Jymm is now known as Jymmmmmm
Jymmmmmm is now known as Jymmm
I cant get my axis enable to work in hal file for my machine configuration
newsig enable_out bit
linksp enable-out iocontrol.0.enable-out
linksp enable-out parport.0.pin-01-out
what are you trying to do, enable a stepper drive?
yeah - there is an enable that will turn all three of my drives on - gecko320s - in TurboCNC I had to set them sep for each axis - the three lines above were suggested by alex_joni
I don't see any pin called iocontrol.0.enable-out
there is an iocontrol.0.user-enable-out
when I start emc from the terminal it says - HAL: ERROR: pin 'iocontrol.0.enable-out' not found
should I use what you just said
right, because the linksp command is looking for that pin and it doesn't exist
do you want the geckos to turn on when you come out of estop? or when you go to machine on?
the one you just stated is used on a line that says estop loopback
that signal goes true when you come out of estop
wholepair: are you still using tkemc?
what I want is: my physical machine E-stop is on pin 13 normaly closed and when I push the big red button I want the machine to stop - I also have pin one wired to enable on the machine - so I gues I want pin one to go low when pin 13 goes high
_and_ you want emc to estop when you push the red button
yeah - there is no need for emc to continue running a program if I hit estop on machine -
well, the preferred way to do an external estop is with ladder logic
Is ladder logic stated in the hal file
no, unfortunately its another bit of config
it can be extremenly powerfull if you have a complex machine
its kinda overkill for a single estop button tho
my machine isnt complex - see: http://www.bikegeeks.com/EMC2_Configuration/
do you think I should skip my external estop and just use the software one? everything else is settup - step and direction, limits, spindle on off.
the last two things Im confused about are enable and estop
you could do that for now
is your external estop a momentary button, or a maintained one?
momentary - you push and release, it pops back out ready to go again
maintained - you push, it goes in and stays in untill you pull it out
it's a maintained one - the kind that you have to twist to revert to normaly open state.
Personally...I think that a hardware E-stop is a good idea always
I agree external estop is important
[02:34:31] <jmkasunich> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl/emcinfo.pl?Sample_HAL_And_ClassicLadder
the link is an example using ladder logic for estop
In a panic its easy to find
ME TOO! - I am sorry to just get you started and then have to go but something has come up - thanks for your help - I will save the wiki url for later :-( - bye
there are actually 4 different ways of doing it in the example, you can pick the one you want
a lot easier to smack a button than it is to hunt down a key
darn.... he left before I realized there's an easier way
(actually before cradek reminded me theres an easier way)
reminded? you mean shot in the dark?
it looked to me like the estop button actually does stop the machine
the comment "emc doesn't need to keep running if I hit estop"
that made me think that the machine would stop anyway, and it's just a courtesy to notify emc
should be "emc damn well better stop running if I hit estop"
if the machine is disabled/depowered, it doesn't really matter
I don't think he has a hardwired estop
from his webpage: E-stop = Pin 13 (Wired with Normally Closed contact)
he also agreed wholeheartedly that an external EStop is important (though that could have just been an external button)
I think he was considering the "find it in a hurry" factor, not the "never trust software" factor
could well be
decisions like "external hardwired estop" vs "external estop button that goes thru software to stop" vs "GUI estop button only" are decisions that the machine builder must make for himself
what is appropriate for a sherline isn't neccessarily appropriate for a bridgeport
not a Mazak ;)
actually, that would be a great class at the workshop - safety
I'm not really sure I even want to go there
it would be great. probably not great for one or more of us to teach it, but still a good subject ;)
I sure as hell don't want to have someone come back to me and say "you said that would be safe"
when in fact they didn't understand what I said at all
that's true. plus there's the - erm - discussion we often have on these lists
(we being various members of the emc community)
the way I see it, if you have to ask me how to put together a safe machine, you aren't qualified to put together a safe machine
but the corollary is also true
if you refuse to learn what's safe, you aren't qualified either
my concern is that people really do have to _learn_
not just follow some step by step instructions from somebody like me who is a thousand miles away and doesn't know their situation
they need to truly understand the system they are building
I wouldn't want to "show" people how to do safe things.
if they need help with details like hal pins, thats fine
I'd want more of an overview, and things to think avbbout when designing a "safe" machine
I think even the tiny sherline drive box has a red estop button that kills power to the drives (hardwired estop, no software involvement)
I'd hope that any turnkey system would have at least a power cutoff for estop
i got a shiny red button from the scrapyard a couple days ago
it'll be a while before i can use it though
i made a tap from some acme threaded rod - boy it is easy and WORKS!
you're just jealous
what are you tapping? plastic? aluminum? brass? steel? hardened tool steel? ;-)
oh, most of my impressed just went away :-)
try tapping 5/8-5 in bronze with a homemade tap!
(it actually works, but I worked up a sweat and got cramps in my hands from turning it with a large wrench)
how long was the taper?
about 6" maybe
03cradek 07HEAD * 10emc2/src/emc/iotask/ioControl.cc: this was printing even with DEBUG=0
i figured a more plug-shaped tap would be appropriate for plastic
[03:13:23] <jmkasunich> http://www.metalworking.com/DropBox/_2001_retired_files/Worm05.jpg
tap is on top
then worm gear gashing and hobbing cutters
why do I see two tapers in the tap?
the first one is roughing
the tooth form there was cut down to have sharp crests like a regular (not-acme) thread
the flat crests of an acme = large chip load
[03:15:24] <jmkasunich> http://www.metalworking.com/DropBox/_2001_retired_files/Worm03.jpg
nut made with the tap
thats pretty sexy
I also threaded the worm gear to fit on the screw, then pinned it
03compile-farm 07Ubuntu 5.10 (2.6.12-magma) * 10emc2head/: build PASSED
that tap did about 5-6 holes before I had to sharpen it
its done about 14-16 total
that worm is very cool
how did you gash the worm without a dividing head? :)
er worm wheel
I have a dividing head
see pic 7
uh.. what is this thing for then?
piece of lab equipment
made 7-8 of them over the past 3-4 years
sell em for about $1500 each
sounds cheap for lab equipment
what is it?
"Crack Isolation Membrane Test Jig"
per some ANSI standard that I don't remember the number
where did you get that number from? google?
of course ;)
there's also a 118.10
it looks like 188 is crack isolation stuff
then there's A137 for ceramic tile
which is what this is for, IIRC
gotta url for the test standard?
hah. those are expensive :)
I didn't think it was on the web
but you seemed to have found something
yep. manufacturers saying their stuff is compliant with ANSI blah blah
thats one of my customers
basically the jig is 2 square plates of 3/4" aluminum plate, with a screw that lets you force them apart to simulate a cracked floor moving from frost or whatever
you epoxy cement slabs on top to be the "floor", then the membrane and tiles on top of that
crank it and see if the tiles crack
[03:26:20] <jmkasunich> http://www.tileusa.com/tstprice.html
I didn't charge enough!
they charge $1450 to conduct one round of tests for a customer
03compile-farm 07BDI-2.18 (2.2.18-rtl3.0) * 10emc2head/: build PASSED
wonder what genius at sherline put 8" leads on the motors for an 18" long lathe
I seem to remember that the sherline drive box has wires and inline connectors instead of chassis mount connectors
(but I could be full of shit, I last saw it a year ago)
that would sure explain it
[03:30:55] <jmkasunich> http://www.sherline.com/images/CNCasby1.jpg
you can see the 4 cables coming out the back of the PC
(they build their drives right into the PC case now)
I guess that saves them an expensive box
and saves their users messing with more connections
is the power supply in there too?
looks like it
also keeps the user from upgrading the computer (not being sarcastic, that's an important thing)
under the guy's left pinky, looks like an open frame switcher
03compile-farm 07BDI-TNG (2.4.18-rtai) * 10emc2head/: build PASSED
they're selling a turnkey system, upgradeing the computer is a non-issue for their target market
darnit, I can't find the old style of keyboard extension
oh, they also have external boxes
[03:33:46] <jmkasunich> http://www.sherline.com/images/8760inside1.jpg
[03:34:04] <jmkasunich> http://www.sherline.com/images/8760pic1.JPG
looks pretty professional
mine's an old UPS case, heh
I've been collecting various cases, external SCSI drive boxes, etc
dunno what I'm gonna actually use
ultimately I want to put a spindle drive in the box too
and maybe just build the PC into it as well
are you going to build your own drivers?
no, using geckos
at one time I wanted to make my own drives
you outgrew that?
then I realized how much time it would waste, and how little money it would save
unless you want something inexpensive like I did
my machine takes NEMA 34 motors, I need more oomph
I have a three axis xylotex board here that I used for experimenting, but its too small for 34's
seems they're too small for much of anything at only 24v
03compile-farm 07BDI-4.20 (2.6.10-adeos) * 10emc2head/: build PASSED
they run nema23's ok, as long as you don't need blazing speed
and the price is right
03compile-farm 07BDI-Live rc46 (2.4.25-adeos) * 10emc2head/: build PASSED
but they're unipolar aren't they?
bipolar chopper microstepping
they use allegro chips
then they're not at all joke like I thought
quite nice really, 2.5A, 24-30V
[03:42:32] <jmkasunich> http://www.sherline.com/8760pg.htm
has pinout and color coding for sherline motor connectors and leads
dunno it that will help
think I have that part figured out
I'm trying to figure out a nondestructive way to do the spindle and switch rewiring
maybe I'll get one of those db25 jumper boxes if radio shaft still has them
1 DB-25 male, one female, soldering iron and wire
yep, same idea
couple of inch long standoffs to space them apart and keep em from flopping
like the gender changers or null modem widgets you see for serial stuff
(used to see anyway, serial is going the way of the dodo)
these zero-resettable handwheels are cool
it's pretty funny to see them on a cnc lathe next to my manual lathe that doesn't have them because I was too cheap
switch em ;-)
"Fred, I swear there weren't any handwheels on the lathe when it got here"
nah, I'd give him mine...
well thats a bug
the encoder is "5-26V BIDI 10P CN"
run the axis into the limit, get a hard limit error
reset the estop, and it tries to move farther onto the switch (getting another error)
I bet that's the same as "I can't turn machine back on after a following error"
its only moving about 0.0005 each time
but thats enough
pretty sure I've seen it do worse
I looked for this once, but the code looked right, so I figured I didn't know what I was looking for.
at the moment its re-enabled, the position command jumps about 0.001
its like there was some residual "desire to move"
I went into hal and turned off the limit switch
the enabled the machine again
and it moved 0.020 before stopping
I wonder if I'm not resetting the velocity in the free mode planner
milled some test boards to see how my cutter worked. It didn't. I ended up making and engraver type cutter. I had some broken 1/8 carbide drills - I ground it to a point. then ground them in half and releaved it. was able to make .020 isolation cuts with it.
I found out I need to make the resister/cap/diode pads bigger.
skunkworks: sounds like a great start
other than that - so far so good. (I had it loose steps a couple of times. I had upped the accelleration a while back and that plus cold seems to make shuttles stall. Need to slow it down I guess
I decided a while ago that I'd rather run very conservatively than ever botch a job
that is what I am thinking :)
so I run my steppers slow enough that's pretty hard to stop an axis if you try
and I never cut that fast
I would have to look at the file but I think the cutting is done at 10 ipm or there abouts
so does your cutter have one "tooth" or two?
1 in effect
then 10ipm sounds pretty fast unless you have a super fast spindle
I cut about .0005/tooth
router - don't remember 25000 rpm
ah you're fine then
it makes a pretty good cut if I don't go too deep.
have a photo?
the relief was done freehand so I don't think it is that great - I could do better.
not yet - maybe tomorrow
how are you holding your boards down?
and have you ever done 2 sided?
sure, lots of times
carpet tape - cool. I was thinking about a vaccum table.
I bet that'd work great if you don't drill
but carpet tape seems easier
it's sometimes a bit hard to untape
the table top for the vaccum would be plywood
plywood won't be flat enough
this is maple die board. hard as hell.
can it hold .001?
I cut .004 deep and .001 makes a big difference in width
it would be milled flat - I would find out pretty quick
can't hurt to try
I'd like to hear how it goes
that is the fun part - I thinks go "ok" I will make a video
hell I don't even know if the copper clad from radio shaft is that flat :)
sticking it down to a flat surface makes it flat
you have to tape every little bit you intend to cut, and press it down to the table
I was measuring some older single sidded board and it varied by a few thousands
which I had not even thought about.
the boards I use don't vary much, they always measure the same .059 whenever I check
but it sure doesn't matter for etching, which they're meant for
so it's a bit surprising they're so good
jees - got sucked into apallo 13 again
I sharpened the carbide like this http://www.antaresinc.net/FactCutterGeometry.html
and seen as how I have a few broken cabide drills - I can make a few more :p
i always wondered how they made the braille dots
I have saved up my broken pcb drills too
well, I found _a_ bug
but not _the_ bug
it no longer moves when I re-enable ti
but I still get a fault
cradek: I installed mysql on a windows 2000 server today. looks promising
the move was because going into the "motion disabled" state didn't clear the velocity command
override limits was working for me earlier - it allowed one jog in the right direction
I didn't even try override limits, I assume it works
but need to try it
skunkworks: while I've installed mysql a few times, I don't claim to know anything about it
neither do I - but I got it go comuicate to the odbc drivers on a xp machine.
so far so good. even made a simple table and coppied a million records to it :)
hmm, the limit switch code doesn't care what direction you are moving (if at all)
I thought it would only trip on positive limit if moving in the positive direction, and on negative limit if moving in the negative direction
that's good, that means I'll be able to jog off limits even if they're all hooked together
no, thats not what it means
it means that even if they're not hooked to gether, you'll need to use the override
(if they are hooked together, you always need override)
I see what you mean
I did have to use override tonight, I thought that was how it was supposed to work
but if they're not, and they are treated individually, you could jog off without override
I agree you could, but I think emc doesn't work that way
it doesn't right now
I wonder if that's a bug
there are a couple things I don't like about the limit handling
well, one thing at a time
I think this "start moving when you re-enable" is a bug that should be fixed in the release
(its a one line change)
any changes to the limit logic should not be in the release, too complex to be confident in without lots of thought and testing
that is not a bugfix
so I'm gonna commit the first change to head and branch now
well it fixes a "can't jog off a limit" bug
you can jog off a limit, you just use override
so it's not a bug, I think that is even as designed
well, I guess its debatable
which makes it a bug to fix for the _next_ release
ok no need to debate it if you're not putting it in the release branch
only thing going to the release branch is resetting vel to zero on disable
great, that'll be nice to have fixed
03jmkasunich 07HEAD * 10emc2/src/emc/motion/control.c: reset joint velocity to zero when motion is disabled
03jmkasunich 07v2_0_branch * 10emc2/src/emc/motion/control.c: reset joint velocity to zero when motion is disabled
03compile-farm 07Ubuntu 5.10 (2.6.12-magma) * 10emc2head/: build PASSED
the limit handling definitely needs work
theres a whole chunk of code dedicated to latching limits, which:
03compile-farm 07Ubuntu 5.10 (2.6.12-magma) * 10v2_0_branch/: build PASSED
1) isn't enabled right now
2) probalby won't work if enabled
what does latching limits mean?
/* Some machines can coast past the limit switches - this makes EMC
think the machine is no longer on the limit. To avoid this
problem, we latch the position when the limit first trips, and we
won't release the limit unless the current position is inside the
latched position. We use motor position because joint position
makes step changes during homing */
/* latching of limit switches is optional - it should not be used if
pos_limit and neg_limit are tied together */
the algorithm described there won't work if we start letting the encoder reset on an index pulse as part of homing
right now there's no way to enable it anyway
the whole thing sounds bogus to me
latch_limits was gonna be an ini file param, never got around to that
if your machine does this, you should fix your switches
easy for you to say
sure, I can say stuff like that, because I don't know squat about it
03compile-farm 07BDI-2.18 (2.2.18-rtl3.0) * 10emc2head/: build PASSED
on some machines (maybe the lathe you have right there) the switches are activated by a small cam
of the table coasts when you tell it to stop, its possible to go right past the switch
the disaster case is if that results in the motor starting back up and continuing to drive the table into a hard stop or crash
the less disasterous case is if the machine stops, but emc thinks you are inside limits, and allows you to jog into a hard stop or crash
I agree that the "right" answer is longer cams
the machine stays off, doesn't it? I mean until the user does something?
I could simulate it by tapping a switch, I'll try it tomorrow
fault handling in general sucks
it seems like it would be hard to jog back inside the limits once this happens, but other than that it's not a bad bug is it?
when we get a fault (ferror or limit), there is a dialog that pops up, with an OK button on it
a person would assume that clicking the OK is a "clear faults" kind of thing, neccessary to restart
but it isn't
clearing the dialog has no effect whatsoever on the machine
so a user could not bother to look at the machine and see that it's outside limits
well what would normally happen is:
03compile-farm 07BDI-TNG (2.4.18-rtai) * 10emc2head/: build PASSED
1) hit switch - machine begins stoppingnumbers on display go red, dialog pops up,
1) hit switch - machine begins stopping
I didn't to that!
it sez you did
2) numbers on display go red, dialog pops up, machine switches from "on" to "estop-reset"
3) machine coasts past limit switch
4) numbers go green again
the dialog is still there, and the machine is still off
you hit f2 to turn it back on
normally the red numbers would tell you that you are outside limits, and you'd hit override limits, then jog back until they go green
in this case, the numbers are already green, but the dialog says you were out of limit
so you have to jog back into your switch, when it'll stop again
that does suck
if you realise you overshot the switch, you can jog back
well, you could hit override limits before you start to jog back
hey, this is just the border of the suckage
if I had that happen, I'd sure say "I need to fix those switches"
on the mazak, because its so fast, we hardwired the limit switches to the estop
to get braking?
so once you hit limits, you have to jumper the switches with a wire to jog back
(and you can't just crank it back)
well, its a safety thing
suppose a dac fails with +10V, or an encoder dies
03compile-farm 07BDI-2.18 (2.2.18-rtl3.0) * 10v2_0_branch/: build PASSED
the thing is gonna run away, and theres nothing emc can do to stop it
so when it hits the limit switches, we kill the power and hope it coasts to a stop before hitting a hard stop and busting something
doesn't estop brake the motors (short them)?
in normal operation you should never hit the switches
that is machine dependent, the mazak puts a resistor across the DC supply, but not on the motors themselves
the software limits are set inside the hard limits, so if you just lean on the jog button, it will hit a soft limit before the switches shut it down
jogging off a soft limit is easy
03compile-farm 07BDI-4.20 (2.6.10-adeos) * 10emc2head/: build PASSED
yeah I even did that on the lathe
and my mill uses soft limits without switches
which is far from foolproof, since the soft limits are relative to machine coords, and unless you have a home switch, where are machine coords?
I home it by eyeball
it has a very easy to see machine origin
lots of folks get out the edge finder, touch off on their part, and hit "home" which is configured to set the current position to zero
(IOW, they use home instead of part offsets)
yeah I used to do that, but this is better
but lots of users don't do that, and therefore don't have usable soft limits
(of course, they're also using machines that just hum if you run them into the hard stops, not crash
for instance I home Z with the tool near the table, and my negative limit is 0
that way if my program is wrong, it aborts instead of drilling the table
good plan ;-)
03compile-farm 07BDI-Live rc46 (2.4.25-adeos) * 10emc2head/: build PASSED
doing that was only obvious in hindsight (at least to me)
do you re-home after a toolchange (if the new tool is a different length)
all the tools I use for a job are in the holders at the same length
ok, thats an exception to the norm
I'd have to use tool length offste otherwise, and I don't want to figure that out
messing with the mazak is a good way to get to thinking about some of these things
the first time you see that table moving at 300ipm, you say "whoa!"
does it have a full servo controlled spindle?
it has a DC drive, but its basically a variable speed drive only
what's the accel at that speed?
03compile-farm 07BDI-TNG (2.4.18-rtai) * 10v2_0_branch/: build PASSED
we have it set to 5ips (300ipm) and 20 ips^2
zero to 300ipm in a quarter second
I bet that's impressive
IIRC we actually ran it at 350 or 400 ipm and 40 accel, but dropped back to minimize abuse to the mechanics
it is impressive
the table probably weighs close to 1000 lbs, and it can do a 10" rapid in 2 second or so
I think that means a full accel/decel takes 2.5"
t = 0.25 seconds
a = 20
1/2 * 1/16 * 20
= 20/32 = 5/8"
5/8 = 0.625
sorry I meant accel and then decel again to stop
duh, you said accel/decel
I read it as "accel or decel"
I was thinking about what its shortest full speed segments are with the current TP
not real short
true lookahead is something we should try to figure out
dunno how hard its gonna be tho
I keep meaning to look into matt's new TP
he says it does that
that would be a good thing to talk about at fest
those kind of discussions are a lot more effective face to face, with pencil, paper, etc
unfortunately he won't be there, but maybe we can hack it into emc2
see what following errors we get :-)
so how much we can shake the machine
early during PID tuning I had a 5Hz or so oscillation get going
who gets to run estop during early tuning?
we had a couple
a lot of what I was doing was late at night, I was the only one there
but we had an extra estop button on the bench next to the computer
speaking of late at night...
yeah, its past both our bedtimes
I'm anxious to work on the lathe more tomorrow
tonight was very encouraging
03compile-farm 07BDI-4.20 (2.6.10-adeos) * 10v2_0_branch/: build PASSED
03compile-farm 07BDI-Live rc46 (2.4.25-adeos) * 10v2_0_branch/: build PASSED
is there any easy way to start a program from a certain program block?
LH: Hello; block or row?
row wouold be best as I haven't written block numbers
If you are using axis; I think is is: click on the row and then the run button
that started it from the start
hmmm, at least I think it use to be like that... then the bad solution run your editor
hmm, too bad :(.. any way to jump to a certain block?
LH: det verkar som vi får vänta på att usa ska vakna
chinamill: Jo, det kan nog stämma
* LerneaenHydra looks at timestamps
* LerneaenHydra notices the extreme activity
This being the old farts channel =)
hello LerneaenHydra / Jymmm
15:20 here :P
[06:17:30] <giacus> 15:20 here :P
wow, wake up soon Jymmm
[13:19:25] <Jymmm> http://www.mcselec.com/index.php?option=com_content&task=view&id=14&Itemid=41
well, air of summer here around
nice, but I hate proprietary solutions :/
It's a BASIC compiler for Atmel uC
AVR not proprietary
not like basic stamp anyway
it is not designed to run under my favorite O.S.
* Jymmm sighs *
I was watching an interesting documentary about 1st mondial war..
its seems so far
lot of crazy peoples
I hated the story in the school
but i'm discovering it right now
I just eard my mother and my father talking abot 1945 war
that was very sad, very very sad
as is all wars
also seen something about 'olocausto'
thats incredible cruel
Jymmm is now known as Jymm
LerneaenHydra: click the line, pick "Run from line" on the menu, then hit run
or "Set next line" or whatever it is
cradek: ah, thank you
cradek: I seem to have strange behaviour, it seems that manual jog goes faster then g0
not much but definetly hearable and visible on an osciliscope
could it be that man jog uses the value in stepgen maxvel?
all coordinated motion (g0) goes 5% slower than manual jog
this was my way of dealing with the stepgen problem, to always go a little under the requested maxvel, but unless jmk does the same thing for jogging, we still need the STEPGEN_MAXVEL hack
like I said before, there's some disagreement over how it should be handled
ah, so when doing a manual jog there isn't a problem with step piling up in front as there can be with coordinated motion?
no, the problem is the same when jogging
so manual jog uses the value written in stepgen maxvel and stepgen max accel, whereas coordinated stuff goes through maxvel and maxaccel (which should be 5% or so smaller than stepgenmaxvel/accel)?
currently if you set maxvel=stepgen_maxvel you will get following errors when jogging but not with any coordinated motion (g0/g1/g2)
jogging uses maxvel, coordinated uses .95*maxvel
and that 0.95 is needed for some math algorithm?
so that steps don't pile up?
ok let me explain the base problem
because of floating point precision issues, stepgen has to allow velocities a little higher than motion will give it
we call this "headroom"
and previously motion would request up to maxvel, so the user had to specify stepgen's maxvel a little higher, like 1.05*maxvel
so maxvel is the desired value, stepgen_maxvel is a tweaked value because of the precision issue
well I don't like this, so I wrote into the TP a 95% hack so the user could put the desired maxvel in the ini and not have to mess with headroom
jmk disagrees and thinks the user should specify the headroom (even though he doesn't like it)
so we have a bit of a mess currently due to the combination of the two approaches
5% souds like a lot though for rounding errors. so actually now there is double compensation, whcih is not too good
I had hoped we could remove the headroom specification because I don't think the user should have to deal with it
so to get as good a speed as possible they should be set to the same values?
no, if you set them to the same value currently, you'll get errors jogging.
doesn't your 0.95 take care of that (oh, ok, a jog would cause errors, but not a coordinated move)
why is 5% needed, shouldn't it be in the <0.1% region?
it's not clear to me how much is needed, and it probably differs with different configs
all of hal uses floats, not doubles
steps/(mm/inch) would affect this?
floats don't have a lot of precision especially when the absolute value is big
yes I think so
a float is 8 sig figures?
we should talk about this when jmk is here
it's not that simple; a float has more precision near 0 and less and less as you get further from 0
it shows up especially if you plot a couple derivatives (vel and accel) near the origin and then far from it
cradek: another thing that is strange, when running for example spiral.ngc i get jerks in the speed when it goes from one line to the other
10% or so, after which the speed stabilises
both audible and visible on an osciliscope
yep I know, I thought we talked about that before
I am running steptype 2 if that makes any difference
you can reduce the effect, but not eliminate it, by reducing your traj period
nope it doesn't
so that effect is well known?
yeah, it's a result of the trajectory planner algorithm
what would a suitable value be (currently 10000000)
that looks like 10 seconds :|
which sounds strange
maybe try 1/2 1/4 1/5 that
no those values are nanoseconds
ah, so 10 mS
not radiation ;)
what happens if the value is too low?
the tp cycle will overrun and miss the next realtime deadline
will I get an error message?
or will I see that it skips steps?
I think hal can tell you how long tp typically takes to run
how do I do that?
I'm not at the right machine to look
maybe ray knows
No I don't. I'm pretty dumb this morning. To much fun last night.
hmm I think it's in `halcmd show thread' but the values are always 0
they are only nonzero for servo-thread
LerneaenHydra: try 2ms, someone else reported that helped a lot
I'll try that
switching halscope between the thread should give an indication.
would it be good to test 1.5ms and if that works then raise it to 2 so that I know it won't mess up sometime while doing real work?
LerneaenHydra: let's ask jmk later why these time values are 0
I bet it takes nowhere near 2ms to run, but it would be nice to be sure
cradek: yes that sounds like a good idea
the thread connection widget in halscope gives actual time
rayh: that's the period: we're looking for another thing, which is how long it takes to run the code per period
you can see it for servo-thread in `halcmd show thread' (time, max-time)
but for traj and base threads, it's 0
Period FP Name (Time, Max-Time)
10006914 YES traj-thread ( 0, 0 )
1005720 YES servo-thread ( 24697, 60809 )
Oh. Okay. Sorry.
I think Time is for the last period
Max-Time is the longest so far
hmm, EMC dies when trying to start with tp at 1.5ms
check your edit again?
it sasy (in the terminal) insmod: error inserting '<magma>/modules/emc2/motmod.ko':-1 op not permitted
it works when set to 10ms
look in dmesg
run the command dmesg and you'll see a better error
dmesg as it is gives far too much text
it overflows and I can't see what it wrote in the beginning
yep you have to dig around to find what you want
you have to run dmesg right after your error
and dmesg -c first?
yes sudo dmesg -c will help
the log is very big, what was that store-text-snippet site called again?
it doesn't like the value 1.5ms
[18541.866405] MOTION: bad traj period 1500000 nsec
any idea why it doesn't like that?
probably would have to look at the source to see why
maybe only wants 1 non-zero value
probably because it's not a multiple of base period
my base period is at 50 IIRC
try a multiple of that
first I'll lower base to 20 as that's what it should be
so TP of 2?
sure try that
or wait, wrong number of zeros
I set it to 0.2ms
which should still work
when set to 2ms it works though
or rather, it loads axis
let's test output now..
hope it helps after all that
it less thats for sure
it still hops around a bit though
looks like something like 50µs
audible but not very noticable
barely noticable on the osciliscope when looking at the waveform, probably not more than 2-5% of the period
what we've done is give the TP a few more runs per acceleration/deceleration
so blending works better
yes, that much I understood
it would be ideal if we could get the output we want from `halcmd show thread' - let's remember to ask jmk
I'm off for a bit
what output are you looking for?
the time that TP takes
it should be there
it's in clock cycles
it shows 0
hmmm - let me make sure that HEAD and RELEASE match
I'm running sim fwiw
only servo thread shows nonzero
are all the *.tmax numbers 0 as well?
09:11:26 < cradek> Period FP Name (Time, Max-Time)
09:11:26 < cradek> 10006914 YES traj-thread ( 0, 0 )
09:11:26 < cradek> 1005720 YES servo-thread ( 24697, 60809 )
sorry for the bad paste
can you paste in the full output of show thread?
not right now, maybe LerneaenHydra can do his
uh, I'll ned some pointers as to how to get that output
while emc is running, run `halcmd show thread' at the shell
oh, another random thing, I get very laggy response for things like feedrate adjust while running the program
traj thread givces 0,0
even when running a program
if your base period is too low, your system will have little time left to run other things like the gui
it might simply be that
LerneaenHydra: can you put the full output of halcmd show thread in pastebin for SWPadnos?
there used to be problems with timekeeping on various kernels
but it's all in cycles, using rdtsc now
SWPadnos_: that wouldn't explain why one thread has it and others don't, would it?
no, which is why I'm a bit confused
I got slow response even with base period at 50
LerneaenHydra: you're running the 4-9 testing right?
I would think so, I used your apt-get script
how do I check emc version?
hey you get base thread numbers, mine were zero (with sim)
base thread I get
which ini did you use? (axis, mini, ...)
emc2 v: 2006-04-09, AXIS 1.2.1
LerneaenHydra: good, that's the last one
well, the easy answer is that nothing is attached to the traj thread
there are no functions, so it takes no time
hmm, I see that
the motion controller is being run at the servo period
but changing its period makes a difference for the TP
I don't think LH imagined it...?
true, but the TP should be figuring out the elapsed time from the parameter passed to it by RTAPI
ah, thats rather non-intuitive
I don't follow
neither do I (that last line)
(not that it really matters, as I can't code... ;) )
all RT functions are passed a parameter that tells the period (or elapsed time since the thread was last run)
though the TP probably has some other stuff that's initialized at startup, based on the expected period
LerneaenHydra: what's your traj cycle time now? is it 2000000 or 1000000?
I'm betting that there's a bug in motion that attaches the TP to the wrong thread
SWPadnos_: maybe so
SWPadnos_: but LH is getting better results when he reduces his traj period
SWPadnos_: that's what I don't understand
ok, that is odd ;)
SWPadnos_ is now known as SWPadnos
ok I really have to go, I'll look back later to see what you guys have figured out
heh - see you
ok - motion doesn't attach itself to anything, which is good
what's wrong with this program, it only does one pass:
I followed the wiki but something is wrong with the program
it doesn't execute the sub until the call, which is there only once
if you add another "o1 call" line, it'll execute twice
oh, how would I make it infinetly looping?
not sure, and I don't think it's a good idea ;)
consider how long it would take for axis to generate the preview plot
this is stricly for testing purposes
what about 10000 iterations?
you can test with a few thousand loops
how would I do that, a while loop?
sure. 1 sec
replace the o1 call line with the following:
#5 = [#5 - 1]
o2 while [#5 GT 0]
that took quite a bit of code
lots less than 10,000 repeated lines of "o1 call"
well, thats for sure ;)
it's not much worse than the C equivalent: for (i=0;i<10000;i++) o1();
it only runs once
man - you know your hard drive is too big for your CPU when Norton Antivirus takes 9 hours 10 minutes to do a scan
well, rather you've got too many small files
still, 9 hours is a while
or a slow hard drive / IDE controller
*cough or bloated antivirus cough*
this is a 5 year old Windows 200 install, that's been ghosted to a new drive twice (once due to partial drive failure)
drive failure... :( not fun at all
nope - had an IBM DeathStar drive
ah, I also had one of those
I got the click-o-death in 6 months
I had a 60. still waiting for my $100 check from the class action suit
that should buy 400G by the time it arrives ;)
I got a newer ibm/hitachi 40gber a while back
that would be much better ;)
any idea as to why the program runs only once?
the looping program?
hmmm - not really
gimme a sec
heh - oops
change #1=10000 to #5=10000
ah, yes that would help
yes that works
silly me - I was looking at the wiki page, which used #5 for the loop var
that would be a seriously scary program to run on my mill
because the top speed may be as high as 180 IPM
it's a Bridgeport
or you who had the high IPM machine?
no, that's not really high speed. you may be thinking of les_w
that may be
his machine is the one that can accelerate at 1G or more
that's a bit
I think he said the Y axis can do 3G, but he's got it lilmited to 0.5
well well, thats rather powerfull
hmmm - it would be nice to be able to watch variables in the GUI
like - how many loops are left?
the subroutine calling doesn't appear to be RT-based
the time betwwen the different loops varies alot
it should be interpreted the same as if there were 10000 repeated lines
but it's not
how are you measuring that? looking at the DIR line on the scope?
timing vary greatly
I have a physical osciliscope attached
and I can hear it
that's the scope I meant ;)
not dir line, as I run steptype 2 (fullstep)
are you running motors / a machine with this?
I have homemade drivers
just motors ATM
I suppose I should hook up a scope to this and take a look
I could take a picutre of it I guess
if that would help
nah - I've got all the test equipment I need for this
wanna buy a 2 analog + 16 digital channel scope? ;)
it would seem best for EMC to render all the lines first
for a small program, that does seem best
large ones could get very big then I guess
but there are g-code files that are over 1Gig
but they don't really tend to use subroutines that much, do they?
and the pre-planning may need to be thrown out if the feed override changes
i'm still struggling with the board in eagle
it would be nice if some CAM program would output code that human could change later
like pocketing code that uses loops and variables
does software exist for placing components automatically on a board layout?
SWPadnos: yes, that woud be good, although it may be hard to implement
especially as the code varies to greatly from producer to producer
SWPadnos: can you try running this:
g0 x0 y0 z0
g3 x0 y0 i0 j10 f1000
o2 while [#5 GT O]
doing a pocket is much easier with loops. you program the outline / hatch pattern as a subroutine, then loop on Z to get to the right depth
and look at the planned path vs the real one? (real one = toolpath the the cone follows)
a 3d pocket is a bit harder though...
the plotted path is sampled in userspace, so it doesn't exactly match the actual machine path
iit was more than slightly off
projected diameter of 508 whereas the real one is 10
are you in mm or inch?
I get an error "unknown word where unary operation could be"
uh, I think I put an O instead of 0 in [#5 GT O/0]
I think you have a machine unit / user unit issue there. it's got the correct diameter of 20 inches on my plot, and the "real" trace matches the preview
where do I set up that?
in teh stepper_mm.ini?
I'm not sure
there's no g20 / g21 in the program, so it's kinda random as to what units are used
still kinda random sounds not too good
shouldn't it complain or something?
that's why you should always put a G20 or G21 in your code ;)
nope - it's perfectly valid
which is which?
g20 is inch, I believe
what does the machine units value in the config.ini file mean?
g21 fixes it
it sets the units for the machine hardware
if you have a mm ballscrew, it makes sense to use mm for the machine
it is a number
currently I've set it to 1
right - it's the number of machine units per mm
1 means mm machine. .0393... means inch machine
machine units... what are they?
it's just the measuring system for the machine itself
if you have a 25mm ballscrew with 5mm pitch, then you should use mm for machine units
so one could just as well change input_scale?
if you have a 1 inch ballscrew with 5 TPI, then use inch
no, that's the number of steps/encoder ticks per machine unit
g-code can be in mm or inch. emc needs to know which one to scale for the underlying machine
ah, ok, but they are both used for teh same calc
one of them will be 1:1, the other won't
oh, so an inch based gcode to be used on a MM machine
then you need both yes
right - either code can be used on either machine
that's why you should specify g20 or g21
it would be nice to just use mm for all machines, but then 98% of americans would get confused ;)
sounds like you use MM too ;)
I use inch, but I'm in the 2% that likes metric ;)
my ballscrews are 1.25 inch dia, 5TPI, so inch is better for the machine
plus I have to talk to americans ... ;)
LerneaenHydra, can you change just the traj period, and see if there's a difference in motor sound?
LerneaenHydra__ is now known as LerneaenHydra
saw the discussion about threads
traj-thread is a red herring, right now the traj code runs in the servo thread (but runs every Nth time, based on the ratio of servo and traj periods
that is also how emc1 used to do it
is there any easy way to see how much time the traj code takes?
its lumped in with the servo code
I missed the first part of the conversation, what is going on?
soo the best thing one can do for now is to test low values, and when there is not enough time for it to finihs one takes the next step up?
please fill me in on what you are talking about
I'm trying to figure out how short I can let the traj period be, to reduce the "jerkiness" when running line interpolated curves (spiral.ngc)
I find it hard to believe that traj is running out of time
are you running steppers (software generated step pulses)?
ok, base thread is the biggest RT load then
what is your base period?
20000nS = 50KHz?
that sounds plausible
I was asking for clarification
clarification around what? why I am messing with traj time?
when someone says 20K, I'm not sure whether they're talkiing about 20KnS (which is a really weird way to specify time) or 20KHz (which is a fairly natural way to specify a frequency)
and servo period is 1mS (1000000nS)?
(yes that is a rather unnatural way to express frequency)
at the moment it's at 2ms
_servo_ is 2mS? or _traj_ is 2mS?
I have not yet tested 1ms
the default for servo is 1mS
the default for traj is 10mS
base period: 20knS
servo period: 1ms
traj period: 2ms
can you do halcmd show thread and paste it here?
also, what is your processor speed? do cat /proc/cpu, there should be a line like cpu MHz : 1667.742
I think I did that earlier, here's the link: http://pastebin.com/675170
its a PIII 800mhz coppermine (I'm not sure of the fsb speed), in a socket-a look-alike
the 800MHz is all I need
that lets us convert the times from bin/halcmd (in clocks) to seconds
first lets look at your base thread
max clocks was 11363, times 1.25nS/clock (800MHz) = 14203nS = 14.2uS
your base period is 20000nS = 20uS
so you never have an overrun, thats good
that sounds like a good value
how much does max vary from run to run, does some code stress the base thread more than others?
the worst case time means that 14/20 or 70% of the CPU is used by base thread
but typically its a lot lower
its not so much the code, its everything else
does software exist for component automatic placement in a board layout?
Bo^Dick: yes, its usually very expensive, and IMHO it does a crappy job
LH: the time it takes to run the code (any code) varies dramatically based on whether its in cache or not, and various other factors
the two numbers in the parantheses are the most recent measurement, and the highest one since you started the program
ok, is there any possibility (within reasonable bounds) of there being code that makes it use 100% cpu, and miss stuff, and if so what happens?
if you start doing stuff while it runs (browse the web, view a video, etc) the max will probably go up
yes, its possible to use 100%
will things like that increase the time used in base thread?
what happens if you use 100%
depends on how bad you go over 100% and how often
lockup the machine
so no max feedrate in material?
time used by the code and feedrate thru the material are only very very loosly related
i'm still intrested to find a software that places the components in a smart way
Bo^Dick: then get out your wallet
there is no chance of 100% cpu usage cause some error that would make it do a max feed?
by no chance I mean within reasonable bounds
LH: you are confusing the hell out of me
are you saying that you commanded a slow feed but got a fast one?
do you know a name on such software?
google "pcb autorouter"
no.. ok, what I mean is that in the worst case scenario when overloading the CPU with RT commands, the machine will only lock up? that there is no reasonable chance that some output buffer will flush causing a queued motion to be done at a very fast rate?
just be warned, autorouter software will _not_ make you into an electrical engineer
i'm not talking about autorouters
autoplace and autoroute go together
no in eagle
I don't think eagle does either autoplace _or_ autoroute
eagle has an autorouter but not an autoplacer
well then you know more about it than I do
what software do you use when making boards?
at work I use PADS
at home, so far I haven't made boards
I don't use autoroute or autoplace, IMHO they suck and I can do better myself
does PADS come with an autoplacer?
I don't know, and I don't care
well then you're a proffesional board designer
no, I'm an electrical engineer
it's funny that autoplacers and autorouters are expensive and does a crappy job
LH: don't know what you mean by "overloading the CPU with RT commands"
funny but true
i still would like to find an autoplacer
if you set base_period too short, the box can lock up, or start having overruns, but that has nothing to do with "RT commands"
I mean that you for example were to set base thread to something like 10KnS, which is lower than the highest recorded
by RT commands I meant CPU time that is used by the realtime subsystem
so i'd be grateful for tips regarding autoplacers
if you set base thread to 10uS, at least sometimes the code would run longer than that
but the results would probably be slower than normal motion, not faster
so it's not possible that the result would be faster (significantly faster)
I never say something is impossible
so it's definetly not good to use 100% cpu
however if thats what you are seeing, there are probably a lot more likely explanations than overruns
no, of course not
no, I'm not seeing that
if the RT code uses all the CPU, theres nothing left for the user interface (or any part of Linux) which is why you get lockups
_probably_, 90% of the time when the base thread runs it uses 3000-4000 clocks (4-5uS), but once in a while (maybe only one time) it used 11363
so if you set the period to 10uS, you would sometimes (maybe only once) get an overrun, but most times there would be 5-6uS left over
at the moment I'm getting slow performance when changing the feedrate while running a program (or doing anything else while running a program) should I maybe lower the base thread period?
if you wiggle the mouse around while emc is running, does the cursor move smoothly or is it jerky?
then I don't think you have a problem with base-period (or servo-period)
however stuff in the GUI moves very jerkily, somethign like 2-7 seconds behind
if the RT code is using too much CPU time, the mouse gets jerky
other GUI stuff gets jerky too
the mouse that is
is it only the EMC gui that is affected, or do other windows get jerky too?
you should be able to drag a window around on the screen while EMC is running without much jerkyness
menus and such seem to be ok, moving windows is ok, but the mouse moves somewhat jerkily
(it shouldn't be much worse than doing the same thing without EMC running)
it's nearly the same
then I don't think its a RT issue
things in the EMC windows however are much slower
I'm running AXIS 1.2.1
you said emc window_s_?
only one instance
is "cadstar" the only one?
Bo^Dick: most pro-grade pcb layout systems (the ones that are likely to have auto-place) are $10,000 and up
I've been running the program a bit more and currently have reached 12733 at most in base thread, which would be 60% cpu
lets back up a little
what is the problem that got this whole conversation started?
haha, I get jerkiness in pulselength while doing line interpolated curves (like spiral.ngc)
i don't care about the expensiveness of the software
what does that mean
Bo^Dick: google is your friend
LH: the step generation algorithm runs every base_period, in your case every 20uS
direct connect is my friend :P
it can _only_ change an output pin at one of those 20uS intervals
uh, when looking at the stepper waveform (grey code) when going from one speed to another (looking at one motor, X-motor for instance), as the speed will rise/lower when doing a circle) there is an intermediate step whose shape is very very far off, somethign like 10% of the step, far more than 20µs. Cradek said this was a known flaw
yes but it changes the output pin far too early/late
any way you can show me a scope photo?
or even a drawing of what you see on the scope?
I'll send a video clip
this is really hard to understand in words
just a few minutes
I dunno if I'm set up for video, but I can try
I can change to a videocodec that you have
I just have to figure out what player and codecs I have
(I don't normally care about video, I don't even have a sound card in this box)
uh, this is a linux box?
if so you can apt-get vlc and the codecs with it, then it has mpeg-4 support
(along with lots of others)
looks like there are a lot of plugins for vlc
which do I need?
they seem to be output plugins, not codecs
output is how you get to watch / listen to things
20.7M of stuff downloading
2 minutes ;)
hmm, maybe the codec come with as default
I found out that I have another program that is more suiable, hence the wait ;)
ok, I have vlc installed
have you got a modem or something faster?
so 5mb or so is ok?
a still pic of the scope screen would probably be fine tho ;-)
I've got both a g3 and line interpolated so one can see the difference
sorry, now the server shoudl be up
pw is basic
I included sound so that you can hear the artifacts too
rayh is now known as rayh-away
remember when I said I don't have a soundcard? ;-)
but if anyone else has one they can hear at least ;)
SWPadnos may have one
oh and for the suspicious: pr0n.zip is not what you may beleive it is ;)
I wasn't gonna say anything about that ;-)
it's an internal thing with a few freinds I have, it's actually quite the opposite ;)
what is the opposite of pron?
cute puppies and kittens?
no, more like really hot! (really, I promise ;) ) people (my friends) in clothing that doesn't suit that gender
hold on one sec - I had to stop the mp3 player to listen ;)
thats why its taking so long for me to download it
SWP was downloading the pron and clogging up your line
no, my upload is limited at 100
the earlier part sounds odd, with sudden changes in frequency
the last part is more like I would expect, with a smooth transition in step rates
that's the line interpolated circular move
the latter is a g3
and one can see the "jerkyness"
in the line interpolation
is that M 1.00 ms the time per division?
so those jumps are very large
those are awfully slow step rates then
it's going at 7.5rps
it might be latency in the hardware
invisible to the code in emc that records time
have you tried the rtai latency test
some motherboards, etc, don't work well for realtime
sometimes you can improve it, turning off power management in the bios, etc
however loweing the traj period to 2ms instead of 10ms greatly lowered that effect
nearly removing it
still there however
mostly only audible
I don't think it was the traj period. you also changed the servo or base period at that time
no, im quite sure I didn't
I can test loweing it and making sure not to touch anything else
LH: do you have your scope handy now? or is the video from earlier?
that video is from 3 mintues ago
[18:18:34] <jmkasunich> http://pastebin.com/675512
so it's still there
dunno if you are doing run-in-place or installed, the difference is only whether you have to specify scripts and bin in front of everything
I'm running EMC2 so that first line won't work, what about the others?
with EMC2 not running,
you should be able to do "realtime start" from the command line
the "scripts/realtime start" is for run-in-place
/etc/init.d/realtime start (IIRC)?
I don't run installed here, since I have several versions at any one time
but that sounds right
yes that works
what that does is sets up an inverter with its output connected back to its input, and to parport pin 2
(you can pick any pin)
ok - here it is. you had tried 1.5 ms at first, but that didn't work (bad traj period: 1500000 nsec), then you lowered base to 20 uS, and set traj to 2 ms
then it worked
so I suspect it was the base thread changing that made the difference, not the traj thread
then for this video I just recorded, i set it back to 10ms, and left base at 20
so with a 20uS period, the pin should toggle every 20uS, so you get a 40uS period, 25KHz square wave
since there are no functions in the traj thread
look at the square wave with the scope and see if its jittery
the video I just recorded was with traj at 10 and base at 20
and servo at 1mS?
base is used only to generate the step pulses
servo is used for the rest of the control
servo is at 1ms
and the traj code runs in the servo thread, but not every time
* fenn wants to embed "units" in inifind so i can say BASE_PERIOD = 20us
do I need to be in halcmd to run the code, becuase bash complains
if servo is 1mS, and traj is 5mS. then the traj code will run once every five times
or 20khz or .00000001 fortnight
the code in that pastebin?
paste the line and complaint here
add #!/bin/bash to the top line
* LerneaenHydra agrees with fenn and wants all units in olden units
I mean, insert that as the top line
I ran #!/bin/bash but it still complains when doing loadrt
did you copy it into a file? (as SWP is saying) or just type it on the command line
ok, thats what I had in mind
actually, the problem is that this is an installed emc, not RIP
so the scripts/ and bin/ need to go (not sure about where realtime is)
he already found and ran realtime
halcmd is in the path if you are installed
ok, so just use halcmd blah blah instead of bin/halcmd
so instead of bin/halcmd loadrt, just do halcmd loadrt
btw, when you are done with this test, do:
halcmd unloadrt all
anyone happen to know the screw thread for computer cases and hard drive mounting screws?
I think it might be 6-32
haL. error: function not.0 not found
yeah, I'm 90% sure thats it
did the halcmd loadrt blocks not=1 line generate any error?
fenn: cdroms are different, I think they're metric
cdroms use floppy screws, I think
jmkasunich: yes they are metric
HD and case screws are usually the same
halcmd show comp
halcmd show funct
halcmd show all -> pastebin
there seems to be a fine thread and a coarse thread
fenn: the fine thread (for cdroms) is metric and size M3
blocks isn't loaded
ok that makes sense
the coarse one, hard disks, slot covers and misc, is english and 6-32
ain't that wonderfully consistent?
hell i mix metric/sae all the time
the IBM PC, and the hard disks, are US inventions, the cdrom drive is japanese
12+3/8 = fun fun fun
(M12 and UNC 3/8 that is)
i've got 8mm bearings on 5/16 shafts :(
LH: try the "halcmd loadrt blocks not=1" again
test works now
got a 25KHz (40uS) square wave on the scope?
still tping the rest
is line 9 really correct?
I shoudl split it, right?
and leave out all the jmk stuff ;)
my prompt is longer than half the commands I type
if you do halcmd show now, it should look like http://pastebin.com/675541
20uS hi and low times, 40uS period I hope
ok, do "stuff" on the PC while looking at the squarewave
what's the MaxTime for the thread?
stuff = open a browser
or anything else that excercises the box
compile a kernel ;-)
how did I check time?
wimpy - play a video, be a real man!
even just dragging around windows
5 years later...
emerge -dU world
anyway, you're looking for any jitter on that square wave
if none, thats good (for you) and bad (for all of us)
well, to being with the squarewave isn't perfect
can you trigger on a pulse width?
its jittering a tiny bit
it's not mine so i'm not sure, but I only know how to trigger on rise/fall
that's interrupt latency variation (jitter), and it'll never go away
if its a few microseconds, thats no big deal
no PC is perfect
when moving around windows I get up to 5µs jitter
the bad ones (some laptops especially) will sometimes just ignore interrupts completely for 10s, 100s of uS, sometimes even 10s of mS
mostly is aourn 1-2µs
occasionly (<1%) 5µs, but not more
that is an indicator of the native realtime capabilities of your computer
put the scope display in infinite persistence mode
you'll see bad pulses better that way
* LerneaenHydra doesn't get SWPadnos
yoah, liek that
under the display menu, there should be an option for "persistence"
I have somethign similar I think
ok -I'm trying to remember what a TekScope is like
it's been a while
i've got somethign called peak detect
the scope in the video didn't look like a tek to me
it says tek in the upper left corner
nope, that didn't help, only helped with peaks in the volage, not in time
it looks like the portable tekscope
(the only isolated scope you can buy these days, it seems)
ok - is that one of the portable benchtops, or the one that's battery powered?
(as I look at the video again and see the big TEK in the corner)
just not used to black traces on white background
runs on 220VAC
the display is slow though so I can see if there is jitter
but the worst jitter seems to be around 5µs
and the jitter we were seeing earlier was more like 100uS or more
the litter amount seems to have a gaussian spread
display -> persist -> infinite
yes, really awful
SWPadnos: that works, thanks
after many iterations i've got lots of 5µs and one 10µs
(lots = 5 or so)
oh, is the parport short-circiut proof?
havn't done anything bad, yet
I'd try to avoid shorting it
probably a short to ground, or from one output to another, won't hurt
but shorting to +5 might
and shorting to +12 or more would be very bad
yes, thats cure
shorting to AC mains would be very bad
it would most likely be a short to groun
shorting to AC would be very very bad, but exposed as is rather hard to find
if its a motherboard parport I'd be extra carefull just because they're hard to replace
yes, it is a motherboard parport, and i've ordewd an external one
you ordered that dual PCI card?
I was in a computer store earlier today, and was amazed to see that you can still buy an ISA parport card
no, a single output pci card
yeah I found some ISA cards in a store too, really surprizing
they took far too much for the daul one
of course the same store was trying to peddle books about "Learning Red Hat Linux"
*new* version 4.2!
and I'll use the external one for testing so in total i'll have two ports later which will leave me with lots of free I/O
"includes CD-ROM with RedHat Version 4"
heh - not far off ;)
I did get some goodies that I want to mess with later
thing that fits in a 3.5" floppy space, plugs into mobo USB conn, and takes 10-11 kinds of memory cards
(including the kind our digital camera uses)
too bad cradek isn't here, he may have wanted too see that vieo, he seemed to know about the jerkiness, although he may have misunderstood me
I've got one of those, and it's even a floppy drive as well
yeah, I may actually take it back and get one like that
he can see the link in the scrollback or logs
look out for cable length though
my server probably won't be on though
I couldn't use mine because the USB cable is about 6 inches long
I didn't realize I only have one 3.5" opening
why would you want a floppy drive?
to be able to read/write floppies
to make floppies?
that I get, but why use floppies at all?
not often I admit, but I mess with older stuff
when you have usb mass storage or CDs
they're still useful, especially when you have legacy PCs around
fortunately I don't
I have machines that have no USB, and booting from CD is very difficult
for instance the compile farm has a floppy drive built in, but for CDs I have to open it up and temporarily connect a drive
dual processor PPro, man. w00t!
that is old
I stopped picking that kind out of the dumpster
I still haven't had the chance to try that memory you gave me last year
hey - I paid $5000+ for this computer
my most recent complete computer find is a dual P3
back in '95 or so
I prefer the ones that somebody else paid $5000 for years ago
I even upgraded to PPro Overdrive chips - dual 333MHz
I can understand that a dual ppro was very expensize
yeah - I need to figure out how to get on the other side of that equation
the case has a triple redundant PS, and 19 drive bays
how big is it?
sounds like some of the HPs they're scrapping at work
and the motherboard is a Supermicro, with spaces for 12 slots
it's exceedingly noisy
think full tower but 16" wide
full server size, but wider
not quite 16" wide
triple redundant p/s.. does that mean when it shorts AC mains to 1.8V you're still good?
maybe only 14-15, I didn't have a ruler
[18:57:43] <SWPadnos> http://www.dealtime.com/xPF-Antec_KS_011B_KS_011B_RPP3002
two of them can die and it will still run
fenn, nope. it has only one power inlet (foolishly)
right, but only at 300W (it's 900W when they're all functioning)
isn't that rather overkill?
* LerneaenHydra meeps back
a ppro probably deosnt take more than 10-12watts
at least I don't think they do
I've been trying to decide if it's worth it to get the SCSI drive array running, or just get an SFF pc with dual SATA drives, and go with that
they have rather puny heatsinks with a 40mm screamer IIRC
the overdrive is a bit bigger
then it may take more power
maybe the 20-40 watt range
jmkasunich: is there anything else I should do with my computer to test?
hmmm. that would explain why the whole machine takes only ~185 watts or so
WRT the latency with the line interpolated jerkiness
it seems like the hardware and the RT stuff is OK
yes it would seem so
and G3 givces a smooth movement
however it's worth noting that lowering the traj periood from 10 to 2ms almost removes the problem
SWPadnos SFF ?
"small form factor"
small form factor
small form factor
small form factor
small form factor
small form factor
SWPadnos as a test bos or will turn into production box?
SWPadnos 3Ware card
sff wouldn't work, due to the need for a tape backup
actually thats what I was looking for at the computer store, mini-ITX mobos
they didn't have any
not at a place that has RH4
I can't recomend them
I have one
SWPadnos oh, you talking cnc or video?
whats wrong with ITX?
the VIA boards are slow, overpriced, and have POS linux drivers
I'd like to get this one: http://www.newegg.com/Product/Product.asp?Item=N82E16856101809
Jymm, file server and backup system. probably a CVS server as well as other things
it would be really cool to have a dual opteron in such a small case
micro-atx (slightly larger, 24.5x24.5) use standard parts and have much better performance/price/linux support
or, really warm, more likely
the new opetrons dont take that much power though
only somethign like 40-50W
SWPadnos be leary of heat and bios in those mini systems
only if you spring for the HE or EE versions (much more $)
LH: I'll keep that in mind, I don't need super tiny
the Tdp is 90W or 110W for most of the "normal" opterons
I want to build a system that doesn't need outside air tho
ITX nothing ...
mini ITX ... expensive for what you get
for use next to the mill
the TPD may be that, but they have the same TDP for the whole range of processors
find a small MB, then use the PicoPower PS
use an old laptop
I was looking at the picopower thing, like it
laptop == shit for RT code
I just dont egt the whole small form factor thing .. I mena who has a shop so small that its an issue huh?
Jymmm: laptops suck
so the ones that most people end up getting only use around 40-50 Watts
robin_sz: I do
SWPadnos you said file server, what power do you need for that?
my shop is big enough, but the NEMA box isn't ;)
I want PC, stepper power supply, geckos, and vfd for spindle all in one box
are you suggesting a laptop for a file server?
actually for servers laptops do have some advantages
built in ups for one
if you run the steppers on 12V a computer PSU may be worth looking into
not if you want tape backups or RAID
variable frequency drive
my steppers aren't gonna be running on 12V
I have a 750VA UPS with a spare battery. the average computer will run for a day or so without power
I dount anyone runs steppers on 12V
nema 34, geckos, probably 50VDC at 4-5A
SWPadnos they have a RAID laptop
SWP: I dunno if cradek mentioned it or not, but cvs.linuxcnc.org is a laptop
I know, but not a tape backup laptop (and FireWire / USB / ethernet is way too slow)
50VDC @ 4-5A, wouldn't that burn the motors to a crisp? thats 200-250W!
I'm talking about a business grade file server, with data protection, etc (this is for my home office)
nema34 motors are 3.4" diameter, 5" long and weigh several pounds
they only draw that kind of power when spinning
SWPadnos oh, you mean an office heater!!! =)
and yes, they can deliver 100-200W out the shaft
Jymm, exactly ;)
800oz-in or so
its lots compared to sherline size machines
but not enough for bridgeport size machines
SWPadnos well shit... why didn't you say that in the first place =)
and here I was playing with my 100gram 2oz-inch motors
(although i will get more powerfull ones later)
Jymm, I thought I did when I said "Dual PPro Overdrive" ;)
so I guess my 112 in-lb (peak) servos aren't in the same class, huh?
SWPadnos eh, that's a hand warmer, not a room heater =)
lb, not oz?
28 in-lb continuous
~4x that peak
SWPadnos Considered? http://www.3ware.com/
nope - I have a SCSI cage and 3x36G drives, plus a 50/130G tape backup
no need for the IDE stuff
SWPadnos 36G <--- ancient
SCSI <---- expensive!
I'm just sayin
I haven't generated 72G of data in my entire life
SWP: are those 10Krpm drives?
I think they're 7200, but they could be 10k
they ones I gave you (or were those 18G)?
or was that someone else completely....
I generate 40gb shooting 90 minutes of video!
jymm: not everybody shoots video
oops - nm
2gb shooting stills
the terminal says that running fsck is possibl unsafe, is it ok to run?
why are you running fsck?
I've got... a bit of data from my camera (pics and videos)
power outage recently
although I know of the journal in etx3
every time I've had a power fail or other un-nice shutdown, it did recovery automatically on boot
what do you mena by recovery?
video, audio and photos are different. this server is there to protect the designs I make in my consulting business
get a message like "recovery required onext3 filesystem /deb/hda1"
oh, Ihavn't had anythign like that
maybe the journal was flushed and I got lucky
SWPadnos offsite storage
and it spends a few seconds fixing up the journals, etc
SWPadnos I kid you not
Jymm, not for customer data
Jymm: you don't get it - video and photos are competely different
unless you're talking about taking a backup offsite
so running fsck is unneeeded? is it safe though?
there's probably an option for fsck that makes it read-only, check but not repair
you can't let it try to do repairs on a mounted filesystem unless you know _exactly_ what you are doing
jymm: when you are talking about source code, or schematics, fpga designs, etc, you can literally put a lifetime's worth of work on a few gig
heh. the largest project I have (of files that I created) is ~800k of assembly code
so big is _far_ less important than reliable
maybe a raid-5 setup with some type of weekly external backup?
even raid 1 (mirroring)
daily backup, weekly / monthly / yearly offsite rotation as well, with the yearly tapes being archived
raid5 doesn't waste as much disk space (N+1 instead of 2*N), but for moderate size, who cares
also RAID, for recovery time
raid1 only needs 2 drives, raid 5 needs at least 3
jmkasunich: yes, for small stff raid1 is enough
jmkasunich YOU dont get it... data is data... you dont treat it any differently as you'll never know what usage will be in the future. For a few GB's DVD-R would be perfect archive, not a $800 tape backup.
dvd's loose their data very quickly
there's no optical recording medium that has archival storage lifetime
neither does any other medium
how are harddisks regarded?
but otherwise, you're right - data is data
LerneaenHydra ~5 years
Jymm: tapes is better in that regard
HD's suck for backup purposes, IMO
jmkasunich if kept in env controlled area
5 years? why so little?
the data is then subject to not only media lifetime, but also mechanical issues, and having the electrical interface change
data is data, _except_ that when two applications differ by two or three orders of magnitude in size, the approach that is right for one might not be right for the other
I don't think you can get an ST-506 interface any more
by harddisk I mean an external harddisk that is off except for the back ups
tape is voodoo... ppl rely on it far too much without testing restores
Jymm: that is very true
makes no difference. the internal drive is still IDE or whatever, and that box better not die
well, if you use harddisks you'll have to buy a new one when both the old and new interface exists and transfer
LerneaenHydra hdd is what I would do. you can even keep a few offsite and rotate them in
Jymm: but i don't think the medium makes a very big difference
SWP: can you get a tape drive to read a 20 year old 9-track reel-to-reel tape?
ValarQ: as long as it's not optical that is (unless it's pressed)
ValarQ My biggest gripe is the tape industry hasn't kept up with the HDD
LerneaenHydra: i'm thinking of the issue with few backup tests
yes, that's true
jmkasunich, not now, but there are companies that can transfer data from 9-track to whatever
for a real backup yo have to test the media often
Jymm: that might be true, but tapes have gotten bigger
though 9-track hasn't been "the thing" for more than 20 years, I think
however I do that now (filehash of all files on the media vs all the files on my harddisk when doing an upgrade)
ValarQ I haven't seen a 300GB uncompressed tape that cost under $1000
somewhere in my boxes of crap I have a tape with the code I wrote in college on a TOPS-20 mainframe
the tapes are cheap, but not the drives ;)
Jymm: me neither, but i have seen 250GB uncompressed under that limit
ValarQ The thing is you could buy 3 hdd's for the cost of one tape, and you need to buy a few tapes.
but if you want a real backup rotation, you need about 10 pieces of media
ohwell there's always time travel!
plus 1 extra every quarter or year
Jymm: yeah, but just because everything else has gotten cheaper doesn't mean that there is anything better for backups
multi-hundred-gig tapes are $40-$60 in a box of 10
you don't get that good a deal on hard drives
a very large (disks and geographically) raid1 system might be a solution thought :)
with FIOS, sure ;)
and some VCS alike filesystem...
one on mars, one on ceres.. how precious is your data
Verizon's 40 mbit fiber to the home service :)
SWPadnos url to these uncompressed multi-hundred gig tapes?
oh, sounds like a lot of pun
well, it can't be better than my now replaced system, 3 PATA disks in a raid0 system and no backups
thats exactly it
the R in raid is for redundant :)
we can talk all day about what is best, but most of us actually have nothing at all
3 pata in raid0
i have a stack of cdr's
thats like begging for disaster
I have a 3-drive SCSI RAID in a hot swap cage, and an AIT-2 tape backup ;)
if only I could install Linux from CD
tapes are great
i replaced it with a 3 sata raid5 when i got the money
the data goes on
[19:39:52] <SWPadnos> http://cgi.ebay.com/LOT-5-New-Sealed-Sony-AIT-SDX3-100C-AIT-3-64K-100-260GB_W0QQitemZ9706205988QQcategoryZ73329QQssPageNameZWD1VQQrdZ1QQcmdZViewItem
I have an external harddisk that I do file hashes against very often
sometime you can even get it off again
SWPadnos: sounds sweet
those are only 100G uncompressed, 230 compressed
err - 260
these days I just rsync another remote box from time to time
260 compressed? wth are they talking about, it depends completely upon what it is you're compressing
i have some folders which i rsync to another location now and then
yes it does
code compressed well whereas a well compressed video doesnt compress at all
(hence the name
LerneaenHydra: try lzip :P
thats a dumb way to rate storage
LerneaenHydra thus why I never use the compressed values
sounds like marketspeak
swp those are $250
[19:42:33] <SWPadnos> http://cgi.ebay.com/Dell-LTO-Ultrium-3-400-800-GB-Tape-x5-NEW_W0QQitemZ8795660342QQcategoryZ3756QQrdZ1QQcmdZViewItem
it's a lot of 5, so they're $50 each
SWPadnos: wheres the $60 for 10 of em
$60 each is what I meant
$60 each in qty 10
tapes are a WORIYAL sort of medium ...
readi f you are lucky
so fuckign true
in my experience, I've only had perhaps one bad tape
(that was noticed, when something needed restoring)
reading from the same drive they were written in, its been OK
buy our 100EiB* drives, now for $100
the ecould have been others that would have become problems if a restore had been needed though
*uncompressed capacity 100KiB, compressed file is a file filled with a recurring single charachter
with multilpe drives .. its been crap
bad tape, bad firmware on the tape drive/library/robot/ or so many issues
the problem is that as density goes up, head alignment becomes a problem
with one drive on a aserver, you are pretty much OK
[19:45:43] <SWPadnos> http://cgi.ebay.com/TDK-10-PK-LTO3-ULTRIUM-400-800GB-TAPE-D2406-LTO3_W0QQitemZ3828891849QQcategoryZ3307QQrdZ1QQcmdZViewItem
with multiple drives across multiple machiens ... well, there are issues with most formats
tape interchange is not something they seem to get right
$89 300GB hdd from fry's
that was true up to (and somewhat including) DAT4 and DLT
jmkasunich: wrt to the jerkiness in line interpolated hilxes, are you going to talk with someone who could help resolve this? or what's going to happen?
it's a lot better now, with AIT and Ultrium
LH: I don't know what is going to happen
unless I can reproduce the problem here, probably nothing
the problem description is just to vague
jmkasunich: and if you do reproduce?
errr, reproduce it
then I can try to debug
spirals are always fun
can you post your config files (tar up the directory) and the g-code program?
uh, any particular server that I should put it on?
or should I just host it myself?
if its not huge, (under 100K) just mail it to me
(I can't access the linux box for an hour or so, im making a disk image)
just teh stepper config?
that's under 100k
whatever config you are using
nothing else I should send?
maybe the program?
yes, the g-code program that causes the problem
wasn't it just spiral.ngc?
all programs cause it
I can't guarantee that I'll get to it today (or even this weekend) but I'll try
anythign is good
$0.30 per GB http://www.netaffilia.com/ad/electronics/frys/i/2006/04/21/16483.html
$90 each for 400G tapes is a good deal
and one you can't match (yet) with HDs
though there is the $3000 investment in the tape drive to consider ;)
rayh-away is now known as rayh
SWPadnos and have you ever noticed how outdated tape drives become anymore?
Jymm, not really. you can still buy a new DAT4 drive. those were introduced ~10 years ago
I've seen barnd new still factory sealed tape drives at the surplus stores.
they're not "new and exciting" for long, but for legacy compatibility, they're a lot better than hard drives
I can still use a MFM hdd if need be
(scarry as that is)
can you buy an ST-506 card, or do you just have a really old PC that still works?
(oh, and MFM/RLL drives were only usable with the controller that formatted them usually - that's why IDE was invented)
but anyway. we can go on about backup methods almost as long as "Estop and what it means to me" ;)
heh... I have (had) an 8bit scsi card around here somewhere =)
heh... I have (had) an 8bit ISA scsi card around here somewhere =)
I think I still have one as well. it came with my 2x SCSI CD-Rom
What I'd love is the archivalness (?) of photographic film with the digital scanning, sorta like barcode
SWPadnos meant to ask you, do you play with AVR's at all? like the 2940's ?
SWPadnos: what do you mean wrt estop?
what he means is that every person has a different idea of what the proper estop circuit is, and we can discuss it all day without ever agreeing
ah, ok. I'd better not open that can of worms then
ding! jmk wins one point
Jymm, yes, I use AVRs a lot, but never a 2940
SWPadnos just the megas?
I don't recognize the 2940 model number
I use the "normal" and megas
SWPadnos AT90S2313 (my bad)
ok. I've used the 2313 before, and the 2343 is the one with analog, right?
SWPadnos Hey, I couldn't remember the PN, remember =) What I was windering is what you thought of them, like when would I be pushing their limits
they're my favorite 8-bit microcontroller
SWPadnos what was the most you've done with them?
remember that 800k project I mentioned earlier? ;)
SWPadnos exploding stuff?
power supply controllers
I've used several chips, from the 90S1200 to the Mega128
SWPadnos you ever looked at the BASCOM-AVR ?
SWPadnos BASIC compiler for AVR's
jmkasunich: what is your mail adress?
everything I've donw so far is in assembler, though the big project may be converted to C soon
SWPadnos Oh, ok. I dont' know C, so I thought I'd try my hand at the BASIC compiler
Ok I haven't touched C in decades
if I wanted to use a HLL, I'd use C, so I never bothered looking at the basic options
SWPadnos Well shit... If I have to (re) learn C, might as well be on a uC
there's a gcc for the AVR
LerneaenHydra: did you get it? (I sent it privately)
nobody ever regrets learning C well
jmkasunich: there, should be send now
cradek no regrets, just commitment to it along with 90,000 other things =)
so if my new espresso maker says it makes 6 cups, is it bad to put that all in my regular coffee mug and drink it?
not not not not at at at all all all
it sure is good
LH: got the config, is the g-code program in there?
cradek: just look at SWPadnos line to see what happens
jmkasunich: *headsmack* sorry, forgot that
spiral.ngc, or one of the loop things we did?
spiral will work as well
eat an orange to get rid of the caffine shakes
cradek: you know how I was talking about the jerkiness in motion before?
eat coffee beans to get rid of that orange flavor
I think my beans are decaf but I'm not sure
cradek green label on the bag?
LerneaenHydra: yes, I'm happy to see jmk is going to look at it so I don't have to :-)
Jymm: they're in a jar, the bag is long gone
cradek: I recorded a video of the effect if you want to see (mpeg4)
I wonder what is worse, powering the drives while the mobo is unpowered, or powering the mobo while the drives are unpowered?
yeah is it online somewhere?
pr0n.zip is not what one thinks it is ;)
are the other temps for me too?
jmkasunich, I'd bet that powering the drives with the MB off would be detrimental more often than the other way around
no, but you can download them if you want
other random stuff I've done
old skool use to say power up ext devices before mobo
that's a good video
I bet the first thing jmk will do is plot the velocity in halscope
jmkasunich: sent the gcode now
also I'd like to know how many traj periods it takes your machine to accel from stop to full speed, I bet it's not very many
no, it's not many
accel is 500 and maxvel is 7.5, and i have scale at 200
I'm running the motor naked (unconnected to anything)
that's a pretty high accel
the motor is run sub-critically and unconnected so it works
naked == cover removed. Man, I was wondering how you were doing that?! lol
I think that's .015 seconds of accel, which is ~ the traj period
should I test accel at 100 or so?
maxvel = 7.5 and accel is 500? that means zero to full speed = 0.015 seconds
(cradek is faster with the math)
blending will be terrible in this case
IMO that accel is just plain too high
that could be the cause of the jerkiness
thats the defailt value
maybe that value can make sense on a higher speed machine?
hmm, that's true
I just checked stepper_mm.ini
I'll test lowering it
maxaccel is 500, and maxvel is 30.48
dunno where 30.48 came from
but that works out to 0.060 seconds zero to full speed
in both the traj section and the axis sections
yes that's what's in the sample config in the release
are you sure you've opened the right file?
[21:04:14] <cradek> http://cvs.linuxcnc.org/cgi-bin/cvsweb.cgi/emc2/configs/stepper/stepper_mm.ini?rev=188.8.131.52;only_with_tag=v2_0_branch
aha, yes in the defaukt
I though you mean my file
so you reduced your vel but not accels proportionally
the load will still be the same on the motor
you said "thats the default value" so I went to see the default
I bet when you put some mass on that motor, it won't be able to accel like that
no of course not
ah, accel is default, but not vel
the defaults (both inch and mm) work out to zero to max speed in 0.060 seconds
maybe try accel of 100
ok, so I should test something like that
yeah, ok. i'll test that
that's still only 6 traj periods, but much better than 1
it's still somewhat erratic
should I change the traj period to maybe 2ms instead of 10?
jmkasunich: did you see the earlier discussion about the traj thread? it doesn't show anything in halcmd show thread
it has nothing running in it
the change before was due to a change in the base_period
I didn't know LH changed that too
remember - there was a problem with the setting at 1.5ms (maybe it was 0.15 by accident), so he changed the base period to 20 us
I wonder if we're getting any smoothing from the cubic subinterpolator then
yes but that wasn't the factor that fixed it
just pointing out that two things changed, not just one
cradek: the traj htread is empty (should probably comment it out and not even create it)
the original code ran the tp in the servo thread, every Nth time
with the base period at 20 and 2ms as traj it is semi-smooth, but with base period as 20 and 10ms as traj (I changed it back *afterwards*) it got all jerky again
N = traj_period / servo_period
how is tp run now?
in that video I sent at 10ms
LH, he's talking to me
currently at 2ms
cradek: its in the servo thread, like before
but it might need looked at
it does a tp cycle every servo period then?
its not called explicitly by a divide by N, its called with the cubic interpolators go empty
ok I have no clue how that works
now that i've lowered the accel it doesn't jerk as erratically, however it still jerks 100µs or so, although it is more smooth
emc2/src/motion/control.c line 1552
that and the surrounding code is essentially copied from emc1
the while should only be true every Nth time thru
Jymm is now known as Jymmmmmm
so if LH wants more TP runs, he needs to decrease SERVO_PERIOD not TRAJ_PERIOD?
and he can see how much is safe using halcmd show thread
Jymmmmmm: is seems there's a nice avr support in linux, but I never try it pratically ..
consider to switch to it
line 425 = end of file?
[21:20:15] <giacus> http://unix.rulez.org/~calver/pictures/curves.jpg
giacus: Loads of avr support under nix, just have to learn C is all.
are we looking at the same file?
yeah vim always bites me in the ass when I use C-g to see the line number like vi
that's apparently the condition that makes the tp run again
then the code in control.c is just plain fscked?
I'm not claiming to understand it
start at control.c line 1547, and read (the comments mostly)
the contents of the while block should run at the traj rate
the rest of it (starting at "there is data in the interpolators" ) runs at the servo rate
I guess that needs instrumented to see if thats what really happens
jmkasunich: how the work is going ? I read something in the mailing list about qemu/vmware
are you building a linux cluster ?
the compile farm is sort of a cluster (except that they don't talk to each other)
I gave up on using qemu, after wasting a whole weekend on it
does it work fine ?
a bit slow, if I understood well ..
more than a bit
wmware server should be free now
but I've no idea if it could help
if my base system was a couple GHz I would try it
but its not
I have a dual P3-600
I was hoping it could emulate four 200MHz Pentium's and be at least as fast as the real hardware
but its not
was a nice idea
[21:50:19] <giacus> http://unix.rulez.org/~calver/pictures/in_class.jpg
anyone remember those holders that held 5 pieces of chalk at a time?
for making music lines on a board?
that's bogus, a C programmer would have written i=0; i<500; i++
jmkasunich: or for writing 500 times/5 on the chalkboard =)
a c++ programmer would have written (for int i=0;1<500;++i);
err - except for the misplaced parens ;)
giacus: Is that a pic of you?
an observation or an insruction?
* skunkorks ment skunkworks :)
hey robin - how is it going?
mostly an observation
so busy these days
nice to chill on IRC for an evening
skunkorks, you can change your name with /nick SkunkWorks
yes - it is nice to remember I now don't need an irc client installed to get on the channel
robin_sz is now known as skunkworks
skunkworks is now known as robin_sz
skunkorks is now known as SkunkWorks
tonights topic is:
lets talk the usual crap ;)
I ground another carbide into a engraving tool (wanted a sharper tip than the one I made yesterday) and promply dropped it and broke the tip right off.
you can buy engraving cutter quite cheaply
yeh - but I wanted one this weekend and didn't think ahead.
"cool, a 000 center drill ... crap"
SkunkWorks like with a .01 flat on the end?
theres a lot of data on engraving cutters
lots of sharpening intrcutions out there
engravers are always tinkering
are you talking about V bit ?
tooling geometry never ceases to amaze me
for some reason, I now have two 1000 litre vats of grinding fluid in my workshop
robin_sz do you even have a grinder?
I had a better link yesterday - I pretty much made one of these http://www.antaresinc.net/EngravPage.html
useful stuff though
we use it for water testing tanks ... its got a rust preventative in it
had a better link yesterday
SkunkWorks: notice how it os not symmetrical
well that didn't work the way I wanted it to :)
SkunkWorks: notice how it is not symmetrical
right - I cut radial releif
it should only cut on the leading edge
if I did it right
the govenrment here vut tax releif. dame thng I guess
it was by hand - you don't want to know
on wood its pretty simple
BUY MY LASER
on alu/metal should be a bit hard
I guess the one I made yesterday worked pretty well - just had too big of a flat on the end
* robin_sz tries subliminal marketing
YOU NEED A LASER
GET A LASER, GET LAID
* giacus want a laser too !
[22:30:15] <robin_sz> http://cgi.ebay.co.uk/Ferranti-450W-CO2-laser_W0QQitemZ7611858229QQcategoryZ26237QQssPageNameZWDVWQQrdZ1QQcmdZViewItem
* giacus for free
I'm looking at a 25W, 10yr old laser engraver right now, just not sure about it
he's asking $7000, but might offer $4000
epilog summit, mfg 1996
he told me a mirror was oxidized and replaced, but when I called T/S they have no record of that repair.
they do have record of a belt,motor, and bearigns replaced in 2001
Well, how did they get the replacement mirror
mirrors and lenses are a consumable
millions of places sell them
also old parts ?
no point, new stuff is cheap enough
laser II-V1 are the biggest
[22:35:25] <robin_sz> http://www.ii-vi.com/
I have no power meter, and it's $1500 for a replacement tube, so I really have no gauge as to the life left in the machien.
I'd be pissed if it went POOF a week later
manual focus, and no air assist. TS said there's no way to retrofit air into it either,
robin_sz: my cousin is always cutting steel by hands :(
he's building beds
with the plasma cutter
should be ready with the cnc table in 3-4 week
he's doind very nice jobs by hands
I've to ask him or some photos
has been very busy in these latest months :(
steels beds here are going a lot
old style steel beds
[22:47:23] <giacus> http://www.vitaliletti.com/
similar to that
Jymmmmmm: that could be a nice business
not difficult at all, maybe
steel and wood
giacus: no place for such a large laser, much less the supply
pff lasers are so 1970's
laserless is the future!
Jymmmmmm: a plasma cutter could be enough for that
giacus: I 'll stick with a laser
laser is nice, what I doubt if is cheap too
giacus: I've seen what a small (<250Watt) laser is capable of, and I'm happy with the results.
I'm not so updated to latest tecnology progress
I just know laser tecnology as very expensive
giacus: Yes, not everything can be had in a surplus store
that mean fo me, also the final product should be very expensive, of course..
you're looking at surplus store
what I doubt is the cost to mintain it after
of course I've not experience about t as robin_sz
its just my point of view
so, not consider it at 100 %
robin_sz: have a great experience about laser tecnology
however, he use it in a factory use
I meant for an hobbist, or a people who want to start a business without spent a lot
alternative tecnology could be good
that's my experience
what I learn in these years is: very often what cost more inst necessarly a better solution
as example, linux i free and is better than windoze
but if you can get a laser machine for a nice price, without issues, its nice.
let me know, i'm interested in too
I was chatting with him in a different channel, and suggested he check this one out
he's a mechanical engineering student
wich channel ? :P
seriously, nice distro
I like it
for what my opinion can be assumed :)
well, mr bush called ms prodi from his air one plane to congrats with him about the victory
mr berlusca says he will never congrats with mr prodi
but we have an hope yet ..
andreotti could be the next president of 'senato' in parliament
of course.. you don't know who is andreotti ..
since FI losted the elections
and mafia don't have any connection to the goverment more
andreotti could be the next wire to connect it
its as an HAL component
but work bad :/
need to be debugged