just seems odd
what are you talking about?
no, it ferrors when I release the jog key
that was just luck
cradek: so long jogs don't ferror?
alex_joni: no, only when I release
damn that's odd..
03fenn * 10emc2/configs/emc.ini: min_ferror was set to 0 for some reason, reverted to .010
halscope shows there is no cumulative following error
fenn: does that fix it?
works for me
cradek: do G0 and G1 work?
alex_joni: didn't get that far because I couldn't jog to home
[00:04:39] <Jacky^> http://cgi.ebay.it/SEXY-FETISH-PERIZOMA-UOMO-CANE-x-natale-o-celibato_W0QQitemZ5644373523QQcategoryZ324QQrdZ1QQcmdZViewItem
alex_joni: I put in fenn's fix already
you didn't have that in the ini..
oops, I lied about that being my ini then
cvs must have merged in that change
I still get my following errors after fenn's fix
probably had some of the core_stepper.hal stuff merged in
I think emc is rewriting the ini with that 0
alex_joni: Self-replicating code, impressive.
I have no idea what I did wrong
cradek : woke up?
btw, the ini jmk submited didn't have a 0 ferror in it
mine's working again
fenn: not sure what you fixed..
i just checked it out
MIN_LIMIT = -2.0
MAX_LIMIT = 4.0
! FERROR = 0.050
MIN_FERROR = 0.010
HOME_OFFSET = 0.0
--- 201,205 ----
MIN_LIMIT = -2.0
MAX_LIMIT = 4.0
! FERROR = 1.000
MIN_FERROR = 0.010
HOME_OFFSET = 0.0
that's the commit log
and my emc.ini (before your change) is reading 0.050
just got it out of CVS
well I got MIN_FERROR=0 somehow
that's a pretty huge default FERROR!
CURSE the webcvs delay
i didnt do anything to FERROR, only MIN_FERROR
well.. seems you commited ferror ;)
that commit log shows FERROR changed from 0.050 to 1.000
damn it ! this is better with the penguin too!http://cgi.ebay.it/ws/eBayISAPI.dll?ViewItem&item=5451580256&ssPageName=MERC_VI_ReBay_Pr4_PcY_BID_IT
I wonder where my 0 came from
* alex_joni points to fenn
hey where did _my_ zero come from then
* alex_joni points to cradek
fight it out
* cradek points to Jymmm
emc does write values to the ini file - look at OUTPUT_SCALE
I have nfc what happened
oh fer fsck sake.. stop pasting that sh*t Jacky^
try this: edit the ini file so that min_ferror is 'MIN_FERROR = 0.050" - include all the spaces
run emc,then exit, and see if the spaces are gone
that'll tell you that emc is rewriting the value
Hi All....Got the analog isolator, and put it on the machine today...so now it has spindle speed control from the PC...
does anyone have anything against removing the IO_BASE_ADDRESS for the second parport from the ini?
it's not used anyways..
SPINDLE_ON/OFF_WAIT don't do anything either..
Is that the address used by bridgeport I/O ?
the INI is cluttered with enough junk that's not used
Well,,, it spindle Reverse/Forward really do not do anything...
hmm.. have a better idea.. I'll leave it as it is
wb9mjn: how do you mean that?
Spindle On/Off sure does....!
not unless the HAL grabs it from ini
is the second parport 0x278 by default ?
Spindle forward and reverse is controlled by the analog voltage in allot of implementations...
petev: I'll start working on subdirs for configs today.. lateron
and I'll start from scratch with an ini
+10 volts is full forward, and -10 full reverse...
you can do that in hal
I removed in order to use 0x278 adress ad first parport and it worked fine
So, Spindle Forward/Reverse is more for old-fashioned drives with Relay reversers...
wb9mjn: yes it is, but you can run that with emc2 too
Wherease Spindle on/off is more for newer drives with computer style enable inputs...
actually, you can lose all the XXX_INDEX and XXX_POLARITY from the bottom of the ini file as well
so you'll have to chose the kind of spindle control you have
I use EMC 1 here....
oh - well then you can't ;)
wb9mjn: too bad for you ;)
A four quadrant drive does not need the forward/reverse...
As a negative voltage will cause it to apply torque in reverse...
that's why io in emc2 has a LOT of ways to drive a spindle
03alex_joni * 10emc2/configs/emc.ini: reverted fenn's change to ferror, as it was unintended.
pc_op` is now known as pc_op
i'm pretty confused about this whole ferror thing
cvs diff shows no difference
cvs diff is stupid..
had the same problem here too
Unless... it was something else that caused the problem
Jacky^: Updating your porn collection huh?
apt-get install pr0n
nah .. was looking for the latest video from iraq :/
Jacky^ when you find it... rm -rf *
Jymmm: yeah ..
* petev is away: eating
the video show italian militars killing peoples..
when aout mission woulde be a 'piece mission'
03alex_joni * 10emc2/configs/emc.ini: cleaned up, removed unneeded IO pins & polarities, added +/- comments reflecting what a user typically needs to change. revert if the changes are bad :)
Jacky^ what, you want to watch a beheading video too?
at the lest, they called so ..
Jymmm: no, thanks
im interesting in understand better I can this war
I watched your video
Jacky^: This or any other war.... there is never any understanding it.
les_w: the audio is rilevant, not the video
fenn, cradek: wanna take a look at the emc.ini I commited?
but is italian :(
i dounbt you hav understand all
yeah could not understand of course
yeah.. who can understand italians..
they have some weird thinking :)
les_w: Nope, not other Italians either.
wonder why the video comes out now ..
Well you missed the k4ts singing concert the other night, and it sounded good whether I could understand it or not. ;)
were near to next election too
jmk_eating is now known as jmkasunich
alex_joni: if i cant trust cvs diff, how do i know wtf is actually in the cvs?
you cvs up :P
I never had problems with cvs diff..
jmk : i still get following errors if i turn default accel back up to 20, with your fix
with STEPGEN_MAXACCEL = 21?
alex turned FERROR back down to .05
not sure what value you were testing with
and with your core_stepper.hal getting its accel from STEPGEN_MAXACCEL, not from MAX_VELOCITY?
jmk was using .05
er, um, i dunno lemme check
jmk: changed a bit the emc.ini, say if you don't like..
reading the commit msg right now
jmk yes it uses stepgen_maxaccel
fenn: stand by
Oh man.... Richard Prior died today
"we bad, we bad"
Jymmm: who's that?
alex_joni what jmkasunich said
best known for catching himself on fire while free-basing cocaine
[00:50:14] <alex_joni> http://www.richardpryor.com/
[00:50:21] <Jymmm> http://www.newsday.com/entertainment/la-me-pryor11dec11,0,6659469.story?coll=ny-region-apnewyork&track=mostemailedlink
rayh, did you look at dave's circle gnuplots?
alex_joni: In his act he'd talk about setting himself on fire.... but he really did too.
heh.. remeber seeing him once or twice
ah another new word
what canwe use that for?
I know it was used for something..
just can't remeber what for..
remeber = memory retentive rubber
I make up words in patent claim texts
ooh nice ;)
then you have to trademark them
what if someone puts a patent on a new word?
why not? it new stuff. Needs new words.
les_w then they can't ref the new word... lol
Gets published in the congressional record too.
then you can TM the new word based upon the patent ref.
jmkasunich: how does that ini look?
alex_joni: look good to me.. all that stuff is in hal now
pretty good so far
* petev is back
thank god! :)
petev is back!!!
fenn: somehow it will get configured a bit nicer, didn't figure out how.. (yet)
I just had to send a congratulatory post back to this guy that made a wooden cnc router:
[01:11:47] <les_w> http://engraving.majosoft.com/html/wooden_hobby_cnc_engraving_mac.HTM
to be honest Ive seen many cheap machine in cnczone gallery too
good website. good marketing skills. Very clever.
uhm.. but slow
width = 72mm height = 100 mm. Engraving time 6 and a half hours.
& hours are a lot
well, later you will have something fast
I think i cant get the job in the url you pasted in about 2 hours
with my actual cnc
that is good
about 28 inches/min
one pass only
I think he will have the same issue I had burning the bits at that speed
yes...must use slow spindle rpm to keep chip load up
but its a cool machine for hobbist use
Dave sent me an email and he can't get above 33 blocks/sec....somethings wrong...
6 hours is awfully slow
well, I was talking to jyymm earlier...he just about talked me into costing up a hobby cnc router. I'll look at it tommorow.
les_w: paint robot?
I'm used to designing industrial machines...but let's see what hapens
no just a 3 axis hobby router
les_w: yeah, but shouldnt you be finishing that alaso?
I have never designed a hooby tool
les_w how about a Booby Tool ?
les_w: nice challenge :)
51 min left
[01:29:50] <les_w> http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&item=7568393590&ru=http://search.ebay.com:80/7568393590QQfviZ1
all those bids are schills though I think
that machine is expensive :)
could I blow that design away?
Beware!!! Deepgroove is who i got the D+M 6 from !
My experience with him is that he was less than contientious with the description of the item, and after the sale/payment less than forthcoming when his mistake (or what have you)
was indesputably identified to him...
well yeah. from the ad copy syntax and the obvious scills the guy seems a boob
52 bids hahaha
les i'm telling you it's all in how pretty it looks
what about the cam idea ?
if the engineering is perfect but it looks like crap nobody will buy it
Jacky^: thanks for the reminder.. i need to write that guy back
hehe :) im like crontab
The problem is as a customer you get really pissed when it looks pretty, but does not work !
Because then you know the people you are dealing with pulled a fast one and wont make good !
oh .. this is really a crap :http://www.ciakilledlennon.blogspot.com/
a pic of this guy appear on a news yestarday in many IT website's
looks like a big crap
cradek/jepler do you know why the view in axis goes blank if you zoom out too far?
oh, why? not really
"don't do that"
how much out?
probably something about clip planes or the limited z buffer bits
alex_joni: a lot
like.. next room?
to where the whole view is only a few pixels across
actually, it seems to happen when the view is 10-20 inches across for me
you mean pixels?
I did dsome testing with G0X10, and when I zoomed out to see the whole move, the line disappeared
not screen inches, scaled model inches
that doesn't sound right then
is your ini inch?
the line thickness amy go below some threshold in the GL package
I'll check that again though
this file is only like 65 inches across (i forgot to put it into mm)
another thing, the whole X display freezes up for about 30 seconds when loading a 25k line gcode file
it sure looks cool when its done though
fenn: it's generating some huge display lists
fenn: with Mesa in the x-server it may cause it to pause
so that probably can't be fixed
however, I get a nice progress bar and no freeze
maybe our X servers are different
probably.. i'm using xorg
bdi is xf86 right?
my machine is xfree86
cradek: are the actual free libs supporting 3D accell. ?
Jacky^: yes, but not for nvidia
Jacky^: a few I think
may Ati ?
ok - roughly 58 inches of total width is the limit, one more mouse wheel roll, and the line disappears
mine is XFree86 4.3.0
yes there are some ATI
there are free, but not open source, drivers for Nvidia
they work quite well
no, not with realtime
they don't work at all
does ATI have RT 3D libraries?
or is it because you can use GATOS or something and recompile?
I think some ATI cards have OSS support in new xorgs
I don't know the details
my hardware and software is all too old
for AXIS software rendering is 100% just fine
maybe I'll see if the adeos / RTAI 64 bit SMP stuff works, and also see if the NVidia driver works at the same time ;)
but not tonight
sounds interesting but a pain in the neck
it is interesting to note that the line disappears at a particular zoom, reagrdless of the window size
how hard would it be to highlight multiple lines of gcode at once?
Video cards hacking seems to be the black beast of free software :/
(my bet is that it's conceptually easy, but would require an array of booleans to tell if each line is selected, rather than a single "current line" number)
Jacky^: that, and installing onto old laptops
swp i was just a list of line numbers
possibly a range would work, if you can lose separate multi-select
yeah, a range would not be too hard
a list can also work
congo the cat is sleeping here in the office on a MSC catalog
fenn: i was thinking ad some game for linux like quakeIII interfacing video card with fast performance
jacky as long as it uses opengl there should be nothing unusual
congo bongo ?
congo is the name of les's cat and I think it's a funny/cute cat name
oh cool :)
nostalgy website lol
mama was queen of the congo..
King of the bongo, king of the bongo bong
whats the right sintax ?
1 print hello
2 goto 1
?syntax error in 1
needs quotes I think
and it should be print "Hello World"
wooow wok :P
although.. it's been 11 years since my last basic program
[04:40] <alex_joni> hello
who's writing basic (and why??)
Jacky^ is, and you're helping
go to bed alex
tring to remember ..
aha, a rectangle that says "Click here to download plugin"
cat stop the program now :/
Jacky^: hit the RUNSTOP key
yeah, you need java
cradek: you need java to emulate a commodore 64 that runs basic
not that easy as you'd say :P
runstop in ibm keyboard ? :(
Jacky^: failing that, hold down RUNSTOP and smack RESTORE
you can't just press RESTORE, you have to smack it, I don't know why
did any of you guys have one of these nightmarish computers?
yes Ive it
i also have a commodore vic 20
I think mine is in a box 3m from me
and maybe a sinclair ZX spectrum somewhere
I used to modem on my C64 and black&white tube (valve) TV
it was a terrible display for a computer but it kept my basement bedroom warm
wow here are the program ready to run too http://www.dreamfabric.com/c64/
java is required
cradek: nice :)
later I got a PC Jr. and after that ... AN XT!
the XT kicked butt
it had ... a HARD DRIVE
woohoo -ST-506 interfaces ROCKed!!
cool , I never used hard drive beforse 386 series
I had a "hard card" in my XT
darn.. can't find that page :/
it was a hard drive that fit in a full-length ISA slot
seen a guy once that built an PC out of 74xx chips
CPU was 4 boards,
I had a 30M drive in college, but with the magic of an RLL controller, it was a whopping 50MB!!!
solid state disk, etc
alex_joni: 74LS I hope, otherwise it would take quite a power supply
yeah.. some 30MHz iirc
wow, that's fast
he even had a network card and surfed the net with it
nah, I'm starting to not believe this
I'd love to see how fast a PC you could make in a modern FPGA
using a fast 8086 implementation
I think they were only in the 10k gate range (+/- a factor of 2 or so)
cradek: that's why I hate not finding the link
alex_joni: I'm finally testing my FO changes
seems right so far
I've always had my max FO at 100%
it will be nice to be able to use it now
yep, I think it works right
well, its raining good time to sleep :P night all
Jacky^ is now known as Jacky^afk
think I have flunked steppers. I have not been able to test emc2 much as my z axis is not cooperating. I seem to be gaining steps. yes gaining steps. randomly. the pusle train to the stepper driver looks good. two different stepper drives both do the same thing.
skunkworks: do you know why only Z?
gave up today.
x and y seem to reposition perfectly.
something stupid I am sure
badly-working steppers can be very frustrating
skunkworks: are you measuring more travel than emc thinks it has output, or is emc outputting longer travels than you ask for?
(ie, for example if you command a 10 inch move, does emc say 10.000, but a dial indicator tells you 10.02, or does the emc display show 10.02)
It isn't emc as I was having the same problem the other day and when I switched out to turbo cnc it did the same thing.
ok (phew :) )
maybe electrical noise
at this point I had the stepper off the machine and was counting rotation - should be 10 rotation per inch.
noise from the motor leads gets back into the inputs and is treated as a step
how many extra steps are you getting?
skunkworks: that sounds like a good way to separate electrical from mechanical problems
could be although on the scope the pulse train seemed good - Using a scope
it is randome
you see noise on the pulse train?
oh I misread
how many steps per rev do you have? IIRC, you have 20000 per inch, is that true? if so, thats 2000 per rev
no - I see a nice pulse train.
I now have 10000
ok, so 1000 per rev
and you're using 5x microstepping (now - was 10)?
you have something attached to the stepper shaft so you can count revs?
maybe put a mark or pointer on the motor shaft next
ha, yeah what he says
just a flywheel with a mark
it is easy to spot that it is off
so the motor shaft rotation is the part that's off?
at 1000 steps per rev, you must be missing or gaining a lot of steps if you can easily see it
skunkworks: does it SOUND right?
that is my thought
you command a 0.1" move (one turn) and it screws up?
skunkworks: what stepper drives are you using?
yes - it screws up. and no it does not sound right. the motor will rock and then I will know that it is off
I tried at a compumotor s6 and now I am using a oem650
are all your motors the same?
sounds like lost steps
same type motor, not same problem...
as in too much speed
no - my x and y are german - don't remember the brand
alex: go to sleep
yeah.. you wish
I also tried two different stepper motors
alex - yes my problem - not emc dependant
skunkworks: so it is ahead, not behind, when it goes wrong?
two possibilities: 1) you are actually gaining steps (noise or something) or 2) the motor is skipping a step during decel
seems to be - although I know that doesn not make sense. - it definatly wasn't loosing a whole turn but the mark was bulding
the mark was what?
mark was gaining.
what stepper driver are you using?
how much does it gain?
are they bipolar or unipolar?
right now compumotor oem650 have tried an compumotor s6
SWP: he answered that question before, that's cheating... you gotta think up new ones
ok - the S6 is the driver (sorry - I'm not familiar with the compumotor stuff)
(the model that is)
it does seem like it might be noise
how is the multiplier (microstepping) set?
skunkworks: you are using step/dir right?
could be the breakout board is home made - yes step and direction
you could try dropping the max accel and max velocity by half
if it still happens, you know its not because you are acceling too fast and skipping steps
I have - I had the accelleration set to .1 - still a problem
he said that TurboCNC also failed the same way
(you gotta think up new questions ;) )
jmkasunich: but it's ahead - that never happens
its gotta be noise or something then
so - how is the microstep multiplier set on this drive?
yes, seems like
never had this problem before. - odd
cradek: if you skipped a step while in decel, you'd wind up ahead
jmk - never thought about that.
not in decel, but in reverse
jmkasunich: maybe so...
but he's already tested multiple accel values, so that isn't it
no - if the motor ever misses a step, it will end up behind - decel doesn't reverse direction
mine just stall if anything goes wrong
.1 accellerats really slow ;)
swp: consider a stepper hooked to a big flywheel
accel slowly till all is spinning
then stop the step pulses
that's not the bench scenario though
I betcha it will be ahead
yep - in that case, I'd agree
so for some value of decel, the motor can skip a step (actually four full steps) and wind up ahead
although you'd be more likely to screw up on accel first
yeah, since it's symmetrical
ok - I see the failure mode now (thanks), but I'd doubt that as the cause when the motor is on the bench
but none of this is his problem
skunkworks: are the snubber diodes in the drive or external?
but drifting off topic is what we do best ;-)
jmkasunich: that reminds me of a funny story
jmkasunich: that was a joke (not funny I guess)
pulled the motor off the machine hooked it directly to the drive. changed the period to increase the "on" time (drives need 200us minimum - I was running 300us - changed it to 500us)
I know it was a joke
I was playing along
oh, sorry, I'm slow
snubbers are in the drive
alex: go to sleep
alex_joni: go to sleep
alex => slower
6am? may as well stay up now
alex goes slowly to sleep
nah.. want to do some coding today
sleep from now till noon, then code
oh - in that case, have some more coffee
screw up emc2, as I took care of emc1 yesterday ;)
jmkasunich: did I tell you I fixed the feed override > 100% problem?
no, howd'ja do that?
the way I proposed the other night
send both speeds down
SET_VEL has two numbers, requested at 100% and max possible
and don't increase over the second
.... ok, now I remember seeing the commit message, you added a second argument to those functions
so now you can set your FO at 999% and it will go your max machine speed always
very nice feature
jmkasunich: remember feed adaptation?
you looked at that when we looked at the FO bug
if you have one axis way slower
if you say so
emc slows down the whole move
guess where that gets done..
cradek: of course you know :P
pick me! pick me!
jmkasunich: so it's not TP afterall
tp is plain stupid in emc2 ;)
not plain stupid..
as it should be
but a bit more advanced might still be ok
I thought roltek would be back this weekend. I was told he might get a kick out of my k&T.
oohhh.. look at the time.. gotta go
skunkworks: have you tried swapping the step and direction leads to another axis?
did that a few days ago. just ran out of time tonight.
might help isolate between motor/drive and breakout board.
what we had thought solved it a day ago was changing to the oem650 - thought that we must not have been driving the s6 hard enough. but now the 650 is doing the same thing.
skunkworks: I don't know if I asked before - are the snubber diodes in the drive or external?
jtr - I don't know. Both drives have opto-isolated inputs.
snubbers are internal
skunkworks: oh ok, I was just thinking a bad one could cause havoc
[04:06:31] <skunkworks> http://www.parkermotion.com/manuals/OEM/OEM350-650.pdf
interesting - compumotor says that you need a 40-60% duty cycle on the step pulse input
(on the S6)
well that ain't gonna work
only at full speed
the 650 requires a 200us pulse
wow, that is long too
200 ns, and also 40-60% duty cycle
sorry, that paste wasn't very relevant to #emc.
did you change your stepgen pulse-width parameter to get the 200uS?
it's not microseconds - the manual says nanoseconds
and 6.5-15 mA, though that shouldn't be a problem
well then it isn't a problem. I think I am too tired.
what's the configuration of your breakout board?
night all - just got last call... Good luck with this one
do you have isolation, buffers, etc, or is it just a set of terminals?
crap - it has been a year since I built it - I don't remember.
it is buffers ttl if I remember right
jtr_ is now known as jtr_away
hmm - so they pull to ground
and you have the + for both step and direction connected to +5V
(they're not internally connected)
SWP: where in the manual did you find the pulse width info?
chapter 3, page 26 (18 in the PDF)
but not from the first google hit ;)
[04:12:51] <SWPadnos_> http://parkermotion.com/manuals/s-sx/s/S_chp3_h.pdf
right - two inputs totally isolated - I have the 2 positives hooked togather to + and the -'s hooked to each buffer ouput. (they may be inverters) what ever it took to get in-out the same. If that made sense
ok - tomorrow, you should try the motor test with several different microstep settings, including 200 steps per revolution (all resolution switches on)
2mS setup time on the direction inputs, that seems really high
I had thought at first that the microstep settings might be flakey, so that sometimes it would move further (thinking it was in a lower resolution mode)
they do talk about indexers all the time, maybe the drive isn't meant for this type of use?
the oem650 can come with or without the indexer.
right, but if the drive is meant to operate an indexer, there would be way fewer direction changes
I also note that there's a "zero phase input" - I'm assuming that that is an index pulse
but it may be some kind of expected feedback per rev of the shaft
I am pretty sure it is something stupid. we have played with these drives before with no issues.
ok - zero phase has nothing to do with my thought - nevermind
(just had to look at hte next page :) )
hmmm - I see 2 ms direction setup time - was that 25 ms a typo jmk?
nevermind -it was an eyeball problem here
maybe I should go to sleep as well
[04:20:18] <skunkworks> http://www.electronicsam.com/video/MVC-341W.MPG
2mS is still very ling
and that could explain it as well
this was done with the oem650's
unless the reversal time is set correctly, any backlash (or correction of overshoot) would cause forward pulses for a little while
right - but turbocnc allows you to set a axis revers pause in ms. No effect.
maybe the driver is screwed ;)
but you tried others
3 act the same way
and only with this motor?
2 s6 and one oem650
tried 2 differnt style steppers
did you set the current for each motor?
I just might be f&cked ;)
there are like 182 current settings, it seems
one is set with dip switches - the other is set with a resister
wait a second - did you try all six combinations of 3 drives and 2 motors?
I think I've asked this one several times, don't think you've answered.... how much "too far" does it go? half a rev, 1/10 of a rev, 4 full steps, 1 micro-step? and how many revs are you asking it to turn? 1 rev (0.1"), or more than one?
(and then both with turbocnc and emc?
how many permutations can we come up with?
12 so far
unless you throw an external pulse generator and emc1 into the mix
we have tried quite a few. don't know how many differnt combos we have tried.
we have tried quite a few. don't know how many differnt combos we have tried.
as far as shaft error - it was random - from a few degrees to 1/4 rotation . now as far as the 1/4 rotation is concerned - I am not really sure if that was a gain or a loss.
a few degrees is a least a couple full steps.... 1/4 turn is 50 full steps, or 250micro steps... this isn't some minor thing
how far are you asking it to turn? 1 rev, 10 revs?
most of the time we where doing inches - 10 turns
nothing smaller than that
if you've swapped the motors, the drives, and even the pulse sources (EMC/TurboCNC), what does that leave?
the parport cable, the breakout board, and the wiring.....
check for loose connections, bad solder joints, etc
operator error ;)
did you try moving the step and dir wires for that amp to a differnet channel (from Z to X for instance)?
when we figure it out I will let you guys know. Like I say it is probably something stupid.
yes - we did change
and still had the problem?
from z to x - that was when we figured it was the drive - same effect
ok, that rules out a defective channel in the breakout board, and most likely rules out cable problems too
is this basically new equipment, or surplus eBay-ish stuff?
its odd that multiple units would be broken the same way, but not out of the question
ok - you may have a motor with a bad winding, and be destroying drives (if they're not protected from that)
can't afford new. - break out board was home made - actually have 2 of them made maybe I will change them out for shits and gigles
you say you've had success with these amps before?
swp - that is scary. didn't think of that.
[04:35:46] <skunkworks> http://www.electronicsam.com/video/MVC-341W.MPG
yes - this is the first time we have had problems
cant view video here, can I have a clue?
just some steppers on a small vise - that was done with 2 oem650
turbocnc at the time
can you go back to that setup and test these amps?
and they work?
I guess I mean we have used turbocnc on the current setup with the same results with z
the old vise setup is long gone
well your problem is one of two things - either your current setup is somehow differnet than the setup that worked (pulse lengths, wiring, breakout board, anything) or the amps are different (busted)
might be stating the obvious, but keep that in mind
try to find the difference
if you can make everything else the same, then the difference must be the amps
oh well. tomarrow is a new day and maybe I will figure out what i did stupid. Like I said I think I have an extra buffer board also to try.
thanks for all your help
good night, and good luck
it is nice having a sounding board
SWPadnos_ thanks for the halcmd -R, I just needed it
I meant to do "while true ; do bin/halcmd show pin ; sleep 1 ; done", but I forgot the sleep
which means when I hit ctrl-C, there is a very good chance halcmd has the mutex
jmk why is the mutex all over the hal code?
shouldnt it be in some sort of read or write function?
because it manipulates data that is in shared memory
what code are you talking about? hal_lib, or halcmd?
well, i've only really looked at halcmd
hal_lib is the read and write functions you want
heh - oops ;)
why is there any mutex stuff in halcmd then?
but the hal_lib API doesn't cover all the things that halcmd needs to do, so it does many things directly
I agree that it shouldn't be there, the API needs to be extended
that is on the to-do list, along with about a thousand other things
what's the granularity of the nutex(es)
usually very short, except when halcmd is doing a show command
it has to hold the mutex while it traverses the entire list, and prints it
so it holds the mutex during printf calls and everything... very sloppy, I know
I mean - how much data is protected byt the mutex - or is there one for "all HAL structures"?
all HAL metadata
its all so interlinked you can't split it anyway
pins are linked into a list, also linked to the components that own them (in their own list) and to signals they are connected to....
I assume it's only a block on writes though
no, any access
no - I guess not, in that scenario
I didn't want to get too elaborate
if you are reading to traverse a list, you could allow somebody else to read, but you can't allow somebody to add or delete a list entry
right, and if you're writing, then don't let anyone read either
until halcmd came along, no code ever held the mutex for more than a couple microseconds
so it was a non-issue
this could be a problem for Ray
the watch halcmd show stuff?
he wants a "watch window" on selected pins / signals
watch only does it once every 2 seconds
that'll be settable I imagine
ok - is there a RT portion of halcmd?
so if halcmd gets suspended while doing a show, and it has the lock, you have RT problems
then you get to reboot - yay!
(another task executes)
no RT code ever asks for the mutex
only init code
right - the mutex is only on themetadata, not the data
ok - phew
(of course you would have had to think of that already)
halcmd has a signal handler that catches ctrl-C, and is supposed to avoid such nastyness
I wonder why it didn't work
how long does halcmd take to get pin data (roughly)?
I think I benchmarked it as about 1/2 the time to spawn a process with bash
what command? one pins worth of data, or show pin (which shows all of them)
whatever you were doing when you needed -R
heh, show pin
with no param?
no filter, show all pins
and I have a lot, ppmc driver is loaded
ok - that's probably still only 3 ms or so in halcmd, plus 8 ms for the process spawn (on my celeron 500)
have you checked spawn times?
do that - time for i in `seq 1 1000' do ; /bin/echo hello > /dev/null ; done
(but you knew that - I think you told me that line)
1.616s, so 1.6mS to spawn
damn - that's fast
where's the typo on that line - it isn't working here
; do, not do ;
ok - phew :0
1.235 ms here, but I have no idea what halcmd would be
anyway - the relative time was much different on my machine - maybe I was only doing one pin in my test
50mS to do 10 iterations of halcmd show pin >/dev/null
did 100 iterations... works out to 1.6mS to spawn, 3.0mS to do the show
anyway - moot point. I just realized that halcmd had to be running at the time, or the mutex wouldn't have been held
not entirely moot
unless you wre fscking with halcmd again ;)
the relative times affect the odds of my ctrl-C hitting when halcmd is holding the mutex, vs when bash is spawning
ok, now I'm fscking with halcmd
added printfs in the signal handler
"got a signal, hal_flag is CLEAR, exiting now"
not getting called or something?
but it didn't, it hung
do you have to do an RTAI_exit() or something?
clean up memory sallocs or references with RTAI, etc?
after it says "exiting now" it calls hal_exit(), then _exit()
and hal_exit unconditionally drops the mutex?
hal_exit() waits on the mutex, but if the flag was clear, who is holding the mutex
hal_flag must not be getting set properly
it's only set once, before parse_cmd
and only cleared after parse_cmd
exactly, and halcmd should only ever take the mutex inside parse_cmd
I wonder if the signal handler has access to data?
it has to
otherwise it can't do anything
true, but is it actually doing anything?
it didn't exit
(or it didn't hal_exit(comp_id)
I put three printfs in the handler
on at entry
and one on each branch of the if/else
I'll add more, after the hal_exit, etc
put in an extra var, initialized to something like 47, then print that in the handler
your machine is an athlon 2000-ish, right?
ok - that's not bad
I think about 1.6GHz actual clock
sempron is AMD's version of the celeron I think?
just curious - the spawn times on my Opteron arent that much faster (also 1.6 GHz clock)
kinda - lower cache Athlon XP
excellent!! problem solved ;)
parse_cmd is called from two places - once in the loop (where we've been looking,and where it is protected)
03paul_c * 10emc2-auto/wiki/ (14 files in 8 dirs): "Auto update wiki from a cron job. Sun Dec 11 05:30:01 GMT 2005 "
the other is where it is called for a single command on the command line
which is the case here
ah - and the flag isn't set the other time
the more common case, even
unless you wrap the single command case in a bash loop
true - the other one is called repeatedly when reading a .hal file - OK.
* SWPadnos_ forgot that interactive mode and .hal mode are the same code
ok, what this means is that a ctrl-C during a HAL cmd (like show pin) is ignored until the end of the command
yes, while the flag is on
of course in single command mode, it would exit then anyway
but the command is un-interruptable
unfortunately, it can't do anything else though
it may be in the middle of changing a signal or something
so if I start my infinite loop of halcmd show pin, instead of hanging on ctrl-C most of the time, it will ignore it most of the time
it will ignore it, until it finishes the current run
in this case, the looping is being done by bash
halcmd handles the signal, and exits (which it was gonna do anyway)
ah - so it'll never stop
you would need the infinite loop script to check the exit status
or hit the ctrl-C while bash is doing its thing
yep - getting lucky works too
hmm, its not luck
while true ; do sleep 1 ; echo hi ; done
is sleep an internal or an external?
what are the odds of me hitting ctrl-C while anything other than sleep is active?
I see a /bin/sleep
and man sleep works, for bash internalls it usually doesn't
anyway, that works fine, the ctrl-C interrupts the sleep and the loop
however, in spite of wrapping both calls to parse_cmd, it still hangs
it says the flag is clear and hangs in hal_exit
gonna try your variable test
ok - just a wild idea there
nope, foo = 56
(that was set at startup)
you know - you may need to exit with a nonzero value to get the script to stop
maybe set done=2 in the sig handler
and exit(1) if it's 2, or exit(0) if done==1
I want to get to the bottom of the flag thing first
the mutex thing
(I know for a fact it is hanging in hal_exit, and never reaching _exit()
ah - and that only happens if the mutex is held by a different process?
there aren't any other processes
just this one
and this one shouldn't be holding the mutex unless the flag is set
ok, somebody is fscking with the flag
I set the initial value of the flag to 23
in the handler it is zero
how are you running halcmd?
while true ; do bin/halcmd show pin ; done
heh - again?
I copied the set flag/call parse/clear flag code from the other place to this place
but didn't remove the original call (right after the clear flag)
ah - so it's 1 then 0 around the call
its 1, call, 0, call again
and in the loop, I didn't notice the double output
oh - oops
maybe parse_cmd should set and clear the flag?
I have a printf'ed version here, want to check a few things first
there's only one return in the function - it would be easy
well thats fscking weird
oops - I mean, what's weird?
the second call is still in there, all I did was add some printfs... and now it never hangs
printf or rtapi_print?
you know, it really annoys me that Paul went sedding thru the code adding a comment to every single #if 0....
the \fixmes ?
/*! \todo Another #if 0 */
ah - I think that's for the doxygen reports
in this case, it was some test code that I if zeroed out
it doesn't need to be in a doxygen report
I thikn you can get a list of todos with that notation (nicely formatted)
not sure though
and all they will say is "Another #if 0"
nothing kate or kdevelop can't tell you
dammit... I put the flag set and clear inside parse_cmd
and I still managed to get it to hang
and you searched to be sure that it isn't anywhere else?
(double checking now)
heh, found a typo, but that makes it even stranger
hal_flag exists 4 places,
delcalred and inited to 0
and you did remove the init to 23?
checked in the sig handler
set to 1 at the start of parse
set to 1 (typo) at the end of parse
ok, added a print of its value in the sig handler
and the init and both assignments now set it to unique non-zero values
lets see which one prints...
no failures... changing the init value back to zero
hm - so if it's nonzero all the time, it works?
halcmd just ignores the ctrl-C, you can only stop the loop if you hit it during the spawn
just for the heck of it, have halcmd print the comp_id just after hal_init
it has to be positive, else the if right after the hal_init call would shut it down
a ctrl-C right in the middle of hal_init() might be able to do screwy things
no, I take that back
because comp_id doesn't get set until the end
but it's not >0 until after hal_init finishes
it's initialized to -1
hal_init() does hold the mutex...
so maybe you should set the flag around hal_init as well
comp_id is initted to -1
and printf will help you to not hit the hal_init call
right - so if hal_init is executing, then the mutex may be held, but the hal_exit is never called, so you have a problem
a ctrl-C during hal_init mught result in the signal handler being invoked while hal_init has the mutex, but comp_id would still be -1, so it wouldn't call hal_exit
yeah, what you said
but thats not the same problem
what problem are you trying to track down?
that would result in the program exiting, but the mutex held, it would hang on the next init call
anything bad when you hit ctrl-C
and I guess that counds
yep - it should be protected by the flag - anything that might grab the mutex
setting the flag before the final hal_exit too...
and try the done=2 thing in the handler as well - it may help in breaking the shell loop
actually, done = 1 is fine
no, because the main loop can also set done=1 if you type quit, right?
so, you want to return a nonzero value if ctrl-C was hit, but 0 if done was typed
I can increment errorcount
sure - that'll work also
once I make errorcount global anyway...
actually, the only things that set done are the sig handler and a tokenizer error, so you could use that anyway (quit or exit just break)
should rename done to "error_exit" or something
maybe - but then again, errorcount is what really decides if it's an error
exit and quit should just set done=1
errorcount counts em, but in -k mode, it keeps going
yep - the done flag still needs to be set
also, the while loop for file mode should depend on done, so you get an automatic break if done=-1
also, the while loop for file mode should depend on done, so you get an automatic break if done=-1==1, that was
if done == 1, I meant
I hate long ugly conditionals in a while
the fgets is bad enough
c'mon - what's a pair of parens, an or, and a !done - it's not that bad ;)
I need another test after the switch, cause if I get a bad state, it would still call parse_cmd
true - if (!done) parse_cmd();
if ( done ) break;
right after the switch
ok, have a test at the top of the loop, after the switch, and after the parse cmd
it seems that you should only need it just after the tokenizer loop / switch
the only thing you may get is an extra prompt printed
if I get a tokenizer error, I probably want to break out of that loop too...
hell, I'm making this too hard...
the tokenizer error can directly call hal_exit, then return(1)
there, tokenizer error never sets done, just shuts down right then and there
that's good, unless you ever change the exit procedure
there should be only one exit path, in the ideal case
when I'm in a function and have an error, I like to do return(errorcode) right there
main() is a function just like any other
that's good for a function that just returns a number, but for a program that does locking, it's not the best method
dammit, where is a goto when you want one
goto (end of function where exit code is)
just put in a label, and voils, goto
I want to be able to bust out of nested loops, switches, etc
yep - that's what goto is for
people frown on it, but it's heavily used in the kernel
label is just "name:", right?
you can tell I use it a lot....
GOTO? in kernel code?
I don't think it works across functions though (which is good)
zillions of them
C only in kernel
they'll shoot you if you mention C++
oh wtf, I've seen everything now.... the shame of it all.
and assembly, of course
GOTO, not JMP... right?
they use it to unroll parts of a driver init, for example, when some error is encountered
that's just so wrong
else they'd have memory leaks all over the place
no - it's the correct use of the concept - not everything can be broken down into nice little functions
that you can correctly recover from etc, etc
the hell it can't. lazy bastards!
Jymmm: you know not of what you speak
it would be possible, but at such a cost in performance, it's not even funny
"not everything can be broken down into nice little functions" Yeah? and why not?
there is a place for C++, and for instance on "no gotos", but the kernel isn't it
ok - I misspoke - it's not always the best choice to bereak everything down into little functions
and C++ has a goto, it's called try / catch
yeah, except instead of executing 1 instruction, it executes ????
if you can answer that within an order of magnitude, my hat's off to you...
jmkasunich: I'm not speaking of shitt switch/case statements at all, more "ternary like" than GOTO's
oh - C++ isn't as inefficient as you think
* Jymmm prefers C, oop is icky =)
in fact, HAL would be much easier to implement (sans the kernel parts) in C++
all of the type stuff that you do manually now would be handled for you, and pretty well, too
get back you fiend!
oops, I mean, C++ and oop are icky ;)
I do oop in C
I know - that's why it's so hard to follow ;)
what, my code?
I do oop in assembly
no, the code is pretty clean
then what is hard to follow?
but there are a bunch of those "is this input or output", "do I use a double derefrence or a single", etc in there
also, the issue with bulk pin export would be easy in C++, but it's nearly impossible in C
jmkasunich probably writes c code like this 'Hello World' example --> http://remus.rutgers.edu/~rhoads/Obfuscated_C/hello_world.c
heh - not quite
that would be the xml version ;)
SWP_Away: If it was the XML version, I'd still take it easily over XML itself.
There is only two instances that I would accept the usage of XML personally.
death and taxes?
1) A server communicating with a shipping carrier live. and 2) SVG (but because I have absolutely no choice )
OK - so a data transport and a no choice scenario.
SWP_Away (and that's what XML was originally for.... transport, not storage.)
yep -I know.
(we don't need to go into it again ;) )
* fenn hands jymmm a lollipop
hmmm, even tho I'm returning failure, the loop keeps going
of course it does.... bash doesn't look at exit conditions in a loop...
only the while condition
then why does a sleep loop break?
fenn: Hey... this thing isn't tequilia flavored!
it somehow knows the sleep ended with a sigint
ok -I guess we all don't know how then ;)
heh.... I could use the done flag to generate a sigint outside the critical region
with the flag clear, the handler would exit
kill -9 me :)
you may be up for a while -you should also go through the loadusr / loadrt / unloadrt fork stuff for this kind of thing ...
Have you guys seen this? http://www.tern.com/g_motionc21.htm
I do disconnect from HAL before the fork, and reconnect after
no hal_flags around those hal_init calls though
don't need them, they're "inside" parse_cmd
Jymmm: I may have seen that at the Tern booth at ESC
ah - right
SWPadnos_: looks promising (if I'm reading correctly)
fairly expensive, but in the same class as MotenC and STG, I guess
4 axis mot ctrl with 80+ I?O, 32 optos and 8 SSR
and not a standard but, and communications over RS232, unles syou buy their 586 board to go along with it
unfortunately, a lot of these things aren't really that useful to emc - all you really need is some hardware that acts like an analog servo
be it a PWM, pulse generator, or a DAC
3MHz pulse and direction signals for driving steppers
useless without drives that can go that fast ...
what geckos can't?
250 KHz max internal rate
after the pulse multiplier, if there is one
hell, that' sbetter than 7khz, that's for sure.
for servos, you'd probably have to spend $1000 for an encoder that can go above 300 KHz
if you can find one
heh, they probably swapped I/O to make the bus special
possibly - it looks pretty PC/104-ish, but they don't mention it
and amd cpu, so it can't be THAT special (like Transmeta)
right - and the 5x86 have terrible interrupt controllers, unless they've fixed them lately
yep - it's the Elan SC520 - the exact chip I tried
"Those Bastards! The killed Kenny!"
03jmkasunich * 10emc2/src/hal/utils/halcmd.c: improved halcmd handling of ctrl-C (prevent shutdown while it holds the HAL mutex)
and on that note, I'm going to sleep
sounds like a good plan, actually
see you guys later
jmkasunich is now known as jmk_sleep
vq is now known as ValarQ
ValarQ: What's happening with the halgui?
rayh: not much lately
What do you think now of making it the way you planned.
i think i will keep the one component is one block style
but i might choose to hide all out-pins and instead only show the signal
that will require a separate dialog for adding signals, but i believe that is simpler
Showing that much stuff is difficult.
I have been playing with a show and configure thing in Tcl/Tk.
Don't know any python so can't help much.
Yea right. At my age that's like looking at classical greek
I never liked python ... all that syntax reliant on formatting
You working still with the Swiss, robin?
on and off
anonimasu: g'day :)
* rayh gets goes for a pot of coffee. brb
* anonimasu is working on a webpage right now
gotten any progress on the TP?
* robin_sz slaps anonimasu
gotten? GOTTEN? ... you are turning american ;)
I AM SO SORRY!
Have you got any progress done on the TP?
* robin_sz pats anonimasu on the head :)
I forget, is TP the task or trajectory planner this week?
I thought the trajectory planner was pretty much fixed and les_w was about to install his 10hp spindle to take advantage of the new found performance??
or did more bugs surface?
There is still some heavy discussion of path definition and passing.
hi ray and robin. A math error was found that improved the path a lot
But there are still issues
but the issues are not from bugs
One was servo rate...but we have a demonstration of 125 microseconds for 3 axes with the vital card
that is plenty fast...so no problem
The other issues is tp/tc inability to velocity adapt
it was not designed to
so we have to be careful not to give emc unrealizable motion commands
most controls slow down until unrealizable commands are realizable
les_w: I don't quite understand the velocity adapt issue. could you elaborate how scale is unable to handle this?
tp/tc cannot do this
hi ray. We it was from my talk with fred.
He just said that tp/tc was not designed to handle no cruise phase segments
so if commands are attempted, the queue will drain.
machine then waits until it fills again
Now, it's not a problem as long as unrealizable motions are filtered out
thats not good
I can think of many 3d surfaces where you might not have a cruise pahse
Well, Fred says it is a simplistic planner and should be dumped
fine if he will write another!
its not that the motions are unrealizeable
they are ... its just the way we handle them i guess
this must have been solved in many ways before now
I asked about modifying the code but he said that would be impossible as he coded it
Sure it was
And the segmentqueue flopped because it was a bit too ambitious and complicated
and Fred did not understand the code
didnt he write it?
the ideas behind segmement queue seemed right, from what i could understand
I guess that I wasn't thinking of velocity adapt as something we would want to do with feedrate override or current from a edm or such rather than something that happens with we try to command emc to do something out of it's abilities.
it was the crazy code world it tied to live in that killed it
What fred asked for are some good papers on the subject. If he understands the algo, he says he can code up a new planner.
Well realistically it appears that the only person in the world that can understand Fred's code is Fred.
I thought that was true of C from most folk anyway.
well, thats why we have a coding standard I guess .. to help smooth the edges
Well some of my code is hard to follow too I guess
but emc has so many glovbal variables and functions
called from who knows where
* robin_sz nods
globals are one of the worst afflictions in emc
Just having dinner and some beers with Fred makes one understand.
He is so damn smart he writes stuff that way just as mental exercise
He can keep impossibly complex interwoven structures in his head
writing it the hard way is not the best way. makes it harder for you, harder for those that come along behind
I don't quite agree with the no globals thing.
well, not no globals
just one or two
globals have their place sure
more than say, 5 in a 5000 line program is a sign of the wrong sort of approach
like $debugMode = 1;
hm not nescessarily..
They'd never have gotten the mars rovers out of the priority inversion if the globals hadn't been available.
* robin_sz isnt sure
one or two globals isnt a sin, using them by default for all variables is
looking at some cutting data dave e. emailed
tight scoping of variables and providing accessor methods that "do the right thing" makes more sense .. but we've done this argument a million times already
I don't see a real problem with variables bounded if we have adequate access to them as needed.
Dave's data is not making sense to me
hey les_w, does your machine run at the moment?
what units are the time axis scale in gnuplot?
some problem with Dave's setup....things get nasty above about 32 ipm
let me check his traj time
is he running EMC?
I thought he said that he was running about 150 ipm
for at least some of the tests.
yeah he ran at 128
I'd like to move my 3 axis router off the old DOS software onto something else,
to get better motion smoothing
on complex 3D surfaces
ok the pictures without an s are quarter degree so 230 points per inch
bet we run 200 to 300 ipm ...
and his traj time is 100 per second
so it chokes?
ok so max he could go is 30 ips or so
before it chokes yeah
ok he just has to long a trajectory period.
Now if that were a motenc card running at 125 microsecond as we have seen, traj could be 500 microseconds (also as demonstrated)
so .. what happens if a path running on that machine has a lot of say, 20 points per inch moves .. and a few 1000 point per inch sections?
so 2000 block reads per second and performance would be much better
It's all servo rate...which determines traj rate since there is no velocity adapt
I think very high servo rates would give good performance even with the existing tp
again, what would happen if in that block were a few (say less than 10) very fine moves clsoe together .. ??
points must never come along faster than the traj rate
it would stutter violently
well, thats of no use as far as I can see
it's not good....but the higher the rates, the less chance of that happening
for 3D work, you often get small sections with lots of small moves, usually due to errors in the scan etc
you dont really care if the motion just smooths over them
not following them exactly ...
other option I mentioned to Fred was to just put a snippet in to throw out points if too close
He didn't like the idea so much...was more of a " throw the whole thing out" ilk
but still, that might be very easy
for 3D contouring work, stuttering is not an option
and yes, cam programs do dumb stuff like that
all the time
* robin_sz nods
how much is the motenc card now?
550 for the light ang 800 or something for the 8 axis
coo. still expensive
but if it really can do 125 microseconds...that is right up there with modern controls
well, that part is yeah
can emc cope with running a 125 us servo loop and similarly quick traj loops ?
traj planning I meant
traj plan is RT and is an integer multiple of servo time. It can be 1:! or higher. the more the ratio, the more cubic subinterpolation you get.
needs to be 4:1 or so to get decent smoothing
so that would give 2000 blocks per second
which I would also consider modern performance levels
m5i20 is about $200 also pci also fpga but field progrmmable
I cannot do that with the isa STG though. Only with the motenc, and I suspect only with an old p3 or K6.
We have already seen some evidence of older proCessors running faster
but Not much testing on that
what's up les?
motenc on 2.8 P4? who knows
not much alex. Just a cold windy morning.
I'd chose a dual xeon, if I had the money ;)
And no dry firewood.
I do wish someone could throw a motenc in a new box
perhaps Abdul has
With that 20 kHz servo rate ad copy on his site
ought to call him.
oh ther m5i20 looks neat too
especially the price!
What did you want to do with a motenc and a new box?
check max servo rate
might actually be slower than an old box
We were running a lite on the Mazak, 800 via.
night not...would like to know
800 mhz via chipset?
Would you need motors and encoders attached to do that test
800 via processor.
I think so
might fool it with zero moves
zero length moves
but you want an intel proc and I run amd.
I run amd too
eiher one really
Well right now the amd64 has a m5i20 in it but I could switch.
I don't know if modern chip achitectures slow down hard RT threads or not
would be great if you could ray
btw m5i20 running emc?
m5i20 is supported by emc2
would like to know max rates on that too
Hey guys -- I have a problem. My mail to the developers list is being rejected with:
22.214.171.124 does not like recipient.
Remote host said: 550-Verification failed for <Kenneth.Lerman@se-ltd.com>
550-Sent: RCPT TO:<Kenneth.Lerman@se-ltd.com>
550-Response: 550 5.7.1 <Kenneth.Lerman@se-ltd.com>... Relaying denied
550 Sender verify failed
Giving up on 126.96.36.199.
Any ideas. As far as I can tell, nothing has changed on my end.
I tried resubscribing to the list, but got the message I was already subscribed.
no clue Ken
should one of us try a test post?
That would tell us something.
takes a little while.
post worked ken
unfortunately for you
I haven't seen it. The last message I saw on the list was from les "Re: might I humbly suggest..."
poped right up for me.
it came through over here
but it takes a while
lerman: what's your current ip?
Les: would you send me a message to: Kenneth.Lerman AT se-ltd.com
different than se-ltd.com?
lerman: can you try something?
Sure... -- it depends.
this is what I would do:
1. open 'telnet lists.sourceforge.net 25'
2. enter the EXACT data I tell you, you won't be able to fix it ;)
by now you should have some info if it's ok or not..
NG already. Got 500 unrecognized command to the mail-from:...
oops.. been a while
The helo returned:
250 mail.sourceforge.net Hello 24-151-6-168.static.nwtn.ct.charter.com [188.8.131.52]
which seems correct.
without the -
I cut and pasted and it seems to hang.
above is what I pasted -- it's still hung
still hung... and no mail from les, yet.
Hmmm. Just sent email to a friend and got the same rejection message. My mail system seems hosed.
sent you one myself now
it came back.. seems your provider has something misconfigured
SMTP error from remote mail server after RCPT TO:<Kenneth.Lerman@se-ltd.com>:
host gaboon.gaboon.com [184.108.40.206]: 550 5.7.1 <Kenneth.Lerman@se-ltd.com>... Relaying denied
Gaboon is a secondary that should not really been in use. I'm not surprised that it is configured wrong. Seems my primary is hosed.
I'm off for now. I'll get back to you when things are fixed. Thanks -- Ken
rayh is now known as rayh_away
hey I have one question....
axis CYCLE_TIME. do the times add for each axis or are they all run at the slowest time?
I think slowest
yet another of those stupid configs ,)
I think so to
the Rt thread does all axes each period I think
which is what I want to hear...
so....125 microseconds for 3 axes....is damn good actually.
Beats Galil's new accelera series
I think it can go faster
Some are reluctant to test the limit perhaps...lockups might trash the file system
But first the gui goes
if careful one can stop there and exit the program ok
yeah.. I usually go till the gui is sluggish
then a bit back
so I can still use it ok
thanks. Well breakfast for me. biab.
ok.. doing some coding now
03alex_joni * 10emc2/scripts/ (emc.in install.run.in): renamed install.run.in to emc.in
Well, I'm now getting mail, the test alex_joni suggested gives (what should I enter next):
[haito](lerman)$ telnet lists.sourceforge.net 25
Connected to lists.sourceforge.net (220.127.116.11).
Escape character is '^]'.
220 mail.sourceforge.net ESMTP Exim 4.44 Sun, 11 Dec 2005 06:43:36 -0800 sc8-sf-mx2.sourceforge.net
250 mail.sourceforge.net Hello 24-151-6-168.static.nwtn.ct.charter.com [18.104.22.168]
What should I enter next to continue the test?
data: sorry about this, I got to test the email
that will send an email to the list if all is ok ;)
03alex_joni * 10emc2/src/ (Makefile configure configure.in): modified to use the new emc runscript, previously install.run
ok.. seems like it went through
Well, I did:
the data line and the line with just the dot and it is just hanging.
it'll run through mailman, and might take a while..
you sure the telnet isn't closed already?
Doesn't seem to be closed -- no new prompt.
okalleo is now known as chinamill
Tried this again with the word data on a line by itself -- no colon. Seems to have accepted the message:
354 Enter message, ending with "." on a line by itself
Sorry about this -- gotta test the email.
250 OK id=1ElSdm-0002Ns-Qa
221 mail.sourceforge.net closing connection
Connection closed by foreign host.
Well, the message got somewhere: -- I just received mail that it was being held for moderation.
want me to delete it?
or let it through?
Try letting it through. I'm still failing -- tried sending another and got:
The SourceForge.net Website is currently down for maintenance.
We will be back shortly
This is a permanent error; I've given up. Sorry it didn't work out.
22.214.171.124 does not like recipient.
Remote host said: 550-Verification failed for <Kenneth.Lerman@se-ltd.com>
550-Previous (cached) callout verification failure
550 Sender verify failed
Giving up on 126.96.36.199.
Not clear what that means -- and the fact that it is cached seems to imply that even if it is fixed, I'll have a problem until the cache is flushed or times out or whatever.
yeah.. sounds like that..
seems the sourceforge site is going through some maintenance
I suggest trying again lateron
I rejected your test email
by getting there it proved that it worked
no need to clutter up the list with it ;)
hope you don't mind..
Don't mind at all -- I hate clutter as much as the next guy. I'll try again later.
I rebooted my firewall and at least I'm getting email again.
ok.. gotta leave for a while
hope it works out for you
But I hate not being able to answwer it.
thanks for your help.
that's what we are for :P
Hi dave-e les had some questions about the scale on your stuff.
gotta run guys.
ray..scale is what ever logging gives me.
the volts are real
who knows what the position axis is
I guess you saw my email...numbers seem to work out
storm and black out all day here :/
We often have blackouts here
but I have an emergency generator
just got it
les..it would be nice to get better performance but mostly from a theoretical standpoint.
I'd really like to know where the bottleneck is.
more practicality for routers though
so you had the same with faster traj times?
or plasma or anything that really needs the speed
I stepped 0.001, 0.0005, 0.00025 and tried 0.000125 where the machine froze
Since Fred told me that emc cannot velocity adapt, the key to good performance is just plain old fast servo rates...because servo rates control the minimum traj rate
not certain where my traj was...may have been at 1;1 not 10:1
well it's a fixed number...if you didn't change it the max blocks per second won't change with just a servo rate change
so I should have tried 0.00025 and 0.0025 [traj]
Hey-- my email to the lists is working again. Thanks for the help alex_joni.
So, now les_w can see my response to his posting. :-)
I'm still slowly working on gettng a motenc up on a smaller machine
STG seems incapable of fast rates, but it's good to know the motenc can go very fast
but it is going slowly
indeed it will go fast...
what is the clock on your cpu?
reading ken's response...got it
I'm 550 or 600 on a P3
ought to do 200 microseconds or so
maybe I can give it a shot later in the day.
ken, read your post. Yeah my decription was a little concise.;)
gotta bring in the bobcat and clean out this place before Xmas...family coming
I just felt we need to have some common definitions though
and yes tp specifies position and all of it's derivatives
Yes, we need the definitions -- but I'd rather see them up on the wiki. It's fine to hash them out by email (or here), but my view is that if it doesn't make it to the wiki, it will be lost (or just hard to find in the list archive).
I just basically used the definitions as they are used in my old college text
agreed on the wiki
What did you study in college that you came across TP. (That was toilet paper in my days.)
I personally split my thinking into "what can we do now" vs What can be done with time"
i'm an aero...we do robotics.
Hell, when I was in college, aero had been invented -- but I'm not sure about robots.
lerman: wow :d
I was at florida...close to the cape. Had some nasa funding
lots of german rocket scientists to teach me my math too.
all very old
there was an interesting progect about autopilot on sourcegorge
[16:28:29] <Jacky^_> http://autopilot.sourceforge.net/
Do you still know how to goose step?
but it seem not available for me..
I get Warning: mysql_connect(): Can't connect to MySQL server on 'mysql.sourceforge.net' (111) in bla bla bla..
oh sf is down right now
i tried week ago too
also with different browser :(
it may is available here: http://freshmeat.net/projects/uav/
the autopilots on the small planes I fly are crude devices
just some limit switches on attitude gyros
Back to Macfarlane -- I'm convinced that we could use the paper as the basis for an improved TP with little risk and large payoff. I'd shoot for a trivkins implementation as a first cut.
yeah the tarball seem available
lerman yes. it seems to be one of the best performance/complexity tradeoffs
segmentqueue was just too complex....with diminishing returns for all that complexity
Jacky^_ is now known as Jacky^
I'm not familiar with segmentqueue -- neither the history nor the functionality
complicated....7 years off an on work.....totally disfunctional.
Now the main issue I see as an observer...not a programmer....is that
Only Fred proctor can change Fred proctor's motion code
I would like that to not be so...
that's partly a Fred Proctor thing
the code looks like a bunch of math and conditionals, with little to no descrition of the reasoning behind them
(not that I'm incluied to delve too deeply into the derivations etc :) )
If one wanted to sugically remove the tp to replace it....I can't really see where it is and isn't
(nor am I inclined to type correctly before I've had my first pot of coffee)
seems spread all over
I think it's pretty much limited to tp.* and tc.*, but I'm not sure
of course, that doesn't tell us the *interface* the replacement would have to match
right - and cubic
I just googled sonja macfarlane and found her web page, email address, and anouther short paper:
[16:39:54] <lerman> http://www.mech.ubc.ca/~sonja/Macfarlane_ICRA.pdf
Haven't read it yet, but it might be a good introduction to the longer paper.
I guess most anything in the motion sucdir is fair game
ideally it would be a box with inputs from cannonicals...and ouputs to the pid controller.
[16:40:24] <Jacky^> http://www.boltontech.org.uk/loonycycle.htm
les_w: that would be ideal
Well, not quite... outputs to hal (which includes the pid controller).
without hal the task would prob be insurmountable ...except for fred perhaps
Sounds like insurmountable -- even for Fred.
might be. That is really old code. He has forgotten a bunch.
So practically speaking hal is probably the only hope
at least it places a clear boundary on one side of the TP
and a clear interface
The only thing I might consider for emc1 is a requested segment rate test
because cam systems are dumb and invariably put close together points
often un needed
you could test that right from the cannonicals
and chopem if needed
actually, there are a couple of numbers that can be calculated from the ini settings, and used in the CAM program (if they allow it)
Macfarlane provides point deletion if they are within the 'tightness' of other points.
Gotta go -- I'm a volunteer EMT and they need me.
nevermind - I need the coffee to sink in a little more
cam programs are dumb though. Even $15,000 ones. Their ouputs can be post filtered though.
hm, could you run a cam generated path through a filter to produce splines?
sure. A TP can be deterministic, so you can do it at anytime
I was surprised I did not get a comment or two about FF also being deterministic
hoping to get beat up a little on that one
we have a piece of software at work that converts illistrator files to dxf. instead of the normal short line segments it fits arcs to follow the shape. you can tell it how far you want to deviate from the original file. I don't remember the name and I don't remember if other files can be input to it. I am sure it is only 2d.
it helped us on our laser for smoother cuts.
I have used such programs
Before our fix last summer it made a mess, because emc pretty much did exact stop only on arcs
now it would work better
however, arc fitting has infinite jerk...
but does lessen the number of points at least
right - don't understand "infinate jerk" though
oh ok. it's simply when acceleration is not continuous
in a trapezoidal planner accel only comes in spurts. Off and on.
In a cubic it's a ramp, but the ramp has to have a corner somewhere.
Fifth is the lowest order where you can specify position, velocity, accel, and jerk at each point symetrically.
that's why macfarlane uses sines, so it's differentiable forever?
Of course the sine is a truncated Mclauren series in math libraries...but plenty of precision.
seems like you don't need all that many derivatives before it doesn't matter much anymore
need at least 4
after jerk, my conceptual grasp goes away fast
thump or something haha
Snap, someone said
oh snap ok
I can't really imagine that physically
let's see...jerk is how hard your haed gets banged in a car crash...snap is uh...the had banging rate?
no not quite...hmmm
snap out of it, jerk
kidding .. ;)
ah...jerk is the RATE you press the gas pedal in a car
snap is the acceleration of your foot on the pedal!!
got it now
how can that be infinite?
and snap is what happens when you slam the gas pedal in a Veyron
I'll snap out of it.
11:03:21 < cradek> after jerk, my conceptual grasp goes away fast
cradek: be more gentle
only if a force instantly stops can jerk be infinite
there is no physical force that can stop instantly
I think you are right
but some can stop awfully fast.
so it's more a theoretical term
yes, and jerk is very high
but not infinite
it might be considered infinite compared to normal values though
yeah - and I'm not sure you can control the next higher order anyway
accel is proportional to torque, once you accout for load
SWPadnos_: depends on how you do the controling
or what the forces are you use
if it's some motion, then probably not
yes, for all practical purposes if you suddenly unhooked a motor lead it would be not much different .
maybe some very slow motion, and the moments the forces change are very small compared to the whole motion
you can control jerk, since that's the rate of change of current in the motor
isn't that accel?
the rate of change of current?
I'm not sure it buys you anything to control the rate that you change the current change rate
generally current is accel
no - accel should be proportional to current fora given load
fixed current is velocity
no - fixed voltage is velocity
current changes velocity
so it's accel
fixed voltage assumes fixed current
hmm the reactive back emf of a motor would be related to jerk
assuming a resistor driver, yeah ;)
you have a resistor (the motor)
doesn't matter how you drive it
the motor is an inductor as well, and that's more signifficant et speed
and even a capacitor
look at the system function
also the flywheel effect
YEAH IMAGINE A Dc motor vs a large capacitor
"charges up" emf as it winds up...
holds the charge...
and can discharge into a load
1/s stuff for sure!
Have system functions for motors in a book somewhere...polynomial of s in both numerator and denominator
So neat....inductive kick in a servo will never get above a certain value if it is used for a jerk limited path
what about steppers
I think same...just with a sine force function superimposed
I suppose that the control side of the driver sees basically the same thing
I just think of steppers as multiphase ac synchronous motors often run with square wave power
uh - yeah, me too ;)
reminds me...I was to cost up a HOBBY market small cnc router for an exercise
and see if I get me too performance or not
skreek engrish, les
ie If I design a hobby machine will it be like the myriads out there?
just another fish in the pond?
right - or will it actually be better?
yeah. I don't know either. Never thought about it. I design commercial machines.
actually, putting some engineering knowledge into the design of a small system should give you a better machine for the same manufacturing cost (or, a 30% better machine for 10% more cost ...)
but it probably wouldn't pay back the engineering effort any time soon
I am just trying to decide what "better" means in that market
how do you out sherline a sherline or out taig a taig
Gee, not that small, I hope.
One thing I note...as far as the hobby cnc routers...many are very poor designs
a few are good
hmmm microstepping hybrid servo?
acme leadscrew with moglice like bearing?
what's the wearability of Moglice (and how expensive is it compared to steel)?
I'd be interested in a small industrial quality machine. Tool changer, threading, etc
My understanding is that Moglice is quite durable.
Well, we have our own. Phylly resins. It's expensive like moglice. I can mix my own (different) compound.
How expensive is Moglice? How much do you need for a small machine?
true - I seem to remember people repairing midsize machine ways with it
forgot but ridiculous price. like $10 a gram or something.
moly disulfide and colloidal teflon in an amine cured epoxy binder
Perhaps a dash of graphite...
Salt and pepper to taste....
I use moly filled nylon a lot...it's cheap
And...having developed conductive coatings, I have two Astra three roll mills!
EMI coatings as well?
I have developed coatings with teflon for switch contacts
Are the heaters for defrosting mirrors?
I developed all the coatings for my heaters yes
yes defrostinf /demisting
that uses a special positive temperature coefficient ink.
Could it be applied to the outside of a glass jar to keep the jar warm?
It took me and my engineers about a year to perfect that ink.
can you say "hot coffee"?
oops - forget I said that (while I sneak off to the patent office ;) )
Japanese have printed simple graphite ink on paper vending cartons for food for decades
damn - we're always behind
i can give other uses
used in cars of course
portable infrared targets
how hot can these "inks" get?
The positive temp coefficient is a big plus for heaters because they can be self regulating.
veries. my lowest is designed to turn on at just below body temperature.
highest is about 150c
yes....if ice is on one spot...only that spot powers up
oh also refrigerator defrosters
duct heaters...print on plastic sheet, roll up in a spiral
How high a power density can you get? (Watt's per sq inch or whatever.)
Watt's -> Watts.
depends on substrate
.5 to 10
car stuff is 2
alex: welcome back -- my email is fixed -- thanks.
lerman: seen your mail to the dev list
Just sent email to Macfarlane asking if she can make additional info (eg. source code) available.
Paul sure poured over that paper a long time.
Is she still in an academic environment?
I used the email address she had on the website: http://www.mech.ubc.ca/~sonja/.
We'll see if it bounces or if I get a reply.
I ask because if she is, given appropriate paper fodder and she might do some work on it
I figured that's why you asked. What did Paul have to say about it?
He read it for days. He asked what the math meant. I tried to teach that. He seemed to understand it well.
was the math correct??
I can find no fault with it.
is the math practical?
(for use on a computer)
Ahhh. Does that mean that YOU follow the math? I got lost / zoned out when it came to the quarternions. I was a pure math major, so we didn't deal with stuff like Laplace transforms or quarternions -- except in passing as specific types of more general stuff.
Macfarlane actually implemented and ran the stuff, but I believe it was with a DSP.
uh-oh -so you're saying that the equation is the answer for you?
I think so. You always want to avoid numerical methods in such calculations
Paul quaternioned his head off.
I do laplace stuff a lot...quaternions seldom
Fortunately, I believe that stuff was in the joint space or orientation part of the paper. So it could be ignored for the purposes of a trivkins only implementation. -- (For now, that is.)
* chinamill is away: eat
yes. inverse kinematics...now that can be hairy. Lots of jacobeans. Often no closed form solution. Poles and zeros.
and alot of anf if ors
Weren't the jacobeans a french political group?
yes they had a platform of partial differentials.
And we won't mention the poles.
Macfarlane is no longer listed in the student directory I looked at.
Her sponsoring professor....
co wrote a couple of those tp papers...
But the email hasn't bounced, so that's good news.
Well, that's a next step.
I should go out to the shop and get to work. I have to see if I can thread the end of a 5/8 inch brass rod to take a 1/4" NPT fitting.
i'll cost up a hobby machine then!
alex_joni_ is now known as alex_joni
jmk_sleep is now known as jmkasunich
03alex_joni * 10emc2/scripts/emc.in: cleaned up the runscript, added comments to various sections
well that sucks
No robin or paul whats happened?
picnet: you've been away for while haven't you?
neither of them is around much lately
I killed them both and ate their liver with with a nice chianti and fava beans
Y about 1 year plus ;)
too bad I hate liver and wine
has anyone carved redwood? If so, how did it turn out?
SWPadnos_: You want help with the verbage don't ya?!
oops, uh - wrong forum ;)
SWPadnos_ btw... you forgot to mention encoder included with your board in your posting
true enough (though I did say that it could be cheaper with a less expensive one)
SWPadnos_: I really think ppl want something quick and dirty. Just a simple alternative to powering up the controller.
we're talking two chips here.
this isn't it - if I'm to not lose money on it, I either need about 50 orders, or it needs to sell for almost as much as a G-Rex
(for more than one axis)
SWPadnos_ Simplify the design (yes I know you already had boards made and all), but what I'm talking about is two chips and a single header
not a fancy header either, let them have a breakout board
not a fancy header either, let them buy a breakout board ( I mena)
you could just sell the PCB, the PCB+components (as a kit) , or fully assembled
yeah - get rid of the switcher (replace with a linear reg), use headers instead of terminal blocks, etc.
and probably get rid of the LEDs as well - the driver is $6
(though they are cool)
could use a invertor to drive them
I don't want to have multiple fast interrupts happening
"manually" driving the LED matrix wouldn't be the best thing
ok scratch the LED, and just use a pot
we are talking 555 and logic chip, but a tad more fancy than that.
yep - the user probably doesn't care what the actual speed is, if they're right there on the knob
you could even do a hi/lo speed button too
no - you probably want some sort of accel/decel, especially for reversals
what atmel are you using?
it acan be super simple, and $5, but if it isn't significantly better than a 555, then there's no reason not to get a 555 instead
that one is the ATMega162, I thikn
it has 3 independent PWMs, with two outputs each
SWPadnos_ That's what I'm trying to say.... ppl would be happy with 555 if they didn't have to wire it themselves.
ok - then you should make a 555 kit ;)
but you could use a 2400 instead
I'm a pro!
no - a software pulse train would be bad, it has to be PWM (and I think the 2400 doesn't have the more advanced PWM, with programmable pulse rates)
what makes you think you need a PERFECT pulse train? This is a completely manual operation, not an arc or anything.
"I just want to hit a switch and make it go 'that way' without having to crank a knob 50K times"
as long as it's slow, it should be fine, but when you go faster, especially with steppers (even worse under load), a ragged pulse train will stall the motors
2400 can work to 10MHz iirc ?
and if it sounds funny (which a 555 wouldn't), then it's a perception problem
8, I think
ok, 8MHz, you only need 30Khz
but if you have a pot attached, you have to read the A/D.
and that's for all three running at the same time
pot is just speed (timing of step pulse)
that can be a manual operation, so itmight only deform the waveshape by a few microseconds
yes, and it has to be read constantly, to keep up with changes
(every ms or two, then averaged to get rid of quantization noise)
spin the knob, the mottor goes faster, right?
why it's a manual toggle switch?
we're talking a pulce RC circuit here
then how do you configure the max pulse rate?
no, 0 to whatever.
ok - a knob then?
yeah, just a pot. for the MAX level, could use another resister or jumpers and bitwise it into the uC
SWP_Away: But, why do you need to "set" a max level... just dont turn the knob that far.
that made no sense to me
SWP_Away you dont turn the volume on your stereo to it's MAX level do you?
but what if you do - the thing can output 100k pps, and your machine stops at 20
SWP_Away ok, it stops, it's not going to damamge the steppers, just stall them.
the active range is only 50 degrees of rotation
unless it's driving servos (G340)
it'll more likely fault the gecko
then you have to have all the reset stuff
SWP_Away, look... this is a manual operation, correct?
(I think it's best to have a limiter in there, even if it's a resistor on the board)
SWP_Away ok, so put one inline with the pot.
or have a 2nd pcb pot to set max level
all I'm saying is that "simple to use" does not mean "simple to design"
or "simple to build"
but all you are doign is having a Fwd/rev switch and a speed control. that's it.
ys, but it needs to work for 1000 different machine configs (somehow)
SWP_Away: step/dir. that's it. it's that simple
and you don't want to break anything
there needs to be an estop input still
you don't want people getting hurt
SWP_Away no estop. Just turn the switch off
no! absolutely not
look, no pulse triain, no movement.
there's still a big dangerous machine there, and I sure as hell won't make something that can try to make it kill you
this is a kit, if you want to intergrate estop circuitry, that's fine.
I think it's a good idea to make a simple to use unit, but it can't be as simple in construction as you're suggesting
but what is it going to do? kill the 5vdc logic power?
it doesn't need to be as complex as my current unit either though
it's just another input that has to be maintained for this thing to output pulses, that's all
your ESTOp can be as simple as those toggle switch covers
but that's not part of the design, that's part of the ext controls.
in fact, that can be used by the CNC to take over the drivers
sure - you'd wire this to the estop chain, so the mushroom switch / limit switches turns thi off as well
that's a relay somewhere else that kills pwr to this box.
could be, in which case, I'd be happy ;)
again, not part of the design itself.
KEEP IT SIMPLE STUPID. I can't say that enough.
yes, you can
SWPadnos_ you've made this fancy thing, but some just want the duct tape and bailing wire.
that's all I'm trying to say.
that's OK, but I don't want to be the duct tape supplier
I prefer to make prototypes that look like finished products, rather than products that look (or work) like prototypes
then supply the bailing wire
seriously, your overcomplicating this aspect of it.
go for both sides of the coin
it's just me, I have to concentrate on the things that bring in the most cash with the least effort
I would love a lil simple box with a FWD-OFF-REV switch and a pot to jog my machien once in a while.
making 100 of something, then answering the 200 questions, all for $5 profiut each, isn't good for me
Right, so go for the $25 kit and a forum
I need to make 10 of something, and answer only 30 questions, but make $50 each ;)
or $10 PCB and a forum
could be done, and I may.
if so, I'll send you the first one
you'll need to write docs than a layman cna understand (or that's what I would do)
That's sorta my style of writing whitepapers.
all part of the design process (or supposedto be at any rate)
the most difficult thing I can think of is the "bypass" of manual/auto mode
andything tri-state like that out there?
actually, it depends on the step inputs
if you don't pull low on the geckos, then you're effectively out of circuit
OH could use logic gates... that would be sneaky
if they're fast enough that is
logic is fast, but actually more expensive than microprocessors these days
well, cheaper than a mechanical switch
you still need the mechanical switch to decide whether or not to tristate
also realize that a depowered CNC may not allow anything else to control the drivers
well, I'm thinking: PC <---> Manualbox <----> geckos/xylotex
so that eliminates that issue, not it's a matter of an A/B (logic) switch.
Hmmm. a QPDT (Center off) might do it.
jmkasunich is now known as jmk_hiding
03alex_joni * 10emc2/scripts/emc.in: fixed some typo's on the ini-finding mechanism
alex_joni is now known as alex_joni_away
alex_joni_away is now known as alex_joni
skunkworks is now known as skunkworks_brb
03alex_joni * 10emc2/tcl/ (TkEmc tkemc.tcl): moved the location of default TkTemc. you can still have one in your own dir
03alex_joni * 10emc2/configs/TkEmc: moved TkEmc to emc2/tcl, users can still have on in their own configs dir e.g. configs/foo/TkEmc
03alex_joni * 10emc2/configs/common/ (6 files): moved to emc2/configs/common, was in topdir initially
ohoh check these out... http://www.infofreako.com/jad/enpitsu-e.html
now, I want my axis to be fully operational now hahah
03alex_joni * 10emc2/configs/stg/ (stg_io.hal stg_motion.hal): moved stg hal files to configs/stg
03alex_joni * 10emc2/configs/ (stg_io.hal stg_motion.hal): moved stg hal files to configs/stg
03alex_joni * 10emc2/configs/ (6 files): moved to configs/common
03alex_joni * 10emc2/configs/sim/ (sim.ini sim.tbl simulated_limits.hal sim.var): moved to configs/sim
03alex_joni * 10emc2/configs/ (sim.ini simulated_limits.hal): moved to configs/sim
skunkworks_brb is now known as skunkworks
A-L-P-H-A: why are you wasting our time with this cr... Holy Shit. Thanks for the pointer. Yup -- does make one want to add a 4th axis.
03alex_joni * 10emc2/configs/ (ppmc_io.hal ppmc_motion.hal): moved to the proper subdir
arrgh, why did I cvs up??
cradek: wait a bit..
03alex_joni * 10emc2/configs/demo_mazak/ (demo_mazak.clp demo_mazak.hal demo_mazak.ini): moved mazak stuff to configs/demo_mazak
03alex_joni * 10emc2/configs/ (mazak_rf.clp mazak_rf.hal mazak_rf.ini): moved mazak stuff to configs/demo_mazak
03alex_joni * 10emc2/configs/ (step_cl.clp step_cl.hal step_cl.ini classicladder.hal): moved to configs/demo_step_cl
03alex_joni * 10emc2/configs/demo_step_cl/ (6 files): moved to configs/demo_step_cl