Lerman_______ is now known as Lerman
[Global Notice] Hi all. One of our larger client servers just dropped off the network for a few minutes. It's back now and hopefully stable, sorry for the downtime and thanks for using freenode!
jepler: any plans for 2.2.8 today?
* alex_joni wants to plan his next 8 available hours
alex_joni: yeah, definitely today
umm, if I have it tagged in about 2.5h will that give you time to build the hardy/i386 packages?
I'll be back in 2h or so, so that would be perfect
I'll start then on hardy/i386, SF and linuxcnc stuff, but we'll sync then
ok sounds good
2.2.8 this weekend? Just curious...
that's what I'm hearing :)
mshaver: yeah, alex and I are going to work on it a little later
jepler: just wondering. I'm sending a some hard drives out tomorrow and I wanted to include 2.2.8 if it was ready. Not a "must have", but a "nice to have"!
jepler: emc reset blending mode to G64 after errors ?
jepler: I've added RS274NGC_STARTUP_CODE = G61 to axis_mm.ini [RS274NGC] and after that your test from friday will show what I've seen (I think it's error)
treat all modal codes as undefined after errors. if your programs are to run in g61 mode, place that modal setting in the part program. If you believe there's an error, please provide a step-by-step way for me to reproduce it and see it for myself
that means, after you get an error do not assume that the path control mode is still G61, even if it was G61 at the time of the error.
RS274NGC_STARTUP_CODE does not run at the beginning of every program, only at EMC startup
that might be contrary to what people expect, I don't know
but, like jepler says, if your programs requires G61 behavior, put G61 in your program
could be RS274NGC_AFTER_ERROR_CODE
EMC: 03jepler 07v2_2_branch * 10emc2/VERSION: bump version for release
EMC: 03jepler 07v2_2_branch * 10emc2/src/configure: bump version for release
EMC: 03jepler 07v2_2_branch * 10emc2/debian/changelog: bump version for release
alex_joni: 2.2.8 is tagged and I'm building the packages I'm in charge of
packages built, preparing to upload
jepler: I'm home now
checking out 2.2.8 now
I hear you checking i tuot
crap.. need to take care of some diskspace first
or something is making that machine make noise anyway
/dev/sda7 8.7G 8.2G 19M 100% /
* alex_joni hasn't started
do you use cvs co or cvs export ?
I meant which one :P
on my two machines (hardy and dapper) I have two different scripts written at two different times
one uses cvs co, the other uses cvs export
or should I rather grab the tgz and build from that?
the one you uploaded to hardy
sure, that one will be there soon
should be the same with cvs checkout.. but still (feels more correct this way)
running repo_put.sh for hardy now
don't rebuild the repo yet
oops, too late
ok, nm it then
shouldn't be too important
I'll let you know when it gets done, it'll be a few minutes at least
but I think (for the future) it's better to finish uploading all packages, then rebuild the repo and upload the new index files
let me know when the tgz is there
I agree, you're right -- I was not thinking all steps through.
as you said.. shouldn't matter with such a small repo as ours :)
on that note.. should we build sim for hardy?
.. still uploading the binary package ..
jepler: are you busy building, or watching a boring upload?
(I have some questions about PCB milling)
jmkasunich: bored mostly
reading the internets
you have one of those ebay spindles, right? and it has a 1/8" collet?
how do you drill holes? do you get drills that have 1/8" shanks?
yes and yes -- "wolfgang engineering" iirc
precisebits.com has lots of drills with 1/8" shanks
in fact virtually all their tooling is 1/8"
most of the tooling is the same length, right? so you don't wind up needing loads of Z travel?
jepler: 2.2.8-amd64 is up
yes. I think that with the depth setting rings all the tools I use extend the same distance from the spindle nose
their website mentions that some of the tools use a different distance from ring to tool tip but I don't have any of those
alex_joni: yep, it's working on the source tarball now
then I need to run one more time to put the -dev deb up
cool.. should be slightly faster
darn, looks like wolfgang doesn't have one listed right now (I'm looking at an ended auction)
e-mail him directly, it was the same thing when I wanted mine
I was just about to do that
is his e-mail on the auctions or do I need to give it to you?
ebay has a "contact the seller" thing, but if you have his direct email that would be better
when you get a chance, could you measure the length of the outer housing? the listing says 5" long, but I don't know if that is the housing, or the overall length including pulley and top of drawbar
jepler: I didn't notice emc2-dev_2.2.8_amd64.deb (does that come later?)
the pulley is designed to be nonremovable btw
12:06:07 <jepler> then I need to run one more time to put the -dev deb up
jmkasunich: I forwarded you the e-mail from the wolfgang spindle guy
got it, thanks
jmkasunich: 5" is the full length of the spindle including the pulley
alex_joni: source package done
alex_joni: about to put the -dev package, but I should be well out of your way by the time you want to rebuild the repo
starting to build stuff now
yeah, I want to test them before uploading.. so it's gonna take me 15-20 minutes at least
cradek: I'm starting to put the dapper stuff, so in 20 minutes or so it should be ready for your signature
jepler: let me know when you finished with the hardy repo
Preparing to replace emc2 1:2.2.7 (using emc2_2.2.8_i386.deb) ...
atomic power to batteries Robin
BigJohnT: do you have your emc2 machine handy?
jepler: finished signing the repo, uploading new packages & indexes now
jepler: have you poked around in that spindle? removed the collet and/or drawbar, closely examined the pulley attachment, etc?
* jmkasunich wonders how hard it would be to make a longer version
should probalby just ask the maker
seb_kuzminsky: To answer your question of yesterday (which I saw hours later), yes, hm2 is still working well for me. I do still have the following error problem which I have masked by setting the max ferror to near infinity. I'll have to get you more information on what's happening before I can expect you to help me. I'll try to do this later today (Sunday), or tomorrow.
mshaver: is this stepper or servo?
I'm interested in that too - one of my goals for the holiday break is to convert my shoptask from software stepping to 5i20 stepping
It _is_ smooth as silk, and I can easily max out my drives 200kHz max step rate.
I'll be reviewing the 5i20 stepper stuff, both firmware and driver, since I have strong opinions on how it should work
Good! If I halscop ferror during a move, it increases a tiny amount each servo cycle (linear ramp), then gains it back at the end of the move (linear ramp).
If the move is too long, and max ferror is too small (like 1.0"), then I die on a following error.
It's probably wrong to have to set max ferror to 100.0" ;)
it sounds like there isn't any headroom
this happens during rapids only?
jepler: should I start doing SF and linuxcnc.org ?
No, any fast move. The faster you go, the more profound the effect.
sounds like an max_vel/max_accel issue
I think you are supposed to run hm2 without setting an max_vel
alex_joni: not if it happens for moves below the limits
if the machine is configured for 100 ipm, and you get gradually increasing errors on a 50 ipm cut, it isn't a limit issue
unless it happens duing the accel/decel phase
I can certainly try it. Do you mean [AXIS]MAX_VEL, or [TRAJ]MAX_VEL?
iirc you always accel at max_accel
it sounds like the driver is running open loop, sending a frequency to the FPGA and not adjusting that as the actual position falls behind (or gets ahead)
mshaver: ask seb for more efficient advice .. I'm just thinking I heard something along the lines
I normally set traj and axis values the same, or similar - make sure that your move is well below both limits
as alex says, the accel will be at the limit, but if it is a long move, you should be able to tell if the error is increasing during the cruise portion, or only during the ramps
I think the hm2 driver has a way to set accel too, so maybe set that higher than the [AXIS] and [TRAJ] values
paste your halscope pics somewhere, imagebin or whatever
I think I'll get some info & scope pictures for this.
like you just said :)
IMO, the 5i20 stepgen should work exactly like the software stepgen, but with a multi-megahertz base thread
alex_joni: sure, that should be fine
ok, doing now
jmkasunich: no, I wouldn't really know what to look for. cradek has the identical spindle but more knowledge, he's probably a better one to answer your questions
* jepler leaves again
yep! so I'm sure it's not a case of not being able to clock the step lines fast enough.
I'll go fire up this machine & see what info I can gather up.
it was non-trivial to get the stepgen module to work correctly, that is one reason why I think 5i20 stepgen should duplicate it as much as possible - this is a solved problem, don't re-solve it
jepler: did SF, linuxcnc.org and wiki (hope I didn't miss anything :)
hm2 stepgen has both maxaccel and maxvel (independently settable for each stepgen instance)
but i think normally they'd all be set to 0, meaning dont limit vel or accel
traj takes care of that
(i may be misunderstanding how this is supposed to work)
i'd expect during the accel phase the ferror increases, and the traj moves the target point further away from the current point
during cruise it should be constant, and during decel it should drop linearly, back to zero at the destination
seb_kuzminsky: what happens if the position command makes a step change and the driver limts are at zero?
iirc, the software stepgen HAS to limit accel otherwise it steps between vels, one step per servo cycle
jmkasunich: hm, in that case it will start stepping at the max allowed by its four timing parameters, until it gets to the dest, then "stop on a dime"
so maybe stepgen.maxaccel is needed
the software stepgen applied both kinds of limits, _and_ it looks at the actual position and adjusts accordingly
since the velocity command to the hardware (or base thread) is quantized, there will be some error, so looking at feedback is needed
the hm2 stepgen looks at position feedback from the stepgen module in the fpga
each servo cycle it computes the position error
then it looks at its current (most recent, before this cycle) commanded velocity
if its time to start decelerating, it does that (at maxaccel)
if not, it accelerates, but vel gets clipped by maxvel
how do you define "time to start decelerating"?
then it outputs the new computed vel to the fpga
"if i start decelerating at maxaccel now, will vel=0 before or after the target?"
if it's near the target, it's time to decelerate
if not, keep accelerating
what is the target? the current position command?
does not compute
i dont really know what i'm doing
so your feedback is very welcome ;-)
well, I should get off my duff and actually start studying the code, and testing it on my RT box that has the 5i20 in it
instead of what I am doing now, which is wild speculation about yet another new project
maybe part of my non-computation is because i've been testing with HAL only, not with all of EMC2
what have you been using to generate the position command?
(crazy idea of the day: build a simulation model of hostmot2 with freehdl or ghdl, and run it on emc2/sim)
jepler: yes :-)
so it is a step function
SWPadnos and i talked about this a while back but no code came out of it ;-)
run that step thru a limit3 block so it actually resembles what comes out of EMC
step from 0 to 1 and see how it moves
ideally, the stepgen will accept both steps and ramps, and do the right thing
another thing i do is get position feedback from an encoder, and net that to the position-cmd of the stepgen
that seems to work too when i twist the shaft of the encoder
a knob chaser
i've got two steppers and three servos, and one servo i just use as an mpg
I should clear some space on my bench and get started
the other four motors try to copy its position
but at the moment I have a feline donut in my lap, and I hate to disturb him
jepler: yay, 2.2.8! thanks man :-)
heh I didn't do much for it
does alex_joni do the .deb builds? I dont really know how the release process works
jmk is either stuck under a cat or clearing space on his workbench
19 year old cat, it would be mean to evict him
wow, that's pretty old
seb_kuzminsky: I build the hardy/amd64 packages and the dapper/i386 packages. alex builds the hardy/i386 packages. alex and I have the signing key for the hardy repository (it's a non-personal key), so whoever uploads last signs the repository. cradek has the dapper signing key (it's his personal key), so he's the one who signs that repo.
seb_kuzminsky: and then there's stuff like uploads to sourceforge, notes on the wiki, and so on
alex's part is probably the least automated
jepler: is there a special blessed build machine for each (distro, arch) pair?
seb_kuzminsky: in principle I believe any machine should give a functionally identical build, but I always build on the same machines
and I don't use pbuilder, though I should..
seems like something the buildbot could help with...
i'd be happy to work on that if you think it'd be useful
* alex_joni uses the machine where he built the hardy kernel
i've been meaning to run the deb builds in pbuilder (to vet the build-depends)
hi alex_joni, thanks for your work on 2.2.8 :-)
not my work ;)
well i'm off to go sledding with offspring #1
see y'all later
* BigJohnT wanders off to the store for provisions as the weather looks nasty here for the next few days
Bitstream generation is complete.
created bitfile 5i20_SVST8_4.bit
jmkasunich: hope I didn't step on your toes / got it mostly right in spec2bit.py
EMC: 03jepler 07TRUNK * 10emc2/src/hal/drivers/mesa-hostmot2/firmware/src/spec2bit.py: run through to bitfile creation
got a bitfile, but don't have time to test it before I leave .. right now
jepler: my toes are happy
EMC: 03bigjohnt 07v2_2_branch * 10emc2/docs/src/hal/basic_hal.lyx: expand on setp a bit
EMC: 03bigjohnt 07TRUNK * 10emc2/docs/src/hal/basic_hal.lyx: expand setp a bit
before I commit is this correct in trunk? "TLO_IS_ALONG_W = 1 Moves the tool length offset from Z to W."
EMC: 03bigjohnt 07TRUNK * 10emc2/docs/src/config/ini_config.lyx: add tlo is along w
EMC: 03bigjohnt 07TRUNK * 10emc2/docs/src/ (docs.xml index.tmpl): add keystick to html
jepler are you around?
I think he's always away on Sunday afternoons
ok, thanks cradek