[02:11:56] <skunkworks> http://www.electronicsam.com/images/KandT/pcboard.JPG
hey that looks pretty successful
so far so good.
is this one of your homemade engraving tools?
I have a crappy video of a drill breaking also :) drilled about 10 holes then broke. Wondering if the collet is out of round.
kind of a crappy holder
what rpm, feed, depth?
about .01 further than the board thickness - around .07
25k rpm and 3 ipm
feed seems slow
it seemed to do it at the bottom - there is a second delay - do you delay at the bottom of the hole?
0.00012" per rev
try more like 10-12
yes but a second is WAY too long
in the words of Buster Poindexter: "Hot Hot Hot!"
at such high rpm you probably don't need any dwell
Remember the gcode generator had a 10 second delay :) but it was probably for a differnt flavor of gcode
anybody here use the m5i20 board?
yeah probably 10 ms
almost there :)
pads are pretty much big enough now for my tastes
I have one, if you need some test run
I'm writing the doc chapter for it
and I realize I have no clue what many of the hal pins do
are you gonna bring your board to fest?
I want to revise the driver to use the canonical encoder interface
in fact, we can see what happens when you load both the m5i20 and the ppmc driver at once
plus an extra parport or 3 (and cl, for even more pins ;) )
skunkworks: this is with my 16-18? krpm spindle
eventually you'll run out of hal shared memory
I think these are the numbers I used most recently
I wonder when that happens?
when it gets full ;-)
halcmd stat will tell you what you have
the most I've seen is about 2/3 full (with a very complex config)
usually it's pretty empty
I guess I'm conservative on the drilling too
never had any problem
SWP: any idea what dac-<chan>-interlaced means?
the DACs are actually PWM, so it could have to do with the timing of the PWM pulses
hmm - a near-default univstep config has 39538/65500 bytes used
I loaded m5i20, and it's up to 55093/65500
parport adds 3.5k
the damn names are stored in shared memory, and they just eat it up
especially since they were increased from 31 to 41 chars a few months back
any magic with 64k? this is a 32-bit system after all
maybe superstition, but I think there were problems with getting large blocks of shmem from the RTOS
could be expanded, but would need to test on all the compile farm slots (at least)
how does one load cl? just loadrt classicladder?
you'd have to run it, I think - not just compile
it gets a bunch of insmod params to set max rungs, etc
yes, actually run on the farm to test
see demo-mazak config for loading CL
servo-thread ( 50975, 74636 ) here.
re the memory thing... I want to get names and other metadata out of shared memory someday
I'd think they have to be shared, since the name is the ID for each "resource"
HAL memory status
used/total shared memory: 63449/65500
active/recycled components: 10/0
active/recycled pins: 422/0
active/recycled parameters: 380/0
active/recycled signals: 42/0
active/recycled functions: 23/0
active/recycled threads: 2/0
univstep + m5i20 + one extra parport + cl_rt module
so about 870 objects
no ladder loaded
73 bytes per object
looks that way
900 is about the limit, I guess
41 (IIRC) of that is the name
more than half
theres a lot of other metadata too, type, links for various lists, etc
I haven't added anything else to the threads either - that would take some space, right?
those are small tho, cause they don't contain the name, just a few pointers
of course, now that I've loaded this stuff, even if I unload the extra modules, I'll still be stuck with this split for pins / signals / etc.
until I run out of shmem
let's add that to the discussion list at Fest ;)
the core architecture needs a major rework
but that is (hopefully) invisible to users, and therefore very low on the priority list
very few people will give a shit (depressing, but true)
isn't Fest about the code?
it depends on who you ask
its also about the users
and the projects
yeah - I'd say it's about the code, and the "user experience"
the hal restructuring will definitely need a branch, thats not a one week job
I think there's enough to talk about, that we need to concentrate a fair amount of time on making a list of TODOs, and a plan
other time for coding, other time for teaching, etc.
very little time for sleeping
"sleep? what's that?"
I'm sorry - I didn't understand the question
damn, I don't know much about the m5i20
lots of pins with "duh, I dunno what it does"
I can look at the FPGA code, but I'm not sure it would help ;)
not real high on my list right now
I just want to get the lists in the doc
they're gonna change anyway
to find out what the pins / params do
wonder what "led-view" does?
I think it lets you put some of the I/O on the onboard LEDs
you select which byte to stick there or something
nope - it's an 8-bit pattern that gets output to the LEDs
technically, I suppose that should be 8 HAL_BITs
its a U16, not U8
I wonder why?
even though only 8 are used
could be, but the hal should be based on what is used
I agree, it should probably be 8 bits
led-0 thru led-7
led.0 - led.7, actually ;)
well, this board is so far from any convention that it doesn't matter
true. but petev likes it ;)
I hope he doesn't like it so much that he gets made when we revise it
I have the FPGA code for an 8-encoder version, which also fixes a bug in the current version
I haven't had time to check it though
hmm, according to the comments in the driver, it _is_ an 8 encoder version
still only four dacs tho
(from pete W, via petev)
hmmm - lemme check that email
haven't checked the #define of M5I20_NUM_ENCODER_CHANNELS, maybe the comment is wrong
apparently the secondary encoders wouldn't have worked, due to the bug
or at least some of them wouldn't have worked
[03:09:50] <skunkworks> http://www.electronicsam.com/images/KandT/drilling.ASF
<really crappy video
03jmkasunich 07HEAD * 10documents/lyx/emc2/hal_drivers.lyx: added info for the m5i20 card, lots of details are still missing (I don't have a card)
SWP: would you mind looking that over, add/verify some of the details?
I can try, but I
I'm not sure how much good it will do
everything I know about the m5i20, I know by looking at the source
the real facts will have to wait for fest I think
* jmkasunich starts on the motenc board
too many damn boards
03jmkasunich 07HEAD * 10documents/lyx/emc2/hal_drivers.lyx: added info for the motenc cards, still need one or two details, but most if it is correct
still need info for the ppmc, but not tonight
Use the source SWPadnos!
SWPadnos: Have you used the 5120 card yet?
Jymmmm is now known as Jymmm
Jymmm is now known as Red70sShow
Red70sShow is now known as Jymmm
Jymmm is now known as Jymm
is it possible to select two objects in the eagle board layout and type a command that displays the ratsnest for those two only?
mention an expensive cad-software that can autoplace, auto gateswap, autorotate and autoroute a design
Bo^Dick: why not autodesign the schematic aswell
oh, and maybe build it too
well the schematic isn't as critical as the layout
i made a... thing. seems to have turned out a bit better than my last spirograph attempt: http://fenn.dyndns.org/pub/camera/DCP_0637.JPG
cool ... thing
nice ... thing
code here if anyone's interested: http://fenn.dyndns.org/pub/irc/3dspiro5.py
based on jepler's spiro.py
awww I am so famous
world famous in #emc
that'll do I suppose
coffee time, I guess
at one point I thought AXIS was getting 1000s of source downloads each month .. then I realized it was mostly due to a few people who downloaded *all* of the source snapshots ever released.
jepler: what's the purpose of that?
like webcrawlers or stupid page-archiving software
I would have thought those are somehow at least pre-intelligent
When Sherline set up the server for emc stuff they included the audio files between matt and fred and were serving a tb every six weeks
hi ray. rainy day perhaps here, so back to the shop
I was going to take off
No rest for the wicked.
this farm stuff just reminds me that we really have poor weather prognostication from NWS. They can't even tell you the weather for the next day.
no, but they usually can for the previous day
"for the next 24 hours, there will be weather
although not very precise
it's a probable weather
"for the next 24 hours, there is a 50% chance that there will be weather"
dark tonight, with scattered light by morning...
we always have weather here
diminishing to 30% in the south
I think they announced 90% for tomorrow here
sometimes its sunny, sometimes its cloudy
always miserable though
Just got an invitation to http://www.3tfuari.com/indexeng.php
I guess they think you're a Turkey ;)
I need somrthing dramatic for the air generator. I was thinking a load of 20 or 30 ultrabright LEDs....
I've got just the LEDs for you
135000 mcd white
at ~3.2V, 20 mA
* giacus just back to home ..
I have a big bag of em somewhere....
how was napoli ?
that's 13500 mcd right?
no, 135 cd
or 135k mcd
they are *bright*
holy cow those must be very focused
nope - 10mm, they throw a spot about like a mini kryptonite light
and about as bright
what size are they?
[13:18:31] <SWPadnos> http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&item=7609370940
10mm LEDs are actually really huge - they're kinda funny looking
it's like the old christmas lights - the big round bulbs, instead of the pin lights they use today
I want some
they would be perfect
can I get your address from your website?
or do you need more than a few?
(if you just want to play around with a few, I can send you some)
then again, it's $30 for the 100 pack ;)
I have about 22v 5VA 40 kHz ac
5 LEDs in series, plus a rectifier
but you need way more LEDs to use all that power
maybe 6, if the batch you get is a little lower voltage
you'd want about 50 LEDs total, 10 strings of 5 or 6 each
would need 78
12 strings of 5
is that 22V p-p, or rms?
ohh that would be big
yeah - I'd get the smaller ones, even though they won't be as bright
22 peak....15.5 rms
yeah i'll do that...lower brightess but perhaps similar total flux
possibly. the smaller LEDs are both smaller and dimmer, so the flux should be a lot lower
5 watts of LED will light a room pretty well I think
the big ones sure will. the smaller ones probably not
I haev a bunch of boards with ~36 5mm LEDs on them, in various colors
they are bright to look at, but they don't illuminate things very well
how many lumens are those?
between 6000 and 14000 mcd per LED, depending on the color
oh yeah - if you sand off the tip of the LED, you get a much better diffusion pattern
the ones I have are red...for automotive. Because they are designed to tail light specs the are lower mcd but high luminous efficiency
are those the clusters or discrete LEDs?
ah - OK
I was asked to write a paper on automotive LED use years ago
so they are old
I can send you a protel layout for a 36-LED carrier board, that's configurable for 3 strings of 10, 11, or 12 LEDs, or 6 strings of 5 or 6 LEDs
similar to the LED I used on the boeng 777 cockpit annunciator
they won't be anywhere as bright as today's LEDs then
14 years ago
yeah I had better get some modern ones
luminous efficiency and flux density have both increased by a factor of 1000 or more in that time ;)
any loss of reliability or lifespan with the new?
hmmm - let's se ewhat happens if I try to run Altium while I still have LabView going ;)
I had better grab some quick....the device has to be in chicago the 15th
nope - better if anything, AFAIK
if you need them fast, just order some and send them to me - I can send you mine
guys, what setting do you usually have for your stepper drivers, full, half, quarter or 1/eight stepping?
they take about a week or so to get over from HK
let me check on it
1/10 - geckodrives ;)
though I use servos, so that's not quite true
clearly the ideal system will command a theta in full IEEE854 double precision floating point, so that this issue of "stepping" is relegated to the past. none of this step+direction bullshit either.
i thought one could adjust that setting in the gecko stepper drivers
winding "A" current = cos(theta) * I max, winding "B" current = sin(theta) * Imax
yep - 1x, 2x, 5x, 10x
i'm considering building a stepper driver that could only do 1/4 steps and no other variant
have you looked at the linistepper for inspiration?
in fact no
have a link?
[13:41:48] <SWPadnos> http://www.piclist.com/techref/io/stepper/linistep/index.htm
SWPadnos: have you used linistepper?
that's a unipolar driver
it seems like such a terrible idea
so several days from best hongkong?
nope, but it looked neat way back when
[13:42:48] <SWPadnos> http://www.piclist.com/techref/io/stepper/linistep/lini_wks.htm
how it works
here's my favorite pic of linistepper: http://www.piclist.com/images/member/RB-ezy-Q33/jelly01.jpg
that's a big heatsink ;)
Jelly Baby Engineers: "Heatsink? You call that a Heatsink??
THAT'S A HEATSINK!!"
that's a small heatsink
there's gotta be a better way to drive the motors, but they have some good ideas about microstepping, I think
but the bottom line is that you really use the 1/10 step mode in practive??
if I were to use steppers, I would absolutely use microstepping
(I've never used anything beyond halfstepping)
the more microsteps you have, the closer to an AC servo the motor bevomes, for the most part
i've heard that there is a tradeoffs with microstepping. otherwise all people would use 1/16 all the time right?
I think the main "tradeoff" is the required step generation rate
the tradeeoff is that you need to generate 16x as many steps to get the same rotation
it's hard to get past 25kHz step rate or so in emc2
with an FPGA based step generator (USC, m5i20), that's not an issue
this limits motor speed
or, you use something like the gecko, where the driver *always* microsteps the motor, and you have a step multiplier on the input
you still get smooth motor motion, and you don't limit the speed
these look ok if I could get em fast
[13:48:38] <les_w> http://www.lsdiodes.com/shop/index.php?main_page=product_info&cPath=1&products_id=19
damn - that's expensive
argh - gotta run
yeah, but I need fast
SWPadnos is now known as SWP_Away
why is there a 25kHz limit from emc2?
Bo^Dick: software step generation
that sucks if you dont have geckos right
Bo^Dick: if base_period is 20us you can change the state of an output pin at a rate of 50kHz
that means you can issue a step pulse at a rate of 25kHz
in fact what i'm considering is if i could build something that can only do quarter steps but i dont wanna sacrifice speed
cradek's maxnc: 16tpi leadscrew, 1.8 degree motor, halfstepping, 30ipm: That's only 3.2kHz
it would only be 16kHz for 1/10 stepping
so with quarter steps there is no way that it wouldn't be fast enough?
sure it would. quarter steps is 6.4kHz which is a lot less than 25kHz
$0.90 is expensive/
i guess its not a 135,000 mcd
but those 5W luxeons are like $35
but this is for president of finishing systems...
(of course, if you design something with quadrature inputs you can generate a full 50k steps per second at a 20uS base_period)
les_w: get a green luxeon led
I see a demo where the air generator is on a table.....hook up little air line and turn it on...and 60 ultrabright leds shine right in the viewers face
blindingly, ridiculously bright
with no moving parts
these led's have heatsinks on them
holy crap the price has come down 10x
[13:57:59] <fenn> http://www.lumileds.com/products/family.cfm?familyId=1
well I need to get some fast. Low price is good, but not required
actually you probably want the star (integrated heatsink) http://www.lumileds.com/products/family.cfm?familyId=2
those look fine
mouser or digikey have them?
A demo like that would knock their socks off
tiny little gadget with no moving parts hooked to an air line....
using only a little air
and blinding light
well i'll get to the shop
have to machine some lexan for the thing
btw, would 1/4 stepping be considered smooth enough?
asking you guys who has got the experience of couse
i guess an experienced stepper driver user would have an opinion
1/4 stepping is what the Sherline version uses.
would 1/8 stepping be seriously better than 1/4 or would the downsides of too small steps be significant
i think i could build a 1/8 stepper driver but is it resonable to believe that it would be worth the effort
the intrest for stepper drivers doesn't seem too big in here. do all of you use servos?
Bo^Dick: I think the vast majority of emc users use steppers
Most stepper folk use commercial drives
i've heard there are some poor commersial stepper drivers too
cradek: the first one you had wasn't very good was it
cradek: i mean before you made the L297/8 design'
Bo^Dick: the original maxnc drivers were unipolar "L/R" type
it was good enough to make parts. it was a commercial product made as cheaply as possible.
what would you estimate the price/unit for those poor stepper drivers
the aluminum box was probably the most expensive component
the rest were some TO220 darlingtons and some TTL inverters
either that or the 20 watt? resistors
the second most expensive part was probably the power resistors
or maybe the transformer
as for what i've understood your stepper driver works well with half-stepping only or do you use it in full step mode?
I use half stepping
my understanding/experience is that full stepping doesn't really ever work
how come there are microstepping drivers when half-stepping works fine?
oh for petes sake
i mean can womthing be better than good enough?
you can make a stepper driver for $5 or $500 and you get a range of performance
why do you keep asking the same questions?
well why would anyone buy a 500$ stepper driver when a 5$ can do the job just as good
there just has to be a catch
to be really honest i'm standing between building a 1/4 step stepper driver alternatively a 1/8 driver
i don't wanna rush into it
Bo^Dick, The answer to your question is in the question itself. "just as good"
A Gecko works well for within it's specs
A Xylotex works well within it's specs.
A unipolar works well within it's specs. In fact I'm proposing the use of a unipolar for a new customer now.
does unipolar mode have advantages over bipolar?
The extra winding on the motor required to match a bipolar costs less than a properly protected bipolar driver.
They are harder to kill.
i think i'll go for a 1/4 step stepper motor since some tables only move 1mm a revolution
else the whole system would be slow
in fact i was somewhat surprised that the emc2 could only do 25kHz for the step pulses
I've run em up to 50k
but then you step on both flanks dont you
but you can't expect a dumpster box to do that.
so in the end it's dependent on computer performance
after all i don't have any insight in how the software generates the signal. i was thinking it had some kind of lookup table that provides information about the step pulses
in my own experiments i managed to stress the parallel port output to 250kHz
without any processing between the flanks
lookup tables are usually very fast
what i'm also thinking about is that lets say that i'm finding out that 1/8 stepping is too slow, could i just add on a pulse doubler?
03rayh 07HEAD * 10documents/lyx/emc2/Master_User.lyx: Tweaked layout
or put in other words, i guess it's easier to turn a 1/8 stepper driver into a 1/4 one rather than the other way around
Right the frequency multiplier used by Gecko is quite interesting and complex.
i do realize that since the doubler must generate a new pulse with a time delay that is proportional to the incoming frequency
does he use some sort of PLL?
so dividing is easier than multiplying.
But using a pic for this is fairly easy also.
I think the pll is one of those neat circuits. :)
is there a wiki or something describing setting up an external parport?
(by external I mean PCI-based)
all you do is plug it in and then use lspci to try to figure out its address
then put that address in your hal file
nothing more than that?
well in fact there is no problem with multiplying a frequency unless you can't wait the first period to get the new frequency
ooh, major update today in ubuntu (or well, I noticed it today)
yeah there was a security advisory regarding X
oh, that one that only affected certain hardware?
How do I get the adress to the normal hexadecimal format?
Is there a quick way to get a listing of the dependencies of a deb package?
check the apt manual
there is a check command IIRC
the lspci output is 0000:01:0c.0
for my parport
Lerneaen_Hydra: that's not the address, try lspci -v?
I take it it's the 32-bit wide adress?
it's the dff0 dfe4 or df80
I think the typical procedure is try each with your fingers crossed
That's the way I've always done it.
then run the system and see if you can toggle bits.
is there any limitation to how many PCI parports you can have?
other than running out of PCI slots
Lerneaen_Hydra: I don't think so
Lerneaen_Hydra: but I doubt anyone has used more than 2-3 so far
and here comes another topic change. the encoder on the lathe currently sends a 1 when there's not a hole and a 0 when there is a hole (or vice versa), is this the type of code EMC wants or do I need to process the signals. (this applies to both the index hole and the 100-holes-per-rev-holes)
emc requires quadrature counts with an index pulse once per revolution
(if you're using the hal encoder module)
if it has only hole/not hole, how does it know which direction it's turning?
Are there two pickups per set of holes
hmm I bet it only turns one way
you could write a new hal module to count this way if you have a clean signal out of it
cradek: it doesn't tell which way it's turning (that is something emc shoudl be able to know, seeing as how it sets the motor dir)
cradek: at the moment it only spins one way
So you need a simple counter not a quadrature counter.
rayh: there is only one pickup per holeset
Lerneaen_Hydra: yes and no - when starting/stopping it seems like you could move one hole's worth in the wrong direction
enough to mess up your count
cradek: the index would come soon though
but aside from that I think you could write a new counter module for it.
cradek: heh, I can't write something like that. I have no experience with any programming language suitable for things like this
I bet you/we could easily copy the encoder module and rip out the quadrature stuff.
a simple mouse has 2 sensors per hole.
is the index pulse in a encoder smaller (a smaller "hole")
skunkworks: there are two photodiodes, both the same size, one for the index and one for the higher-res stuff
skunkworks: a mouse doesn't know which direction it's spinning though
yes it does
I meant the opposite
I guess I was thinking of too many different things at once
cradek: which language is that module written in?
cradek: with the rt kernel - would a celeron proccessor run the same as a III of the same frequency - if realtime cannot use cache?
skunkworks: I don't know
03rayh 07HEAD * 10documents/lyx/emc2/Master_User.lyx: Tweak to TOC
just thinking out loud
cradek: hmm, I can read C somewhat (as long as it's relatively commented), so I may be able to do that
Lerneaen_Hydra: after looking at encoder.c, I think the best way to get a non-quadrature mode added to the encoder module would be to pine for it in front of jmk
pine for it? (must be some expression I guess)
like whine for it perhaps?
more like talk about how nice it would be if only there was one
but seriously, you won't find better-commented code than his
I take it that's no major task?
I don't think jmk likes winers :)
I found stepgen.c to be very very well commented
Lerneaen_Hydra: several of us could write this driver; my advice is don't worry about it not being there yet - just go on with your conversion
[15:53:12] <cradek> http://cvs.linuxcnc.org/cgi-bin/cvsweb.cgi/emc2/src/hal/components/encoder.c?rev=1.16
here is the file
cradek: alright, singlepoint threading is somewhat low on my list of things to do, my first and foremost task is to gut it and try to find out which IC's i need to jack into
the brains are in update() which uses the lookup table lut
ok I understand
modifying that file is beyond my cababilities
or rather, modifying it and making the result work
that's ok, don't worry about that
okay, I'll concentrate on the mechanical/electrical stuff
[15:58:29] <cradek> http://dictionary.cambridge.org/define.asp?key=pine*3+0&dict=A
Silly question... why do routers do 30K RPM? (not being funny here)
uhm, for smoother cuts?
* bill203 makes up answers.
so is 30k always the best speed?
AFAIK faster is better
depending on the material of course
wood and aluminum come to mind
wood in this case
who's played with the puppy-emc2 cd??
I booted it once but never did any real work with it
I haven't seen the author in ages. He used to be on irc with the nick 'cncuser'.
* Jymmm thought Ray created puppy.
cradek: where you able to view this? http://www.electronicsam.com/images/KandT/drilling.ASF
It is pretty crappy
And you can hear someone say "whow"
HOLY SHIT!!!! http://owmv.com/index.php?topic=856.0
Talk about having a fucked day.
hah I saw those awhile back
skunkworks: no, it doesn't seem to play for me
I don't remember the last one (second truck falling into the water too)
yeah I don't think the last picture is real
OK I was doubting it a bit too
all the people are gone
notice the last picture has the same bacground as the 5th one up
the guy in the white shirt and the lady next to him amoung other things
Yeha, notice the rig in the water... it hasn't moved at all.
just made fuzzy
holy crap is soooo tru... LOL
But still... I could imagine that happening.
Murphy is a dead man when I catch him!
it seems clear in win amp
ive seen a guy come in drunk.. i know cuz i was drinkin with him.. and lose a 110 ton ore truck over the bank in the road cause he drove it to the bunk ouse to get his sunglasses...
explain THAT one
I have bunch of "Why I got fired today" pics... like cargo ships loosing their entire load into the ocean.
that bites... if they ARE yours...
yeah jymmm loses a lot of jobs
so do I... is there a pattern... ; )
they hacked 5 guys at our place yesterday... 1 with 30 yrs 1 wit 26...
others were expected...
1 just wouldn't shut up... even with me.. and i wear the name of the company...
i am not a spy.... but isnt that what a good spy would say??
dmessier: Not mine =)
we are 5S advocates and keep our tool shops in RE-building fixtures... cause we threw them out and the job is back
just like the CAT
5S, you have that too?
it every where ...its everywhere.... lean to
here in SKF sweden 5S boils down to basically trash, trash, trash, trash, trash. (trash your stuff)
(throw it away)
works with good planners... doesnt work without them
trash it... re print it when you need it...
saves trees... right??
i have 20 blue binders all marked with stuff i switched... LOL im an asshole...
follow mee... i dont think so... ; )
i cant even remember the 5 s's
that whole 5s thing seems to be just standard practices riddled with corporate speak
we are based in France.. and it falls here like a BRICK
i am in Canada
im the worst proponent... i pile things...
but know were it is
[19:10:49] <skunkworks> http://www.electronicsam.com/images/desk.jpg
heh, that would not be 5s certified (neither would my stuff be, and I'm glad about that)
[19:11:50] <skunkworks> http://www.electronicsam.com/images/desk1.jpg
what with the CCTV camera recording yourself?
cradek: the last one is the computer you dialed into to work on the 1gb problem. (the one on the desk with the cover off)
Lerneaen_Hydra: No that was taken off of one of the machines that people monitor from home on the weekend.
jepler: yes - an old term :)
was it using a 300 baud modem?
hayes? or manual-dial acoustic?
No - he came in on the cable modem :) I just still use that term because I am odd.
you typed "odd". Did you mean "old"?
I don't think I am that old - 32.
ancient IRT to my age ;) It's all relative really
not really though
old enough to remember the 300 baud modems
what's up ?
oldest I remember is 5.25" drives
err, 5.25" floppies, not CD/DVD and the like
floppies? How hard I try to forget
Lerneaen_Hydra: I've written, but not tested, a counter hal module that will work with your spindle sensor
cradek: ooh, sounds interesting
cradek: I take it that it was a minor thing to modify the origianl file?
Lerneaen_Hydra: I realized I can use it on my lathe too, and get half the resolution (and therefore twice the speed) by using just one of the phase inputs
cradek: well then you'll have an incentive too, which is a good thing ;)
about how minor it is, I'll have to let you know after it's tested
cradek: because it only turns one way ever?
at least I should be able to test it
cradek: how are you doing the index then - or do you just keep counting?
that idea was rattling around in my head recently too
it has an index input just like the encoder module
skunkworks: Heh, I remember that printer
I got it for free - and have printed about 2 sheets off of it.
makes it so people can't see what is on one of my monitors :)
when the walk into my office
NOW we know what the camera is for =)
honestly - I think that is the only reason it is there :)
cradek: so he will need the encoder pulses plus one index pulse.
who is he?
hmmm, and I can get an input back so I can hook up the home and limit switches properly
cradek: I still haven't gotten to see the lathe in operation
skunkworks: Aren't I special. Yes, that should be correct
besides that video of course
I could do pseudo-knurling - I wonder if anyone does that with cnc lathes and thread cutting tools
In respect to steppers... which thingy in emc allows for the use of encoders?
cradek: will the input be one pin for index (1 or 0 as index) and one for encoder (0 or 1 as hole or no-hole)
cutting matching "threads" in both directions would give a knurl pattern
oh, you mean in borkborkbork langage what's called "lettring"
sorry, my borkborkbork is rusty
[19:27:16] <cradek> http://en.wikipedia.org/wiki/Knurling
yes that's lettring all right
Jymmm: lots of special hardware counts encoder pulses, but in emc2 you can also hook your encoder right to the parport and use the encoder hal module
never thought about knurling that way cradek
skunkworks: I hadn't either, but it seems like it would work - I have a terrible time getting a good knurl on a small lathe like the sherline
cradek: 3axis machine (no encoders yet, us digital?), would I need a 2nd paraport?
Jymmm: you have 5 inputs and lots of outputs, it really matters how many switches etc you have
cradek: I've never mucked with encoders, so I dont' know how many signle are beign returned. vcc, gnd, pulse (I guess).
an encoder with index pulse uses 3 parport inputs, two are quadrature and one is index
Ah, so five wires.
Jymmm: an encoder usually has A and B channels (+2 gnds)
if it has an index pulse, then another wire
are..uh..what are they called, the switches that activate when you've reached the max accepted value in X+/- and Z+/-, really needed (for stepper driven things)
Ok, and when using the encoder hal module... do the steppers follow the encoder or the other way around?
you can use the hal modules any way you want
I mean for error tracking (missed steps)
Jymmm: cradek is using an encoder to read the spindle speed
not as feedback for stepper motion
spindle position, more precisely
cradek: er, yes, sorry
I guess what i'm talking about is "closed loop"
you could use encoders with steppers with emc2 but it's not a typical configuration.
Jymmm: you can do that, but don't use PID
as that will kill/stall the steppers
alex_joni Ok, what would I use instead?
I don't know how well it will work. When steppers fail they don't just lose a step, often they stall and don't move again until you go through an acceleration phase again
you'd use the position from the encoder as feedback to emc, rather than feedback from stepgen
use stepgen and save your money on the encoders
use stepgen + encoders for feedback (might work, not tested)
that's true, interesting
SO that would make it closed loop (but untested)?
well all it possibly gets you is a following error when your stepper stalls
if that's your only goal, it would probably work fine
but no way to correct it huh?
Jymmm: steppers can't be corrected
yes, fix your program or machine so you don't get stalls
once they are stalled, you can't fix them
you probably need to stop, and restart
no, I mean like stepping back a few line of gcode, etc
Jymmm: that's ok from the GUI
but not automatically
alex_joni I'd be happy with that, if I can maintain the reference -vs- lost step positions
Jymmm: then encoders are ok
It's just when I lose position, if I can't reestablish the correct position, then the material is scrap
then you need home switches more than you need encoders
if home switches give high enough precision
a machine that loses position in anything but a catastrophic case is just broken or misadjusted IMO
or being used improperly I guess
Well, shit happens. I'm just trying to work around it.
for me it seems like any event that causes position to be lost destroys the work (and tool?) anyway and is usually a program error
If I do short run jobs, no problems. Get into the 2 or 3 hour jobs, I get issues periodically and I haven't been able to narrow down the problem yet.
that sucks, sounds like a pain
have you run the rtai latency test for a day? maybe your machine has glitches
he doesn't run emc
the other thing to try would be lowering accel and maxvel by 5%
in that case .. you could try emc for a day :-P
* cradek whistles innocently
Yeah, but I was thinking that if I have some way to "regain postion" I could compensate for the stalling and not make more trash.
I have BDI, but I can only get 90IPM out of it. TCNC allows me 180IPM, so that's why I've been using it.
I needed the faster feed rates when workign wiht plastics.
Jymmm, that is a part of the cost of multi tasking.
rayh I understand that, I'm not complaing.
Just trying to come to some viable solution at the stalling issues w/o costing an arm and a leg
I have had better luck with emc2 than turbocnc for max feed rate.
skunkworks what drivers?
skunkworks: No, I mean gecko drivers?
oh - no z has a compumotor drive on it and x and y have burger lahr 5 phase stepping drive/motor
Ah, ok. I'm using Xylotex right now. I might have to buy geckos as they compensate for stalling
emc1 I had issues with - emc2 workes much better as far a pulse train for some reason.
these are industrial stepper drives. I really like the oem650 compumotor drives and you can get them pretty cheap on ebay
they are 6 amps max and do micro stepping
70 if I remember right
very simialar to the geckos.
[19:58:55] <Jymmm> http://cgi.ebay.com/COMPUMOTOR-OEM650-RC-LNC-OEM650RC_W0QQitemZ7596238165QQcategoryZ97184QQssPageNameZWD1VQQrdZ1QQcmdZViewItem
we have gotten them for under 50 if you keep an eye on ebay
Ok, not much up there today
[20:00:45] <skunkworks> http://www.parkermotion.com/manuals/OEM/OEM350-650.pdf
16 jumper selectable motor resolutions (200 - 50,800 steps/rev)
no step multiplier though
geckos have that iirc
we usually run them at 1000 steps per rev
not all geckos
I usually don't mention it because we don't need the ebay compition. but we have quite a few stock piled and really want to move to servos.
skunkworks: Well, come on man... lets here the numbers =)
we probably have about 10 total
not including some other stepper drives that are lower current.
Reading about their resonance tuning guide now...
Jymmm: about step rate -- how many steps/second is 180IPM? Does xylotex give you any choice of microstep resolution?
jepler yes, Xylotex give 8 step
which is what I have it set to
steps/sec iirc is 16000
Jymmm: I have never tried to use it, but recent bdi4emc versions have something called "quickstep" which can use base_period way under 10usec. alex has done bas_period down to about 7usec in emc2.
Jymmm: so you might want to take another look at emc2 or bdi4
10 usec BASE_PERIOD = 50kHz theoretical step rate
jepler: I'll have to see what I have installed on the laptop, not sure which BDI4 version it is.
you have to do extra work to get quickstep enabled
you're using a laptop to run it?
I only had one lap top that worked with tcnc
I would never trust a laptop without extensive running of something like the rtai latency test
all the ohter ones injected pauses in the pulse stream
IBM ThinkPad iSeries 2611-552 366mhz Celron
Jymmm: that might be your screws-up-every-few-hours problem
cradek how so?
the stalling primarily happens on a rapid too I have noticed.
in both X and Y.
you should run the rtai test. I've never used it but I think they even have a bootable cd that will run it
cradek is this "rtai test" something completely seperate from emc?
cradek why did you say the laptop might be the "every few hours" fubar?
because laptops are often unreliable for realtime code
my old one would glitch when the processor fan turned on or off, for instance
maybe your screen blanks after an hour, or your hard drive spins down, or who knows what else
ah, that makes sense.
I have APM disabled
you need to extensively test a laptop before you trust it for pulse generation
but that doesn't mean anythign these days
often that doesn't seem to matter
[20:26:02] <cradek> http://issaris.org/rtai/howto.php
you want test 1, kernelspace latency
and what kind of results should I be looking for?
hmm on later versions the tests show "overruns" which is the important thing
I'm going to try it here and see what it looks like
any aprticular version I shoud ld download? 14 15 16 ?
I guess I got 16
is keystick working in emc2 by chance?
yes in emc2 head
also xemc which is probably almost as fast
What I was thinking is running emc2 on a desktop box, but controlling it from the laptop... not enough room for another monitor
that would work fine with any of the guis
and NOT running X on the desktop box
ok, latency min/avg/max: 2023/1095/6062242
ok, latency min/avg/max: 2023/721/6062242
then you've found your problem
I think the max should be more like 20000 ns
maybe I need to confirm apm/pnp is disabled... let me try.
Ok, changed settings in BIOS, power cycled and running tests now...
latency min/avg/max: 1876/1128/26811
that's more like it
now let it run overnight
latency min/avg/max: 2102/1142/6105945
was it 26000 for a while and then jumped to that?
that machine is unsuitable then
Ok, let me try my real laptop for shit and giggles =)
latency min/avg/max: 1212/2239/211810 (APM & PNP are enabled)
what do those numbers mean?
skunkworks: RTAI test results
Jymmm: I think 211us is also unsuitable, but I don't have a machien to test it on
Jymmm: it would be nice if someone could run that on a known-good machine
cradek: Yeah, that's what I was thinking. I have a test box here I could try it on.
03cradek 07HEAD * 10emc2/src/ (Makefile Makefile.inc.in):
new hal module that's like encoder but without quadrature. for lathe
spindles that have a simple pulse train for rotation and a reset pulse.
it should also work with just one phase of an encoder. this is completely
03cradek 07HEAD * 10emc2/src/hal/components/counter.c:
new hal module that's like encoder but without quadrature. for lathe
spindles that have a simple pulse train for rotation and a reset pulse.
it should also work with just one phase of an encoder. this is completely
cradek: What mobo cpu video are you using?
good night all
skunkworks: Ok, you got some pricing on these drives yet?
03rayh 07HEAD * 10documents/lyx/emc2/ini_config.lyx: first try at user level ini explanation.
Jymmm: I just tried it on my known-good laptop and get a consistent 16xxx max
I think anything over 20xxx is abnormal
Jymmm: I've run on a variety of systems
cradek I'm running it on my p3 800 now, getting about 20287 for max
jepler's known good machine gives 9487 max
an hp a couple years old I think
I won't muck with my good know laptop, it's still virgin =)
by known good, I mean known good for rtai
are you using a desktop box too?
my mill is on a desktop
what mobo cpu?
it's an oddball server class machine with PIII-666
video/network as cards or on board?
I had to disable the onboard video on it and use a pci card
strange... I'm suspecting chipsets now
what do you mean? your p3 is fine
cradek: Right but why a P3 666 is getting 9487, and a P3 833 is getting 20000 seems a little strange
the 9487 isn't a p3
20000 is fine, don't worry about it
your other machines were 600000, worry
This is an intel mobo, I can't disab;e onboard video, just specify which is primary AGP/PCI
if you're getting 20xxx it's fine, don't mess with it
but do run it for at least a few hours to be sure
03rayh 07HEAD * 10documents/directory.map: Added wiki and emc2 layout
yeah, I'll let it run over night. I'm sick and tired of this stalling-after-a-few-hours crap.
any of those 600000 glitches will stall steppers if they're moving fast
I'm surprised you can't hear the glitches
But the funny thing is I can manually jog... no problems at all
have you ever jogged for hours?
seems like it some times =)
well good luck finding working hardware, you're on the right track now
Gawd, I hope so...
you've been using that laptop all along? the one that gave 600000?
SWP_Away is now known as SWPadnos
that glitch is 12 pulses at 10kHz
it's no wonder it didn't work
and ubuntu is a dawg on this box (the P3 833)
it runs perfectly fine on my 666
has anyone noticed if ubuntu is significantly slower than BDI 4.30?
would be surprising - they're very similar systems
yeah - that's why I was wondering
they both take forever to install but then run fine
my celeron 500 with BDI 4.30 seems fast enough, though not "snappy"
On both, I immediately ditch the desktop in favor of icewm
I scored a new PIII 667 for it though, so we'll see what happens ;)
and the terminal in favor of urxvt
it really makes a big difference in interactive "feel"
yeah, gnome-terminal is terrible
I'm interested in seeing the difference in Dapper, but I haven't finished an install yet
on old bdi (4.2x?) "konsole" was terrible too
(I have decided that Gentoo isn't worth the trouble though)
eterm is nice
I tried gentoo once and was not impressed.
I use konsole when I log in remotely with CygWin
it would be good with a 37 GHz machine and 100mbit internet access
itherwise, it's not worth it
cradek: with freebsd don't you also compile most of your own software?
37 GHz quad processor?
jepler: not usually, unless you want the very newest versions
37 GHz 8 x dual core processors, with 1G on-chip cacche, and 128G RAM per CPU
jepler: with ports you can choose to compile or install binaries
you could then install e.g. firefox as fast as a binary download
You got a waterfall in the backyard to cool this?
and the sun to power it ;)
got an epa permit to boil water in the lake.
SWPadnos I'll take two please
that'll be $3 gazillion, in small bills
* Jymmm ebays for 10,000 monoply games
<SWPadnos> you could then install e.g. firefox as fast as a binary
about gentoo ?
gentoo compiles everything when you install it
it took close to an hour to compile firefox (on an Athlon 2200), after downloading a source tar that was already 2x the size of the binary
how much packages gentoo have ?
I thought about using Gentoo once ...
then got smart, huh
for a server maybe is nice
I'm not sure for a workstation
I decided to put my fingers in the office drawer and slam it a few times instead
it was cool for a while, but since CPUs are so fast these days, there's not really any reason to do machine-specific optimization for most installations
overall, less painful
this is a machine that's been around since the middle 2.4 days, and was used as a testbed during the kernel 2.5 cycle
it makes like, 5% difference
it would be interesting if there was a way to do an internet-wide "ccache" which somehow proves cryptographically that the object file you get is right
03cradek 07HEAD * 10emc2/src/module_helper/module_helper.c: new hal module `counter'
it's been at least a year since I've upgraded the software, and there's just a bit too much to do now, so I keep getting emerge errors
it's not worth it, when I have a shiny new ubuntu CD sitting here
the machine is off - I got tired of the fan noise ;)
and evolve into a debian user :)
I like the ubuntu brown - I'll stick with that for a while
if only I could reliably burn CCDs with this stupid Windows machine
i forgot what the difference between ubuntu and debian is ...
that's the difference
lots of #ubuntu folks end up on #debian asking questions about why <balh> doesnt work .. b ut it does work in debian, the ubuntu people broke it .. so we send them away
leader of Ubuntu project is an old Debian developer
if I recall well
robin_sz: yeah I got a pretty frosty reception once when I was trying to get help with build-kpkg or module-assistant or some such thing, when I let it slip that I was actually using ubuntu
it made me want to give a big "fuck you" to the #debian channel, but I managed to choke back the impulse.
make-kpkg and m-a are debian warts, but I'm not surprised they don't want to take credit for them.
well, they dont keep track of what the ubuntu people have done, and typiclally it works fine in debian, but not on ubuntu, so it not their problem
but when you say "I'm trying to find out more about setting up my module to build with m-a and by the way I'm using ubuntu" I still think they could point me at the documentation (if, in fact, it exists)
and sure include a caveat "we don't know what those central-african dogfondlers have done to it since the Debian developers shat it out as a perfect diamond"
as well Debian user calls theyr OS Debian GNU/Linux
were Ubuntu not
why noit just ask on #ubuntu in the first place?
that's the relevant difference
I tried on #ubuntu but got silence
it's all "I can't install ndiswrapper, waaaah"
that's what they get for having users
its a bit like buying your car from kmart and then taking it to yor local dealer to ask him to explain stuff
back when we still got bdi users on this channel I always tried really hard to be helpful to them
I wish the people on #debian would have done the same
I know I can't demand it
justg pretend you have sarge then :)
demand satisfaction, or your money back!
staff. grrr, pain in the ass/
ok bedtime for me .. exhausted
see you robin_sz
* robin_sz waves
03cradek 07HEAD * 10emc2/src/hal/components/counter.c: fix some documentation and error return codes