Have about 3mm clearance on three sides
ok.. so 2xLS7266; 16 x I; 16 x O; (how bout 4 DACs ;)
All on one PC104 card ??
don't think they can be fitted
IO can be SM
all can be sm surely?
DAC's maybe too
there must be some connectors
I'll give it a try...
Hmmm.. 82C55 surface mount, and two LS7266 in 24DIP package
you want 82C55 for IO?
I did the first one with plain 74HC574's
I'd be suprised if it needed a whol pc104 for just that ... should be a snap in smd
one little FPGA could do it all..
robin: hmmm... there are a lot of chips...
but the LS's are better
7266's are available smd
well.. I'll call you to change them when they are broken
without breaking stuff around them
But don't use the laser to get them off the board.
easier than than changing through-hole
although that's not such a bad idea
I always mount through-hole on sockets
anyway, why should they break more than any other chip?
because... they are connected to the outside world
and it's such a cruel world out there...
I usually optocouple the LS's as well ;)
but I found that they usually brake if you don't get voltages right (forget to tie one unused pin and stuff like that)
having done a little production engineering on this sort of stuff, I just think smd these days. through hole is too costly to consider in most apps
* alex_joni is cursing STG's arrangement of ports
not for prototypes...
or small series
true, but its expensive in board space
you gotta bear in mind one major factor I have to consider is availability
I ain't got no Farnell around the corner
they deliver europe wide
down here it can take up to 2 weeks delivery on a chip
wheres down here again?
and small stuff I get delivered only 500+
or even 50+
when I just need one damn piece
I think my last resistor order was 150K pieces :)
suffice it to say, farnell where a 'little' more expensive than the others at that volume
stuff from farnell is usually 50% more then others :D
well, like double or more in those quantities
I'm paying around 1.60 a thousand for 1206 size resistors at the moment
remember how you keep leaded resistors noce and tidy, because they cost money?
well, chip R's you can pop em out the tape feeders just to check they feed nicely
run out 50 or more and its still less than 1 leaded resistor
03paul_c * 10emc/src/emcmot/ppmc_aio.c: Ran this file through dos2unix to get rid of M$ line terminators.
Note for any budding driver authors - Don't use a single source to build different variants. It won't work with kbuild.
what do you mean with diff variants?
#else // build a servo module
there are tricks to work round the problem, but it may break in later kernel versions
did we agree a #if/#endif protocol for the code?
however, you can build kernel modules and user space apps from the same source files.
like 'more than 5 levels of nesting, shot at dawn
8 levels, hung, drawn and quartered
shot on sight for nore than two levels.
harsh, but fair.
5 levels will get you removed from the list.
I fear we may have a very small list :
also, while we're on ...
8 levels, and you'll be forced to work on an XP box for the rest of your natural.
now now, xp aitn that bad
ah that aitn that bad either :)
win 95, 98, ME ..
with USB keyboards
now .. thats BAD
paul_c: Installing 3.01 now.
wanna copy of Win-1.2 ?
I need ..
a bridgport stepper motor ..
one of the big finned things
rayh: Duff burn ?
looks like a pineapple on acid
bigger I think ...
Yep. Still can't read it on the box I wanted but others will okay.
old series 1 thing
only ever seen them attached to a couple of tons of scrap iron.
Might be something with the via 800 MiniITX mb.
rayh: It's probably the boot kernel not set up for the VIA chip.
Seems like the installer asked english three times and standard or custom install twice.
will need to recompile without some of the Intel options...
That could be but rc45 works fine.
As does the 2xx stuff.
ooh, Interact 720 HV? any good?
2.6.x is compiled with IOAPIC enabled
and 686 processor
looks nice, 20 station tool changer
rc46 and BDI-2.xx are 586 only as I recall.
my last laptop crashed when local-APIC was enabled
* paul_c makes a note to recompile the boot & install kernels without PAIC support.
or even APIC
gotta love PAIC
the proc says that it's a 6.
I think you did have a 5x86 def in 45.
what's the mod sign in .c again?
I like the watts furniture logo.
Who's indexer is that -- cutting gears?
Hal May - photgraphed by Dan.
Oh. I thought it might be. Great shot.
Can add more to the slideshow, but do need to watch the overall size
too many large images, and will run out of space for the installer.
25 minutes remaining.
install is a little slower than an RPM based distro
paul: is the new BDI available for download?
nope - Ray is testing a very beta version.
I see... I wanna test too... ;)
Don't want to appear rude...
but downloading an ISO hits my bandwith bad.
ohh.. I see..
if you need mirror space....
I've got to do a bit more arm twisting with the guy in town.
At the moment, I'm doing a 34Gig update of my Debian mirror
If I can get him to do downloads, perhaps I can get him to put it up on his t1.
Space isn't the problem, it's bandwidth I need.
I could put BDI-s at college (on a 150 MBit)
bit faster than my 128K up.
deb mirror is there already
When you are ready for a trial release, I'm thinking maybe we can put it on his company site.
Will have to rebuild a couple of kernel packages, and then roll EMC in
The RTAI package is now in two halfs...
I did notice that on the via box, I saw a deb page on boot. Huge lettering.
On the athalon box it did not show -- only text.
Why did it not come up on the athalon.
what graphics card is in the Athlon box?
All of these boxes are video in RAM.
Let me look.
03alex_joni * 10emc2/src/hal/drivers/hal_stg.c: Started the STG driver. this is pre-alpha right now. some encoder feedback should be available. The rest (DAC, IO) will follow.
Trident cyberblade on the via.
The sis works fine for 46, Suse, and knoppix.
We'll see how 3.01 looks in about 10 minutes.
ok... guess I'll call it a day
Which SuSE version are you running ?
9.1 enterprise server.
with a stock 2.6 kernel ?
Yep. No real-time yet.
I'll take a look at the config on my laptop and compare with the boot & install kernels.
Don't believe the time estimates - It is probably way out.
"An unhabndled exception has occurred....
fsckit - Can you save the debug to floppy ?
I can try.
Nope won't save it. I can transcribe it here.
Nope it died when I pressed cancel.
The last line ?
could not find GRUB runtime files.
OK - Sound's like grub didn't get selected for install.
I'll boot up knoppix and have a look round.
There may be a couple of logs in /root
Knoppix uses LILO but I'd bet i could add 3.01 to it.
in /root there is anacdump.txt (1.2MB) install.log(21.7 KB) and install.log.syslog (1.2 KB)
anacdump.txt is the debug log
the install log is just complaining about not finding some postfix.sendmail stuff.
The other two will also have some usefull info about the package selections
install.log says 456 packages.
Is grub in the selected package list ?
grub not found.
Any chance you could zip those three files up and mail them over ?
I can try.
tgz, bz2, or zip...
I'll look at it first thing in the morning.
time for me to slide off to bed.
I would like to say thankyou ... to all the zoos and circuses whose kind loan of monkeys has made the testing of EMC2 possible
hm .. I had installed bdi-46 last week, and was able to run sherline_inch without any actual hardware connected
I reinstalled bdi-46 and running sherline_inch says something about "amplifier fault"
I really only need to run in sim mode, but there's no sim.run / sim.ini on bdi
any idea what's going on?
well, I copied sim.* from another machine and got going
huh, pretty quiet night. Morning all.
alex_joni: what's new?
it's a free day here
national day ;)
I started a stg driver for emc2, but lacking a STG board I can only write blindly ;)
Why write a driver for a board you don't have?
because it'll help emc2 ;)
and because the encoder part is pretty similar to the one on my board (I already wrote that driver)
emc2 is pretty nice, but without hardware support it's pretty useless for a lot of people
What Linux distro do you use? I'm working on my new backplotter, and I've been concentrating on getting it to run on RedHat 9 and on BDI-46 (because I have easy access to them), but really I have no idea what most people are using.
(and on fedora core 2, as a ngc-file previewer)
I am using SuSE 8.2
but that's because I had troubles installing BDI on my hardware
and at home (also SuSE 8.2) it was easier for me to patch the kernel and install EMC, than reinstall everything
but I plan to switch to the next BDI (deb based)
do you know what version of Python is installed on that system? Do you know if SuSE has a python-opengl package you could install?
give me a moment
and a few othe python packeges (around 10-15)
taken from ftp://ftp.iasi.roedu.net/mirrors/ftp.suse.com/pub/suse/i386/8.2/suse/i586/
python-opengl-188.8.131.52-395.i586.rpm on the SuSE 9.1
jepler: short question, did you play with tar ?
* alex_joni is wondering how to make incremental backups using tar
I don't know much about tar
when I make backups, it's of the form "burn my home directory on 20 CDs"
I do have a tar -cvzf in cron.weekly
alex_joni: and the version of Python? python 2.2.x? 2.3.x?
alex_joni: thanks for the info
cd /home && for i in * do tar -cvzf /backup/$i.tgz $i done
use j in place of z and replace tgz with bz2
as long as you have bzip2 installed...
is bz2 better than gz?
I really like gz because I have gz support on M$
no bzip2 on M$ (at least not standard)
on SuSE 8.2
sorry it took so long but I just copied 12 GB of backups
bz2 is 10-20% better than gz on many types of files.
-rw-rw-r-- 1 anthony webmaster 7840762 Nov 30 03:18 Python-2.4.tar.bz2
-rw-rw-r-- 1 anthony webmaster 9198035 Nov 30 03:27 Python-2.4.tgz
alex_joni: for anything more complex than what you have, you (in my opinion) should be using dump and its built-in incremental support
dump -0af - /home |bzip2 >/backup/home.dump.bz2
(assuming /home is a partition)
use dump -1 for incrementals
for interactive restore mode:
cd /home; bzcat /backup/home.dump.bz2 |restore -rvf -
actually, several dump -1 will be differentials, not incrementals, but that's probably what you want if you only want to have to look through two files (the 0 and the latest 1) for your data.
Seems that SuSE started a mail transfer agent and I'm getting some mail in two places.
don't understand why grub failed - It is installed as part of the initial setup
Got your most recent mail in /var/spool/mail
* paul_c is no expert with postfix/exim
I agree that most of these issues would be resolved with a completed install.
I was installing to hda5. I wonder if extended ext3 might be an issue?
I got it going by modifying the knoppix LILO.
If it had been xfs or reiserfs, it may have been..
Had a very bad experience with reiser a few years ago.
at least with ext3, you can mount it as ext2 and recover data if it goes tits up.
Let me dig out a hd and try a clean install.
cradek: that is some very valuable information
I'll look at dump
don't seem to have dump installed... there is a dump2efs..
make that dumpe2fs
dumpe2fs is not what you want
I guessed so...
I found a dump.sourceforge.net ... seems more like it
now I gotta compile for 2.2.13 ;) and some ancient glibc
it should come with your OS distribution
look for an OS package first
gcc 184.108.40.206 ;)
-> rpm -qf /sbin/dump
for SuSE 6.3 ?
well, I don't know anything about SuSE
I'll look .. maybe I still have the CD's ;)
but every unix that's not an abomination has dump/restore
thanks again for the hint
* alex_joni is going home now...
cool, my modern dump has a -z for built-in compression
I could pipe the output through gz.. right?
looks like you also need to use dump's -u flag for incrementals or differentials to work
logger_aj .. are you getting this?
paul_c: Same problem with the clean install.
Testing another build here with a new kernel.
cradeck: confirmed abort bug on june 2004 emc build
so it's not a new thing
I don't know whether I think that's good or bad news.
les: What happens if you put a % on the first line ?
paul_c: doesn't help
it returns to 0,0
% seems not to matter
the same for a comment block ?
did not try...but non motion prep codes did not seem to help right?
I think that did help
if your first real program line is g64, for instance
ok...I was in a hrry and did not try them
well that's the workaround I guess...and the reason I never saw it before
it's nice to understand the problem
for safety all my programs have lots of prep codes
all I knew before is "sometimes" this happens
well it needs to be found...I will be looking as I get the chance
emcmot.c and what else?
More likely in the task interface or the interp.
emctaskmain.cc for starters.
Could you guys get me up to speed on this error?
everything we know is in the bug tracker on sf
[16:37:49] <cradek> http://sourceforge.net/tracker/index.php?func=detail&aid=1074914&group_id=6744&atid=106744
here is the url
bits are flowing slower than gear oil at the pole.
sorry phone call
I will look those files
yikes got the custom router bits ordered
they are cnc ground
I almost started ordering diamond wheels, c2 carbide, and silver brazing paste
I could make em...but it would be a learning curve
and diamond wheels ain't exactly cheap
and I think they're easy to ruin
especially when dressing them
Dress a diamond wheel?
yes...the resin bonded ones are dressed to true them
the plated ones are not dressed...but they are less accurate
I've got a pile of diamond drums out in the shop for stone work.
Never heard of dressing any diamond cutter.
Take that back. I've got a 30" diamond saw that you dress by mashing the diamonds back into the babbit.
A diamond grinding wheel or a diamond cutting wheel?
I did some research on wheel manufacturer sites
* paul_c has a diamond/CBN wheel that needs to be dressed after mounting.
diamond grinding wheel
Also got a bunch of diamond bands.
"crush dressing" is the best method
but truing with a diamond point works too
it just breaks the soft resin bond I guess
Never to old to learn.
but I have not done it...and dressing the wheel the first time would be scary
but then shelling out the $500 for 5 router bits is a bit unpleasant too
Where you getting those bits?
[17:04:31] <les> http://www.southeasttool.com/about_us.htm
bet that machine costs a buck or two
ah..just got a call from them...talked em down to $350...that helps
a few words on the phone can be worth some money!
I told them the price had to be low enough that I am not inclined to make em in house
Your search - cnc etch-e-sketch - did not match any documents.
didn't someone do this once?
seem to recall it yes
Yeah, I saw a picture of a cnc etch-a-sketch
I'm pretty sure I've seen it too
now what I need is one of those robotic vacuum cleaners
wonder if they work well
neat idea but the etch-a-sketch really needs a Z bit
I've often thought about one of those robotic lawnmowers...
but what I need is a brushhawg
When automating large equipment with whirling blades, one must ask oneself:
"Just how good is my programming?"
haha for sure
I have cats, you see :)
but cats know to stay away from whirling things
kids, however, aren't so smart.
I've had some pretty stupid cats
I designed a lawn mower guidance system but never prototyped it
This litter is pretty smart
litter of kids or cats?
used rotating laser beacons
need to see at least two at all times
dif GPS is too slow
I've played with several ideas...it usually came down between precision and cost
well this boiled down to product liability
beacons linked with rf to servo lock...each sends an ID modulated on the beam
I think my best best for a "hard" safety limit would be one of those electronic dog fences
shock em if they get too close?
But I live out in the sticks, so it would've been a couple miles of wire to bury :(
asdfqwega: I am seriously out in the sticks too
And I have too much "statistical noise" for beacons
that is why they are modulated
I'd have to put beacons every 10 feet or something - so we're back to miles of wire for power :(
I'm in ohio - not enough solar
And being in a valley on a hill doesn't help - and that's not including the trees
well...put em on a stake...beacon can be removed, taken in, plugged in to wall wart
for trees and stuff must always see at least 2
might need several
anyway the real problem is mowing grass but not fingers
Actually, the best solution: Get a Gravely - It cuts a widers swatch, so several acres doesn't take long
and it cuts heavier grass, so I don't have to do it as often
ha...I have a gravely model L
And if you REALLY don't want to cut your grass:
Get 3 sets of lawn ornaments
with those wires to stick in the ground?
Get one with 1' wires, one with 2' wires, and one with 3' wires
When the grass gets higher, just pull them up and put in the next set
That way, people think you mowed your lawn ;)
I had better like cutting grass...I will be building the two hole golf course next spring
mow green DAILY!
yay, I fixed the abort bug
That quickly? Wow.
by digging through the layers and layers of embedded switch() statements in emctaskmain.cc
it also fixes RUN / ESC / RUN
this used to give a strange "File is already open" error
yes have had that
bringing up emctaskmain.cc to have a look
no line numbers (using Word)...could you give me a search term?
the one in the middle of case: statements
what was happening was this:
it would do the abort message, which cleared interp_list and closed the interpreter.
then, since it was in state READING, it would go on to read the next line, which reopened the file and skipped to the first gcode
this would leave the file open, leading to the error if you hit run again
if instead, you hit abort again, the abort would do the right thing since the interp is no longer in READING state
got multiple finds of return retval...what is the preceding line?
this also explained why the repeated lines were required (so the interp was still READING when you hit ESC)
previous line is retval = emcTaskIssueCommand(emcCommand)
heh, I think my fix is way wrong
cradek: next, fix the problem with step at the beginning of the program not setting pause when it's done (I think that was it)
* cradek looks at the surrounding code
oh, well, fix this first
I bet I should split out ABORT
but my two minutes of testing said it was fixed right!
hmm still can't find it using previous line...emctaskmain.cc is the file?
what you broken ?
I'm still trying to get a good compile for the motenc..off the cvs
vitalmod.o is about 105.5K which is different from what ray gets.
tkemc comes up with wierd numbers ... just like it did before you fixed the type for the encoder
leme see what I have...
Build from Oct 28th
it also changes numbers when I come out of estop.
vitalmod.0 105410 bytes
that sounds about right
I'm busy moving computer, etc from the machine to my test setup.
the most recent build, vitalmod.ko is 110349, but you don't want to know about that..
off the current cvs...rc47?
dave-e: have you confirmed you are using the same compiler? you can't compare .o sizes otherwise
what ever is native to rc46
the latest build is for 2.6.9-adeos
ie. brand-new install of rc46...don't screw with anytthing...just load the cvs and move the rcslib/etc
just need the rtai & linux-rtai defs.
To be honest, once you have rcslib compiled, you don't udate from cvs.
that makes sense since it hasen't changed.
ok, I take it back, I don't think my fix broke anything
going to go shopping and see if I can get the experimental rig up.
see ya later.
must have had a power bump when out in the shop...need new ups battery I think
not much, but I fixed the jog bug
really? what was it?
well, all the symptoms we found were right, and it just did the wrong thing in that case.
after the abort, the interpreter tried to read the next line from the (closed) file, so it reopened it and read the first line
so are you confident of your fix now? I thought you said earlier you found something it had broken, but didn't elaborate.
but only when it was in the READING state
jepler: I'm pretty sure it's ok.
did you commit the change?
please test it!
if my memory serves me well this was for emc1 too.. right?
yeah, which tree did you fix it in?
nice going cradek ;)
thanks, but it really wasn't hard once I decided to actually look for it instead of complain and hope someone else would
yeah... usually I get there too
paul_c: are you around?
yes just got back
les: paul said you have a STG?
yes stg 2
or don't I remember it right?
I started a emc2 driver for STG
stg1 is not made anymore I think
but there are a lot out there so a driver is needed
very different stg2 from stg1?
just some index latching changes...they are very similar
the only functional driver for the current stg II and emc1 is newstgmod
oh and newstgsegmod for segmentqueue
....but segmentqueue has problems
Okay, I have a question about motion blending - if anybody wants to here it
Can it be enabled and disabled, and can I change how much?
Or is this something that must be done at compile time?
yes G64 enables it
yeah but newstgmod is not GPL
right...like emc it is public domain
well emc2 is supposed to be GPL
sounds ok to me...if possible when using public domain emc elements
not using public domain elements
cradek: I'm going to try your patches now
* asdfqwega needs to cvs co 'documents'
I was down for a while...doing a fix for the abort thing?
* paul_c scrolls up...
alex_joni: Yes, I have an STGII card.
cradek: the problem is gone .. of course, I can't comment on any new problems you introduced..
paul_c: actually I wanted to ask you about that PC104 board
need optocoupling on the IO ?
That would be wise
hmmm.. then I don't think it can fit
maybe with SM optocouplers
so on the simulator, I've noticed that "step" goes a little bit into the next motion each time
alex_joni: You make boards?
once in a while
That must be handy
paul: don't seem to find any SM 82C55 (only DIP and PLCC)
I've tried, and it looks like butchery :(
maybe SM socket for PLCC
it's pretty easy actually ;)
but you need a good autorouter (I have a very good one )
he's doing this for the last 6 years ;)
plcc is sm
I have a feeling it has something to do with motion blending
alex_joni: Farnell part No. 54064
but I agree there are sm sockets for plcc
sm opt-coupler in an 8 pin sm package
hard to solder
2 opt-couplers inside?
* asdfqwega just tried to use lyx 1.3.4 - no go
8 pin is kinda a waste for single ;)
* asdfqwega urpmi'd back to 1.3.3 - go
anyways... I'll see tomorrow what I can get around here
les: thanks - G64 sounds just like what I was looking for
Part No. 143480 - Dual opto in an 8 pin sm package
sounds better ;)
can you try the hal_stg some day?
alex_joni: Does the machine you use to autoroute PCB's also drill?
Oh, wait, I didn't read that right
I don't use a machine to autoroute PCB's
I was talking about designing (the autoroute stuff)
that does a friend of mine
when the design is done I take it to a local board manufacturer
alex_joni: nope - Not until jmk produces a sane config method.
masks & stuff
halcmd and the baggage that goes with it.
hmmm... seems you don't like it ;)
That's one way of putting it.
do you have a better version in mind?
of halcmd ?
of whatever bugs you
halcmd is ok if you want to test just some parts
e.g. connect a halmeter to a encoder count board
and see if it works
I think a halgui would help a lot
OK... Now try taking a rank newbie through setting up and configuring his machine with halcmd
I agree on this...
it's not that easy if you don't know it...
and who do you think will end up having to provide the support ?
hmm.. some very good manual :D
just RTFM all over the place ,)
and how many read the manual ?
Manuals? If I need help, I just whining to Paul :D
Gr, can't type or speak today
you could read a manual...
how to type or speak ;)
ok... so there should be a more intuitive version of halcmd
not comand line?
* paul_c hands alex_joni the "Regular Expression" handbook
Yeah, something X-windows, with pastel colours and big, gay WinXP-style icons
And the talking paperclip for help
Much of the data required to configure hal is already in a stock emc.ini
some of it is implied data...
about all that is missing is pin assignments for RT IO
so a halconf utility that takes values from emc.ini
sets up hal connections
if I got you right?
pin assignments also in emc.ini ?
Not for step/dir, home, & limit pins
those in a .hal file?
that's not very different from now...
For the current demo emc.ini, core_stepper.hal AND standard_pinout.hal are required
4-5 new lines in emc.ini
and only 41 lines in the two *.hal configs
virtually doubling the number of parameter required to get emc running.
and you want to remove some?
or make them implicit?
Many of them can be inferred from the stock ini, yes.
I think there can be a compromise between what you want, and what is currently done
maybe... But there are other issues that need to be fixed first.
for example - Run the emc demo
and from another console rmmod the parport driver, or any of the other "components"
better still - Set up a loop that rmmods and insmods the parport driver.
I see.. no lock on modules
It can get much worse than that...
at some point, the computer will crash
well without lock it's pretty obvious it can
on the other hand.. locking a module because emc is running is also bad
if emc hangs... rmmod won't work
If emc hangs, a reboot is required.
then I guess a lock would be ok
problem is.. who sets it, and who can read it ;)
Once emc is configured, a lock is mandatory
is there a common way to lock modules?
There is the MOD_IN_USE macros
but due to the emc/hal structure, that won't work too well.
there is another method...
I understand the security/safety aspects - but when you're making chips, who's going to be poking around with rmmod?
in a single user workshop, perhaps.
* asdfqwega thinks of all the goofballs he's ever worked with...
In an industrial environment, look at the average ape.
I take it normal separation of root priveleges isn't enough
Stability, security, and robustness.
I defer to those who know those three better than I
Loading and unloading of modules requires root priviledges
however, there are ways round that... 2.4 & 2.6 kernels have some neat tricks for hotplugging that could be used.
I understand (from using 2.6 kernel) that modules can have a 'user count' - being in use can make module removal impossible (depending on kernel compile options)
that feature is available from 2.2 onwards
So how do you lock a module, when there appears to be several ways to circumvent that?
replace rmmod with a symlink to a script?
nope - It needs to be done within the kernel module
An internal lock
yup - A variation on the MOD_IN_USE counters
Paul, I'd like to tell you that you're clever - but I probably don't know enough to full appreciate everything you do
Now, I'm going to be chased out of my own shop - the fumes are getting think in heree
paul: how about "The 2.6 kernel, however, knows not to unload a module that owns a character device that's currently open."
that sound usefull for 2.6
2.6 modules can also be forcably removed (depending on kernel configs)
but the MOD_INC_USE_COUNT, MOD_DEC_USE_COUNT sounds useable
only 2 macro calls...
those macros work fine if a hal driver is locking hal.o
but the drivers need to register with hal.o and be locked themselves.
there is a can_unload function
that can be defined
"For more complicated module unload locking requirements, you can set the can_unload function pointer to your own routine, which should return 0 if the module is unloadable, or -EBUSY otherwise."
ok.. me going to bed
Need to be carefull with some of the kernel level functions - Some of them are not available in all releases