50 and keepin the dizzeal on the down low.. right dog?
when I started programming - all we had was 1's
AND WE WERE THANKFUL FOR THEM
I've seen a few of those before
two clearing passes with 1/16th?
that'll look really nice
I think 3 - your clearance path starts at the traces I assume
oh right, I thought it was better to cut away as much as possible with the big tool
might save the point on the little one
in fact I usually set the end mill a tiny bit deeper than the V tool
so it just cuts air a lot of the time
The only thing I see - is it cuts outside of the physical board.
not a big deal. I will probably cut the boards to actual size.
halscope remembers the last setup it had. i changed the names of the threads to more readable versions, now halscope wont load saying "halscope: could not find thread: 'THREAD thrd1'", how do i tell halscope to 'forgeddaboudit' ?
a cnc engraved plaque for a buddies dirt track late model race car.. in memory of his brother. <- just got a call for that one
*more of a dash plate really
tomp: rm .scope_cfg or whatever the file is
it be .scope.cfg ;)
if i changed the ini and added desktop shortcuts in my live cd and i install will it thransfer all that when i do the one click install?
*no big deal i only had to tweak the par. port pins and the scale
*and max velocity
...if only there was a gui for those two things
and a reset ".ini" button
or reset automagically when the settings are saved
would that be a pain in the rear?
cradek: so i killed the file & got "halscope: config file '.scope.cfg' could not be opened" (bangs head)
killed = rm
make an empty file named that
and make it rw
*just an idea
that isn't an error, it's a status message
.scope.cfg will be created when you exit from halscope
but there's no halscope to exit, it didnt run
1st msg = "halscope: could not find thread: 'THREAD thrd1'" 2nd line was "Floating point exception"
ok, floating point exception *is* an error
hokay.... so i didnt ask it divide pi by zero....
well, i touched an empty .scope.cfg, and now get the "Floating point exception" only, & immediately
i stop realtime & unloadrt all beforehand...
is the scope.cfg file in the cvs?
no, it's created when you set up scope channels
no recycle bin?
whats the proper way to cleanup after such a mess? ( will look for a raw .scope.cfg but believe it's autogenerated )
scripts/realtime start; bin/halcmd show shows nada, nobody running
how can i/do i need to inspect if anything is stale in shmem?
if i rm the dummy .scope.cfg, i'm back to 'cant open' error... this is why we reboot, we cant find a way out
cradek: i had to manually edit the .scope.cfg and change the thread name to a valid one from the new set of threads , it finally works again
i have a question about six axis lathes
does the sixth axis refer to the sub spindle, or to a Y axis
i think you can make it what you like in you kinematics, and have no idea if any 'convention' exists for 6 axis lathes
i think this particular lathe is referring to an indexable live tool
the mazak website says it actually has seven axes
wahh! overheated my l298 waaah! thermal shutdown
maybe individual turret posts can move ( like turreted spindles? )
instead of a turret
it's a mill spindle
with a toolchanger
the turning toolholders apparently index to 0 degrees
it has two of these
so the turning tools have 360 degree indexing, Y axis motion, and the main and sub spindles also have 360 degree indexing
apparently the toolholders have an additional rotary axis
so i think this is like a 10 axis lathe
oh mazak, what won't you think of next.
that would be a fun post processor
i like to know that, if for some reason, i wanted to make a 4 position turret
on one toolholder
that this lathe will allow me to index it as such
hmmm, not thermal shutdown... need physical scope ( and a bigger workbench )
have you seen the entire engine block machined out of a solid block of alum.?
yeah, that thing was a
4 posn on 1 toolholder , poseidens' trident / prahphut's 4 faces, a helluva tool table entry
i wonder if you could use the toolholder indexing as a mechanical linkage
for a static tool
like, adjusting a box tool or something
the possibilities here are blowing my mind
what only a cool $million?
oh i have no idea
yeah the no price listing means.. don't ask i can't afford it
i've made tools physically bump a counter to fake tool life on machines without such, and wire a plc to limit their use accordingly
it's on the mazak website
they don't list any prices
( justa thought many computer scope interfaces save numbers rather than pixels, the number sets can be fed back into the scope for viewing or comparison, maybe halscope could do that.)
ok i hooked up my hd and rebooted back into my controller box... the touch off works fine
so something i run screws it up?
**usually only emc2/axis, firefox, gaim (for irc), a stopwatch app, file browser and a text editor
**i have gaim log into my yahoo account also, but i never had an im come through to this box with it.
**the machine is running samba through to my windows box where i draw and create the g-code, then save to the emc2 box
03cradek 07TRUNK * 10emc2/src/hal/utils/scope_disp.c: label the traces
03cradek 07TRUNK * 10emc2/src/hal/utils/scope_disp.c: don't leak these whatever-they-ares
how might i convert differential ttl encoder signals to single ended? using just A and B phases is too noisy ( spurious counts )
a differential receiver
it's a special chip
or just an RS422 receiver chip, I believe
special, but not too special
any numbers? ( was that the hp chips? )
err - damfino :)
they were probably available at Radio Shack, back when they sold components
03cradek 07TRUNK * 10emc2/src/hal/utils/scope_disp.c: show a few more digits
do you have index as well?
nice - Digikey has a single-chip 12-channel receiver
one of those can handle 4 axes with index, and it's $8.76 in single quantity
SWPLinux: no index on this encoder, but i dont have to use all the pins ;), i remember using the hp1488/89 sender/rcvr pairs, but it seems there's (too) many different ones available now ( and the Digikey brick came again today )
heh - I noticed it's more of a cinderblock these days
all tore up , mailman tried to force into farm sized box, gave up, then chucked on porch :)
waah ! i cant hook the encoder up tonight
old tamper-proof equipment = hard to salvage parts from
are 4 wire steppers good for anyhting?
SWPadnos: SWPLinux you awake?
* Jymmmm will take that as a no =)
Hi there anonimasu, how did your plasma table work out?
Dallur: Are you around?
03jepler 07TRUNK * 10emc2/src/hal/utils/scope_disp.c: since gtk2 is a requirement now, all this gtk2-only stuff can be enabled instead of conditional
03jepler 07TRUNK * 10emc2/src/hal/utils/scope_disp.c: make value shown in TIP area match the spot where the circle is drawn on the trace
make round peg fit round hole? ;)
hi skunkworks how are you today?
Good - I finally played with the eagle script you guys did. Really works great.
[13:00:58] <skunkworks> http://www.electronicsam.com/images/gcode.png
now you need to optimize the g-code to get rid of those rapids
(you didn't use this script for the first board you milled?)
No - I used a pcb to gcode http://groups.yahoo.com/group/pcb-gcode/
this software takes the output from the eagle gcode script and reorganizes it to reduce the distance in rapid movements: http://axis.unpy.net/downloads/01103508580
Cool - more stuff to play with. How do you run it? thru terminal?
I think you have to compile it first
that would be great - the pcb to gcode software was pretty random also.
I think you "sudo python setup.py install" to compile & install it, then run it as "gcode.py < example.ngc > example-opt.ngc"
the README doesn't seem to be too helpful in explaining that part :-P
I bet it really speeds it up. I guess I can see in in emc how much time the unsorted and sorted gcode will take ;)
the README is funny -- it must have been from before chris made the improved drivers for his mill'
it says the optimization would have saved an hour if rapids are 10IPM -- what a laughable speed
03jmkasunich 07TRUNK * 10emc2/src/hal/utils/scope_disp.c: rename function to match what it actually does, don't pass args that aren't used
now friends that's why we need machines capable of 1g
Just this once .. here's the log: http://www.linuxcnc.org/irc/irc.freenode.net:6667/emc/2007-05-03.txt
Martin_Lundstrom: im around now :)
jepler: doe gopt.py sound correct?
I did a gopt.py ampb.ngc ampbopt.ngc
still running :)
skunkworks, did you forget the < and > ?
I need them?
it's probably waiting for you to type in some G-code to optimize ;)
jeez - that worked
much faster with the redirects :)
so not a big improvement - 32.8 to 28.9 minutes. I suppose I have tons of cutting compared to shuttles
that's a 12% savings in time and something near that in machine wear
err - closer to 11% I guess
* skunkworks needs to hook up a tool set switch now :)
12 if you round up ;)
looks a lot better.
[14:30:33] <skunkworks> http://www.electronicsam.com/images/gcode.png
[14:30:36] <skunkworks> http://www.electronicsam.com/images/gcode1.png
yes it does
I wonder if I should try it on my big G-code file
do it - do it
well, it's 21meg, 550000+ lines of code
lots of rapids in there, so it would probably help a lot
so? you have a snappy computer
I don't know the big O of the optimizer ;)
I suspect it may be N^2
SWPadnos: It's not particularly efficient .. I think it's O(N^3) and also takes lots of memory
how many lines of gcode in that PCB you just did?
your gcode file must also have special comments -- the program doesn't understand gcode, just these comments
ok, so if I wanted the computer to spend a lot of time on it, I could just put one block and section header / footer into the file
but I guess it's mostly meant for 2-D routing, not 3-D contouring
a file with one block will either break my program (because it's a corner case I didn't investigate) or go very fast (because the only possible order is optimal)
ah right - it reorders sections, not gcode lines
03jepler 07TRUNK * 10emc2/src/hal/utils/ (scope_disp.c scope_files.c scope_usr.h scope_vert.c):
add "AC coupled" display. Click "Offset" and turn on "AC Coupled". Now, the
Offset will be (re)set at each redraw so that it is the average of the captured
samples. May give weird results with triggering, because the "AC coupling" is
only in the display portion, and is unknown to realtime
it is about 8500 lines
jepler: very nice - wondered about ac coupledd
ok, so if I could insert all those section comments, I'd have ~60x the number of sections, which at O(N^3) would take 216000 times as long to process
Dallur: Are you still around?
I should take a look at the optimizer and remember how it works -- I might be giving incorrect information
Martin_Lundstrom: yup, im here, at work though
Dallur, I will try not to bother you too much then, I will study the config a bit more before questions.
step one uses prim's algorithm the adjacency matrix representation which is O(V^2) according to wikipedia. The resulting minimum spanning tree is traversed in a particular way to give a path no worse than twice length of the optimum "travelling salesman" path. After that, an optimization step is carried out which gets rid of crossing paths. This is O(N^2) for each pass, but the optimization step is repeated until it makes no further improvements.
(N, V are both numbers of vertices)
but I don't have any idea how many times on average the optimization pass will run
* Jymmmmmmm refers SWPadnos to bigO.com
hmmm - I think I'll pass
I seem to recall that running it on 6000-dot stippler files was tolerable but much larger than that and it really started to suck
(that's 6000 blocks in one section)
Wich is the latest version of EMC avalible threw repository? Mine says 1:2.0.5
2.1.4, I think
but you have to manually change the repository name (and modify ini files) to upgrade
SWPadnos, so wich repository shall I use?
[15:05:58] <SWPadnos> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?UpdatingTo2.1
Dallur: what version are you running? Do you think it would be possible to use the deb package version for the config?
Dallur, I guess not!?
Dallur: maybe you prefere another time to discuss this, since you are working
Martin_Lundstrom: probably better to catch me at home
ok, no probs, would it be a good idea if I try to prepare something?
Martin_Lundstrom: There are a couple of things which are incompatible with the packaged version, but I hope to submit patches real soon, hopefully they will make it into the next version of emc
Martin_Lundstrom: as to when it will be released,, anyone ?
Hello lerneaen_hydra, how are you?
17 years, 23 days, 16 hours, 42 minutes, and 19.3 seconds
Martin_Lundstrom: 'lo there
Jymmmmmmm: would that be the time since you last got some? ;)
soeey, too good a chance to miss
lerneaen_hydra: Yes, so WHERE'S YOU RALIMONY CHECK AND WHY HAVENT YOU COME TO SEE YOUR BABIES!!!
Dallur: right now there's not much of a schedule regarding 2.1.5 or 2.2.0.
O_o, dude, I seriously doubt I was virile at 3 years old
most likely be after the cnc-fest?
some time after..
lerneaen_hydra: : If there old enough to crawl, there old enough to ball!
I'd like to see 2.1.5 (bugfix release) before fest
Jymmmmmmm: err. ookay...
* lerneaen_hydra edges away a bit
One of my humours co-workers sayings
but yeah I doubt 2.2.0 (new features) will be out before fest .. fest is a time where we conceive and begin to implement new features, and if we released 2.2.0 before fest it would make it longer until users would get those features
Dallur: so if im keen on trying, would you think there is a realistic idea for me to arrange so I can compile (thats no probs for me) and then arrange the stuff I need for the config?
(to start I would settle for a subset of the functionality)
mmmm rigid tapping mmmm
Martin_Lundstrom: Compiling is pretty easy following the guidelines on Wiki
Martin_Lundstrom: I also created a simulated config, which should help you see how it works for you
Martin_Lundstrom: But I do want to test it way better before it's used to operate a machine day in day out
I have g0x6 g0x0 m02, how can I repeat that 10 times?
one easy way would be to copy the two "g0" lines 10 times using your favorite text editor
Dallur: I would like to be a happy tester ;) I have to go now, I hope to catch you later, I will se if the woman will aprove of such ;)
jepler: is there a loop of sorts?
[16:03:30] <SWPadnos> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Oword
copy-n-paste it is!!!
Now, the G0 speed is indepedant out anything on screen within AXIS?
Jymmmmmmm: if you change feed override to less than 100%, G0 moves will be slower
rapid speed is specified in the ini, and is modified (down only) by feed override
ok, cool. Is there a way to pause for a moment after the x0 by chance?
* cradek gently suggests Jymmmmmmm pull up the gcode quick ref
[16:10:38] <cradek> http://linuxcnc.org/docs/2.1/html/gcode.html
is p0.5 ok?
Is the velocity display within axis combind of all axises?
Jymmmmmmm: yes, and it is not a particularly good estimate of the velocity
I have jog speed set to 118ipm, right now, but it doesn't seems that that is right.
it seems the axis is only moving around 40 or 50ipm
Not speaking of the velocity display
but actual speed
check your [TRAJ] limit too
cradek: ah, so AXIS can say one thing, but the ini can still say "no way"
I don't understand
yes, you can configure AXIS so that the jog slider goes way up into unattainable speeds
Even thugh AXIS allowed me to set the jog speed to 118IPM, the REAL speed is limited to what's in the ini?
Anyone in the US need TIG power supplies or a plasma welders ? , http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&ih=004&sspagename=STRK%3AMEWA%3AIT&viewitem=&item=140112289740&rd=1&rd=1
It's a chance but for $500 it's hard to go wrong
jepler: FWIW, I'd be content on it displaying the commanded velocity (in response to: jepler: Jymmmmmmm: yes, and it is not a particularly good estimate of the velocity)
why not add a radar gun to get the actual resultant velocity?
stop watch and a ruler.
radar guns are cheap and so much more fun to play with
Jymmmmmmm: neither the commanded nor actual velocities are provided to the GUIs in the "stat buffer", so what AXIS does is estimate the velocity by computing distance(new_position,old_position)/time
Jymmmmmmm: you can get a much better estimate by performing that calculation in HAL, but the window on this IRC client is too small to describe that in detail.
has anyone heard of this? It's hard to judge from a screenshot that might be a mock-up anyway, but it looks interesting anyway: http://juergen-riegel.net/FreeCAD/Docu/index.php?title=Main_Page
jepler: It's all good =) Just sounded like you weren't too pleased with it is all, thought I'd offer my two cents =)
jepler: woohoo it has a deb pkg =)
whoa a windows/mac version too
Jymmmmmmm: let me know what you think -- I can't actually try it out right now
jepler: me neither, my vidocard only has 4mb! lol
I like that it has a CLI
03cradek 07TRUNK * 10emc2/src/emc/motion/motion.h: make velocity magnitude available to the guis
03cradek 07TRUNK * 10emc2/src/emc/kinematics/tp.c: make velocity magnitude available to the guis
if you see robin_sz tell him im looking for him
anonimasu: will he need a bullet proof vest?
well, wheres' the fun in that?
Hi gfuys, got couple honeydoos out of the way, and ran the emc_install.sh
Now trying to configure it, but it keeps telling me the stepgen max velocity is too high, by about a factor of 3 at 44700 steps/sec
what did I miss-remember?
03cradek 07TRUNK * 10emc2/src/emc/motion/ (motion.c mot_priv.h control.c): make velocity magnitude and running line# available to scope/meter
03jepler 07TRUNK * 10emc2/src/emc/nml_intf/emc_nml.hh: put current_vel through to the stat buffer, and use it instead of the estimate in the AXIS DRO
03jepler 07TRUNK * 10emc2/src/emc/task/taskintf.cc: put current_vel through to the stat buffer, and use it instead of the estimate in the AXIS DRO
03jepler 07TRUNK * 10emc2/src/emc/usr_intf/axis/scripts/ (axis.py emctop.py): put current_vel through to the stat buffer, and use it instead of the estimate in the AXIS DRO
03cradek 07TRUNK * 10emc2/src/emc/motion/control.c: show vel due to free planner when not coordinating
03jepler 07TRUNK * 10emc2/src/emc/usr_intf/axis/scripts/axis.py: get rid of no-longer-needed code related to old velocity estimate
03jepler 07TRUNK * 10emc2/src/emc/usr_intf/axis/extensions/emcmodule.cc: get rid of no-longer-needed code related to old velocity estimate
gene: does this help? http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?UPDATING
or more likely this: http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?TweakingSoftwareStepGeneration
03cradek 07TRUNK * 10emc2/src/emc/motion/motion.c: typo
Dunno Steven. Its slowly falling into its old habits, but I'm still getting errors from the stepgen claiming the maxvels are 2-3x too high, and nothing I've done so far has changed the setting the error displays.
if this is an upgrade to 2.1.x, then that's a new feature :)
it now tells you that you're trying to do sometihng that can't be done
The machine is moving reasonably well, accels a little slow yet, but we'll get there. This is a brand new, clean install.
Using the script to do the install, trhen a reboot
with your old configs, or stock configs?
the configs as copied to my home dir, and edited. The old configs I couldn't salvage as I couldn't mount a usb key to store them on, so thats all history
ok. can you paste in the actual error message you get, and also stick your ini file on pastebin?
I've been snooping around the wiki, but most of the stepper tuning stuff there is pretty old, still valid, but old.
theres two pastebins, which one?
either, as long as you tell me which one you used ;)
works better than http://pastebin.com
pastebin.ca works better most days
lemme see if I can get into pastebin.org, brb
looks like it will be the .com version
[19:36:02] <Jymmmmmmm> http://pastebin.ca
I went thru all the motions at the .com, and it boiunced it back at me, so I'll try the .ca version next
this is running .... http://cpp.sourceforge.net/
it's a pastebin
do I have a tool that I could use to just drag-n-drop that file on the pastebin window?
Its about 20 copy/paste ops other wise
If I was to get two encoders, is it possible to "track" theri movemen, then be able to compare that to actual gcode? then (visually maybe) plot them both and see what/where things start to go wrong?
gene: is it one file?
Open it in an editor, hoit CTRL+A (select all), CTRL+C to copy, CTRL+V to paste
or are you doing CLI ?
[19:46:18] <gene> http://pastebin.ca/469509
gotta go see if theres any caffiene left, brb
like a yawn, its catching
wow - that is the longest TRAJ or SERVO period I have ever seen
do you have a 1 MHz CPU in that machine?
That coul;d be a one too many zero's typu
but it happens twice :)
naw, amf athlon, boot says its doing 1600mhz
ok, then change both TRAJ and SERVO periods to 1ms (1000000 ns)
there's no reason to have them any slower than that
huh? Lemme look again, ok, brb
(though it won't help the step generation error)
they're 9 and 90 ms right now, unless I've slipped a digit
1.8 (inches per second) * 32000 (steps per inch) = 57600 steps per second is a very high rate for software step generation, requiring a base-period no bigger than about 8680ns. it looks like emc is correctly warning you that it cannot attain the requested step rate.
read the wiki page that was suggested earlier: http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?TweakingSoftwareStepGeneration
use the attached spreadsheet to find a step rate you *can* attain
fixed that to .09 and .9 milliseconds, stiil the error, I'll check that page out, thanks
Humm, the machine can go faster than 30000ns, but getsz pretty laggy by 20000
0.09 ms is waaaaay too fast
set them both to 1 ms, and that shuold be the end of that problem
jepler correctly pointed out that the system is correct in telling you that you can't go as fast as you want with software step generation (with the base_period as it is)
damn no coffee =(
now we are really in trouble! Letncy test runninmg
now we are really in trouble! Latency test running
hmm, looks like the worst case there is about 15,000, is that reasonable?
Can emc "plot" from using encoder input?
even if it's just recreating the gcode
03cradek 07TRUNK * 10emc2/src/emc/motion/control.c: fix teleop vel display
Hmmmm, I just heard my Z axis stall =(
Jymmmmmmm, the AXIS plot shows feedback position, I believe
that gets overlaid on the preview plot, which is the exact programmed path
no, it shows commanded position
Oh, so I could edit something and use encoderinstead?
cradek, is that commanded position on the backplot?
SWPadnos: yes, the backplot is commanded position
the DRO can be commanded or actual position
oh - nevermind then :)
Ok, a bit of tweaking and I'm getting 28ipm, with the z axis screw whip apparently the limit. But I'm still getting those errors from xy at startup.
Do I reduce the traj or the axis default vels for that? not the maxvels but the defaults?
just for grins I tried to run the axis logo, way too fast, realtime following error! joint 0 & the machine actually bounced!
jog speed can be dialed up to 45 ipm, I need that at about .6 of it, or around 27ipm, which var diddles that?
gene, your X and Y axes have double the resolution of Z, so that's probably why you get the error for X+Y but not Z
You'd need to reduce the axis MAX_VELOCITY for each axis, to match (or be a little lower than) the actual max the machine can do
have you gone through the step generation tweaking page (and spreadsheet)?
as best as I understood it. pplaying with the axis logo itself, I now have that running except for a realtime error when I click the GO arrow, but it works just fine after that.
realtime error like "unexpected realtime delay" or realtime error like "stepgen can't do that" ??
unexpected realtime delay
aqnd then it runs just fine
and then it runs just fine
unexpected realtime delay is a relatively bad thing
this install is from the EMC2 liveCD (then updated)?
But I do have another concern, the function keys can't control it, they seem to be ignored. No, install from kubuntu live dvd
and then the emc2-install.sh script was run, then rebooted
you're running the "normal" EMC? ie, not a self-compiled run-in-place copy?
supposedly the 2.1.4 release
ok, that should be fine - you're running a realtime kernel then. you may need to disable some things in the BIOS
what kind of hardware? on board video? keychain drive plugged in?
no, nvidia card, keychain plugged in but not even mounted.
[21:16:12] <SWPadnos> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?TroubleShooting#Unexpected_realtime_delay_check_dmesg_for_details
if you're using the accelerated nVidia driver, that could be the cause of the problem
if you're using the open-source nv, it probably isn't
nvidia card: try using the vesa driver, even "nv" causes trouble
I sit corrected
see if you get the real time error without the keychain drive plugged in
I can paste that, but I expect you've seen those. I think its using the nv driver
bummer - hte APM/ACPI BIOS settings section has no data :(
but the nv isn't loaded, fbcon and vga16fb & softcursor, and that looks like trouble.
brb, gonna go look at Xorg.o.log
SWPadnos: I added some very short suggestions
heh - ok :)
yeah, Xorg.0.log says its nv, but that softcursor still bothers me, should it?
FWIW, nv ran just fine before on about a 2.1-HEAD version
err - no, it shouldn't bother you
Do I need to lengthen a loop?
are you saying that the cursor (and by extension the system) seems laggy?
not really, it feels good
ok, then I'd say your BASE_PERIOD is probably fine
running a 23000ns inner right now
do you get the error everytime you start emc - or is it random?
that should be OK. 20000 works on my celeron 500 (and I think I've gotten it down to 16000 without crashing)
when I start running gcode, like in this case, the axis logo itself
jepler: can I display actual, command and backplot?
gene: did you look at the numbers in dmesg?
yes, they're indicating a somewhat lenghtened response but its not gross, as in 5x, more like 1.1x normal
* skunkworks still thinks gene should unplug his keychain drive.
jepler: well backplot and actual (via encoders)
diesable what you dont use in BIOS.... usb, secondary controller, serial ports, etc
I can't reach it without getting stepladder :)
gene: try disabling the keyboard
sounds like a poor design ;)
I bet it will run fine after that
oh, and the mouse too while you're at it
disable the CPU - that's sure to stop the realtime delays
SWPadnos: that will cause one big one
yeah, but it will get rid of the error messages :)
I'd rather disable scheduling
Hey RT Tests show ovl max as lower with even secondary controller disabled
As a first impression, I'm not too upset, its actually running better than ever. Congratulations guys!
did you have anything connected there?
gene: the 1.1x is actually an arbitrary numer
so reports like yours might help finding a proper value
lemme paste from dmesg then
The BUS has to poll those devices, if there's no device to poll, less latency
emc1 had something like 150% instead of 10%
In recent history there were
[ 9523.547802] 126303, 126173, 126739, 144673, and 107900
[ 9523.547804] elapsed clocks between calls to the motion controller.
[ 9523.547816] This time, there were 132827 which is so anomalously
[ 9523.547818] large that it probably signifies a problem with your
[ 9523.547820] realtime configuration. For the rest of this run of
[ 9523.547821] EMC, this message will be suppressed.
I think EMC1 had 1000% (10x) for a threshold
I'll go find some more coffee, the ladder, and get rid of the key, bbl
that's kind of a funny message, since 144673 is > 132827
I thought so too. The usbkey is gone now
SWPadnos: the trigger is current sample > (threshold) * [previous N samples]
just yank the damn video card
easier to do in hardware ;)
That must have been it, no more error.
USB polling REALLY sucks
ok, so the samples before the ones printed were probably also in the 126000 range
the long one (14) caused a short one (10) and the next relatively-normal one, compared to the short one, triggered
this case looks like it's right on the edge of not being a problem
good to know, thanks
it's hard to say exactly what threshold is a problem and what's not
At any rate, the axis logo just ran, displaying as much as 23 ipm, a speed record by a good margin for this machine.
did you reduce the axis limits? the 1.6 IPS max may be a bit aggressive for that machine ;)
03cradek 07TRUNK * 10emc2/bin/.cvsignore: new binary
(I think it was 1.6)
Yup, thats faster by far than I have horsepower in the spindle to actually do :)
wow that sounds fast for a mini mill
that also eliminates the stepgen error messages
and the velocities are now down into the .5-.6 territory
ok, much better :)
Thats a micromill
can the mill handle the sudden deceleration?
only if you wire-tie it to the bench ;)
it seems to be happy with settings of 8.0 and 8.8, my psu is from hell, weighing in at about 30 pounds
chuckle, one of its jumps moved it about 1/8"!
good night all
its sitting on a steel kitchen cabinet top, one of the porcelan coated things from yesteryear, probably 65 of them even
good night and thanks for the help
gene: I didn't do anything.. as usual
thats my specialty these days alex
There is a cold front moving through and I'm in summer shorts so I think I'll retreat to the house, unless someone can tell me how to configure the keyboard so the function keys work with emc2 again
do the function keys work in Linux overall?
hmmm. was it kubuntu 6.06 LTS that you installed from?
yes I think so, lemme see if I can get a tty shell, no I canb't, and yes 6.06 live dvd.
isn't a ctrl-alt-f2 supposed to get a tty shell?
I think it should
is the correct keyboard selected? (in preferences or system setup)
deader than the star trek guy.
then I'd say it's not an EMC issue ;)
Ok, I'll go ask on the kubuntu list
I'm not sure where the menu item is in KDE, but there should be a control center panel for keyboard configuration
just install a VT100
you could also try killing X, to see if it comes back (ctrl-alt-backspace)
yeah, experts and that list are an oxymoron. It does come back, with a new login request.
ok, so log in and see if the F-keys work
that xdm replacement thingie is running
(or gdm for gnome)
well, that was a quick exit, and it didn't restore anything.
and the fkeys are still dead.
do you have another keyboard?
I'm sstarting to shiver, so I think I'll retire for some food, its dinner time here & I have NDI who's cooking
heh - have fun :)
many thanks guys
NP. enjoy dinner
I need to sample from the paraport at 16KHz
is that doable in EMC2?
HAL can do that. do you need EMC as well?
32KHz to be safe
this is for adding encoders (on the phone with usdigital)
I'd say yes
on the emse part
SWPadnos: Lets say yes in the EMC part
well, EMC can't sample the parallel port at all. but some of the HAL components can
if you want feedback (or a DRO) for your system, then you need a base period of 30000 or so (to get the 32 KHz sample rate)
that'll give you roughly 16000 counts/second (8000-10000 would be safer)
all you need to do is connect the parport pins to the software encoder counter HAL module
MY Base PERIOD right now is 17000
So the quadrature encoders can work on my system?
He's saying minimum of 16KHz, safe bet being 32KHz
a 17 microsecond period gives ~60 KHz samples, or 15-20Kcounts/second (to be safe, max limit of ~30k counts)
it all depends on how fast you want to spin the encoder
if you only want to go 1 RPM, then you can count very slowly
Well I told him max of 120IPM and I'm running input scale of 16000
17120 base period
that's a bit fast for software counting
the 120 IPM/16000 scale
well, I cna slow it down to 80IPM
do you need the encoders to be the exact same resolution as the steppers?
also, are you planning on putting the encoders on the motors, or on the screws?
the encoders were looking at are 2000, thats in case I buy geckos that do 10microstepping, xylotex does 8
he encoder resolution versus stepper resolution is irrelevant. there's no control system that can use both at the moment
Hmmmm ~$70 AVR32 boards are real
are real what?
Jymmmmmmm: why are you so bent on using encoders to cure another problem?
Jymmmmmmm: I think you should try borrowing a gecko off someone, and trying it on your most problematic axis..
anonimasu: The encoders are not to core/control anything at all.... The intent is to collect data so I can determine when/where/what is happening when it does stall. I can't fix the problem, if I can't repeat it.
SWPadnos: ah, gotcha
anonimasu: My z never stalled, till today when I was running cds.ngc and I couldn't repeat it out of 4 times
The machines has been on for 4 hours to get it up to operating temperature too
03cradek 07TRUNK * 10emc2/src/emc/motion/control.c: send free and teleop vels to the gui too
Can I change the latout a bit in AXIS? there's a lot of whitespace on the gcode readout and a lot of greyspaceabove the feedoverride
do you mean can you, or can you easily?
k. my biggest thing would be what I changed would get overwritten at the next upgrade.
yep if you are using packages, it will
also, there may be some things that aren't visible in your setup, but would be on others, like spindle override ...
yes if you hook up more things in hal, you get more widgets to control them
03cradek 07TRUNK * 10emc2/src/.cvsignore: ignore
03cradek 07TRUNK * 10emc2/configs/sim/.cvsignore: ignore