[00:02:19] <skunkworksemc> http://imagebin.ca/view/pxM7L0N.html
ok, not much change
I'll see what happens here with offboard video
seems to help ;)
ok - one more.. http://imagebin.ca/view/MoU_f_H.html
it is a bit sluggish at the moment ;)
skunkworks FSCK U
skunkworks and the mobo you rode in on!
don't hate me because my latency is so low...
skunkworks Fine, I'll hate you for every other reason then! ;)
the grass is green, I hate skunkworks
don't hate me because my latency is so low...
I mean ;)
damn up arrow
now - do I reboot? maybe that will break something...
Evening all - question on BASE_PERIOD I was reading Andy Ayre's blog on tuning steppers - http://www.britishideas.com/2008/04/29/configuring-emc2-for-three-axis-stepper-motor-control/
In there he has a SLA7078 chip - I am using the same driver Chip.
If my latency is 18,000ns, and the CHip requires 12,000ns, then is it correct to say I can use a base period of 32,000ns?
one way to look at latency is that a pulse can vary from the nominal length by +- latency. nominal pulse length = 1 base period = 32000ns; 32000ns - latency 18000ns = 12000ns actual pulse length
Ok - So I could then use a base of 30,000 or a bit more if I wanted some room as an initial starting point for testing?
Running my stepper until they stall and then recalculate the base?
you would change the maximum velocity or acceleration to fix stalls; the BASE_PERIOD can remain the same.
Ok - I saw in his post that he would recalculate the base depending upon the speed attained at stall
what recalculating BASE_PERIOD does is free up a little more time for stuff like the GUI
I take it that the base should depend on the latency test and the driver specs and call it good
His was much worse latency (92,000ns) so then I assume that would make sense to do
fwiw I downloaded an SLA7078 datasheet and I am not sure where Andy Ayre got a 12000ns time for step pulses
I saw that number on it - there is a point where is talks about 7 to 12 ns depending upon the stepping level
my datasheet page 4 says "maximum response frequency" "fclk" "clock duty cycle=50%" "250kHz"
I would read that as saying you can drive it with a 250kHz square wave, making each half of the square wave only 2us big
if it turns out you're not maxing out your steppers at a 30000ns BASE_PERIOD, you might try going lower..
if not, no reason to tempt fate
yeah, I see the number you are referring to on the next page, but I'm not sure what it means
aren't datasheets great
I saw that freq note as well - not literate enough to understand its application
When I originally set it up - I thought the timing was related to the actual stepper response - took a whlie to realize you were all talking about the chip - I got there at least )
as an FYI - that is the driver chip used on the HobbyCNC boards made by Dave Rigotti
hm, whoever assembled this table on our wiki also came up with the 2000ns figure I did, but I have no idea if he was able to test it, or just did it from datasheets. http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Stepper_Drive_Timing
That seems quite fast
it doesn't mean the steppers will ever run that fast, of course
but they will probably stall that fast
I am moving my router in from the cold and will be properly tuning everything this time - I will report my results
Probably other tings on the baord which will slow things down
were you able to get the skew correction code set up?
(am I remembering the right person?)
I have it all loaded up on this machine - Just haven't mmoved the box to the basement yet - once I get the router set back up (tomorrow eve hopefully) and tweak the steppers, I will try and run it.
you recalled correctly
getting it to compile and install is the big hump
Someone mentioned I will need a home switch to use that - is that correct?
Thats already done
for a simple skew correction, the setting home should be enough on open loop
no -- you'll probably want to "do" homing, but that's as simple as jogging to a spot you can see by eye and hitting HOME
no switch required
I already do that )
Then my manual home switch will work.
huh that's a surprise
* jepler is trying out millkins.comp in the simulator
jogging doesn't do quite what you might hope
MDI and program mode do, though
the jog keys jog the motors, not the axes -- so as you jog with the up arrow, motor #1 turns and both X and Y change on the DRO (because due to the correction a movement by only motor #1 actually is moving in both X and Y)
with $ you should be able to jog either joints or axes
cradek: millkins.comp lies and says it's KINEMATICS_IDENTITY, so you don't get to choose.
I never considered doing that.
so it still run kins both directions?
it should, since data still needs to be copied in both directions
even for identity
it must be; axes are updated while jogging, and joints get updated in mdi
yes, if you change it to KINEMATICS_BOTH then you get $ to shift betwee joint and coord mode, and you can jog axes (but of course not incrementally
does it work right? I know teleop/rotaries are currently bogus
I just popped a starlite mint in my mouth, and it's not mint, it's cherry
that's what you get for popping strange named things in your mouth!
well, I just hope it was intended to taste like cherr
no doubt =)
you just ate the MILLION DOLLAR mint =)
it was more like dental cherry, so maybe it was the quarter million dollar "mint"
cradek: I didn't try rotaries at all
does this look clearer to anyone? http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Installing_EMC2#On_Ubuntu_6_06_or_8_04_from_source
(disregarding the fact that most here already know how to use CVS)
has anyone tapped Acrylic before?
only by hand
did you succeed?
did the threads stay intact for more then 48hours?
pipe thread, for liquid fittings
well, I didn't leave things connected for long
IIRC, it did seem brittle
that's what I am worried about
ds3: I have, successfully, once
ds3: for an acid bath, and it's still intact (been a few months)
acrylic is cheapbut I want to use screws... so it is either tapping it or pushing my luckand trying to install an insert
you have to learn to tap acrylic for maximum watercooling bling
jst_home: cool... did you do anything special to prevent cracks?
ds3: it was pretty gut wrenching... I remember I used a lot of oil. A LOT
what about trying to install a press in insert in acrylic?
ds3: and these were taperd threads too
jst_home: Hmmm... would you do it again or would you try to design around it?
ds3: with tapered threads... Not sure, if I did, I'd do it before finishing the rest of the part :)
ds3: non-tapered threads I think I'd feel a bit better about :)
and I suppose you'd also keep extra stock around? ;)
ideally, yeah :)
anyone tried pressed in inserts in acrylic?
anyone here know anything about low power wood burning lasers?
how low of a power?
night night guys
oh hey ds3, what's your class on again?
SWPadnos: got 2... which one?
ds3: < 5W, or whatever would be available for less than say $300 or so
SWPadnos: oh... heh... summary title is - resource constraint linux and the other is bluetooth basics
but more than 150mW :)
ok, thanks. night
oh... those DVD lasers do burn, just slow
ds3: I've got one of those sitting right here, and very little luck with that so far.
ds3: but it is entirely possible that I'm just not focusing it well enough
jst_home: paint it black with a sharpie and try again?
ds3: the wood?
black stuff works for me
oh, that's an idea... I'll give that a try!
ds3: I really do need to by those goggles everyone recommends... :)
wonder why ;)
I been look for one under $25 and no luck :(
ds3: yeah, seems silly to pay $50 or something for cheap plastic
but then again, it'd be kinda lame to blind yourself too
ds3: you have one of those in a cnc setup?
a laser, that is
no, not yet
I stopped after getting headaches from playing with them
been looking for a reasonable price google for over a year now
ouch, that's probably not good :)
I'm planning to do a rotary one here at some point, but there's quite a few steps left...
to "print" on dowels etc
oh rotary in thatsense... like a 4th axis
yeah, only it'd probably start out being a rotary plus only one other axis
i.e. rotate, and move back and fourth over the dowel
and then see how that goes :)
I been meaning to finish my 4th axis so I could do that mechanically
well, I've gotta go make some metal chips here for a bit, bbl
[Global Notice] Hi all, I'm going to need to take down services (Nick,Chan,MemoServ etc) for a few moments to apply a tiny little fix. It shouldn't take more than a couple of minutes and I'll let you know when they're back up again. Apologies for the inconvenience and have a good day!
[Global Notice] Hi all, as promised I return services to you, I'm all done and they're nearly as new! Have a good day and thank you for using freenode.
how do i know if i'm useing the new gforce video card instead of the m/b ? ....http://pastebin.com/d4f8255c0
it's an agp card and i have apg selected in the bios
i'm still getting the realtime error after giving the controller card its' own power supply and a new video card
piasdom: obviously you connect the monitor to the other VGA output
alex_joni: i have the monitor connected to the apg card .... nothing is obvious to me :)
if it's connected to the agp card, then I'd say the AGP card is working
alex_joni: ok great..but why the continuous realtime error ?
did you check other causes?
[13:00:27] <alex_joni> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?TroubleShooting#Unexpected_realtime_delay_check_dmesg_for_details
i don't know any other causes...i just worked on the controller card and got off the m/b video...what else is there /
this is my dmesg ... http://pastebin.com/db7a975e
piasdom: good, now pastebin the result from a latency test
and i don't know all the effects latency...i'm a machinist and new to this computer control
[13:07:24] <alex_joni> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?TroubleShooting#Unexpected_realtime_delay_check_dmesg_for_details
ok...last time it ran around 60000
how long did you leave it running?
a few hours
but just started test and it's at 1800000
i now that's can't be good :)
sounds worse like the onboard card
try rebooting without the AGP card, and see what happens
change to pci in the bios ?
that setting is irrelevant
but probably PCI to be safe
could try disabling USB
brb...hopefully :) ok..no usb
cant imagine why he consistently gets crap results
not generally crappy, just RT-crappy
he has been at this for some months now
nah, there are loads of celeron RT-crappy
archivist: what can I say...
make one want to send somebody round to see whats really there
it's a bit far from here
if my latency results are around 25000..how much higher do i set the base ?...1000....10000....or a hundred million thousands ? :)
sorry - no latency test results this morning...
Title: Genetic algorithm based optimization and post optimality analysis of multi-pass face milling http://arxiv.org/abs/0902.0763
piasdom: around 30000
i don't have any results yet..was just wondering what to do with the results..so if it 28000, still set to 30000 or set to 33000 ?
piasdom: start with 30000, let emc2 run
if you get an error, increase it until it goes away
some headroom is sensible
once you have a value which doesn't cause problems, you can start messing about with speeds & accels
cool, thanks alex_joni
i left test running and closed all apt. except ksirc,test and opened adept...before opening adept results were 41000..after 1000000
would a wireless mouse effect latency ?
it's usually video, especially if the latency jumps when you're opening/closing/moving windows
so it could be nvidia ?
say.. i'm too lazy to read docs.. why does sim/axis want to home on startup and then says "cannot home while shared home switch is closed"?
home before you jog
the simulated config has simulated switches, and the current machine position is "on" one or more of them
it's a symptom of trying to simulate too many things (it has home switches you can't see)
well, i disabled usb and used nv driver and i got 500000 within a minute
had envyNg and removed that...still 500000
rebooted a number of time and now it's 420000
piasdom: use visa driver
i'll try that brb
yea! my 5i20 and 7i31 arrived
ooh new toys
what are you going to retrofit?
I'm testing a THC using a custom board that Peter is making with the 5i20
opened openoffice doc and it jumped to 200000 useing vesa driver
I don't see any docs for the 7i31 on the mesa site :(
any of you guys use it before? (the led card)
I used it once
does it need power to the small plug?
you just plug it between the 5i2x and daughter card, and it tells you what's happening
no, I don't think I did anything with that plug
there is 5v available on the ribbon to power stuff
sometimes there is a jumper for internal/external power
you don't have to supply external power
be careful if yo decide to - don't use a floppy cable. the 12V line will be a surprise :)
yea it has a twist in it :)
I am not using those external power plugs on any of my setup
you have twisted floppy *POWER* cables? :)
oh I thought you were talking about the flat cable
nope - the flat 4-pin power connectors fit, but are wrong
* BigJohnT searches for his "heh" key
SWPadnos: that was a mistake IMO
the connector should be compatible with a floppy power cable, but with 12V disconnected
or it should be different
(or heck, use 12V on some boards, with a regulator)
thanks for the heads up SWPadnos
that's one place where a couple of screw terminals would be very useful
SWPadnos: I assume, but don't know, that the "12V" pin would be a no-connect on the mesa boards. Doing otherwise is just asking for trouble
I'll have to remember that before I naively hook one up
the two center pins connect together, as do the two outer ones
and they connect directly back to the 5ixx card too
so the ground reference should be PC ground (of course, it's before the isolation)
so basically you would short the 12V and 5V rails of your PSU together
and connect that to the 5/3.3V supply from the anyIO card
I'm not sure which wire would vaporize first
let's try it!
I bet the PS would just not come up
you'd have to power on first, then plug it in "live
I'd be placing my bet on the (probably short) trace between the pins on the mesa daughterboard
then if you're really lucky, the blown trace leaves 5V disconnected at the daughtercard and 12V connected through the ribbon cable back to the PCI card
which is set for 3.3V
no wonder my mesa boards don't work
I think I collected all the smoke
I figure when technology advances a little more, we'll be able to get that smoke back in those chips
did jmk have the same problem considering his blog name :)
like the people that have their bodies frozen after they die in hopes of being cured later
what's his blog name?
I'm sure he turns electronics into smoke as part of his job
magic smoke http://www.jmkasunich.com/cgi-bin/blosxom/photography
I was supposed to have built a 3 phase drive that would let the smoke out of the power devices slowly
haven't done it yet
there is no slow when the smoke wants to get out
halcmd loadusr hal_input -R 0 seems to work when moving the mouse to the bottom right, but dies when moving upper left (negative X or Y values), with this error:
TypeError: Integer expected, not long
I still remember the day I told a student to test the short circuit protection on an amplifier....it failed
in line 31 of hal_input i printed self._drive.items() and got ('0.rel-x-counts', 36L)
basically it adds an L on to the current value when you go negative
er.. when you move left
with the mouse
* fenn hates english sometimes
fenn, TRUNK or 2.2.x?
fenn, you forgot to put the actual error in the bug report
I only see the reaceback
TypeError: Integer expected, not long
this is the error, I assume
this broke it http://cvs.linuxcnc.org/cvs/emc2/src/hal/user_comps/hal_input.py.diff?r1=1.16;r2=1.17
"Patch from Brad McLean: Handle HIDs that give low-resolution signed values for relative axes. Allow suppression of duplicate events from devices that generate them unnecessarily."
you could try setting the precision to something other than 32
but I don't know if that will help
well, there's no reason for this not to work:
import hal; h = hal.component("test"); h.newpin("number", hal.HAL_S32, hal.HAL_OUT); h.number = 1L
+ sign = ev.value & ( 1 << (prec-1) )
precision defaults to 32
that may be one too many
oh, -1, nevermind
huh, what a coincidence!
fwiw I think halmodule should be changed so that 1L is acceptable
I'm about 50% done with that
Hi, I'm just jetting by on a lunch break, but I'll take a quick look.
you guys are awesome
eric_unterhause1: says who?
you should have patented the idea while you were ahead
what's so green about cardboard?
it's usually made from recycled paper
recycling it is pretty nasty from my understanding
i dont get why they have the usb ports sticking out at a nice snap-me-off angle
alex_joni: did I have this idea? it's such a stupid one, I'd rather I hadn't.
one of the grad students had an external usb hard drive case in a cardboard box
yea it's way more environmentally friendly to just have the motherboards lying on the desk in a pile
jepler: you had a thingie in a cardboard box with a red light ontop ;)
bolted to the side of the desk like one of the founders of Sun did
I recycle computer cases fairly frequently
too bad they keep changing standards
I just put the new computer in the case I already have
I don't remember the last case I bought new
alex_joni: oh, that -- it filled me with shame
I just bought some new ones
(or the last computer I bought new for that matter)
I think it was a dual pentium 166
I got an HP one new because it was so cheap
I usually buy new mobo, memory hd for my main desktop
all other comps are recycled
dual core athlon £115 ish needed a bit more ram though
[17:57:38] <alex_joni> http://isc.sans.org/diary.html?storyid=5797
Okay, here's the story, I think (I'm not at home, this is a quick code read). The last line of my patch, 162, is:
self.set(code + "-counts", self.get(code + "-counts") + value)
Once things go negative, some of the code above can cause the 'value' variable to become long. It appears that self.set() is not restricting it's value to S32, but is accepting that long.
A fix would be self.set(int( ... ))
But I can't test right now, or this evening, unfortunately.
self.set(int(code + "-counts", self.get(code + "-counts") + value))
ValueError: int() base must be >= 2 and <= 36
ok that appears to work when i put the set in the right place
self.set(code + "-counts", int(self.get(code + "-counts") + value))
incidentally, this doesn't fix the Contour jog/shuttle problem
ah no, it doesn't do the right thing after all
"It works on my machine" :)
now i only get movement to lower right
What are you seeing with your Contour?
fenn: I'll have to dig deeper then, next break I get. I'm on the road on a consulting gig at the moment.
bradmc, yeah - the jog/shuttle (non-Pro) reports both the jog wheel and shuttle "disc" as relative axes, but they're not
so when either gets to a zero position, there's no report (since relative 0 = nothing useful)
I think there's a quirk that can be set in the kernel inupt code, but I haven't gotten that to work yet
Did you set -absolute to True in the HAL?
x/y relative motion works OK but wheel-counts only goes more positive
there's no report when it reaches zero
Okay. I'll chew on that. Gotta run. Thanks both.
so it almost works, but it goes from +1 to -1 or the opposite
ok, see you
I fixed up the hal module to accept python longs for s32 pins as long as they're in s32 range, but I don't know whether that'll make hal_input work as desired, or just stop it throwing python exceptions.
i just had the strangest rounding error on a conversational control
the machine apparently rounds down, universally
so there was a .0001 difference between an arc and a line it lead into
so instead of cutting the line, it decides it was going to rapid across it
this machine requires user input to start any and all rapid moves
not this one
ouch, rapid instead of feed is bad news
strangest thing ever
and the feed rate for the line was 4 ipm
i have NO idea how it decides to go to full rapid
but i changed the direction of my rounding, and now it's running peachy
but i just don't understand how it did what it did - if you enter a disconnected line, it pauses and tells you it has to rapid to the start of the new line
this thought it was connected, because using regular rounding logic, it was
strangest thing i've seen so far
aside from the horizontal mill we have whose Z axis decided to take off at 800 ipm straight forward while it was just sitting there one day
bring a spare pair of underwear!
and don't put your hand in front of that axis without the e stop on
also, since you guys have pretty varied interests, does anyone know how big a hybrid rocket (of any variety of your choosing) would have to be to provide 20K lbs of thrust for three seconds?
something big enough to hold enough fuel for 60k lb-sec impulse
but how big is that, in diameter/length/weight etc
i am scouring wikipedia but no avail
I'd guess 1-2m diameter, probably 5m long
maybe smaller if it's only 3 second
hmm.. thrust-to-weight ration is 4.0 up to 136.66
the NK-33 provides 339k lbf average thrust
and weighs about 2700 lbs
so maybe scale that by 10
300-400 lbs the engine + fuel
that is awesome
but remember that the NK-33 has the best thrust/weight ration in the world :D
uses oxygen rich fuel
* alex_joni reads wikipedia for 2 minutes and starts to talk about rockets like he knows anything
section: Energy Efficiency
"Oxygen-rich engines were never successfully built in America. The Russians however perfected the metallurgy behind this trick"
i am getting around 10 cubic feet of fuel that has an impuls of 440 lb-s per liter
to get a 60k lb-s impulse
and that is a very conservative number if i am reading the charts for impulse for various fuels right
all this leads me to believe that it is feasable to run a 3 second quarter mile in a 2000 lb car using a rocket
not that anyone would, just wanted to see if it was feasable
good night all
hmm.. 10 cubic feet in 3 seconds..
the rocket-turbine (turbonique) only managed 7.3sec
10 cubic feet of solid rocket propellant
and how much liquid?
and i figure in order to get that high a number, it's probably not feasable to use just one rocket
no idea, i haven't picked a specific fuel/ox combo
if you're strapping a rocket to some thing with wheels, perhaps you could get rid of the engine
and reduce the amount of mass you have to push
this would be a missile with wheels
and a seat
no engine, drivetrain, etc
so the 2000lb requirement kinda goes away
well that's my conservative, "is this possible" guess
yes it's possible, the question is how many G's you can handle
i calculated 9g's
for a 3 second run
put a nuclear bomb in a well and it'll shoot the wellhead into orbit
er.. escape trajectory
600 mph @ 3 seconds is the trap speed
9g's to get there
and then probably 4-5g's on the way back down
there might be a way to design the fuel casting to have enough internal surface area to get a 1/3rd volume burn per second
that would be easier than trying to synch and throttle 2-3 engines
in large solid rockets they cast the fuel grain in a lobed shape, rather like an internal spline
oh hot, that's what i was thinking of
so as it burns the surface area decreases (giving constant thrust with declining mass)
so maybe one rocket is still theoretically doable
a long, thin motor with a lobe/spline cross section
not that this will ever happen but it's fun to design this kind of thing
find the burn rate, divide by 3 seconds to find the thickness
that will require picking a fuel combo
and that will also reduce the amount of fuel required because i picked a low impulse, most fuels have higher than what i used
well gasoline/nitrous should be straightforward to find
er.. you said hybrid so what's that, polyethylene or something?
i was hoping to use a solid/liquid instead of liquid/liquid
hybrid is solid/liquid
yeah but what's the solid
anything you want, which is the problem
i think they use polybutyl rubber usually
i am having a hard time finding information on hybrid fuel combos
rubbers are the most commonly listed, with nitrous
but you can use a salami
...the lunch meat?
i don't think i'd be able to lift my arms much less eat a sandwich in this thing
[21:42:41] <fenn> http://kwc.org/mythbusters/2006/04/episode_51_myths_reopened.html
"This may look like a salami, it may smell like a salami, it may even taste like a salami, but it's rocket fuel."
that is incredible
there isnt very much "real" rocketry data because of ITAR
international traffic in arms regulations treaty
well maybe some testbed rockets would get what i needed
haha now i'm talking like it would happen
toastatwork: what are you trying to do?
blow himself up
this man speaks truth
I figured that already
but I thought you might have a point to it
i wanted to see what rocket design would take you down a quarter mile in 3 seconds
an F16 might
20k lbs of thrust for 3 seconds is what i calculated
so whatever has that, it should do it
oh wait, i didn't factor in drag
oh well, i'll look that up in class tonight
ty all who helped look stuff up
F16 doesn't have that much thrust.. that's why they have a giant steam catapult
anyone know how to rotate two angles at once in opengl?
can anyone tell me how to add a page to the wiki? I've done the "Basic Steps" and have edit links at the bottom of the pages, but I don't see an "add page" button or link.
mshaver: just edit the URL to say the page you want
to add a new page make a [[link]] and click on it
yeah, or that
when you go to a nonexistent page, it shows an empty page and lets you add stuff
actually, that's pretty obscure, isn't it
fenn: not at once, but they add, so you can just glRotate twice with different vectors
i have azimuth and elevation, or a and b like in g-code
sort of, but since the primary way of getting to a page is via a link, it makes sense that you put the directions to the page into the wiki before you put the actual page in there
that's just two glRotates isn't it?
but when i do glRotatef they apply at the same time instead of adding
try putting them in the other order
i've been stuck on this problem for quite a while, it's why i wrote the tracking-test.hal
what vectors are you using?
glRotatef(az,0,0,1) and glRotatef(el,1,0,0)
or glRotatef(a,1,0,0) glRotatef(b,0,1,0)
someone's code on the web has el,1,0,0 then az,0,1,0
i dont think it matters?
the order sure does
sigh.. coding by exhausting all permutations
does that mean you got it?
i havent yet gotten sufficiently systematic :)
by exhausting some permutations randomly?
if you want to try your hand... http://fennetic.net/pub/irc/vismach.py
sorry, I'm working on ... anything else at the moment
and then figure out what /dev/input/event your mouse is on and put that in tracking-test.hal
SWPadnos: cat /dev/random > curl http://wiki.linuxcnc.org/$1