Is anybody left?
didn't think so, hows Stephan today?
03cradek * 10emc2/src/emc/rs274ngc/interp_convert.cc: remove debug message
Stephen is fine though ;)
well, if you don't, I'm not sure who else I'd ask...
Got a halscope problem
I don't know if its something I did to my .ini file or what but it won't start now, after trying to start it before emc2 was inited properly
OK. I may be able to help, thoguh that's questionable since I can't run it right now
it hsould start, but the settings may have been lost (or they may be invalid)
In which case I do what?
I think the settings are astored in ".scope_settings"
I'm not sure what dir that will be in for an installed EMC
Ah, I'll go see if I can find such a beast, back in a few. I ran updatedb as root before I shut the monitor off & came in to thatw me out.
heh - OK
search for Xilinx
SWPadnos live p0rn huh?
multiple perspective pr0n - heh, heh
hmm wouldn't it be funny if xilinx had an app note that did what you wanted
take a look to my toy :P
I've seen appnotes for SDI, JPEG/MPEG compression, gigE and 10GigE ethernet, etc
giacus: you need to do some down-sampling on that video
also stuff like IDE, SCSI, DDR bus, LCD controllers, etc.
ah so you "just put two and two together and get 75"
fenn: should be low enough 640
put two and two and five and 17 and pi^e together 75 times, yeah
as well, I used the sony camera :)
giacusL you finally build a sex robot huh?
working on that
if you call snapping your fingers and turning your head "sex"
it should be able to spy with his webcam
good angle from down there
yeah, it follow the sound very good
a bit slow ..
im thinking to remove some gear ..
SWPadnos: Doesn't seem to be such a beast as .scope_settings,
is there a .scope* anywhere?
Bunches, prolly 25 or so
right - thanks (I just found it in scope.c)
found it, should I just rm it?
There was also an error about unable to remove shmem 01 too
you could take a look at it, but removing it should set you back to square 1 (ie, no traces defined, like the first run)
Humm, 2nd line has MAXCHAN 0, and thats been one of tis error msgs too.. ???
I think that was 4 before the missfire
And it was set to show 4 channels
yep - that should be 1,2,4,8 or 16
you should be getting an error "could not set record length"
reset it to 4 and alls happy again. Now to the real problem, following errors
even at crawling along at 10% speeds.
again - man, will you quit it? ;)
Well, If I could figure out whats doing them, I'd be very very happy. Its intermittent
you have steppers right?
as in its running a loop that incs x for each pass around the y/z loop, and after about 2.475+ inches, boom, axis 2 error stops it, but when you try to re-enable, axis one goes into an infinite loop of errors
what is it doing, geometrically, exactly when the following error happens?
And, watching it with halscope, axis 2's following error comes and goes as a spike at about .050 down in the z motion
Its cutting a path, downward and forward in the G2 or G3 mode. This following spike, when it occurs, is 0.050 to 0.1 down in z from the top, start of the cut.
is it at a transition between two program lines?
the first is an arc, is the next a g1?
No, its well into that motion
Let me snip just the loop and paste its short.
are you running cvs head or testing?
cvs head, about 4-5 days old
n0120 #8 = [#8 + #7]
n0130 G19 G2 Y#2 Z#5 J#3 K#6 X#8
n0140 G1 Z#4 Y#1
(n0141 G1 X#8 Y#1)
o101 while [ #8 GE #9 ]
ha ok I can't make heads or tails of that
you've got some g2 and some g1
are the g2 helical?
they're in the YZ plane but also with X offset?
The x pos, is #8, the increment is #7, and theres a x thrown in else its a two piece g1 after the curve, and no, not helical, only about 100 degrees of an arc
so it's a plain old arc in YZ and it ferrors in the middle of it?
And I was having trouble before I made the x move part of the g2
Thats a pretty good description Chris
I still don't understand the arc - is it a flat arc only in a YZ plane, or are the two endpoints at different X coordinates?
totally intermittently, feed override to slow things doesn't seem to effect it
it looks helical to me, since X is moved to #8, which has just been incremented
ok that's what I thought too
* fenn wonders how hard it would have been to write code for real variable names in the interpreter
yes, the two endpoints are different x coorords, but remove the X#8 from the g2 move, and uncommenting the other line doesn't make any diff
fenn, not too easy, since G-code is specified to ignore whitespaec (or lack thereof)
I'd expect trying to interpret strings would slow it considerably
maybe in 1943
#var# could probably work
#var wouldbn't though
izazat a typu? 2043 maybe?
gene: ok so let's say your arc goes from 0 to 100 degrees - at about what degree does it ferror?
my guess is somewhere around 10 degrees. And it seems to move back and forth if the feedrate is changed, slow it way down and the error moves closer to the start of the move
ok, that's a sure sign of a blending problem.
have you plotted velocity and accel, to see which constraint is being violated to cause the ferror?
heh everything's a nail when you've got a hammer
And, like I said, its intermittent, it will show a low amp on the scope trace twicve in a row, then look great for 5 passes thru the loop, weird
And I haven't figured out howto do those plots yet.
ok, it's a little tricky, because you have to set up differentiating modules in hal
there is an example in the distributed sim.ini
but another approach might be to update your cvs to try the new trajectory planner I stuck in there yesterday.
not much use in spending time debugging the retired one
so.. is the new tp ready for prime time?
eh, guess it needs more testing
well it's in head
it seems to work better than the old one
only xyz are done, no rotary axes
I can do that I think, but lets make sure I'm using the right cvs up variation first
cvs up -r HEAD
use cvs up -dPA
use cvs up -dPAC
ah that's a good one
use with caution!
oh it saves to a backup, nevermind
I've got the -rHEAD running now, but its running silently with lots of bandwidth used.
I have a torrent running for ubuntu but its about petered out I think, upload share is nearly 5.
-rHEAD is still grabbing stuff, in src/libnml, doing one file at a time like its replaceing them all.
many files changed
Oh, and the -rHEAD is correct?
probably it's fine
I usually use -A, which means remove all sticky tags
gene: how much accel and vel stepgen headroom do you have configured?
Lots, it helps the following errors, a lot! Its building again now, but when thats done I can ge tthats tuff out of stepper_inch.ini if the HEAD didn't overwrite it.
the old planner would sometimes violate the acceleration constraint by 2x, so if you had MAXACCEL=10 you had to have STEPGEN_MAXACCEL=20 or more to be really sure it would never ferror.
now you can actually set them the same, but I recommend the stepgen settings be about 110%
(jogging may still ferror if they're the same)
From the TP section:
DEFAULT_VELOCITY = 0.0267
MAX_VELOCITY = 0.65
DEFAULT_ACCELERATION = 3.0
MAX_ACCELERATION = 5.0
ok in an AXIS section what is STEPGEN_MAXACCEL?
From say axis 1, all the same:
MAX_ACCELERATION = 2.9
STEPGEN_MAXVEL = 0.68
STEPGEN_MAXACCEL = 4.1
Somewhere around .67 is max rapid
ok I recommend you set all your MAXVEL the same (TRAJ and AXIS sections)
also set all your MAXACCEL the same (TRAJ and AXIS sections)
set SETPGEN_MAXVEL to 110% of MAXVEL
Ok, hang on a few
set STEPGEN_MAXACCEL to 110% of MAXACCEL
(on all axes)
maxvels all at .67 now.
ok so stepgen_maxvel should be about .74
Ok set MAXACCEL to about 3, you can hear them accel and decel that way. Any slower and its a slow start, then snaps to speed.
ok, so stepgen_maxaccel is 3.3 then
gene: there's documentation for the new blend tolerance mode at http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl/emcinfo.pl?TrajectoryControl
Ok, got 2 copies of that headed for the printer, darn that was slow took a minute to start
Now, back in the TP section default=2.0 maxvel=3.0 Right?
(don't think default is even used)
I've gotta run... Let me know how it works out! I'll try to check in later.
Ok, but that .74 may cause stepper locks, I'll test it and thanks a bunch Chris. I'll probably have to go feed the missus here at some point myself.
I've been meaning to ask again, how come I have beautiful fonts on this screen, and such shitty ones out there?
fonts are rendered by the X server on the machine whose display you're using, not the one the program is running on
cradek pointed out that the DisplaySize setting is important for small text
the large text should be taken care of with proper font setup, (which I can't help you with :( )
This is the readout numbers, the really big ones
Selecting helvitica after the startup helps, but even then its blocky.
well. all I can say is that it's the local X server settings, not the remote ones
beyond that, I'm not so good with X font configuration
(in fact, I could be wrong about that as well, but I think I'm not)
And, carving virtual metal, its quite happy now. And TBT, I've loaded this machine up with half a GIG of really nice fonts from Goldenweb.it, great site.
do they work in other apps
specifically, opther tcl/tk apps, if you have any
Yes, even in OOo composer or draw, made my xmas cards with them
Theya re installed in teh systemwide location, so they're all useable AFAIK but haven't tried in kmail yet.
I wonder if the exact font you have on the emc machine is installed correctly on the inside machine
Don't know, I've not removed any from the FC2 install though.
hard to know what's default these days
you also said that one is XOrg and the other is XFree86, right?
What scales the update rate for the backplotter? Its doing lots of bad traces as the virtual machine speeds up?
the AXIS backplot?
Yes, this one is Xorg-6.8.1-901
No, the tcltkj
No, the tcltk
I think it's CYCLE_TIME in the task section
Humm, if I can spare the cycles, that needs cut to about 10% of its current setting. Any other side effects?
none that I know of, just more CPU time for eye candy ;)
Also, I appear to be at a feed limit at an override=100%, good, I like eye candy, its about the only kind I can use now, with my sugar problems
feed limit as in "the motors won't go any faster, captain"
or feed limit as in "I can't move the slider past 100%"
And here I sit, sucking on my 2nd beer of the day, yes, they appear to be tapped out even without any power, and I can move the slider to 250%
Low cal beer even, rocks gReen Lite, 83 cals a bottle.
but from 100% to 250% is a no change in how it runs.
yuck - beer (or even near-beer)
ok. what are the programmed feedrate, and the max_velocity?
Well, I do like black jack too, but I long ago figured out I was an alky, so 2 a day is my self imposed limit.
In this little proggy, 110 I think IIRC
I have about 2/year, unless I'm on an airplane (Bailey's and coffee)
I've a half gallon of black jack my successor gave me the day I retired in 2002, still have about 1/2 of it, Baileys is the missus drink, when she does which isn't very often
a fifth lasts her a year or so
I guess I'm kinda like the missus then (in that regard :) )
a small amount is good for you the dr's claim, but more than 2 a day doesn't sem to have any more benefits
according to the sawbones anyway. I've seen some that took 2 6packs to make them 'right' too.
I've seen some that think they're right after a beer or twelve, but they're almost always wrong
It finished that program, which carves a total length of 6", .025" at a time with no more ferrors, this is great!
And, from the looks of the backplot, its moving some faster than my motors can keep up when power is applied I believe, so this is a marked improvement
One guy I used to like for his music, sang in the Long Branch, a bar in RCSD back in about 1961 when I was building titans, took him about a case to find his voice, and then he was great, till then just hoarse
I used to wait till about 10ish to go in because it made that much difference.
Called his band "John Paul Jones and his 4 5ths".
Damn, 45 eyars ago. That bar was gone by the time I actualy moved to RCSD for a decade.
or "10 5th's"
hey - that's before I was born. :)
He had 2 or 3 pickers and a drummer too. But not that much before I suspect, I've seen pix of you, 30ish I'd say?
heh. older than that. 37.5
but thank you anyway (such a nice boy)
I thought I detected a trace of grey in them locks :)
working on it - gotta catch up with the wife ;)
And here I am working on my 72nd, chuckle :)
Sitting here reading up on the TP, from that link back up the log now, sweet! Any of this yours?
all cradek so far
It sure seems like he has a good grip on this, thats for sure.
yep. and it should be simpler for others to get a grip now (though sadly, none of us has done so)
that I know of. I know I haven't taken the time
I expect I'd better go see if the missus is hungry yet. I looked at it a month ago but got rather glassy eyed in short order, too many vars whose duties were well explained, but thats just me I expect. Slowing down as the years sneak up on me, darnit! :-)
Talk to you all next week if not before, thanks, a lot, both of you.
lerman_ is now known as lerman
cradek: I don't know if you can fix this, but the thing I get when I click on the path tolerance example image is still the small image.
hm. With "O" words you can write an infinite loop g-code file. AXIS will never finish loading it (I assume, haven't checked) with no way to abort
cradek: I don't know what happened here, but it sure looks suspicious. At the time of the weird blend, I was moving the feed override slider, from 100% down to about 30%. http://emergent.unpy.net/index.cgi-files/sandbox/HEAD-funny-blend.png
the program is http://emergent.unpy.net/index.cgi-files/sandbox/circles.ngc
and I can make it jump circles repeatedly.
what's special about that particular circle?
fenn: Nothing; it's the circle where I moved the feed override slider
i mean, there shouldn't be anything random in the blending
fenn: it was made with love
cold, calculating love
it's not random, it's something that happens consistently when I use feed override
and it's nothing to do with O-words, I wrote the equivalent non-looping program (200 lines) and it works just the same away
it jumps from one circle to the next?
yeah it jumps to the next circle .. but then it does the circle again, the next time around
the jumped-to circle
can you reproduce it?
I haven't tried yet
and I notice it stays yellow
W T F
are those G2's or G1's?
they are g2
you can see the whole program at the bottom
there's a G1 at the "9 o'clock" of each circle, to move over to the next one. but the circles are circles.
I know what it is
you're triggering a premature blend of the upcoming g1
do you know if your circles are large enough that the accel constraint limits the feed?
the config file is /home/jepler/emc2/configs/sim/sim-axis.ini
I can put it somewhere if you like
I bet there's a bug in the blending criteria - it's based on velocity which might require some rethought for arcs
what's the difference between yellow and purple?
yellow is a blend
purple is an arc, red? is a line
ok so why is purple so far off the mark right before the misblend
to the left of the green X
mostly because GL can't draw arcs
notice how they're not round (the white lines)
yeah that's why i thought it could be G1's
unfortunately gl doesn't know anything but lines
hey cradek...couple of emc2 questions
ah cradek hides.
I'm waiting for the questions
sorry I guess I don't know the protocol
the easy one how to connect the dac to spindle
can't wait for the hard one
I'd ask fenn
er, which servo card is it?
* fenn fumbles around
general syntax will do
ie. what is the spindle called
dave-e: I have to make a part here, so can you ask me the other question first?
in the latest download.. emc2 tkemc displays G21 but executes in inches....if I declare G20 it says everything is out of range
the spindle is iocontrol.0.spindle-speed-out
ah... ok will hav to try that.
dave-e: what download exactly? a package, or cvs?
backplot looks like it is in mm
oh you're using the latest testing deb, I remember
wow, I can't imagine what broke that
what are your user units in the ini file?
well it sure is weird
dave-e: something like linkpp iocontrol.0.spindle-speed-out stg.3.dac-value
but when you start tkemc it thinks it's in mm mode?
within a couple of orders of magnitude
you haven't touched it recently? (I mean, you're sure it's not an order of magnitude off?)
mine is 0.03937007874016
ok that sound right...but I can check
* fenn thinks the whole .039.. thing is silly
there are four of them, one in TRAJ and one in each AXIS
check to be sure they are all the right number
I'll check them all
offhand I can't think of anything that would have changed this behavior.
I'll go down and check and report back
tkemc hasn't been touched, and the units handling of the interp hasn't either.
ok I need to make chips here (dust actually)
oh, it is probably something I did...finding it is always a problem
you never know, it could be any of us.
yes...I need to make chips also which is why I'm so uptight about this.
also check that all the AXIS are TYPE=LINEAR
bon soir as tous... ; )
et as vous
at&t announces its plans to acquire bellsouth
we are persecuting cradek for this: http://emergent.unpy.net/index.cgi-files/sandbox/HEAD-funny-blend.png
any news on the v22 osprey nose landing gear??
where's the blendin'
Messier-Dowty exhibits V22 landing gear at Paris Air Show - but that's old news eh
or maybe future news
mains... we've made all a along
way cool alu pcs....
we scraped 1 the last day of work last yr... bited
if we had all stayed home it would have been cheaper... ; (
shit for brains all of us... driven by schedule
and yr end
in "bon soir as tous" is "as" plural because tous is plural?
72 G USD
or is the s just because you cant write an accent
its formally correct
what makes the raw part cost so much?
alu forging .. with certain properties...
and machining technology required to make a cylinder out of it
its triangular in shape
the bore is triangular?
no the finished part is... the bore is UP 1 leg
wanna see something cool?
[03:29:51] <fenn> http://www.newscientist.com/channel/mech-tech/dn8802.html
dmessier....any connections with Dowty-Smith
cradek....mez culpa ... TRAJ value was still 1.0
fix that and things work better...
tk would not come up with the spindle dac command.
dave-e: oh that's great news
and spindle maxed out and stayed that
yes it is....hate to think I make those kind of mistakes but...
now to get the spindle going.
I looked at the demo_mazak stuff but that didn't help much
dmessier...are you familiar with the reusable release bar...for catapults?
who knows maybe I've got a typo in the spindle command line
it may just need another line for scaling
i'd check the signal chain starting at the VFD and go back thru hal and then EMC
well the dac to spindle drive is OK, it has not changed from emc1
so we probably need to check my typing and also figure out how to scale it so 10 v = 4K rpm
Anyone have suggestions for 1/16" spiral downcut bits?
dave-e: try gets iocontrol.0.spindle-speed-out
will do next time down to the shop
"watch halcmd gets iocontrol.0.spindle-speed-out"
or watch --interval=0.1 if it doesn't update fast enough
be back in a while
I haven't actually cut something with emc for some time, that sure worked well
jepler: this pcb is for a TSOP (?) chip with smd pads on .8mm centers
cradek: what's the project?
for ted's project
it's a carrier for the smartmedia chip he desoldered
03jepler * 10emc2/src/hal/utils/ (scope_horiz.c scope_usr.h): allow dragging on the horizontal area to move the plot
what's the horizontal area?
the place that shows what part of the plot you're seeing
can you lose a scrollbar then?
about the wiki page: alex put that there - I have no idea how to use images in the wiki
except that what this does is manipulate the state of that other widget (scrollbar, scale, or whatever it's called)
that's all for tonight.
fenn I found some more stuff in the demo_mazak hal file...now to try to understand it. ;-)
some blocks for scaling
[04:34:33] <A-L-P-H-A> http://cbs5.com/watercooler/local_story_059005903.html
check out the videos. heh.
nope. trying to get chatzilla to load the exact set of tabs I want at startup, and log the way I want
any idea why they say 'end mill' and 'router bit' here --> http://www.store.yahoo.com/drillcity/10comhobkit.html
router bit makes sense, end mills doesn't
ah, wait. the end mills can bore deeper, since the shanl is the same size just about up to the depth collar
the drills can't go that deep, and the router bits can be used for routing, which the drills aren't meant for
But the list all thee in that kit: endmills, drillbits, AND router bits.
But they list all thee in that kit: endmills, drillbits, AND router bits.
right. it's a combo kit, so you can drill, route, and err - mill
what sthe diff between endmills and router bits?
ty, at least I'm nucking futs
well, that remains to be determined :)
in this respect.... we all have to be a lil crazy to survive in this world
If you look closely at the pictures the end mills have flutes and the router bits have a dimond knurl type pattern on them.
generraly an end mill is used to cut a closed bottom pocket where a router is used to cout out the board or cut slots for mounting large compunents.
well... do you think the 'router bits' would shred wood?
[05:57:30] <Jymmm> http://us.st11.yimg.com/store1.yimg.com/I/drillcity_1883_534793
I have found that the router bit style cutter has a tendance to plug up on soft material. no place to the chips to go. They are ment for harder material.
SWPadnos is now known as SWP_Away
ah, ok. that makes sense. I just need 1/16" uncut anyway =)
err downcut I mean
down cut is ok if you are milling through the item it gives a clean to surface but if you are milling in a slot with a closed bottom there can be a problem clearing chips.
are you milling wood?
cutting out pieces or carving?
need the downcut for clean finish adn to prevent the material from being lifted up
a lil of both, mostly carving
I use standard 1/4 inch shank carbide router bits. 2 straight flutes. I have them down to 1/16"
by straight i mean no spiral.
I've been having good luck with CMT spiral downcut; but they don't go lower than 1/8"
They are available in square end or ball tip.
the CMT's? square.
how do you secure your work that you have problems with it lifting up?
generally how big are the pieces you work with?
up to 24" square, but on avg 8" x 16"
Iahve had good success with a combination of clamping and double sided carpet tape. The carpet tape is very thin and holds down the piece.
I'd be going thru a lot of tape =)
I have carved some "image to gcode" type pictures using tape alone.
carpet tape is cheap. don't buy it at the hardware store, make friends with a carpet installer.
I hate using tape... the adhesive residue is a PITA to remove most of the time.
It's bad enough I have to use it to cut lumber w/o chipouts.
funny. the tape we have available here does not generally leave a lot of residue. as long as you don't leave it for to long. It is not foam tape. some kind of plastic with adhesive on bothe sides. generally comes off clean.
I have some here, great stuff... too good sometimes.
I think I maght have asked you this before...what do you use for image2gcode???
I haven't done any with EXCEPTIONAL results (yet), but http://imagetogcode.com/
and BMP2DIN http://engraving.majosoft.com/html/body_free_software_tools.html
I like the "imagetogcode.com" program. USed up my 30 trial. trying to decide if I want to register it.
I used it and never used it! lol
[Global Notice] Hi all. Just a reminder: if you're experiencing the DCC router exploit with a Netgear or Linksys router, you can connect to freenode on port 8001 instead of port 6667, and you can disable SPI. Either should work.
[Global Notice] Both fixes should render the exploit ineffective. Also, please remember to identify to nickserv again; since our problem during the weekend, we had to reup all of the servers and some of you are not identified. Thanks.
what tooling were you using with image2gcode?
I have used 2 bits for image carving. 1/8" 90 degree "V-Groove" scoring bit and a 5/64" 2 flute straight router bit.
I have all three of these. http://www.dimarcanada.com/products/product_details.php?pLineIDVar=2&familyIDVar=11&categoryIDVar=151&groupIDVar=191
I'll chekc out the rest of them tomorrow... getting ready to hit the bed casue alex_joni is here now, and well, he'll just bore you to sleep
Jymmm: go sleep already
* Jymmm plays a alex_joni
good night Jymmm
giacus is now known as giacus_afk
Does anyone know about stepper motors here?
hi jepler :)
anyone tried the livecd ?
i heard that the graphical userinterface doesnt launch on most hardware :(
I haven't yet.
the only place I could try it today would be qemu and I think you already tested it there
giacus_afk is now known as giacus
cncuser: what abut the HW didnt work ?
amd processor ?
glacus: grafix subsystem
glacus: the probing of the hardwrae gives kernel oopses
know more about video card ?
glacus: try it :)
Is 75 ohm considered a high resistance for a stepper coil?
it works with qemu and vesa, matrox and xorg, nvidia and xorg
I've sis video card
this is 3 setups out of 10ths ot thousends
gacus: you can add your working setup here: http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?CncUser
at the bottom there is a section withs etups
ok, I've do download the iso and try it yet
Bo-Dick: it seems high
giacus: Will this make the motor weak?
what is bipolar ?
are you sure about the coil youre measuring
how many wires ?
They are stepper motors from floppy disk drives. I've wired them bifilar that means eight wires coming from them. Each cois is 75 ohms.
Bo-Dick: If they're from floppy disk drives, then they're designed to be run from 12V at most. That gives 1.8W, which is a pretty low power.
you won't run a very big table off of floppy drive steppers
What would happen if I gave them more than 12 volts?
You'll heat part of the motor out of spec.
Ok. Lets say I've got current limiting.
Bo-Dick: the yours is probably a very small motor having that resistance
You're still stuck with <2W no matter whether it's 12V all the time, or a higher voltage (24V?) that is on only part of the time due to the current limiting.
should be normal
the mine are 2 A , 1.1 hom
from a copier
giacus: do they have a voltage rating?
The prob is that the cheapest motors are 30$ and up. If I want a three axis system the price of the motors alone yields 90$. That sucks.
skunkworks_wrk: no voltage specified there
incidentally, the motors sherline sells for their minimills are 3.2V, 2A == 6.4W
giacus: What is the wire gauge inside the coils? Mine are 0.12mm thick.
4 phase 2 A. 1.1 hom, 200 step
giacus: 1.1 ohm?
[14:08:29] <giacus> http://www.micromed.it/Elettronica/stepper.htm
look at model M22
I bought 3 of that E. 30
around 70 hom you can find the M19 (very small motor)
Bo-Dick: try to search for datasheet
it could be somewhere ..
I think my motors are like this except that I've connected the coils separately (bifilar): http://www.doc.ic.ac.uk/~ih/doc/stepper/others/example2/
if your'e experimenting ok .. but remember that usually bipolar native motor has better perfomance
with just 4 wire
giacus: You mean 4 wire are better than 8 wire?
there are motor you can use as bipolar or unipolar
usually 4 wire motor has better performance
giacus: Do you know what "bifilar" means?
dunno what you mean
2 wires ?
or maybe youre talking abou coils construction inside
bi i italian mean two
[14:17:20] <Bo-Dick> http://www.cs.uiowa.edu/~jones/step/types.html
but isnt that I suppose
[14:18:03] <giacus> http://www.vincenzov.net/
[14:18:46] <skunkworks_wrk> http://www.allelectronics.com/cgi-bin/item/SMT-87/400600/STEPPER_MOTOR,_1.8_DEG__STEP_.html
[14:20:23] <skunkworks_wrk> http://www.allelectronics.com/cgi-bin/item/SMT-85/400600/STEPPER_MOTOR,_LIN_4218L_.html
skunkworks_wrk: Looks like a floppy drive stepper. They could have rewound the windings with thicker gauge wire.
[14:23:20] <skunkworks_wrk> http://www.alltronics.com/stepper_motors.htm
Bifilar motors are always best because they makes it possible to connect the windings in parallel.
Applied Motion Products NEMA 23 Stepping Motor
Applied Motion Products part number 44A501711 Bipolar stepper motor. This motor is a NEMA size 23 (approx 2.25" sq x 2" deep) the shaft is 1.25 x 0.25" approx. Specs are 60 in/oz, 200 s/rev, 2.3V, 2.3A. Large quantity available.
25M002 Unit Price : $7.95
that is a nice motor. I have played with them. cheap and 60in/oz.
skunkworks_wrk: Wish my local dealer had such cheap prices!
Unfortunately this isn't the situation.
where are you located?
allelectronics.com says they ship to foreign destinations, but I don't know the details.
I think my motors have this tiny wire gauge just because they are specifically designed to floppy drives and are intended to be weak.
When measuring the resistance in the coils with a multimeter the instrument shows 75 ohms on each coil.
This would mean that I would have to apply high voltage in order to get some torque from them.
Do you think anyone has ever used a floppy drive stepper in a cnc setup?
Yes. Succesfully, no.
and stepper motors are usualy ran at 20 to 25 times their rated voltage.
for cnc operation.
Jymmm: So you think my small stepper will blow if run at 25 times the rated voltage?
Bo-Dick: If it's really designed for 12V, then you're looking at 300V. Nobody runs steppers with a 300V supply
Gecko drives stop at 80V
What if I give them 60 volts?
Thats 5 times the rated voltage.
The purpose of using a higher voltage and a current regulator system is to increase torque at high step rate.
but it doesn't increase the holding torque or the low-speed torque
Does it decrease the holding torque or low-speed torque?
if your steppers are rated at 12V, they probably have some high induction on tham as well.
I don't know the inductance of the coils but the resistance is 75 ohms.
12vdc/75ohm = 0.16 Amps
Could I get some performance if I current limit at 160mA?
I guess the only advantage with these motors were that they were very cheap since they come from scrapped 5.25" disk drives. I've put some effort to convert them to bifilar so that I can connect the coils in parallel. I've heard that the motors are slightly more effective this way than if theyre connected in series.
03cradek * 10emc2/src/emc/kinematics/tp.c: fix premature blending caused by reduction in feed override
again - nice work cradek. (sorry I didn't get a chance to test it this weekend)
The problem with a real successful stepper cnc setup is that everything has to be high end technology in order to work flawlessly. The driver must support microstepping in order to get a smooth ride, the midband-resonance must be suppressed somehow and the motors must be powerful enough in order to get a good result. It feels like as long as one tries to compromize the result will be crap.
I've really underestimated the complexity of a successful cnc system!
Is g-code the same as "gerber"-code?
new AXIS release candidates. These are needed for the HEAD of emc2. http://axis.unpy.net/index.cgi/downloads/01141663025
Hi jepler, Is Chris around?
He was here earlier
Chuckle, caught ya!
how's your machine gene?
cradek: Its working great but the comnsole is drowning in a debug msg
something about straight feed? that's fixed
I fixed that as soon as cvs started working again last night
so I should do a cvs up -rHEAD?
sure, if it bothers you
are your following errors gone?
If you're already using HEAD, there's no need to specify "-rHEAD"
-r is "sticky" -- remembered from run to run
so use it when you want to switch from HEAD to TESTING or vice versa, but it's harmless to leave it off most of the time.
Seem to be, I see. My copy of the cvs docs sitting on the shelf is about a decade old :)
oh, I don
't think they've changed anything in at least that long.
;) That would be out of character
so a -dP would be fine than?
gene - so you are running cradeks tp?
And that aborted with no msgs. Yes yesterdays version
I have been itching to try it but time is an issue. ;)
I just did a (non-anonymous) "cvs update" of emc2, and it completed just fine.
he also has tollerance path following also - Cradek: is that in the head?
itch itch itch
Other than the console drowning in a debug msg, I'd guess it was moving a virtual cutter at 1.5-2x what my motors can do last night. Betterscratch that itch, sounds spretty erious
gene: I missed it - did you say the following errors you had are gone?
yippers Chris, thanks
I aven't tried the G64 thing yet, but will when I can see the machine next
I'm having trouble with anon cvs though
jepler: good grief
but cvs still doesn't like me either
I think sourceforge doesn't like cvs
Logging in to :pserver:firstname.lastname@example.org:2401/cvsroot/emc
cvs [login aborted]: end of file from server (consult above messages if any)
twould appear so, yup exact;ly the same here
Is srcforge running out of iron?
that is normally what I get - and that was a week or more ago
same here, it doesn't work for 2-3 days, then will for a couple of hours. PITA
cradek: what file is that in & I'll take it out myself
gene: find src | xargs grep "convert_straight:" will probably tell you
gene: I'll only tell you if you don't complain about merge errors later
gene: remember how this always ends badly? :-)
;) merge errors?
cvs merge conflicts
Humm, yes, it did the last time thats for sure but now I know about it
Its not that big a deal as long as emc isn't running anything when I want to start halscope
gene: maybe anoncvs will be back in a bit, it might save you trouble to just wait
Who would we tickle at src frg
Who would we tickle at src frg?
I think you can submit requests to a bug tracker type thing, but I'm not sure how
it looks as if a -rHEAD is going to work finally. I wonder if cvs has a max user copies limit of some kind?
looks like it patched:P src/emc/rs274ngc/interp_convert.cc
gene: yeah I made a few small changes to halscope .. nothing spectacular
But theres certanily some cold mollasses in the update progress
Done finally then ./configure --enable-run-in-place && make clean && make && sudo make setuid
couple of warnings early on, lemme see if I can get them out of the history when its done
about a kajillion warnings about a visibility attribute thats always been there, and this:
hal/utils/m5i20cfg.c: In function `findpcicard':
hal/utils/m5i20cfg.c:447: warning: implicit declaration of function `getline'
otherwise clean build.
that warning has been around a while...
what var might speed up the update of the backplot display, and can I go below 20 u-s now?
There haven't been any changes lately that would affect the useful BASE_PERIOD.
for tkemc and mini, [DISPLAY]CYCLE_TIME is used to control the frequency of updates
AXIS uses a value that is hard-coded in the source.
Ok, got that, reset for .1 seconds. I haven't tried axis in about 3 weeks now, lots of new goodies?
nothing much since 1.2rc1, just small tweaks.
ok, is this correct?
#- Cycle time, in seconds, that task controller will sleep between polls
CYCLE_TIME = 0.010
And I'm gonna try this again
#+ Base task period, in secs - this is the fastest thread in the machine
BASE_PERIOD = 0.000019
#- Servo task period, in secs - will be rounded to an int multiple of BASE_PERIOD
SERVO_PERIOD = 0.001000
the .000019 errored the last time
the syntax looks right
whether the settings are good for any particular machine, I don't know.
If .000019 didn't run reliably a few days ago, I don't think there's any reason to believe it will today.
jepler: I notice in axis the strings "directory" "file name" "files of type" and open/cancel are not translated
I mean, on the file chooser
can we mark those, or is it a tk problem?
right, from var/log/messages:
cradek: Yeah. The layout and text of that dialog comes from Tk. Perhaps there's some Tk thing that needs to be done to make it be translated, but I don't know what it is.
Mar 6 12:26:54 shop kernel: [1162929.346070] MOTION: init_threads() starting...
Mar 6 12:26:54 shop kernel: [1162929.346076] HAL: creating thread base-thread, 19000 nsec
Mar 6 12:26:54 shop kernel: [1162929.346113] HAL_LIB: ERROR: clock period too long: 19276
Mar 6 12:26:54 shop kernel: [1162929.346118] MOTION: failed to create 19000 nsec base thread
Mar 6 12:26:54 shop kernel: [1162929.346127] MOTION: init_threads() failed
Mar 6 12:26:54 shop kernel: [1162929.346147] RTAPI: WARNING: module 'HAL_motmod' failed to delete shmem 02
Mar 6 12:26:56 shop kernel: [1162930.827761] RTAPI: Exiting
Mar 6 12:26:56 shop kernel: [1162930.827805] RTAPI: Exit complete
gene: Based on that, I'd conclude that you can't use a base_period that is that low.
but maybe 0.0000193 would work, based on the number shown in that syslog message.
weird message... confusing even when it says its too long :)
I think I looked at the source code for you last time we discussed this
19000 is the amount you requested, and 19276 is the value that rtai offered. Since 19276 is "too long" compared to 19000, it doesn't work.
my home machine works fine at BASE_PERIOD 10usec
Ahh, I see. Interesting way around the rosebush there. And it worked.!
gene: if you're after performance in the gui, reducing BASE_PERIOD is the wrong thing to do
Well, I figured I had enoug cojones to run faster than 20.
gene: you should only reduce that if you need a faster step rate, and there's no sign you do
[16774.265948] RTAI[sched_lxrt]: hard timer type/freq = 8254-PIT/1193180(Hz); default timing mode is periodic; linear ordering of timed lists.
[16774.265962] RTAI[sched_lxrt]: Linux timer freq = 1000 (Hz), CPU freq = 598266000 hz.
[16774.265971] RTAI[sched_lxrt]: timer setup = 2010 ns, resched latency = 2687 ns.
Its running that code again, following errors are under a millivolt on the two main movers
Oh, the scale says 500 u-v in the halscope screen, litteral interp :)
gene: you're using halscope to find this? The sampling frequency of halscope is unrelated to [DISPLAY]CYCLE_TIME, and can be much smaller (right down to BASE_PERIOD)
The following error is probably updated each SERVO_PERIOD?
Actually its "500u/div" and its in inch
So I should call that what, anstroms?
So I should call that what, angstroms?
I'd call them microinches
microinch in this case then, ok
less than 1000th then, cool
You are using steppers right? Should it only be off the amount of your step amount? (following error)
And the backplot is a lot smoother too, might play with that some more though one corner is really rounded
yes, steppers, xylotex setup
gene: that's a feature. You can use path tolerance mode to get the roundness you want.
What I'm seeing is an artifact of the update rate for teh backplot, slow it down and the corners suare up, otherwise it cuts across them
cradek - what is the command for that? g64 p.010
skunkworks_wrk: yes, http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl/emcinfo.pl?TrajectoryControl
gene: when you slow down, the actual path's corners do get sharper.
gene: assuming you're talking about feed override?
Yeah, but this is equ to a .075" error visibly, yup, and I have nothing over 100%, the slider moves, but the rate is maxed out.
gene: you can easily get a .075" blend depending on your vel vs accel settings. If you want less, use the tolerance mode
I don't think I am, else the edges of this cut would be quite high, and they're perfect in white pine
you're right that it could be the gui update speed fooling you too.
No use wasting a $10 casting when optimizing code I say
The efect is reduced noticeably by speeding up the display time.
gene: you can increase the gui update usually
at least for tkemc and mini, don't think axis uses the stuff from the ini to set the gui refresh
Been testing the tp.
Thats what I did, cut it in half, and its tkemc. It apparently does get it from the .ini
and tp works great AFAIKS
Yes the refresh rate is settable from the ini for both tkemc and mini
With accel constrained as it is we should be able to increase the rate of pulses with stepper.
By golly it wasn't all the backplotter skip, I put in a G64 P0.0005 and the corners are now quite sharp in the backplot
I assume its slowing down at the corners and that give the plotter time to do it right?
In which case I should move that command to earlier in the code, the first corner is rounded about 1/4"
Hey, I might get this gcode thing down pat in another few years :)
Help. I've got input scale set to 16000. Max vel to 180 ipm.
Is it really spitting out 48K pps?
at 0.000020 yes
I think I'd need better amps and more voltage
rayh: I went up to 50-60 kHz before the RT fix (the fix JMK did a while ago)
BASE_PERIOD = 0.000010
yup, looks good for something around 1GHz (maybe even lower)
Dunno here, without stopping emc, it takes about 4 seconds to open the gcode file with gvim, thats getting a bit laggy.
sandwich time bbl
Really. That is an awesome improvement.
gene: maybe your BASE_PERIOD is too low
rayh: yeah, but it's not TP related, it was before that.. now the TP can keep up with those figures too ;)
alex - do you have a tarball of head?
Right. Before the new tp we had to waste overhead for accel.
skunkworks_wrk: I can make one, if you need
Chris: Does a g64 setting carry over when loading a new gcode file? 3D_Chips is running perfectly, but a LOT slower now.
alex - when I mouse over the top menu bar on wiki.linuxcnc.org - it disapears. IE
another question - when I try to do a cvs up -dP
it wants me to use cradeks password ;)
How do I reset it to work annonomouse? I did a cvs -d:pserver:email@example.com:/cvsroot/emc login
and is still asks me for cradeks password when I try to do an update.
alex_joni: are you there ?
cncuser: nope ;)
alex_joni: could you please remove the link or add a notice about it running only on a few grafics setups ?
alex_joni: im allmost finished backporting the whole stuff to puppy 1.0.4 :) back then puppy ran everywhere
ooh.. ok, I can do that
skunkworks_wrk: don't think you can.. use a fresh dir
ok - thanks
seems to do it on every page <-when I mouse over the top menu bar on wiki.linuxcnc.org - it disapears. IE
the penguin stays though ;)
huh? what browser?
when I scroll the top off the page - when I scroll back the menu reapears. but the second I mouse over it it disapears again.
skunkworks_wrk: there you go..
that must be the problem
format c: ?
allways a good idea
skunkworks_wrk: get firefox or opera
they work perfectly on doze, lots better than IE
I don't have that problem with firefox for linux :)
arachne is also a good choice
never heard of it?
i used that back in the days with a dos bootfloppy containing dialup, tcp, an ircclient, a mailclient and arachne, the graphical webbrowser for dos
skunkworks_wrk: I can replicate it on my IE too.. (what can I say.. it's IE.. buggy)
[18:34:42] <cncuser> http://home.arachne.cz/
izzat jurassic or what? :)
should still work ;)
If most people run ie - don't you think it should be fixed if we are trying to get them to run emc?
Feed that page to the w3c consortuims check page, I'l bet it will get a tummy ache.
[18:36:03] <cncuser> http://home.arachne.cz/netdos/
people who want to run emc2 don't run IE
hey - it took me a long time to try emc - I could not bring my self to try to muddle around a new operating system. I think a lot of people come from turbocnc (dos) to try emc.
i have a ibm ps/1workstation whos harddrivecontroller isnt supported linux. thats the only box id use dos :)
Controller not supported? Let the folks on lkml know, it will be next week
Can you get its signon ident?
Does this box have a cd in it?
skunkworks_wrk:Does this box have a cd in it?
the machine you're trying to run emc on...
gene: don't think the lkml folks care about an IBM ps/1
that's ancient ;)
- I have emc2 running - on multible machine
Thats possible I spose, and the reason I said jurassic :)
skunkworks_wrk: still need the tgz for HEAD?
How many ps/1's are left running today, bet its not many
[18:44:42] <alex_joni> http://lists.debian.org/debian-user/2004/02/msg00587.html
I am working on it alex - thanks
That reminds me of the mercedes commercial from a few years back, 1 million miles, 30 rusty years old, nothing but spark plugs (unless it was a diesel) & still ginmg after the groceries
darn, need to make a big picture .. wonder if I can
2x3m at 100 dpi at least
gene: 3dchips is probably slower now because the old planner violated your acceleration constraint. If your machine can handle more acceleration, crank it up in the ini file.
Well, the important thing is that it actually runs without following errors. IIRC the ini file says around4.1 max now
skunkworks_wrk: did you figure out the cvs problem? I suggest nuking that directory of my developer checkout, and getting a fresh one
cradek - I just csv-ed the head into a new directory on ubuntu. Now when I try to start stepperinch with axis I get a "cannot activate display program usr/local/bin/axis"
gene: if the machine can handle it, turning up accel will make a big difference. Mine runs fine at 20.
skunkworks_wrk: is this run-in-place?
skunkworks_wrk: you'll have to build axis from source too
been there. ;)
skunkworks_wrk: normally you could use the one from the package, but it's incompatible right now.
I might try that, I can hear it accelerating now quite easily
ok - thanks
gene: a good test is making a gcode file that goes back and forth in one axis, and cranking up accel until it causes the steppers to stall on reversal. Then reduce it somewhat (30%?)
I had given that some though, but hadn't located my missplaced round tuit :)
cradek - I can't get sudo apt-get install xlibmesa-dev libxmu-dev python2.3-dev
skunkworks_wrk: try sudo apt-get build-dep emc2-axis
skunkworks_wrk: that tells apt-get to automatically get the dependencies required to build the package
skunkworks_wrk: your ubuntu has python2.4, not python2.3, so that list of dependencies is outdated (maybe it's for bdi?)
cradek: does your ubuntu script put the apt-src line?
jepler: it didn't at first, but I changed it quite a while ago.
I'll change the instructions to mention 'apt-get build-dep'
I borked it.
Can't execute DISPLAY program /usr/local/bin/axis
Shutting down and cleaning up EMC...
Could not find pid(s) for task classicladder
it did let me run the sudo env EMCROOT=/home/sam/emc2sam python setup.py install
from the axis directory
I still think you don't have configure right... let's try it again
that I un-tared
go to emc2sam/src
look in Makefile.inc for RUN_IN_PLACE
I suspect it's no, but it should be yes
If you're configuring emc2 to install in /usr/local, then you should use EMCROOT=/usr/local
jepler: he said he was going for run-in-place
cradek: oh, I missed that, sorry.
jepler: the /usr/local thing makes me think it's compiled wrong
RUN_IN_PLACE = @RUN_IN_PLACE@
that's expected for Makefile.inc.in
RUN_IN_PLACE = no
sorry wrong file
so you want to do ./configure --enable-run-in-place
ok so I did a ./configure - then make then sudo make setupid then recompiled axis. still a no go. So I must be doing something wrong still
did you configure with --enable-run-in-place?
yes - I tried just doing that - no go. I still have to recompile axis right
[Global Notice] Hi all. Just a reminder. There's an exploit floating around that will affect you if you're using certain Netgear and Linksys routers. Someone will type 'DCC' followed by a string on an open channel, or via message, and you'll get knocked off freenode.
then you're in /home/sam/emc2sam and you run scripts/emc right?
[Global Notice] To render the exploit ineffective, please reconnect to chat.freenode.net on port 8001 instead of the default (6667). You can also disable SPI (stateful packet inspection). This may affect you, and be resolved similarly, if you're running a Norton firewall packet.
[Global Notice] There's a bit of info on my blog ( bloggage.org ) which you should feel free to link people to, though the site will bog down easily.
After ./configure --enable-run-in-place can you run with the other front-ends -- tkemc or mini?
[Global Notice] Also, for unrelated reasons the network was reupped over the weekend, and you may no longer be identified to nickserv. You should probably reidentify or restart your client to resolve this. Thanks!
(&& make && make setuid)
oops sudo make setuid that should be
cradek - I was not in the emc2sam dir. Used to the default ubuntu emc2 install I guess.
sam__: you should see yellow blends.
Humm, going from 3D_Chips back to my code, the scaling is AFU, what do I add to my code to restore that?
Said differently, whats the gcode to put it back in inch mode
gene: you should specify your units at the top of every program
Thanks, my copy os the rs-274 stuff is out in the shop
Thats better, Now, one other problem, using a 1/4" ball nose, I don't seem to be able to load that radius from the tool table, has no visible effect on the carving
But now I'd have to re-add it, the file has been reset I see.
cradek: - AWSOME.
that's good to hear - did you play with tolerance mode?
I ran it both ways. pretty cool
here is a strait g64 and then a g64 p.010 at 50ipm http://www.electronicsam.com/images/KandT/Screenshot-1.png
that yellow is ugly but I'm going to leave it in for a while - it's very helpful if there's another bug like the one jepler found
I don't have a problem with it. Kinda shows that it is working as it should.
what bug did jepler find?
skunkworks_wrk: Something to do with blending arcs and changing feed override
here is jepler's bug (I already fixed it) http://emergent.unpy.net/index.cgi-files/sandbox/HEAD-funny-blend.png
at the moment the trace turned yellow unexpectedly, I was fiddling feed override
(the bug has nothing to do with how badly the white path is followed; that's an axis bug/shortcoming)
ah - it is funny that in g-code you can't do a circle in anything but the 3 axises planes - yet cradek is making cool arcs where ever he wants. (blending)
I'm told they're actually parabolas
skunkworks_wrk: I think the planner actually supports arbitrary arcs/circles/helixes, it's just that you can't write them in gcode.
skunkworks_wrk: the circle normal is a vector, not an axis specification
I'm sure a line-line blend makes a parabola - an arc-arc blend makes ... some other shape
either way - pretty damn cool.
Maybe it's time to extend the interpreter again. g17.1 i- j- k- to set the normal vector used for arcs
jepler: I'll do the planning part if you do the rest...
we're departing from the kramer document more and more... I don't necessarily see it as a bad thing, as long as we keep up on our documentation
jepler: :tag PmCircle
I am sure that if I look at the interpreter's arc handling, my eyes will scab over and I'll never be able to develop axis again.
hey I bet alex can do it
* alex_joni runs screaming ;)
haha hi alex
arbitrary arc support in gcode
hrmm.. not XY, XZ, YZ plane?
I'm with Alex and getting a headache
cradek: I always wondered why it is limited to plane ;)
we'd call g2 "left handed" and g3 "right handed" and have a way to specify the normal vector.
But its not, I'm making a small x move while doing a yz arc right now.
but I think that actually we need arbitrary helixes
gene: that's not an arc, it's a helix
and in the grand scheme, correct
cradek: I think we could do this
alex_joni: I'm not sure how you would fully specify the helix.
but I'd rather that we start after moving testing again
or having a release already
like you said, G<whatever> specify the arcvector (normal vector)
That might not be too bad an idea, this is working great right now for me.
then use I,J,K to specify the helix
just like it's done now, but generalizing it
sample question: if a draw in xyz is done, how would one detect that the tool is trying to reach under an uncut area but its a straight sided tool and would then cut away the overhang you want to leave?
=rain on parade
emc can't do that, since all it has is g-code
gene: I don't think there's a general solution because your program doesn't specify the part you keep, only the part you cut away.
gene: that part is done by CAM usually
CAM software might do it by casting a ray along the spindle to the tip of the tool, and noticing that it intersects the "part to keep"
yeah, CAM has additional knowledge that gets lost on export to g-code
As any program does. So no, there's no general fits all solution. But cutting 45 degree slants the other way would be a little easier
right now, the only solution is a tilting z axis under 4th channel control. More excedrin headache trying to track the cutting tip in a 4th dimension.
Besides, I didn't buy the tilting post :)
gene: (MSG, Now please flip the workpiece)
4th dimension = time ?
gene: see it's easy to do that in gcode
(MSG, no, not like that, the other way around)
it even works if you put it the wrong way in..
(MSG, I don't think you got it back in quite the right place)
(MSG, keep trying till the little green light turns on)
Thats provided you have the workpiece in a shuttle that snaps in place to the required precision ROTFLMAO
I can see that in a war of wits, I am unarmed ;)
4th dimension= z angle (in radians just to make it interesting)
that reminds me of programming in polar coordinates on the robot
ugh.. that sux
seems like r->p is a pretty easy thing to do in software... I could even probably write that
yeah, it's easy to do.. but hard to grasp when using teach in
I can easily move the robot in joint space (the only problem I don't have enough fingers), also carthesian is a joy
but something weird like part-assigned coord system, that's yucky
I'm out for a while
cradek : Here ya go... a 10,000 Year clock http://www.longnow.org/projects/clock/prototype1/
logger_aj: | grep les
I'm logging. I don't understand '| grep les', giacus. Try /msg logger_aj help
giacus: do that in /msg
Jymmm: that is sweet
alex_joni looks beautiful, don't it
like a piece of art
alex_joni: is logger_aj a python bot ?
Jymmm: indeed it is
a bit poor
didn't bother to find a better one
if you know of one, tell me I'll change it
phenny is written in python
performe google search too
reminder and other
leave msg for peoples are not connected
that can be done with freenode
don't need a bot for that
how can I know latest time les_w has been here ?
or message him ?
some message to NickServ, can't remember
alex here is a strait g64 and then a g64 p.010 at 50ipm http://www.electronicsam.com/images/KandT/Screenshot-1.png
sorry giacus, it's MemoServ
try /msg MemoServ help
always if.. les_w identify to server when enter ..
skunkworks_wrk: ain't it sweet? ;)
that is unreal. very nice.
[21:19:00] <alex_joni> http://www.koders.com/info.aspx?c=ProjectInfo&pid=FYNZBP1UYCT185R6GQ651VFDBD
lol.. at 412704 lines of code
given a labor cost/month of $5000
using 90% of functions required (10% bad code)
crap - how are we going to get into the black? ;)
gives a development cost for EMC of $1,857,168
2 million dollars ;)
now that's realistic, I'd say
wasn't mach based on emc? I don't know if that would erk me or not.
well it is open source..emc that is
heh (51.41% - C; 30.09% - C++; 10.4%- Java ; 8.02% - Tcl ; 0.07% - Perl)
mach, in addition to the embedded code on the Gecko G-Rex, and I believe DeskCNC as well
SWP_Away is now known as SWPadnos
skunkworks_wrk: those all use the rs274ngc interpreter
dave-e: was just saying that :
23:19 < alex_joni> lol.. at 412704 lines of code
23:19 < alex_joni> given a labor cost/month of $5000
23:19 < alex_joni> using 90% of functions required (10% bad code)
23:20 < alex_joni> gives a development cost for EMC of $1,857,168
well, not I'm saying that, www.koders.com is saying that ;)
that's an interesting tool there
it really is.. rocked me pretty hard when I found the sources in there..
it is not very up-to-date
but search works nicely
I think you still have to count the %10.
cause every project writes code that later gets discarded.
actually, that includes code for emc, emc2, rcslib, and rtapi. there's probably some duplication in there
bill2or3: the 10% is documentation and other crap ;)
yeah, the other crap I mentioned :D
probably a bit more than 10% though
but even if 50% of it is needed, it's still a $1M project
still lots of money
it says it runs non realtime on nt - is that still supported?
hey , guys... as usual I need some help
yes, for emc1 (*not* BDI or emc2)
I'm trying to connect my spindle and scale it
skunkworks_wrk: I have a doze package if you want that..
dave-e: ok.. where's the problem?
no thanks alex - I think I am getting used to linux ;)
figuring out the hal pins and getting them connected...
dave-e: ok.. you are using stg, right?
dac out is stg.3.dac_value??
dac out is stg.3.dac-value??
ioconnect does the M3/5
here's what I would do..
connect the spindle to dac 3
then define a new hal signal (newsig spindletest float)
link it to the dac (linksp spindletest stg.3.dac_value)
set it's value to 0 (setp spindletest 0.0)
make the spindle drive active
slowly increase the value (setp spindletest 0.1)
and see if it's working
you make it sound easy
(note - this works in emc1 on the same machine, accordingto dave-e)
after this is working ok, link the signal to iocontrol (linksp spindletest iocontrol.0.spindle-speed-out)
you should need some scaling in there, I'd think
then you can in emc BDI, use S000, S100, S200, etc
I need 0.025/rpm
the output at the iocontrol.0.spindle-speed-out is in RPM
OK. good to know
so, for the scaling, use halcmd loadrt blocks scale=1
you need to scale it, there is a scaler available in blocks
dave-e: the 'loadrt blocks scale=1' loads a realtime component named blocks which will start a scaler
the blocks is a component full of goodies (scalers, comparators, and's, or's, etc..)
what's your max spindle RPM?
is it limited by the STG output?
at what voltage?
ie, 10V = 4kRPM
then it's OK
yep, no limiter needed
alex_joni: stg DAC's dont have a scale parameter?
fenn: can't recall :/
hey, they do ;)
they have a gain ;)
erm - w00t
so how do I get too it ?
gain and offset (same as the scale block)
dave-e: sorry about that.. you don't need scale (although that's also a way to go)
* alex_joni writes a sample HAL
hold off a sec alex
remember that jmk recently changed stg to use a divider rather than a multiplier for scale
you might just be able to use 4000/10 = 400 RPM/volt as the scale
and that sould be in the changes I got
but not for dac!?, that was for position
zero offset, 400 scale should do it, I think
gain gets multiplied
OK - I'm pobably wrong ;)
well since it is open loop I can test w/o hurtng anything
OK. just use the volts/RPM as the gain, 0.025
either "setp stg.3.dac-gain 0.025"
or put an ini var in there (there is probably one from the emc1 ini already), and use that
dave-e: I wrote a small sample HAL
"setp stg.3.dac-gain [MOTION]SPINDLE_VOLTS_PER_RPM
want me to paste here?
paste and I can print the page
it was MAX_VOLTS_PER_RPM in emc1.. right?
MAX ... for positive, and MIN .. for negative, I think
and MIN_VOLTS_PER_RPM (no idea what that did, offset?)
reverse speeds use MIN... I think
is that even valid in G-code?
SWPadnos: yes it is
ok. there you go
dave-e: got it?
got-it both as a file and print....now off to test...
dave-e: hope it works, but be carefull
if it doesn't complain ... if I'm not around, drop me an email at (alex_joni AT users.sourceforge.net)
alex_joni: do you have the URL for the latest puppy emc?
jepler: I do, but cncuser asked me to remove it temporarely
[21:50:17] <SWPadnos> http://www.cncgear.com/EMC/
alex_joni: OK, oh well
jepler: seems puppy >1.0.3 has issues
he's currently redoing it to an older version of puppy
I suppose I should leave in the version information there, huh?
SWPadnos: you might want to remove it for a few days..
alex_joni: He said something earlier about video card problems..
jepler: yes, it's very bad on HW detection he said
I can still give you the URL if you want..
do you think I shuold put in an older version there?
I didn't put any version, he said it's gonna be a few days, not more
ok. I can change that now.
do you know the nature of the puppy 1.0.3 problems?
HW detection problems..
1.0.3 is fine though
1.0.4 and older are getting worse cncuser said
1.0.4 and newer
1.0.4 and newer?
but it should work OK if it detects your hardware correctly?
he said on 10-20% of machines it worked
erk - that's bad
Erc is good
nice irc client for emacs :D
what erk mean ?
or "oh no"
the soothing sounds of strangulation ;)
sounds like python to me
in some comics I seen eeeek
eeeeeek is more of a scared sound
somethingh like aarrghh
just the other day a coworker was using erc and someone typed a commandline in irc of the form "... -- -t ..." and it translated it from morse code
Erc is cool
eeeeek is the mouse version of aarrghh
ok, so I put a note on that page saying that a new ISO should be available in a few days
mumble mumble ..
mean thinking ?
giacus: you know that?
in the comics
italian comics sound like thinking
oh, yeah.. thought you read about it somewhere..
depends on context, I think
could just be "kvetching"
SWPadnos: was talking about the latter part
hey alex-----worked first time off
no joke....now I'm up and running
so you're at the same level of hardware functionality as you had with EMC1 now, right?
thanks to a lot of helpful people
dave-e: you're about the second one (think the mazak was the first one) using spindle speed :D
I wonder if we should use your experience to fix up the stg sample config file
emc2 is just too rich to come up without help
dave-e: there is another issue I am concearned about
there is no such thing as a time between spindle stop, and brake on
the ini/hal files need changing, due to the stg position changes
how usefull/needed is that?
how did that work in emc1 ?
Mazak doesn't have a spindle break
any speed feedback?
ok, so I guess it doesn't matter ..
there is support for feedback ;)
spindle is servo with tach feedback
dave-e: you could use an ADC for feedback
it grunts and come up or down
and see it on the GUI
it would be much easier with encoder feedback
tach needs a bit of boost/conversion probably
nope plenty of votage on the tach
voltage divide for adc
ok, then conversion
it might be pretty easy to do..
scale and S/H
HAL takes care of most
the NML/emc stuff I've already done (need to search my brains for that ;)
adc doesn't convert well on a dynamic sig
I've not tested the ini ... MAX_VOLTS_PER_RPM yet...but will do so next trip to shop
going to switch systems here..bbs
ok, seems I didn't remember well.. spindle_speed_in is not used :/
* alex_joni will do that these days
Are the blueprints for the geckodriver available?
Bo-Dick: Don't think so..
Bo-Dick: no way ;)
I always tell people looking for a stepper drive to consider http://pminmo.com/l297-8/l297-8.htm
although.. someone is manufacturing a gecko ripoff
but they sell it for more money!
fenn: ROFLMAO, really?
heck why bother with manufacturing the thing if that's the business plan?
alex_joni: tormach used to sell it
I mean, why try to figure out the best way to design a stepper driver when the answer might be found out there.
it doesn't hurt to understand what you're doing
but i think you've been asking all the wrong questions
Absolutely not but I've realized the best would be to just take a working concept and rip it.
fenn, is it that "keingling" company?
i dont remember
robin_sz was talking about it a year ago
When having a stepper driver that supports microstepping, does the cnc software percieve that as a higher resolution stepper motor i.e. more steps/revolution?
the ProDrive 2000? (or was that the Tormach unit?)
so sometimes that's not good, as you need higher pulse-rates
so the very smart stepper drivers do some additional multiplying
That would mean that the only effect of the microstepping is to smoothen the movement?
ray was saying he got some pretty good speeds on steppers with the new motion module
SWPadnos: yes that looks a lot like a gecko
Bo-Dick: right, the precision is not something you usually need
alex_joni: i think it's stupid to multiply pulses inside the stepper driver.. it should just interpolate smoothly between them
fenn, it should look the same, the board was basicaly photocopied from a Gecko ;)
fenn: no, it's good to drive the motor with microstepping
alex_joni: Is there any rule of thumb to how much speed you can gain from microstepping compared to (non-torque-compensated) half-stepping, assuming that you don't run into trouble with the max step rate emc can deliver?
* alex_joni knows little about steppers..
oh alright then
alex you should have a hal file in your email soon.
jepler you might try MAriss paper - might mention something
do you want the ini also?
dave-e: can't hurt..
ok....on the list for the next shopping trip
gonna run...too much to do.
dave-e: glad you got it working ;)
THANKS FOR ALL THE HELP! :-)
jepler: It's hidden (again) --> http://www.geckodrive.com/photos/Step_motor_basics.pdf
what a nice guy
glad his machine is working
cradek: Oh, that wasn't sarcasim, huh?
he was a big help when we were working on Roland's Mazak
Jymmm: what? of course not
jepler: i dont think microstepping gains you any max speed, just reliability
cradek: Come on now... look at who the regulars are in here.
cradek you see the 10,000 year clock?
* Bo-Dick is gonna need an advice from the pro guys... here it comes
I need some help. The biggest effort I've done so far is the polarity shifting bridges for all stepper motors. This is the last step that drives the current to the stepper motor coils. Sometimes I've considered starting over with the stepper driver since there are so many details I haven't taken into account. My question is, does this implementation limit the opportunity to make this a more advanced stepper driver? What I mean is: Is this bri
cradek Ok, thought you might like that =)
Jymmm: I was a bit worried about showing it to cradek
it's very nice
wish the pictures were better
alex_joni how come?
in the end he might have ended up building one, and not working on emc2 anymore ;)
cradek: There's a link at the bottom to request hi-res photos
* Bo-Dick realizes this might maybe be a little advanced...
Bo-Dick: your post got cut off
Thats weird. I can see it in full length on my own screen. That must be a bug!
I guess I have to start from the beginning then :/
I need some help.
The biggest effort I've done so far is the polarity shifting bridges for all stepper motors.
This is the last step that drives the current to the stepper motor coils.
Sometimes I've considered starting over with the stepper driver since there are so many details I haven't taken into account.
My question is, does this implementation limit the opportunity to make this a more advanced stepper driver?
What I mean is: Is this bridge compatible with the PWM concept, microstepping or anti resonance etc?
Heres a picture of the bridge: http://www.carmi.se/misterstarshine/Projects/polbridge.gif
that looks fine
Jymmm: any idea about how to cook a cat ?
giacus: let it sit on your hot stepper motors
one isse might be the switching speed of the bridge.... pwm reqires fast speed switching to keep the switching losses low...
usually fets are used these days
fenn: ouch.. mi drivers have heatsink and fan !
you also need to be concerned with turn-off time being slower than turn-on
Why is that?
because otherwise you're shorting the supply directly to ground when you swap polarity (both transistors on one side are on at the same time)
wow i wonder if they used EMC to make that clock
hello miss K4ts :P
Not for common emitter fortunately :)
K4ts: we are talking about how to cook a cat
two provlems with that configuration
K4ts: any recipe ?
K4ts: really, it's very appreciated in thailand
1) darlington connection means that you always have at least 2xVbe forward voltage, which means high conduction losses
hello mr John
2) fairly slow transistors meand high switching loss
The datasheets says the bandwidth is 3MHz for them... :/
K4ts: did you understand ? :)
K4ts: cats are very appreciated in thailand restaurants
i no mongoly! ah ah
Bo-Dick: bandwidth is an analog thing, switching speed is different
oops: disregard my #2
K4ts: I suspect someone wnat roust you
because the emitter follower connection never saturates the transistors, they _will_ be fast
switching loss will be low
(but conduction loss will _really_ suck
Is this a behaviour for transistors specifically or due to a poor design?
mostly a result of the darlington connection
So the darlington does more harm than good?
there is a reason why MOSFETs have replaced bipolar transistors for all but the very cheapest, smallest application
jmkasunich do you finally bought the sound card ?
I don't like noisy computers
not yet ?
Seems hard these days to find a motherboard without integrated audio.
Bo-Dick: darlingtons are pretty outdated
dont' you like nice music ?
Does the PWM controller need to have access to all four nodes in a switching bridge?
now that you mention it, my mobo does have audio
but no speakers connected ;-)
Bo-Dick, not neccessarily
jmkasunich: Im' a singer I'm thinking to write a song for EMC
So the only thing that sucks about this design is that BJTs are used instead of switching mosfets?
so, got it popoular
jmkasunich: can write the notes
yes giacus ok
and you sing
oh .. K4ts you now we got a vip this month ?
stallman comes in sicily island in 27 march
want to go to see the conference ?
Does anyone know the design concept of the geckodrives switching bridges?
I know he write a song fo Free Software
Bo-Dick: I got 3 geckos ..
whait I go to sing
K4ts: wait :)
I sing stadio
Is it servo or stepper?
Bo-Dick: I know a pirated version of geckos
it is online at the official website !
at www.pirategeckos.com ?
tehy explain why pirated geckos are NOT as original
now I can't tell if you're kidding or not.
got a link?
depend on what youre thinking
aha - protobyte is the shithead pirate company
I seen it time ago
this may be the document: http://126.96.36.199/search?q=cache:znxautiwTSMJ:www.geckodrive.com/whyg201.htm+site:geckodrive.com+&hl=en&gl=us&ct=clnk&cd=6&client=firefox-a
jepler: yeah :D
it doesn't seem to be on the geckodrive site anymore
they probably took it down during the lawsuit
what Bo-Dick is tryng is the reason for what I spent 1 year searching ..
hope I'm missing
jmk, you're invited to hear me in paltalk
I think the secret is in the software
probably not readable
there should be no software in a motor drive, IMO
(if that's what you're talking about)
K4ts: jmkasunich don't have any sound card yet ..
SWPadnos: why do you say that?
SWPadnos: better change that to "there should be no softward in a stepper drive"
SWPadnos: I mean PID and other
are implemented by drivers
in the case of gecko
or a servo drive, I think (though I might be convinced that PID could be usefull in a DSP or FPGA)
darn, no copies of the image on the "wayback machine" either.
the USC is a pulse generator, not a motor drive
the EMC driver that runs the USC is good to have in software though ;)
* giacus likes K4ts ideas about EMC song :)
jepler, I think there's a PDF of that information somewhere, with the images
EMC should be more popoular
i sing for EMC sure
then, we'll write a song
and publish it everywhere
yep - the pdf is called "G201 vs. PirateDrive_2000.pdf"
and gcode to play the on steppers.
damn bill20r3 beat me to it
oops - no perios after vs
come la ascolteranno pi?
K4ts: can you sing in english ?
you have to learn ..
i think it would be better in italian
fenn: yeah.. but she acn do it also in english :)
it should really be in Gcode
if we write the text
she can sing
I will paly the guitar
just need a drimmer
use the Z axis, peck cycle ;)
who can writhe the text ?
I sing for emc i sing for emc whit base di singing on the rain
K4ts: you agreed to sing the song ?
"CAUTION: The following scope picture captures the death of a step motor drive. These graphic images may be unsuitable for children and sensitive adults. It is rated for mature viewers only."
now we have to write the text for the song
who can do it ?
K4ts: Jymmm is sleeping ..
I think he have not good idea's
dou U like california sound ?
bob dylan ?
let me think ..
the probelm is the text for the song
I've the base ready
we need to think about the chorus
Image there's no EMC ..
no develeopers around ..
something like that
imagine no rtai ..
uhmm no ..
K4ts: Image is the only song you know in english ??
mike buble ?
I want to go home !
and move to Gx20
colud be nice
pink floyd wish you ....
K4ts: any other song ?
too much old ..
allora ci penseremo domani
go to bed ?
K4ts: a domani
K4ts: ok,, night :*
se you tomorrow
see ya, alex
sorry guys.. just kidding
* Jymmm OTP
I'd get erasable if I were you
OTP can really suck ;)
otp == on the phone
Jymmm: pen and paper and write the text for the song :P
Jymmm: song should be talk about EMC
every turn 4 phrases
youre very inspired there :D
* giacus chechking if inspired is correct ..
should be :)
Jymmm: just start, we'll debug it after ..