EMC: 03cmorley 07TRUNK * 10emc2/src/emc/usr_intf/stepconf/stepconf.py: Added overwriting checks and copying of prebuilt pyvcp panels to usr config dir. Redid how the ladder program copies files...next id to have stepconf hook HAL up
hey cradek, have you tried tort.ngc or tort2.ngc with cutter comp?
(there were two, weren't there?)
SWPadnos: they won't work because they switch planes
hmm. so they do
I could write something if you thought it'd be helpful
not sure how it would be possible to spot something that's wrong
maybe if you did the uncomp path too, like my test programs
but if it was all overlappy, it would still be hard to see
g41 / O<nastypath> call / G40 O<nastypath> call
flowsnake would be a cool one to do that way :)
with several offsets
aren't its moves really short?
for higher order ones, yes
order 3 or 4 would be deep enough for offsets though. maybe even 2
yep. level 2 looks like a good line/line example
for both left and right comp
isn't flowsnake really just many instances of the same few cases
yes it is
60 degrees left, 60 degrees right
it's a better demo than test really
q: why isn't it working? a: I didn't compile
[03:05:09] <cradek> http://timeguy.com/cradek-files/emc/flowsnake-offset.png
what happens when the offset is a bit more, such that it can't enter the pocket formed by the 4th, 5th, and 6th arcs?
it will gouge the "part"
that's when we try to convince jepler to finish offs :)
it is not cam...
ok - only looks ahead 1 segment
yep, it does one corner at a time
it's not quite true that it only looks ahead 1 segment. it can queue up arbitrary number of things
but it is true that it only considers one corner at a time
you could quickly lose your sanity contemplating many-move lookahead
imagine moving into a corner and then moving up and down 100 times while changing feed rates and turning mist on and off etc
(this actually works)
it has to save all that crap up until it gets the move that completes the corner
this is all done in userspace, right? no connection to the motion queue
ouch. is that queue dynamically sized?
that spot in t3.png gets a lot more complex if the moves are in Z also
for example, starting low and moving higher each segment
that works fine
(it has always been able to do that)
there is a regression test, something-z-changes
well, it's harder to know whether you'll "gouge" in that case
comp is basically a 2d problem isn't it?
there could be intentional gouging to make openings that aren't full height
SWPadnos: yeah I don't know whether I should error on the gouges I can detect (the local ones)
currently it doesn't error, it just follows the path
dunno - hard to guess what the programmer meant
to some extent I think 'there is a good preview - he should look at it"
is probing allowed with cutter comp on?
ok, and drill/peck cycles?
well I should check that last one... I didn't see it
makes sense - you probably want to specify the hole center
EMC: 03cradek 07TRUNK * 10emc2/src/emc/rs274ngc/interp_cycles.cc: forbid canned cycles with comp on
because that couldn't possibly be what you would want
short downtime for cvs
2 seconds of notice :-)
well, the only time I could see it being useful would be to make a pocket with holes at the (comp determined) corners
but that's kinda dumb
you can sure z- z+ in the corners
oh, like I'd want to write a peck macro when there are perfectly good G7x or G8x ones already :)
but maybe I should have tried it instead of disallowing it
if you want to test it and it works, we can reenable it
but it would be a very strange application I think
well, I would, but I'm trying to see if opto22 input modules don't work with mesa in general, or if it's just mine
why wouldn't they?
but the one I tried didn't
steves_logging is now known as steve_stallings
I am pretty sure that Chris Helgesen is using an Opto-22 rack or a clone of one for inputs in his Hurco conversion.
I wouldn't be surprised
actually, it's likely that the module I tried is bad - the LED didn't light up no matter what I did (tried applying 5V in both polarities, toan IDC5 module)
Chris ran all his limits, homes, etc. thru the Opto-22 stuff. He has one of the 5I20 boards from our group purchase.
cvs service should be restored
ok, I'll keep plugging at it - the modules were in the racks when I got them off eBay ;)
cradek: nice dodge on saying exactly which of those things might not be allowable that you know are not allowed
jepler: confirmed that cvs is working
ok, it works better when you actually activate the power supply output
uh - well, I turned on the power supply that was supplying the opto22 board, and now the input module actually has enough power to work :)
no, I meant it's always those little details that get you. like troubleshooting for an hour before figuring out I reloaded the old code...
the annoying thing is that the voltage display showed something - 4V from the Mesa card (through a couple of cables)
[04:09:30] <cradek> http://timeguy.com/cradek-files/emc/lathe-concavecomp.png
it's just that it has an output disable switch that's hard to tell the state of
EMC: 03cradek 07TRUNK * 10emc2/src/emc/rs274ngc/ (interp_queue.hh interp_queue.cc interp_convert.cc): fix placement of concave corners on lathe
this will make lathe programming so pleasant
like, just program the part outline and then use cutter comp? :)
um, yeah - isn't that neat?
sounds like it
I'll let you know after I try that electrolytic rust removal thing :)
you can pretty much turn on comp when you load the tool, cut the outline, move away, turn it off, load the next tool
EMC: 03cradek 07TRUNK * 10emc2/src/emc/rs274ngc/interp_convert.cc: remove comments of doubt
seb_kuzminsky, something I started doing is using backquotes ( ` ) instead of double-quotes ( " ), that way it usually doesn't matter if people use them :)
(referring to your user list email)
run "lsmod | grep hostmot2" (without the quotes).
chris@rover:~$ `lsmod|grep ide_cd`
bash: ide_cd: command not found
what do you mean it usually doesn't matter?
(I wish there WAS punctuation that didn't matter)
right - nevermind :)
I got bash: CVS:command not found :)
could use something they don't know how to type: «lsmod»
that dude's email was totally confusing
sure, it expanded your directory into a command line and then tried to run it
most people can't find the backquote key ;)
ha, that's believable
took me a sec to figure that out
seb_kuzminsky, yeah, I didn't get it either, so I didn't bother reading it again (and again)
as a computer snob, it's amazing to me people can manage to download a bootable CD image and install and OS from it, yet fail to compose a legible email
I'm sure you'd do as well in Chinese ... ;)
is quoting etiquette different in chinese? if not, i guess i would do poorly
steve_stallings is now known as steves_logging
seb_kuzminsky: that people can't make legible quotes is half the fault of shit email clients, I suspect
and character set issues
only half, mind you
he was using yahoo web mail, that ought to not shit the bed i'd think
wasn't there just a (huge, OT) thread about how bad yahoo is?
I wouldn't know, I don't use any of that crap
yes, there was
sometimes at night I cry about web 2.0
I also don't know how the quoting is handled by the client - mozilla makes nice gray bars, but it figures out where to put them based on > at the start of lines
and it sometimes gets that wrong
yeah, and someone else might say that is the fault of us morons posting in ascii
instead of nested <blockquote>s or somesuch
but screw that person
how about that newly popular (as of OutLook something or other) way of quoting text in a narrow column under the response, with the responders "vcard" or whatever on in the right column
SWPadnos: thanks for helping dave - I thought he had done this before, but I guess not
(or have you not seen that?)
I've never even once used outlook
I wonder if I'll be helpful enough
but I get messages from people who do
and mozilla does try to display html messages
even from MS clients
I probably do too, but I see the text alternative part
do you know what i hate the most?
alex_joni: because it's confusing
grr tab completion
I'm sure it will be ;)
btw my county is on fire: http://www.dailycamera.com/news/2009/jan/07/i-70-closed-over-vail-pass-avalanche-control/
ignore the randomly-named url
my house on Old Stage Road caught on fire but didnt burn down, in the AAA-God fire of 1990 or 1991 or whenever it was
does this happen often?
lightning and wind start a lot of fires around here, plus the odd hippie
we get them from trains sometimes, but it has to be very uncommonly dry
AAA-God lit his matress on fire, then threw it out the window
what is AAA-God?
this crazy mountain hippie dude went by that name
so that fire wasnt due to natural causes, it was an act of god!
thank goodness his config runs
EMC: 03cmorley 07TRUNK * 10emc2/src/emc/usr_intf/stepconf/ (stepconf.py stepconf.glade):
EMC: Finished HAL connections for pyvcp. Added checkbox to allow/disallow automatic
EMC: HAL connections for pyvcp and ladder.Need to test on a machine with a parport to
EMC: make sure HAL connections are right.Probably need to work on the pyvcp panels
EMC: and evaluated if they should have different features.
EMC: 03cmorley 07TRUNK * 10emc2/configs/common/configurable_options/pyvcp/xyzjog.xml: sample panel -kinda miss named...
EMC: 03cmorley 07TRUNK * 10emc2/configs/common/configurable_options/pyvcp/blank.xml: need this too
EMC: 03bigjohnt 07TRUNK * 10emc2/docs/src/drivers/pluto_p.lyx: split out pluto p
EMC: 03bigjohnt 07TRUNK * 10emc2/docs/src/drivers/images/ (6 files): split out pluto p
EMC: 03bigjohnt 07TRUNK * 10emc2/docs/src/ (Master_Integrator.lyx Submakefile docs.xml index.tmpl): add pluto p to html
EMC: 03bigjohnt 07TRUNK * 10emc2/docs/src/hal/drivers.lyx: add pluto p to html
jepler: ouch, that's a nice example ;)
jepler: especially since interpreting is halted when M66 is encountered
I wish ken wouldn't feed the troll
I wonder if you interpret my email as troll-food :)
although I didn't mean it as such
any response to someone who just wants to whine/bitch/complain is troll food. but I see you didn't respond to paul so it's probably not.
"spiral" can be run with compensation; I assume it couldn't before since the commanded moves became very short near the center
neat. does it look right?
did you try both sides?
inside should fail as the arcs get too small
I'm not 100% sure of the last move "inside"
[16:42:19] <jepler> http://emergent.unpy.net/files/sandbox/fishy-lastmove.png
actually I guess it's the last several moves
highlighted to show what one move is: http://emergent.unpy.net/files/sandbox/fishy-lastmoves.png
arcspiral does error "near line 1000: Tool radius not less than arc radius with comp"
is that the same line where line spiral goes wrong?
I could detect these 'reversals' and give the gouging error.
but is the preview showing what's wrong more useful than a gouging error?
in line, are the compensated segments immediately before it veers off approaching zero length?
in sim/axis_mm, load http://emergent.unpy.net/files/sandbox/step1.ngc
(gives an error) then http://emergent.unpy.net/files/sandbox/step1.ngc
(doesn't) and look at the funny movement right at the center
then hit ctrl-r and see that funny movement disappear
just in case you can't get it from that descrption I can show it to you
when I load it, I get an error, but still get a preview
the second file to load is http://emergent.unpy.net/files/sandbox/step2.ngc
load step1 then step2
at the center there's a very odd arc (helix, really)
that is not there when you hit reload
I bet I'm not clearing the queue
I think it could be something like that
EMC: 03cradek 07TRUNK * 10emc2/src/emc/rs274ngc/ (interp_queue.cc interp_queue.hh rs274ngc_pre.cc): this fixes a problem where a correct program loads incorrectly the first time after getting an error with compensation on.
thanks, good catch
but your dro font sure hurts my sensibilities
I was trying to help someone get a truly grotesque font
I mean, large enough to be legible
EMC: 03cradek 07TRUNK * 10emc2/src/emc/rs274ngc/gcodemodule.cc: avoid "cannot do G1 with no feed rate" after switching units
EMC: 03cradek 07TRUNK * 10emc2/src/emc/rs274ngc/ (interp_convert.cc interp_queue.cc interp_queue.hh): allow units changes during cutter comp
EMC: 03cradek 07TRUNK * 10emc2/src/emc/rs274ngc/interp_convert.cc: three different things could previously cause the mysterious "gouging" error. replace all three with more meaningful errors.
* alex_joni hates "smart" phones
actually stupid phones, trying to seem smart
hmmm. so why exactly does the spindle speed value go negative?
because scale.gain is negative
SWPadnos: to drive a dac?
M4 causes motiuon.spindle-speed-out to be negative, and M3 makes it positive
I thought that's the most general interface to a +/-10V DAC
alex_joni, sure, but I wonder if it's more common to have a direct DAC vs. a 0-10 + direction setup
it's easy to do fabs() on that
way easier than trying to get the negative for reverse
(e.g. scale -1 * ..)
and a mux
not way easier, just easier
I think the way to "fix" this is have another pin coming from the motion controller which contains the abs value
I wonder if it makes sense to add a pin like motion.spindle-speed-abs
ok, good idea ;)
who/what are you trying to serve?
(not in this room)
so you're 20 also ;)
if it's gs2_vfd, fix gs2_vfd
a bit older :)
[21:27:00] <skunkworks> http://www.cnczone.com/forums/showthread.php?t=71072
it's motion, not the HAL stuff that actually makes the spindle turn
well.. it's actually electrons
no it's the electroweak force *carried* by the electrons ;-)
is that more like the aether or the phlogiston?
of course that makes sense
alex_joni: ah yes! that makes it all clear :o
seb_kuzminsky: obviously because it contains the Yukawa interactions between the fermions and the Higgs field
[21:29:50] <alex_joni> http://upload.wikimedia.org/math/9/5/b/95b992dfdd5c0b364636996f62688067.png
<- like that
bah, the higgs field is a myth old ladies tell their grandkids to make them eat their broccolis
I think that cradek's lathe and the mazak are two examples of machines for which a signed spindle command makes sense
while solving this problem, please make all the spindle speeds in per-second instead of per-minute
I think that guy asked that question once already, and skunkworks answered it
yep, he sent him the url :-)
STFW and RTFM all in one
jepler: I agree it's not a real issue having to use abs(), but I think having an extra pin which already does that is helpfull, and doesn't cost much
you'll get no argument from me as long as you switch everything to revs per second at the same time
that will break configs ;)
(although that hasn't stopped us before :)
perfect, do it now before 2.3 comes out :-)
seb_kuzminsky: no other time possible
are you saying all of time is before 2.3 comes out? :-P
all of time is now
marris posted the scematic of the g320 after someone posted a copy http://www.cnczone.com/forums/showthread.php?t=70788
hm now I want to know whether his quadrature counter (3 NXORs and two RC filters) works right, because that's not a combination I've seen before
(assuming it is correct... ) ;)
it is correct. the RC filters give you oldA and oldB, so it's easy to get step/dir from that
I see A and oldA, but the other RC is not giving oldB, it's giving something else
oh. maybe I should look at it then ;)
direction = B xnor oldA, cnt = (give an edge for every transition of A or B with 1/2 the phase shift between A and oldA)
I sure can't speak to the analog section, but using up/down counters and adders to compute a digital error and then converting to analog with an R-2R ladder sounds like the kind of thing that would be invented independently many times
hmm.. a lot more "stolen" stuff in the hungarian ftp linked from that thread
I notice the "copied" schematic uses a COTS chip to convert quadrature to step+direction, and in doing so adds a x1/x4 feature as a bonus
I am sure surprised that encoder input circuit is right
(surely it is)