[Global Notice] Hi all, It appears we were visited by the ghost of a not entirely uknown bug. Affected users: approximately all of you. All I want for christmas is a new ircd. Sorry for the inconvenience and thank you for using freenode.
score 1 for kenneth
hmm.. not sure that helps
even if we all feel that way
one of those - ignore it and it will go away
nptd isn't running it seems
gonna pull the cable
wrong channel :)
but it works in here too
hmm.. maybe you're getting ntp broadcast packages
even if ntp is not running
"A broadcast server sends out a packet about every 64 seconds."
well, it still had a 300uS latency after I pulled the wire
do you have any USB stuff plugged in?
I could disable the mobo usb controller in the bios
"Periodic SMI rates available on many systems are every 64 seconds (approximately one minute), 32 seconds, 16 seconds, 8 seconds, 64 milliseconds, 32 milliseconds, 16 milliseconds, 8 milliseconds, and one millisecond. These periodic SMI rates are provided as an example, and some hardware embodiments may provide a much wider range of periodic occurrences that may be used by the USB support software and the BIOS."
(already disabled the sound in the bios, and power management is off
"The SMI is initially set by the USB support software to be generated at a slow rate, such as, for example, every minute or every 30 seconds, so as not to unnecessarily impede performance of the system during the boot process by causing unneeded processing of SMIs. The USB support software then starts the USB host controller such that the USB host controller processes USB device descriptors using the obtained portion of memory, the USB
hmm, I've heard SMI in connection with RTAI latency problems
lemme reboot and turn off USB in bios
was afraid you're gonna ask what SMI is
system management interrupt
ok, just read that too
[09:02:58] <jmkasunich> http://www.captain.at/rtai-smi-high-latency.php
any wireless on that pc?
ok, the SMI sounds like a good guess
arg - disableing usb didn't fix it
(I didn't expect it to, after reading some webpages about SMI)
seems you need to disable it at kernel build time
hmm.. it seems you can make that rtai module they talk about
jmkasunich: RTAI comes with a kernel module, which disables all or certain SMI interrupt. You have to set the SMI options before you compile RTAI, or compile the SMI disabling module individually as I did.
that means I need to get all the rtai source, right?
apt-get source rtai ?
whatever I named the package, right
* alex_joni is getting old :D
dpkg -l | grep rtai
I think rtai_3.3
[09:19:34] <alex_joni> http://dsplabs.cs.utt.ro/emc2/dists/dapper/emc2/source/rtai_3.3-1.tar.gz
rebooting again, then I'll get it
getting the source nwo
btw, this is the pc I want to use as my machine controller
ok, got sources
I don't have a gcc yet
which version did you use for the kernel
alex_joni: still there?
was a bit busy
4.0 I think
but apt-get build-dep emc2 should pull it all in
I sure hope I can get this to work
800MHz P3 should be pretty decent
wonder if you can't make it work with the existing smi-module
matrox millennium in AGP slot, 4 PIC (one with a 3-com NIC), and 2 ISA
there's an insmod parameter "smiReset"
I would try 'insmod smi-module smiReset=1'
then rmmod smi-module
the unloading should do something to smi (rthal_smi_restore)
where is smi-module? I only see rtai-smi.ko
the smi-module is the source
it seems the version I compiled already has the rthal_smi_disable() in it
I'm gonna let the apt-get build-dep finish ( a minute or so)
then gry the smi module
gonna need the build deps anyway ;-)
it seems the captain.at page adds smiReset=1 by default
which you can do at insmod time
but they also change this define:
#define CONFIG_RTAI_HW_SMI_ALL1 // <<<=== IMPORTANTE
in the rtai 3.3 I packaged it's set to 0
I got this in dmesg just now:
[ 889.710176] RTAI: Intel chipset found, enabling SMI workaround.
[ 889.710197] SMI module loaded
when I loaded the module
running the test now
maybe (hopefully) it's any better now
did you pass smiReset=1 ?
try unloading the module, see how it behaves?
I didn't rmmod the smi module yet
still no good
ok, then the last hope is the #define in the module
where is the rtai source installed?
in the dir where you did the apt-get source
ah, my home dir
* jmk-st reads README.SMI
there's a setsmi utility it seems
huh, one guy seems to think that an ACPI enabled kernel helps with this
you did turn off all APM from the BIOS.. right?
ok, I need to edit #define CONFIG_RTAI_HW_SMI_ALL0 ?
change it to 1?
I would try that
* alex_joni curses at the rtai API
[09:47:09] <alex_joni> https://www.rtai.org/documentation/magma/html/api/
* jmk-st curses at apt-get source - all the files are owned by root
it's from 18 Jan 2005
jmk-st: you still need root to install the module, so don't bother changing the owner
just use su -
err.. sudo -s
but I need to restart my editor with sudo
ok, where do I invoke make? the rtai dir?
oh, theres a makefile in each dir
just run it in the dir where I changed the source I think
that's a bogs makefile.. unless I didn't clean up the package
it's KBUILD.. so not that easy
get the Makefile from captain.at
that one should work without too much hassle
[09:51:06] <alex_joni> http://www.captain.at/rtai-smi-high-latency.php
the Makefile you got there assumes you have the sources where I had them
but that was because of the way kpkg builds additional modules for the kernel
I can tell that I'm tired
because I keep falling behind what you are saying
I'm seeing the changes to the file on the captain at page
lemme make those first
* alex_joni slows down
looks like one of the changes was already there
(the call to rthal_smi_disable() right before the return
that's what I said a while ago :)
see, I said I was slow
you only need to change the #define
the smiReset you can pass at insmod time
(I also changed the default so I don't have to specity smiReset at instmod time
ok, changed and saved
now makefile stuff
mv the Makefile out of the way
the one in i386/calibration?
and make a new one with the contents from http://www.captain.at/rtai-smi-high-latency.php
ok, the one thats there now is a symlink
to a place that doesn't exist
lrwxrwxrwx 1 root root 68 2006-12-22 04:25 Makefile -> /usr/src/modules/rtai-3.3/base/arch/i386/calibration/Makefile.kbuild
-rw-r--r-- 1 root root 210 2005-12-05 12:33 Makefile.kbuild
yeah, that's where I had the sources
remember apt-get source gets you the exact tree I was working on when I packaged the stuff
mv Makefile Makefile.old
the download is done, would it make sense for me to go to the right place and apt-get source again?
try the Makefile from that page first (way easier if it works)
ok, I have a makefile now
just run make? or do I have to pass it args?
make is enough
it calls "make ... modules" by itself
make -C /lib/modules/2.6.15-magma/build -I/usr/include/asm/mach-default/ SUBDIRS=/home/jmkasunich/rtai-3.3/base/arch/i386/calibration modules
make: Entering directory `/usr/src/linux-headers-2.6.15-magma'
make: *** No rule to make target `/home/jmkasunich/rtai-3.3/base/arch/i386/calibration/smi.c', needed by `/home/jmkasunich/rtai-3.3/base/arch/i386/calibration/smi.o'. Stop.
ok, I see the problem
edit the Makefile
obj-m := rtai_smi.o
instead of just smi.o ?
rtai_smi-onjs := smi-module.c
rtai_smi-objs := smi-module.c
rtai_smi-objs := smi-module.c gets added right after the previous line?
make: Entering directory `/usr/src/linux-headers-2.6.15-magma'
LD [M] /home/jmkasunich/rtai-3.3/base/arch/i386/calibration/rtai_smi.o
/home/jmkasunich/rtai-3.3/base/arch/i386/calibration/smi-module.c: file not recognized: File format not recognized
rtai_smi-objs := smi-module.o
so now I have a nice shiny new module, in my home directory tree
gotta install it?
mv it to /lib/modules/ etc
root can do that ;)
save the old one though
now you tell me
oh well, we didn't need that old thing anyway
insmod the new module, and try the latency test again?
I think it worked!
80 seconds so far
now go to bed
in a minute
I know I'm sleepy enough :D
didn't you just wake up?
2 hours ago
but I'm dead tired .. not sure why
Just this once .. here's the log: http://www.linuxcnc.org/irc/irc.freenode.net:6667/emcdevel/2006-12-22.txt
jmkasunich: is it on rtai that rtapi_get_time() is real slow?
I guess I will ;)
steves_logging is now known as steve_stallings
hi alex, I really like the nice drawing for Santa with full compliance to regulations, printed copies for some of my friends who call themselves "engineers" but are really paper pushers 8-)
thought everyone in here would like that
SWP - any need for me to settle up on the group buy today? headed out for several days holiday trip this afternoon
Not immediately. I still need to finalize the shipping costs and such, so I don't have a good total for you yet
it looks like they want to do one charge, but they may drop-ship, so thet'll make things even more interesting :)
we could pay you the board costs now and the shipping cost in meal(s) at that mexican restaurant next fest
alex_joni: intense rtai stuff lastnight/this morning. is that going to be worked into main kernel? (don't really understand but look like you guys where able to disable another latency causer)
I still don't know if I'll be re-shipping or they'll drop-ship. there are around 12 addresses, which may be too many for them to drop-ship to
skunkworks: that one was mobo specific
I don't want to put that into the generic packages
ok - makes sense
because I have not much clue what it does on various boards
steve_stallings is now known as steves_logging
make_shortcut is defined and used in pickconfig.tcl -- hard to see how it could be undefined
yeah, he is pasting the entire error message in a /msg to me
you are either a saint or a fool for trying to help him
I have not acknowledged that I'm even getting the /msg (and I won't)
maybe he used make install and it's broken
oh, defined and used in the same file, hmm
I told him to put it in pastebin, and until he does I'm not doing anything
I want to see all the output, not just the error message
so I can see what version he built, etc
flo-h has reported some problems with "make install" (not this problem) and I think they're now all fixed
I should try it sometime
alex_joni helped me fix my latency problem last night
I saw that
I didn't try to understand the details
its a known problem that apparently affects a lot of "newer" Intel systems
newer in quotes because this isn't that new
[16:55:00] <jmkasunich> http://www.captain.at/rtai-smi-high-latency.php
fsck - if he can't figure out pastebin, he's not gonna be able to figure out emc
how can you survive the endless "help me" posts?
* Lerneaen_Hydra is well aware that he has done the same thing himself
Lerneaen_Hydra has _not_ done the same thing
ok, not the same thing, but still asked for information of which was probably on the wiki
same here. ;)
I don't know how much of the problem is simply language problems, but unless this guy can communicate better we can't help him
but then if one of us gives a URL to the wiki, can you figure out what to do with it?
too bad someone doesn't know portugues on here
ah maybe it would occur if you don't have a ~/Desktop directory
wow, that's obscure
cradek: if I get the link today, yes, probably, if I got it the first time I ran emc, probably not
what is make_shortcut doing anyway? making an icon to start emc?
IOW, he doesn't even need that part
who gets to explain "cvs up" to him?
(I bet alex has been helping him a lot if it wasn't any of us)
please work, please work.....
why the f--- wouldn't he have a Desktop dir?
damnit where's my coffee mug?
jepler_ is now known as jepler
I wonder if he actually got the fix?
if he checked out the branch tag, instead of the branch itself, he wouldn't get any updates I don't think
there's no fscking way we can teach this guy cvs
there's only a branch tag - there are no version tags on that branch
cat: cvs/*: No such file or directory
grep Revision tcl/bin/pickconfig.tcl
# $Revision: 184.108.40.206 $
you could try going this route to find out the version he has
normally I hate Revision tags but maybe today it will be useful
but I don't want to overload him, lets see what his CVS says
cvs stat works just as well
oh hm -- I didn't know that command
yeah I use it all the time
usefull, but a crapload of output for guest to parse
darnit wrong channel
he shouldn't have to rebuild for that change should he?
no but if you want to encourage "build every time" rather than teaching a complicated rule...
weird - OOcalc has no "binary" display format (that I see)
oocalc isn't for people like you
ah - there is a function called BASE taht can display a number in any base
I'm trying to figure out why Peter likes the bit-reversal method for doing interleaved PWM
cause bit reversal is trivial in hardware?
that could be
I'm trying to prove to myself that it works
digit reversal isn't a common mathematical operation
I don't see how it would work
it will I bet
consider 4 bit values. 0001 is less than 14 other values. 1000 is only less than 7 other values.
normal PWM - inthe time it takes the counter to run a full cycle, N of its values will be less than the target value N, and M-N will be more (M is the modulus of the counter)
bit reversal won't change that
it will only break up those N and M-N into smaller chunks instead of having them all adjacent
perhaps I didn't understand the method then
do I keep the PWM value of 0001 and compare it to successive counter values -- but the counter goes 0000 1000 0100 1100 ... 0111 1111 ?
the counter itself doesn't get reversed until the comparison, I think
counter and comparator are both the same for both cases
ok then I misunderstood
so the counter still goes 0000 0001 0010 0011 ...
you just bit-reverse the counter value between the counter and the compare
he probably has a single master counter
but the comparisons are 0000 1000 0010 1100 ... (reversed every second time, I think
the bit-reversed counter is just a permutation of the 15 values
then a block that either bit reverses or not
yes, there's a reference counter
the output of the "optional reverse" block goes to the comparator
strangely, the Mesa FPGA only reverses the top 8 bits, and the lower 3 or 4 are left alone
maybe to keep the output frequency from being too high
maybe he doesn't want the output frequency to get too extreme
what you said ;-)
that could be ;)
* jmkasunich hides from the happy Brazilian
jmkasunich: is thank you enough or do I need to bless you too?
ah yes - 2 posts down from the one jepler linked Peter notes that he can control the output frequency using bit reversal, but not with the overflow method
err - 3 posts down
that's one happy (and possibly quite religious) Brazilian
that bit reversal thing is very clever
have to remember that one
it's in the pwmgen.vhd file under hal/drivers/m5i20 :)
sucks for software implementations tho, bit reversal isn't pretty on a general purpose CPU
it also means he needs a magnitude comparator instead of a simple equality comparator for each channel
it seems fragile though - if you reverse on the "odd" counts, then you'll always have a value above 128 (which may be desirable)
for a long word that can get either complex, or slow
well, consider the 3-bit sequence 000 001 010 011 100 101 110 111
you always reverse for PDM, and never reverse for PWM
ah, ok. now I get it. thanks
you put the high frequency bit in the highest value position, which gives you a high frequency change
duh. that is clever
I was stuck on the vhdl code, which looked like it was only doing the reversal some of the time
I wonder what the FFT spectrum of a bit reversed counter is
[18:10:05] <jepler> http://emergent.unpy.net/files/sandbox/br.py
I bet its nearly ideal for running thru a lowpass filter
we were discussing ripple earlier - dunno if you have any corrections for my ramblings :)
from a signal processing point of view, both PWM and PDM can be considered one bit quantization of the carrier signal
for PWM the carrier is a triangle
correction, sawtooth probably
if its a simple counter
* alex_joni is back home
for PDM its that bit reversed signal
some PWMs are triangle -- AVRs have such a mode
yeah, we always use triangle for drives
sorry about guestXXX, shouldn't have told him how to get HEAD
fwiw that bug was in 2.1 and HEAD
and now its not!
I told you it was alex's fault
[18:14:13] <alex_joni> http://pastebin.ca/288899
cradek: I got my share
that's only a part of it :)
when he /msg'ed me I didn't answer
jmkasunich: same here, because I was away
its easier to just fade away on the public channel
but /wc works great :D
alex_joni: thank you _very_ much for the help last night!
jmkasunich: glad you got it working
the only thing we should do is add that piece of info into the wiki
I beat on it some last night (after the netsplit) and more today
no latency more than 15-18uS
did things like "dd if=/dev/hda1 of=/dev/null" and scp'ing an iso from my other box to this one at the same time
so disk reads, disk writes, and network activity all at once
plus much dragging of windows and such
jmk-st: run a few glxgears too.
how do I do that>?
glxgears & glxgears & glxgears &
I see gears, they don't turn fast
that kills latency on my machine
drag them around, maximize one
while (1) ; do glxgears& ; done
this is a matrox card, probably no hardware accel
the gears are updating at 1-2 frames /sec
but the latency is fine
jmk-st: I can plot the fft of the waveform I get, but I don't have a clue how to interpret it...
jepler: put it online
[18:19:48] <jepler> http://emergent.unpy.net/files/sandbox/fft.png
this is fft of what?
I am using numarray.fft.fft([bit sequence generated by the algorithm]) and then printing the magnitude of the resulting values
I feed that to gnuplot
I'm not familiar with raw fft output
it's an 13/256 duty cycle
is that a complex FFT or real?
usually I'm looking at something from spice or whatever, and they neatly convert it and give a frequency axis
what does jmk-st stand for?
it's a complex fft but I write the magnitude of the values
you are looking at the one bit output (the PDM) or the bit reversed count value?
it's "normalized" to the number of points you use
st = shoptask
I'm looking at the one-bit output
this is the computer I hope to use for my machine
right, so you get only 128 on the horizontal azis, because you can only get to half the sampling frequency
jepler: I was talking about the reversed count value
jmk-st: oh -- I can plot that pretty easily
[18:23:33] <jepler> http://emergent.unpy.net/files/sandbox/fft2.png
ah, shoptask. IRC while milling?
keyword "hope" to use
lunchtime here .. bbl
I'm working on the conversion
the computer is pieces on my bench right now
I went digging thru my boxes last night to see what I could build
[18:25:26] <jmkasunich> http://jmkasunich.dyndns.org/shoptask/pc-bench.jpg
FCC approved for emission of electrical noise!
page cannot be displayed
not bad - 20G drive, looks like a G200 video card ...
is it just me?
must be ;)
nice small board you got there
regular size mobo
looks a bit short - normal width though
like 9x13 instead of 11x13 (or whatever)
8.5x12 I think
I'm about to call Mesa - can you email me the spreadsheet as it stands?
more stuff to go in the box; http://jmkasunich.dyndns.org/shoptask/geckos-sink-ps.jpg
jmkasunich: did you add my latest boards?
I have them if he doesn't
jmkasunich: nice paper
alex_joni: no, tell me again
(1ea 7i31 7i33, 7i37)
1x7i31 1x7i33 1x7i37
I guess that's not too unreasonable if the stepper and some other part (leadscrew maybe) become resonant
so it swings back and forth
you mean the one that jepler found?
geckos? I think that is cheating ;) step and direction - whatever
the one jepler posted, yeah, silly me
hmmm - you may want to have the 3 geckos facing the same way, and stagger them so you can adjust the trimpots withoug removing them (though that would interfere with a fourth driver)
the machine has mounts for NEMA34 steppers, I'm not gonna make things harder for myself
oh - stepperzs. nevermind ;)
SWPadnos: I had those thoughts yesterday
I want to keep the 4th drive option
I'm pondering whether I should sell my G340s and either (a) use real analog drives or (b) wait for he CPLD version
there is one pot per drive for low speed smoothness, but I determined that I can adjust it ok
you can always make a right-angle screwdriver ;)
alex_joni: your boards are added to the list now
how much are geckos each?
Lerneaen_Hydra: 120 something
$114 for the non-pulse-multiplying ones, $144 (I think) for the ones with the PLL
one stepper per gecko?
and I think $150+ for the ones with integral caps
134$ 1 piece for the G202
pulse multiplication sounds sort of stupid
Lerneaen_Hydra: not at all
unless you can soft-toggle it on/off
it is if you have hardware step generation, but not if you're using software only
can you soft-toggle it?
skunkworks: the G203V is 147$
Lerneaen_Hydra: no, but you output 10 times less pulses for the same speed
it's a jumper-set multiplier, x1, x2, x5, x10
say you have a gecko with 10-microsteps
so emc outputs 1 step the motor moves 10?
and the same smoothness, since it's still a microstepping drive internally
if you use the 10x multiplier, you command one step and the drive performs 10 microsteps
just call it sinosodial voltage instead :p
so you end up with the step resolution, but microstep smoothness
skunkworks: available from today (the G203V)
He called them vampires because they can't be killed ;)
the G203 is the drive to buy, if you want something that can't be killed
back in a bit
jmkasunich: what's the thing in the far upper-right of that photo -- it has some kind of weird coiled connector on it
er, weird coiled wiring
vampire BNC network connectors are teh shit
[18:52:33] <Lerneaen_Hydra> http://en.wikipedia.org/wiki/Vampire_tap
jepler: the pc photo?
jmkasunich: yes, I was joking at seeing a corded phone
I'm a cordophile
I hate cellphones and cordless phones, gimme an old fashioned handset anyday
* Lerneaen_Hydra nods
its funny, that phone is a cordless base station, we used to have a cordless handset for it that we kept in the bedroom because ther was no phone jack there
the handset died
the base station works as a corded phone, even without power
I recently cut off the power cord and removed the antenna because they were a nuicance ;-)
damn - I just found out that kelinginc.net is closed until the 28th
I wanted to order my steppers today
I've heard advice to buy one corded phone, so that when your electricity goes out you can still call 911
we have two
what is the standard for phones in the US? 70-ish VAC and signal?
48vdc loop voltage, 50-100V ac at something like 20Hz for ringing
ah, sounds similar
ok. this time I really am going away for a bit
not a chance
you'll never leave
damn, foiled again
hmm.. I remember how it feels when the phone rings
was installing a phone socket for someone, and not having what to do with the wires, I put them in my mouth
let me tell you, that's not what you want to do when it rings
that's painful. the ring can be up to 200V
100V I think
at least 100V, no more than 200V (or something of that sort)
I almost could feel the CLIP
jepler - I assume it's OK to ship your stuff to cradek?
hmm, how much current can you pull out of the telephone cables?
SWPadnos: he's already said that's fine
Lerneaen_Hydra: enough, but it's not legal
alex_joni: who's going to know ;)
it is legal for phones though
voltage will probably sag at 100W though
that's nearly an amp
10W or less I think
through those tiny wires
2A when the phone isn't ringing
oh, not much
that's at 48V
2 A @ 48V?
no - I was pointing out that 100W is ~2A at 48V
not saying that you can get that much
hmm, now glxinfo says direct rendering yes
I wonder what made it change its mind?
the biest I tell 'ya
jmk-st: probably you didn't have enough video memory for a depth buffer before
I'm guessing you probably have a 4-8 meg card
8 I believe
how do I tell it no direct rendering?
try it with first :-)
you want it if it doesn't kill latency
gears spinning nicely
latency 14uS so far
put a screwdriver between the gears
see if that kills latency
it doesn't like running two of them
and only one runs at a time
same with my matrox
* skunkworks thinks alex is having too much christmas spirit
if it works for one window, AXIS will be fine and you should leave it
one set of gears spin for a couple seconds, then stops and the other starts
mine has been that way forever
thanks for the help
* jmk-st is not an X pert
the crash was (I hope) a one time thing
you'll find out...
lol.. that's a vote of confidence
just hope your rambus ram isn't bad
I've restarted glxgears about 10 times with no problem
I have ~2 Gig of that stuff
it might have been something in the software rendering code that isn't being used now
alex_joni: that stuff is very costly
cradek: I got 2 workstations each with 1G in it
pain in the ass to find now also. we still have 2 motherboards with rambus
dual Xeon @ 1GHz
drat - I already removed the cdrom drive, otherwise I'd pop in the CD and run memtest86 for a few hours
for about 200/piece
jmk-st: it's on your boot menu
I was confusing that with the install CD validator test
ok, gonna do that, bye
does anyone else have windows xp as one of their boot options?
skunkworks: what's that?
skunkworks: I have about 14 kernels and one win2k
I was just happy to get my duel boot working.
this is : dapper, dapper, bastard, 2k
when you say that I always picture the windows and linux fighting to see who gets to boot
bastard is something between breezy and dapper
I think cradek wants to duel ;)
have xp here
skunkworks: dual-booting Dapper and XP here, but no realtime.
I have realtime - but it doesn't play well with my portable.
this is a laptop as well.
I should say - it runs but I get overuns.
I started with the live CD, but the laptop locks up during boot. I question the value of putting realtime on the laptop since there's no pport.
there are parports though for cardbus
jtr: you can run emc2 without realtime if that's what you're interested in.
the cardbus parports should work like an PCI board
so it could (in theory) drive something
every 60 seconds or so I get 5 overruns
it's 64 seconds :D
well, hadn't thought about a cardbus port. Was thinking to get familiar with HAL, etc on this unit.
oh - you want to help me do a new kernel? :)
skunkworks: not a new kernel
you only need a new module
which jmk can probably email to you
that would be neat - my computer has a printer port.
I can send you a revised rtai_smi.ko
I would not mind trying it
how hard is it to install?
sudo cp ...
sudo cp ;-)
well, at the moment the system with that module is running memtest
that would be cool if it works
thats ok - when ever you get a chance
skunkworks: you can easily do it yourself too.. as an excercise
you have my email address
get the rtai sources, set them up, change a file, make a new Makefile, compile, copy to proper location
I was following you two last night - that was scary.
lots of knowing where stuff was - I still have a hard time editing xorg.conf :)
* alex_joni never thought he can be scary :D
I'm writing up what we found last night on my blog, if you think its wise I can post the modified module there, and link to it
jmkasunich: I'll put it on the wiki
and then there is the damn case sensitive linux :)
didn't see that in the distro line
all damn small linux
only damn small linux
but I am learning. With lots of help :) (thanks again everyone)
was that a recent find? being able to disable the smi?
yeah, from this morning
you mean at RTAI?
I think they know about SMI for about 3-4 years
but what works on some boards, doesn't work on others
[22:13:41] <jmkasunich> http://www.captain.at/rtai-smi-high-latency.php
that documents the issue
can I tell if this thing has an intel chipset?
says a isn't a valid option
alex_joni: your link tries to open in real player ;)
xp is fun
maybe your XP doesn't know how to read text
and it needs real player for that
that would be cool if that made my portable usable for realtime. icing on the cake
I should try a more recent emc2 on my laptop
same result here (another xp box); rebooting...
my video memory isn't shared.
got it.. THere is no settings in the bios for powersaving
ok, it looks like Mesa will ship to me, and I'll break it up for everyone
they have everything in stock, which is nice
more addresses than they expected I guess
actually, I think it's because someone I hadn't talked to before got the email
were they happy about the order?
I think so. $4k is a reasonable sum
SWPadnos: wow that will be a lot of work for you
SWPadnos: put mine with the lowest priority ..
eh - not too bad. I've mentioned to my wife that she shouldn't be throwing any boxes away ;)
I can wait 1-2 months :D
I'll call to see what the inbound shipping charges will be, then check with UPS/FedEx for outbound. I can get totals to people next week (they won't ship until after Christmas anyway)
wow - thanks again SWPadnos
no problem - now all we need to find is a motor supplier that'll give us the same deal ;)
and to talk to mariss
I think I'm more interested in "analog" servo drives these days. I suspect you get a better finish
to get him to do an servo driver :D
although I asked about the G340
and you can turn it into pure analog
SWPadnos: I know someone who will sell you a hydraulic K&T mill
he's got reasonable discounts also
oooh - uh, yay?
or at least some "analog" parts from it :-)
it's really annoying that I have (a) no space and (b) no forklift
"analog" == "fluidy-ish" ?
sure - a great deal.
not too leakey
we actully have 4 hydraulic servos if you need ;)
the only hydraulic thing I own (or really want to own, at this point) is a reasonably nice Greenlee chassis punch
I have hydraulic brakes in a couple cars
analog but electronically controllable, how's that"? ;)
I was looking at a hydraulic press for clamping connectors
I'd hate to replace those with digital
cradek: I hope on all your cars
I have a friend that drives that way
like UP/DOWN output from stepgen
no coasting allowed
* alex_joni wouldn't trust stepgen into his car
jmkasunich: no offence :D
not with solenoids on the accelerator and brake, thank you
actually it's the PC I'm concearned about
* jtr is back
bbl - out to dinner.
ok - time to run off to dinner with my wife. I should have a shipping estimate when I get back (inbound, anyway)
alex_joni: no offense taken
theres lots of things I wouldn't trust to software of any kind
it would suck to have a big red mushroom on the car
for situations when it goes berzerk
hey jmk - did you get the email about VMWare 6.0 beta?
two features you'll like: 1) Automation APIs (VIX API 2.0) - You can write scripts and programs to automate VM testing.
you guys think there's no software in your car? ha.
2) Headless mode - You can run VMs in the background without the Workstation UI
cradek: I know there is
cradek: LOTS of it
cradek: there is no software between my foot and the brake pedal, and between my hands and the front tires
thats what counts!
with pid loops everywhere :-)
unless you're in a BMW 750
jmkasunich: that might not always be true
jmkasunich: ABS (a software system) can disable your brakes.
also I have ESP and it brakes wheels by itself
well at least I can still steer myself
the BMW 7 series is drive-by-wire, similar to a Boeing 777
they haven't messed with the steering quite yet.
not in the BMW
you won't be catching me driving a BMW anyway
they made it entirely drive-by-wire, so they could seal off the cabin from noise ...
not my class of vehicle
you'd be driving too fast ;)
yeah, I don't think they make a pickup ;)
ok, now we're really leaving. back in a while
another magic smoke picture: <A HREF="/pics/db-unit-transistor.jpg"><IMG SRC="/pics/db-unit-transistor-thumb.jpg" HEIGHT="200" WIDTH="200"></A>
[22:55:35] <jmkasunich> http://jmkasunich.dyndns.org/cgi-bin/blosxom/db-unit.html
I think that's a bit more than just the magic smoke
magic dust and chunks too?
yeah the escaping magic smoke took a few parts along with it
magic smoke and magic smoke container
I'm hoping the main power transistor is OK
I'm almost sure the resistors are OK
yay, the transistor is OK
13mA base current, gives 5A collector current
that's a very high gain
looking for ratings now
hFE 75/100 minimum
600V should be great for the spindle
datasheetcatalog.com doesn't want people passing URLs I see
[23:09:43] <alex_joni> http://www.ortodoxism.ro/datasheets/MitsubishiElectricCorporation/mXvrury.pdf
there is a diode of about the same rating\
hmm.. is it only half a year since dapper came out?
feels like _way_ longer
jepler: yeah, I know :)
two resistors of about 100 ohms each, a couple hundred watts each
a 5K resistor probably 50W or so
we have 15 ohms on our servos I think
and the PC board with the smoked SCR and a few other paryts
that's one 15 ohm resistor for 6 axes
it's connected to the PS
300V on the older ones
this is probably designed to trip at about 375V
200V DC bus for internal axes (smaller)
300V DC bus for external
at 375V, each resistor will dissipate 1400 watts
external meaning positioners and manipulators
for a while anyways...
thats why they have thermal switches attached
heh.. yeah.. don't touch it :D
15 ohm in series with the servos?
skunkworks: no, this is connected to the PSU
when there's an overvoltage it gets shorted through there
or something like that :D
the servos are about 4-5 ohm / winding .. so 15 ohm would be HUGE
oh ok - i think I see
That is what I was wondering :)
but when you're breaking the motor will generate electricity
which goes back in the drive
and from there back into the PSU bus
increasing the voltage
at one point the drives will pop
good night all
jmkasunich: what is your website again?
there is a link from there to the actual blog url which is longer than I want to type
I can handle that :)
um - I get a web hosting company www or http://
[23:42:46] <jmkasunich> http://jmkasunich.dyndns.org
I can't spell dns
thanks - again
was it cradek who said my electrical box was plenty big?
gecko power supply9 x 6 x 5
gecko heatsink 9 x 7 x 1.5
mobo 9 x 12
power supply 6 x 6 x 3
hard disk 6 x 4 x 1
spindle drive 10 x 13 x 3 (without fans)