Won't make menuconfig break my existing kernel (at the very least change the name, which seems to break EMC2)
No matter, I installed the kernel from a .deb, so I don't have the option
Ignore that, my newbishness was showing.
yay, michael h also reports threading works right in git-master
you cannot imagine how tickled I'd be to never have to "fix" that code again
I can imagine
However, I don't seem to be able to use 2.4-pre :-(
(dare I ask?)
Because my kernel doesn't include parport_pc
that kernel is a curse
Yes, but it does give me 2.5uS latency rather than 25uS latency
does that actually make an important difference in your application? 25uS is usually quite fine
that number only matters if it matters
Yes, well, that's a fair point, but I paid for that CPU, I don't like it sitting there idle.
oh it's just a matter of aesthetics!? :-)
Having the RT stuff running in a dedicated CPU just seems right.
I definitely sympathize
however it's an extremely huge pain building that kernel and getting it right
I know, I wasted months trying to do it
that's dedication! I'd quit after only days :-/
I am puzzled why 2.3.4 specifically excludes parport_pc and 2.4_pre requires it...
andypugh: because 2.4 is going to be smarter about enumerating and sharing the parports in the machine
no more searching for the right base addresses
for instance, "would you like parport #1 or #2? I found them for you!" in stepconf
it's a huge usability improvement, and it even allows people to use their parallel printers etc.
So, if I am feeling brave. Can I compile an rtaismp kernel like the one I have now, give it a new name, boot into it, then rebuild 2.4-pre on that kernel without breaking my existing one (that will work with 2.3.0 installed from .deb)
I should have said rtaismp+pport
yes one can certainly do that
Would anyone sane so that?
whether you can do it remains to be seen (sorry, just being honest)
you can have any number of kernels installed
I have built several kernels. They all booted, but none yet have run rtai
if that's the only change you make from the working .config, it is likely it will even be compatible
I am guessing that just adding a device driver to a kernel that is rtai + smp should be straightforward
yes, adding parport should be only one or two things you change in the config
Should it be included or modularised?
I'm surprised it was excluded - I have never done that
it should be a module
Hmm, where do I define the new kernel name? I am nervous about the prospect of over-writing
And it won't make. "No rule to make target /include/config/auto.conf"
[00:43:48] <andypugh> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Debian_Etch_Compile_RTAI#Configuring_kernel
Says to use make-kpkg. Is that advice still current?
anypugh: hi im back. any questions for me?
Yes and no. I found that my kernel does not have pport_pc
ok you need that on. ill figure out a way.
and make-kpkg is the way to build a deb
and you'll have to. thats like trying to add ext4 support to a kernel which only supports ext2 in the config
new option in kernel = recompilation
Well, I think I have had enough for tonight.
just give me a sec
I am doing everything with VNC and it looks like editing the menu.lst to reboot from the normal, non-smp kernel hasn't worked, so I can't get to the CNC PC without a trip to the garage.
device drivers > parralel port supprt (Y)
then set it to compile and enjoy your break
Yeah, I did that, but neither make nor make-kpkg run
make clean first
I did that for make-kpkg
post output to pastebin or 3 > lines here
As I said, the linux machine is in the garage and not booted..
And it is 1am
Oh, what the heck, I won't sleep anyway...
seems as though your in the wrong directory and gcc cant find the makefile hence "No rule to make target"
if you want to compile from different directory, you will need to do something like make -C (or lower case c i forget) and then the directory of which you would like to compile
Well, the plan to see if EMC would run with the standard kernel (with parport, presumably) fell at the first hurdle. I had forgotten that EMC will only run on the exact kernel it was compiled on
So, I go to the usr/srs/linux-headers.....rtaismp directory and sudo make menuconfig?
Add in the p-port..
the kernel source itself. not the headers source, or anything. the actual linux-2.6.XX.X kernel
Yeah, sorry, misremebered while waiting for VNC to wake up
No, actually, I only have headers directories in usr/src. Where should I be looking?
oh that explains it. you need to download the actual kernel source
if you are planing to use RTAI, you'll want to download 184.108.40.206. you can get it from kernel.org on their ftp site :)
No way. I have already wasted months of my life trying to make a patched rtai kernel from a vanilla one
220.127.116.11 is vanilla
but if you know C pretty well you can port the RTAI 18.104.22.168 rtai patch toward other versions as well (besides 2.6.24, and some others)
Yes, and you stay up till 4am then find that the RTAI libraries contain hard-coded paths and will never work unless your kernel is exactly the right name and in exactly the right place.
Been there, done that.
And I have never used C in my life.
the kernel source just needs to be named 'linux'
Not my experience
well you need 22.214.171.124 or some version of a kernel that is supported by RTAI if you don't know C
As I said, I spent months when I first stated with EMC trying to build a kernel that supported RTAI and SMP, and never got one to work.
cradek earlier said it was a fiddly and thankless task...
there are many kernels already supported by rtai
i can see him saying it would be a thankless task myself
btw, here is my HOWTO on compiling an RTAI compatible kernel: http://neo-technical.wikispaces.com/emc2-arch
I can apt-get emc2 v 2.3.4 and boot into the non-smp kernel, live with the 25uS latency and forget about the wasted CPU
Or I can build 2.3.4 on my smp kernel from git. (which ought to work)
a kernel from git master definitly wont work
I don't think kernels come with emc git pulls?
out of all the total kernel releases, only very few are officialy supported by RTAI
I have two of them,
but you need PC_PARPORT
so you can either modify your kernel source that you already have somewhere that supports RTAI or get a different kernel and apply the RTAI patches
well.. not modify the source, just the config and re compile
I have the 2.6.24-16-rtai that comes with the live cd and a 2.6.24-16-rtaismp that works with smp and any version of emc that doesn't need parport_pc
its not necessary required IIRC
No, you only need parport_pc for the 2.4-pre version of emc2
but you might want PC_PARPORT... andypugh> Can I install parport_pc by hand?
previous versions specifically unload it.
you cant unload something that does not exist
such as the PC_PARPORT module
I think it is in the 2.6.24-16-rtai kernel, but for some reason not in ....rtaismp
check the config file in /boot
cat config-2.6.24-16-rtai | grep CONFIG_PC_PARPORT
Yes, it is present in ....rtai but not ....rtaismp
so if you want smp, you will need to download a package like rtai-kernel-source-smp or something
and then re compile that exact source package
Perhaps there is a reason for this, as I never managed to get a working smp rtai kernel built with a parport installed (I assumed it was required)
i got RTAI working on arch without parport
Yes, the question is if it is possible to get it working _with_ parport and smp?
ie, did whoever created the files at http://linuxcnc.org/experimental/hardy/smp/
have a good reason for excluding that module?
maybe it was a mistake
try turning it on in the smp kernel
I am not saying it is impossible, I am saying that I am incapable.
i must switch computers. ill contact you again in a bit
Perhaps I will give it a go tomorrow.
I used http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?EMC_With_Custom_Kernel
but only steps 6.1 onwards using the precompiled .debs.
you cannot recompile a kernel with only debs
you need to the source and you will need to generate your own .debs
Yes, that is what I am saying
I have then compiled a patched emc2 to run on that precompiled kernel.
I am not all that far from meeting in the middle.
Thinking back, my problems last time might have been attempting to run EMC on my custom kernel without recompiling it.
(EMC, that is)
i didnt need to re compile emc
I assume you didn't change the kernel name then?
the API string? no
installed emc will only run on 6.12.24-16-rtai
Whatever, it's late..
alright good night
People who say Linus isn't usable by consumer-type users seem to have a point
I am finding this web of interdependability hard to fathom as a long-time Mac/windows user
andypugh: if you used the stock emc install....
But the stock emc install doesn't have the new, working, threading patch
but if you used the stock install - then git it... (I am not a linux person and do it all the time)
Yes, that ought to work, and is the plan.
It just seems odd to have no SMP support when nearly every PC nowadays is at least dual core
like coordinate rotation... http://electronicsam.com/images/KandT/testing/rotate1.png
wtf kernel doesn't have smp
what kind of crazy distro are you using
for the emc2 kernal to work for most hardware - they make the kernel 'safe'
Yes, you can either install from the CD or with apt-get and have no SMP, or patch your own rtai kernel and build emc from source...
And you can upgrade a working installation on an smp kernel with apt-get and totally break it. I did last night.
However, booting into the normal ...rtai kernel and apt-get installing emc2 2.3.4 should get me going again when my metal order arrives, but in the meantime the postal strikes are giving me time for fiddling.
there are postal strikes?
in the UK, yes
Anyway, if I don't sleep soon I won't be paid enough to afford any metal, so I best be off.
I can't afford any metal, but at least I'm on IRC at 3am \o/
and watching plasmacam vids
momma got a box of computer junk from the auction. 10Mhz vellemen scope, yay. me grab
has anyone ever demoed EMC at any of the trade shows such as Fabtech or Manufacturing Week?
when you want to open up a .gnc file, how do you make emc default to a certain directory?
emc keeps trying to open up the .gnc files from /usr/share/doc which makes no sense
there's a setting in the ini file
ngc files are in a directory specified in the ini file [DISPLAY]PROGRAM_PREFIX (i think)
NC_FILES_DIR or some such, which should point to either ~/emc2/nc_files, or /usr/share/somewhere/nc_files/
right, program prefix - that's it :)
not sure if we have the emc-environment set up right
don't recall running the emc-environment script
ok, you only have to do it once (per shell), even if you recompile several times
I'm trying to figure out HAL now
you're an EE, right?
(or you work with some)
i didn't do --enable-run-in-place ...if that matters
did you make install then?
yes, I get the structure, I'm trying to se how it fits and plays with EMC
SWPadnos, yes. to /usr
ok, then you shouldn't need to source emc-environment
then again, I don't know if the paths are correctly set up by make install (and of course you'd need to either re-login or reboot or something to get them set correctly in your shell
or.. just run ldconfig as root :)
program paths don't get changed by that :)
std emc from cd has .ngc files in /usr/share/emc/ncfiles (which doesnt make sense to me but its de facto )
that is where "samples" officially go
we were just making sure the demos ran correctly
newer installs may have them in /usr/share/local/examples or some such
or is that the configs
heh, I lost track too
it's documented somewhere, like at the LSB website
is there a way to see how our odd latency anomaly will behave by running the servo_sim?
no, I don't think so
I'll try it on real hardware tomorrow
the sim configs (even the ones that run realtime) are synchronized - all the HAL operations happen in the same threads, so the timing appears to be ideal even if it really isn't
some older Athlon boards are working fine, jitter is 10uS with the live 8.04 CD
On my MS-7267 Intel Core 2 Duo im stuck at 65uS heh
all have ATI AGP gpu cards
Alex got an Athlon board of some sort down to a stable (but slow UI) 7 uS base thread
it may not have been doing any real work in the base thread though
wierd but when he runs glxgears the time drops to ~1uS
ah, a core 2 duo with an SMP kernel, right?
yes with 1 core dedicated to RTAI
in that case a do-nothing shell script would also help the latency a lot
like random output going to /dev/null?
like this: while true ; do echo "nothing" > /dev/null ; done
not random, that will stall eventually since you won't have enough entropy in the system
I saw that on a core 2 duo system I made as well
SWPadnos, that command solved it too :)
the final product actually had that script in the initscripts
yeah, the latencies went from ~2000 to ~200 ns when I did that
with spikes to 1600 or so I think
any possible way i set the CPU to be always doing something in the kernel?
run that script
like changing SLUB to something else or changin IO delays, etc?
idle=poll might do it
hmmm, no, it doesn't - I think I tried that
1.6uS? are any of these systems ( few microsec latecy) documented anywhere so other people could assemble the same?
I can tell you what I did, but it's unlikely to be useful
since it had no network and no video in that system
maybe if i pre empt the kernel lock (cant spell it)
(either would cause ~16000 spikes)
SMP should already be preemptible
the RCU is still set the classic though
what is RCU in this context?
ah - read-copy-update
any suggestions for my setup with 7i43+7i33+7i37, running two linear servos from point A to point B, only 1 servo runs at a time
no vid no network, how did you get the data? serial link?
L84Supper, with the 120 uS latency from time to time?
tomp, yes, that's why I wrote the original modbus driver
swpandos : yes
tomp is now known as tom3p
just use it, with no base thread
it should be fine, unless you see anything much longer than that (PID phase issues might start to show up then)
make menuconfig > general setup > RCU subsystem
[03:32:56] <Neo_The_User> http://git.kernel.org/?p=linux/kernel/git/tip/linux-2.6-tip.git;a=shortlog;h=refs/heads/core/rcu
ok, with some tweaks we saw 300uS, but even the 8.04 cd was never over 130uS
I'd let that run overnight
or for a week
we have arch behaving at ~120uS just like the live cd
have you tried this test yet? http://marc.info/?l=linux-kernel&m=124469647405168&q=raw
(on a non-RTAI kernel)
with some nasty hacks put in by me. pulled some code in from zen, set some dangerous compiler flags, etc
ok, take a vanilla kernel, apply that patch and run the test (overnight maybe)
that will tell you the latency that is caused by non-OS-related things
if that peaks at a number significantly below the 120 uS or whatever you see with the RTAI kernel, then more hacking is in order :)
if not, don't bother
with AGP Fast Write disabled the jitter did drop to 80uS , so I have a felling that the onboard GPU may be the problem
could well be
too bad we cant kill the vga bios
incidentally, I haven't ever run that latency test, it was posted to the list by someone who is familiar with kernel development (nostly a BSD variant I think)
my coreboot rom image didn't support apic yet on this board
8.04 CD only had xt-pic going, with the custom ARCH kernel we got APIC going
which appears to not have had much effect
I do have the award bios tools, I could pull the vga bios out to kill vga
what version of coreboot did you try?
it seems that v3 doesn't support much yet
(by way of chipsets)
V3 is being merged into V2
yes, V3 would have required too much driver work
we have Kconfig going now
also serial-ice for coreboot testing and development
cool. I'll have to take a look at it again
I'd love to have a bunch of Atom330/i965 boards use it
I'd have to fix up a mptable to have apic working with my coreboot patch
someone just submitted the first patch for Atom + 945
it dies right after ram init
945 has been supported, ATOM is just starting
ram init it usually the hardest part of the whole port
so it's actually good news
I'll have to look into it again - it's been a while since I toyed with the idea of using it
AMD 780/700 support is just around the corner
but I'd love to see these machines boot in <10 seconds or so
even better (I didn't want to be too demanding :) )
power-on to login in under 5 sec easy
3 sec should be the norm for the 780/700
even faster when there's no X or login screen (all remotely operated)
since coreboot is not a BIOS, ram and chipset init is under 1 sec
the rest is load time from flash or HD
these have moderately quick Patriot SSD drives
and there is a complete open BIOS called seabios all written in C now that can run as a payload
Intel hasn't been very supportive
only when it comes to mesa 3D and drm drivers, intel is way more ahead than anybody
they have their EFI agenda
plus most laptops have the DOS backdoor and "phone-home" in BIOS now
coreboot would take that power away
well, I'd want it for a bunch of Intel D945GCLF2 board
not a netbook
(though it wouldn't hurt there either)
that board should be workable
is that the Atom 240 board?
Oh nice... what is that going for/
and 64 bit, unlike the Z5xx or N2xx
$89 with CPU installed
the Atom is what needs the work
the 2xx, single core versio, is also 64bits
we need Atom docs
WHAT? where? I was looking at the single core version for that price
the 230 is, not the N270/N280
oh newegg BAH
one of the devs has Atom docs under NDA but he's too busy to finish the patch :(
yeah, was $65
also available for a similar price from mini-box.com, and they may even discount for qusntity
does one have a PCI slot?
wow Foxconn Atom + 945 mioni-itx 57.99
hmmm tempting to use it to replace my dual PIII
should pay for itself in a year
[03:56:38] <Neo_The_User> http://www.intel.com/products/processor/atom/techdocs.htm
[03:56:39] <L84Supper> http://www.newegg.com/Product/ProductList.aspx?Submit=ENE&DEPA=0&Order=BESTMATCH&Description=atom+mini-itx&x=0&y=0
these entire systems (2G RAM, 330 CPU, and 32 GB SSD) take ~26W under load
that's with 2 copies of GLXGears running, plus some compilation and other stuff
measured from the wall, using a picoPSU
VIA has the appearance of being helpful but they only follow through part way
they are already testing the ARM cortex9 quad core soc's with Ubuntu 9.10
so about 36W with a real HD
I think a single PIII-800MHz runs at about 36W by itself
probably not that much - you can get 1TB or larger drives that use 24W for spin-up, but are then ~4W operating
but the point is I am going to go from about 150W down to about 40W
which makes the payback at about 12 months
let us know what the latency is on the Atom boards
well, that's a good point :)
one of the best in the wiki table
and I don't think a 1.6GHz atom iis that much slower then a 800MHz PIII
like 8 uS
I was trying to get rd of our last VIA mini-itx boards but I can jump to Atom if it's an issue
I could use some of the VIA mini-itx's
ds3: you'll regret it
ds3, what would that machine be for?
SWPadnos: low power stuff that need a PCI interface otherwise, I'd use an ARM for it
I mean the atom board to replace the dual-P3
was thinking of using an older VIA board but that is a ATX board 9big)
SWPadnos: just random stuff.. it is my mail/dns/shell box
the PIII is morethen fast enough
there's another atom-based board that can probably be used as an HTPC as well
it is just burning a hole in my budget
but it's $170 or $185
that's too long of a payback
yeah, unless you avoid getting a DVD player :)
or a Tivo
Oh no... I can use an ARM for that :D
got a cnc, a lot of time and tons of money? build this http://www.kirkhammotorsports.com/book_aoe/
PCI is one of the last reasons for me to even stare an an x86
the other board also has wi-fi onboard
much rather run hard wire ethernet
I could use an ISA slot though
wow. one model even has a DC-DC converter onboard - no picoPSU neede
oh sure - it's got gigabit LAN too
a DC-DC converter would be nice
a gel cell UPS
and passive cooling - no fans at all on the board
i really wish the x86 crap would do away with the complex power supplies
[04:11:52] <SWPadnos> http://www.newegg.com/Product/Product.aspx?Item=N82E16813500027
it would be nice
just give it a 5V in on a 1" square copper bar and be done with :D
I've seen power supplies like that (except it was more like 8 sq.in)
more or less a 2x4 made out of copper
oh yeah, some of the "super computers" are wired up like that
there was one that was a servo power supply
yeah, the Atom mini-itx boards could easily be +12VDC in only
a 2000V/2100A servo power supply
VIA pico-itx have +12VDC only
I wish I could find a small board (significantly smaller than mini-ITX) that has at least 4 hi-speed USB ports and gigabit ethernet
and enough CPU power so it doesn't completely bog down with all of those pumping data
SWPadnos: stack a few OMAP3 boards together? :)
[04:16:37] <L84Supper> http://www.via.com.tw/en/products/embedded/ProductDetail.jsp?productLine=1&id=730&tabs=1
then there's the "significantly smaller than mini-ITX" problem ;)
SWPadnos: 4 gumstix would still be smaller then 1 pico ITX :)
do those have gigabit?
probably cost less
L84Supper, that looks nice, thanks
vx800 chipset + 1GHz C7
there is doing gigabit and then there is passing data at gigabit speedss
there is alo a model with vx855 for HDTV support 1080p
bummer - looks liek the via only has 3 USB ports
[04:20:10] <L84Supper> http://www.via.com.tw/en/products/embedded/ProductDetail.jsp?productLine=1&id=950&tabs=1
ds3, yeah - that's the problem
the P720 has gbit + 6 usb 2.0
linux accel graphics support will be the only issue
not an issue with this application
time for bed, thanks for the links
thanks for the input on EMC as well
ouch p720 = $297
[04:26:05] <L84Supper> http://www.via-itx.com/epia-p720-10e.html
[04:26:29] <L84Supper> http://micro-xpc.com/items/boards/via/s-rie-p~px/32820-detail.htm
or p700 for few $$ less http://www.via-itx.com/epia-p700-10l-development-kit.html
SWPadnos, one of these? http://www.amazon.com/D-Link-DI-624S-Wireless-802-11g-108Mbps/dp/B0007A6P4S
if it runs OpenWRT, you're in business
I think his requirement was 4 usb 2.0 + 1 Gbit LAN
that has 2x usb 2.0 and 4x 1Gb LAN + free wifi
that was my first dip, there are lots of cheap routers
would 2 USB be enough with a hub?
yeah, they are they guys that keep Mips and ARM socs cheap
they keep the mips factory open
they just started moving to the ARM bandwagon
I think the PS3 emulates mips when it's playing PS2
although they probably haven't even started making PS2 in brasil yet, but when they do it'll be homemade chips
375% import duty in brasil. I stopped complaining about UK prices when I heard that
VIA makes PC's there
I bet they didn't pay 375% import on the plant
they make some high tech PCB's in Brazil now, micro via stuff for cell phones
taxes are for little people :)
i think VIA teamed with some local co to get around the issues
why are handheld plasma cutters so much smaller than cnc ones? duty cycle?
or is the GiantTech just a compact design?
hush now child and finish your wooden table
any devs here in the far east?
I'm just about done cutting & drilling the wood, and I've worked out how to build it in a reversible way with the ribs under the table glued on, by having open slots instead of holes for the captive bolts
pics to follow, but I haven't photojournalled this at all. I'll fake it all later
has anyone used 2 sets of shelf sliders at right angles before? that's the only original idea in this afaics
like shelf sliders http://www.instructables.com/id/Low_Cost_Hobby_Servo_XY_Table/
and more than 2 per direction sounds over-constraining, like one of 3 must be binding to some degree
naah, they are springy in the Y axis, about 2mm sprung play
or Z axis in that design, but Y relative to the slider
I'll tweak them for minimum friction, as that's when they'll be best aligned
and I've got them opposed, so when one pair is fully in, the other pair is fully out (minus the overlap in the tool bearing region)
so they add some support in the other axis
I reckon 2 pairs will be way better than one
boy, and I complain when I have to use a screw drive vs a linear motor
although that seriously erodes the cost advantage
I was thinking about linear motors again today. if the pole pieces were shaped to fit in slots sawn with a circular saw in plastic, that would be cheaper than assembling stacks of precise-thickness iron & mica
high temp plastic
yeah good plan
something cheaper than potting it all in high-temp epoxy tho
that's the one :)
sintered glass powder?
glass loaded epoxy would probably do for the economy model
how much is the cheapest linear stepper so far?
you can get glass sheet down to ~10 mil thick, it's actually flexible
like the covers for glass slides on microscope, they bend, just a little
not as resilient as mica tho, nor as good a dielectric but I suppose that not a problem.
how many thousand volts were you planning to power it with ? :)
I think of laminated poles, and I imagine mains current eddies
ah one day I'll have some nice rollers and a plasma cutter. assembling stacks won't be too much work then
ptfe tape the dielectric constant is ~2
pcb core and prepreg in sheets down to 2 mil thick is pretty cheap
varnish would do, if it's a constant thickness and the poles are OK with minimal spacing
and that probably depends how high the 'rotor' flies
Lucite has an Er of 2.5
I'm not an electrical engineer, so I'll just suck it and see (one day)
jst_home_ is now known as jst_home
MarkusBec_away is now known as MarkusBec
this is interesting: http://www.orchestracontrol.com/
seems like quite a complete CNC controller
never heard of it before though
seems the commercial was supposed to be 2007, is it being supported properly
it's kinda opensource :D
not proper open source though
right.. looking at some manuals now
they have: nurbs, bspline, inv. kins, fieldbusses (ethercat, profibus, sercos), etc
also compensation (out of skew, backlash, etc)
I suppose people should read the sources to see if any has come from opensource :) and "take notes"
[11:05:00] <alex_joni> http://www.orchestracontrol.com/Portals/0/RespositoryManager/Press/documents/Orchestra%20Policy_v2.5comm.pdf
page 3 is interesting
seems very biased to make it free for the developer, and then charge/make money from end users
not quite free for the developer either
depends on the definition of use commercially :)
we were testing...honest
one sure needs to test for a couple years in production before saying it's ok
in the M6 section it says "The axis positions may be modiﬁed" I wonder what that means?
jthornton: that probably means that the toolchange may cause axis motion
to the toolchange location
thanks alex_joni, that makes sense
has any body done a spindle orientation in hal? or has a min to look over my hal
[12:29:38] <robh-wrk> http://pastebin.ca/1646387
the mazak sample config has spindle orient
i borrowed some ideas from it but quite old having to transfer to hm2 also from it
just wanted to make sure it looked ok before i tried it
loads ok, just throught might be better way to do some things now or not
the only thing I can see (in my pre-caffeinated state) is that you may need to be sure the orientation has reset (from the encoder index) at least once before enabling the spindle orient
MarkusBec is now known as MarkusBec_away
otherwise, there will be a large negative wind-up in the PID, trying to get from 72 million-ish to something under 360
i did wunder if not resetting before enable of PID would lead it to try "unwind" position
I guess you can try it on the machine and see
i guess as i set index to reset on next pass it would go from 1000count to zero right away and confuse it even more
yep, that would be more confusing for it
I don't know what if anything was done on the Mazak to account for this
i think in maxak they used there tool-change bit to enable the index-enable to reset only time i could see it linked in
i have a speed cap soi cant see it shooting off any where fast trying todo, if it does not work i guess i will find away to reset counts before enable the PID
I guess the worst case is that it will try to go backwards quickly, starting just on the wrong side of the index, and will have essentially one full turn to accelerate in reverse
and in which the PID will be completely saturated towards the negative
at that point, the position will reset, the index line will go low (since the index has been found), and the PID will then have to recover from the negative saturation
it'll really depend on the spindle accel I think - not so much the top speed
(since the best case would be that it would get back down to zero speed in just about one revolution - the same time/rotation it took to accelerate)
if you have trouble with windup, consider pid.N.maxerrorI
cradek_ is now known as cradek
looks neat (seb linked it on the wiki)
hi cradek just ur reply, i have put in maxerrorI
it'd be right up jmkasunich's alley, he has used pov in the past for modelling stuff like this
that is cool
MarkusBec_away is now known as MarkusBec
I read the licensing scheme for that software. It seems pretty sneaky. You can contribute to the software but then if you want to use it commercially you have to pay them ...??
Sort of a one way street it seems. They make it sound very open but as Alex says - look at page 3. Seems deceptive.
re: openscad, I wonder how it is better than brlcad? I speaking from near total ignorance here as I've not used brlcad either...
better and worse, it seems
openscad can import DXF files and manipulate DXF entities easily (it seems)
brlcad has a graphical editor (hard to use, but since CAD is an inherently graphical application ...)
I don't know which one has better constraints management, if either actually has a solver at all
I seem to remember that it looked like a pain to use dxf with brlcad...
what's a solver?
openscad may, since you can declare variables and use them in formulas when creating entities
heekscad is slowly getting constraints and a solver
well, I'm thinking of constraints like "this hole goes on the same horizontal line as that one
like solidworks or other sketch/manipulate parametric modelers
I see. I haven't used any of them.
Solidworks is nice (and expensive)
I've used various CAD packages off and on for the last 20 years or so, so I have sporadic knowledge at best :)
I definitely like parametric modeling though - you change a dimension and everything that depends on that dimension gets "rebuilt"
as opposed to using tools like trim, which allow you to fit an entity to another, but doesn't help if you move the trimming entity later
parametric modeling makes it seriously productive
(draw a square, then move one edge to make it a rectangle - the "sides" don't extend to the new length)
I have CadMAX, which has the basic feature set of SolidWorks, but looks like a 1980's program
and is $300
I wonder if it runs under WINE. it probably does
I keep nagging the heeks guys for solidworks-alike
that's probably not helpful :)
that's a bit like kids asking "are we there yet?"
(no, admmit. we'll stop the car when we get there!)
well the user interface at the moment is not too useful
someone here was working on a parametric constraint solver for heekscad, I think
I test and feed bugs at the moment
ok, shower time. bbl
MarkusBec is now known as MarkusBec_away
MarkusBec_away is now known as MarkusBec
MarkusBec is now known as MarkusBec_away
MarkusBec_away is now known as MarkusBec
re: recent talk about fieldbus vs sercos http://elettrolinux.com/HW-driver/sercans-iii-gnulinux-driver-for-sercos-iii-master-pci-sercans-iii.html
EbiDK is now known as EbiDK|AWAY
hello :-) -- I am looking for a program where I can design some piece with holes and after drill them on my new CNC. I am new to CNC and I am not mechanic. Can someone tell me about a program where I can make STL files to GCODE, for drilling? --
if you know the coordinates where you want the drills, you can easily write gcode by hand that will do the drilling. Each hole is only one line of code.
cradek: but I was thinking in design in Blender and after use some program... no?
I have installed pycam
blender for holes seems overkill =)
If you want a hole at X=25mm, Y=25mm, 25mm deep, drilled at 100mm/min, you could just program G81 X25 Y25 Z-25 F100
yeah but you can render the finish in the holes :)
but I have no idea how to do it using all that software, sorry
all words that = that above
hmmm, but I am using Linux, no Windows here
cas what you want to do is simple, you can do it with just a text editor, even with windows notepad
you need to learn just a few magic words, the 'gcode' language
but I want to start doing something simple, to learn such kind of a program
I wouldnt use blender for cadding
notepad is simple
if thats the goal
the goal is to be able to design in some CAD program, and after machine it
he said I am looking for a program where I can design some piece with holes and after drill them on my new CNC
isn't blender a sketching/drawing program? those are usually unsuitable for cad because they can't make a feature a certain size
I saw alread FreeCAD and Art Of Ilusion, as for 3D CAD
drilling a few holes can be easier without cad
but he probably wants to do more later too :)
but if the point is to learn software, this advice is no good
(and also, I don't have any good advice then)
PROEngineer is nice, works in linux if my mind treats me right
cas you dont need cad, but if you want to do it that way, look around for cad that will output gcode (you would learn more with text editor, THEN be ready to evaluate cad)
well, later I will use my CNC as a EMCRepStrap machine, a 3D printer, and then I need to design in 3D, so, I was thinking if I can reuse the same softwares...
solidworks is nice with plugins, but its windows
I just can use Free Software, in Linux, no Windows here
you will know how to choose software after learning basics in text editor
cas you may notice for hole drilling 99% of use write gcode, artistic 3d works with cam
it suppors 3d-printing
for reprap you are making solid its not the same
you work from stl then
but is also gcode, no?
what about heekscad? is it better then pycam?
same as what... let him learn >what< the simplest way, then get fancy
and, could I convert some drill code from STL on pycam?
there are many simples to learn
cas you cant evalate better w/o the basics
cas : you might want to browse http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Cam
L84Supper: I already that that page a few times ;-) -- thanks.
I am not looking for a fancy, expensive, complex program - just a simple one, for Linux, Open Source.
any user of pycam here?
lack of replies implies most hand code
the language of cnc is gcode, the programs you speak of make gcode from complex forms. understanding gcode, at least a little will help you and will ntake 1 afternoon to get a good basis for later
'make gcode for comlex forms'...
good point, most of us use a text editor, why not take that as an experienced way to do it?
if is like that... I was expecting some software to do it automaticaly...
for a list of hole locations to drill, a text editor is probably as productive as a CAD -> CAM workflow
if you want to make more complex items later, then CAM will probably be necessary, especially if you want to do 3D contouring (which it sounds like you will)
you can do some automatically but it needs hand holding
drawing a solids model with holes in it and then expecting software to figure out how to drill it is a very roundabout way, and one that it will be hard to do with free/open source software
QCad may be the best bet for that
I think you can just say "put a hole here", and it outptus the G-code to do that
but there are very different programs for doing different levels of CAD/CAM
many beginners expect to draw a solids model and then tell the cnc machine "now make that!" but that is not how it really works. To a large extent you need to learn to be a machinist first.
what you want now is 2d (actually, maybe not even that if it's just hole drilling)
at least try the basics before getting fancy
2D is more or less "route this pattern, drill these holes"
then there's 2.5D, which you may be able to use the same tools for
QCad can output gcode?
that's more than one layer of 2D machining (like a case with a pocket, and maybe different things at different levels, but each level is flat)
I think so
maybe I'm thinking of GCAM or something like that
what you want for the future is full solid modeling in the CAD package, and 3D contouring for the CAM package
cas I do "complex" like gears, but actually hand code in 150 or less line gcode
or possibly 4 or 5 axis CAM output
but QCad is not Open Source, right?
these different levels of usage don't usually use the same processes, or the same programs
cas: i dont think qcad outputs gcode at all, not the free vrsn
oh right, it is GCAM
cnc requires some human effort, you cannot leave it all to the machine
for 2D/2.5D, you might be able to look at the code and see that it's correct by inspection
that's a lot less likely for 3D contouring or 4/5 axis
and there are many many other factors that need to go into actually machining a part: material, cutter material, speeds, feeds, coatings, coolant ...
ok, EMC shows the code at same time it is running -- I will start looiking/learning it for sure
read a little about drilling cycles
I will try GCAM, last time tested it, it was crashing a lot :-(
it is pretty new
once you set the motion mode, each new hole is literally a line that just has the X and Y position of the next hole (I think)
or a line of evenly spaced holes can all be done with one line of gcode
just installed gcam segfaults back to gedit ;)
on the other day, I made a cube on Blender, after used pycam to convert it to gcode and looked like a very uncorrect way to mill the cube! If it was me, I would do like pycam did...
does gcam do pocketing?
I want also try svn hecckscad
you machine a cube by placing in the vise six times. If you want it to end up cubic, you have to think carefully about what order to cut the faces and how to place it correctly each time.
this is the kind of thing where you need machining skills, not software skills, to accomplish the task
or round bar in a 5 axis with one slitting saw :)
no software ever made can make a mill just spit out a cube
archivist: I bet there are many ways to do it :-)
thats the touble, there are many ways and the programs dont know the machine or tools available
Do you have a shaper? [OK] [Cancel]
Do you have a face mill? [OK] [Cancel]
src for gcam rqrs gtkglext-1.0 dev pkg
you would spend hours answering dumb questions
is your machine up to a good bite or not
will it climb mill without throwing a chunk at you
i just compiled gcam, started it, then hit new project and it seg faulted... nice
cas if you have never machined, go to night classes and learn on a manual machine
archivist: I did that! It was a blast.
that's where I learned to say "There's a reason they call it boring."
klick0: did you got the SVN version?
no, i recompiled it in debug mode, was about to debug it... not sure i feel like it now
i'll grab the svn version
klick0: the SVN version looks more stable
new compile runs, segfaults at render
mine segfaulted at same spot
who knows, probably somethign simple
i run gentoo
ok, run a sample, it 'worked' ( wher ethe hell is the gcode output??)
ok file export gcode as emc
klick0, I used to run gentoo all the time. maybe I can help you ;) SWPadnos: when you sent me this patch http://marc.info/?l=linux-kernel&m=124469647405168&q=raw
that goes against the vanilla kernel (Non RTAI) what kind of test did you mean? IIRC, Latency-test depended on RTAI
that test halts all the CPUs, then tells you what latency *still* occurs
so it shows you latency that isn't caused by your kernel, and can't be configured out
some may be BIOS-caused, or SMI, or whatever
i don't think gentoo is the issue, just figured i would mention it, everythign else i have runs fine...
gcam output is running in sim, some profile zigzagged down in z.. seems valid enuf... ok, so use gcam, but have a drink ready!
can you post the segmentation fault or the error you are getting? SWPadnos Thanks! :)
sure - let us know what you see
tom3p: but the gcam outputs a .cnc file, not the same for EMC2?
cas. when you save it, just choose emc format AND change the extent to .ngc
gui_menu_file.c line 131... hmm
can you pastebin it please?
tom3p: I did it and got an error like "," invalid character, when opening on EMC2
cas i'd need to learn more about gcam before criticizing it much, just like you should learn how to write simple gcode in text editor so you can evaluate cam systems.
[19:22:08] <klick0> http://pastebin.com/d63c68f93
cas: then your gcode output has an error
that's from svn trunk
cas: and if I had to guess, it's got commas in it :-)
ahh, i see the issue
endmill, i have no endmills
initialize to first endmill, but i didn't program any endmills, so it's going to crash obviously
cas from the beginning: i installed the libgtkext1.0 dev package ( used sysnaptic), the i downloaded the tgz file from gcam.org and opened it. the i ran ./configure, then make, then sudo make install, but... it didnt really install, so i ran from the new directory ./gcam <enter>
i didn't do a make install, i just ran from the directory, perhaps the install puts in some default end mills.. i dunno, i'm sure i can define something
klick, yes try running form the build dir
and there are no drill, but it has drill cycles/macros
well make install installed the default end mills
which fixed the crash
i used the 1/4" flatbottom endmill, dunno which tool it used
do you exec from the install dir? or is the app known on the sys path?
I am running the gcam from the folder I used to build it
well you must have end mills defined somewhere
if you've done make install EVER, it should be alright
kudos to Justin Shumaker, nice app
cas yes that seems neccesary, i asked klick0 if he did the sae, or managed to get a real install
it installs by default into /usr/local/gcam/bin/gcam
just not in your path
hmmm, I import one STL file and after export gcode, change the "," to "." and load on EMC, but gcode seems nothing more then Initialize Mill and Stop Mill
duh, build instructions on http://gcam.js.cx/index.php/Manual
tom3p, Justin Shumaker is twingy when in IRC, not seen him in awhile, used to work full time on BRLCAD
so, for what can we use gcam? it can't control a CNC, like EMC, right?
gcam makes gcode
emc runs gcode
cas way too complex for me to investigate, the leap from 'drill some holes' to 'import stl' is a big one.
cas you are expecting too much, you have to learn some
archivist thx, thats right, i remeber running it before ( 2yrs ago?)
well this has no dxf import, i thought it did
bout that long since her worked on BRLCAD
cas gcam doesnt control, it make a map up of where to move, emc is a control, it drives along the map
hmmm, so looks like I can use gcam to make the holes I want... and learn gcode :-)
cas learn gcode with text editor. the appreciate gcam by looking at the output wit the text editor
klick0: better than dxf, it has svg import ( just kidding )
hhehe, yea, i read on the forums he hasn't done dxf because it's more difficult..
i thought svg was more difficult
i've wrote many dxf importers
no svg though
svg there are many common libraries though, so not sure why I would write one
there are some dxf ones, but each has it's own quirks imo
yes dxf gets murky as devs are on vrsn xxxiv and acad comes out with xxxv
dxf is messy :)
dxf is as complicated as you want to get it, it just depends how many entity types you want to support
I did a dxf repair prog so we could send to a laser company
ok, back to hooking up the dancedancerevolution game pad to usb to joypad to hal to emc ;)
i don't even want to know why you're doing that eheheh
i use hal to debug it
but i can jog with fancy steps :)
hehe, right, that's what i was guessing
before i close the ffox pages for gcam, he's written a nice manual
another question: I am getting an error with another EMC config: emc2/configs/zenbot_emcrepstrap/repstrap-extruder.hal:5: pin 'halui.machine.off' does not exist
cas maybe ask at reprap, they might have made that config
tom3p: the author of that files do not answer to me... :-(
but, what "pin 'halui.machine.off' does not exist" means?
andypugh, welcome back
I don't have the hardware attached to PC, can be that the problem?
cas code tries to us ethe pin, but it doesnt exist. its not hardware problem right know, its ' i didnt pack that in my bag' problem.
So, I have decided to try building a kernel with parport_pc included. I am wearing a heavy-metal T-shirt, spectacles repaired with elastoplast and I have ordered in Pizza. Can you think what else I should do to make success more likely?
you mean with SMP right?
andypugh: if you're trying to build a package, set up ccache before your first try, instead of after your fourth
some /me tries not to picture the sight
cas post the hal files on pastebin.ca, put the url here, and someone may look at it... sorry but i gotta get back to work
can someone look at this hal file? :: http://pastebin.com/m18440f7d
:: and say why I am getting the error "repstrap-extruder.hal:5: pin 'halui.machine.off' does not exist" ?
Neo_The_User: I have a kernel with SMP, but it doesn't include parport_pc
andypugh, OK. Would you like me to write you a HOWTO on patching the official ubuntu kernel source with RTAI and a bare bone (base) config that you can tweak / modify to your needs?
Thanks for the offer, but I have already found such a document on the wiki
So I will try that first.
OK. Before you start compiling a kernel, download rtai 3.7.1 and go to base/arch/x86 and see if your kernel is supported. if not, check base/arch/i386
then apply that patch in the folder to your kernel source (not header source, nor anything else) and patch the kernel via patch -p1 < /your/path/to/the/rtai/patch
while inside the kernel source directory
cas: Have you got a reference to halui in the machine .ini file?
ie the line HALUI = halui ?
andypugh: yes, I think. Here is the .init file: http://pastebin.com/m59e2e620
cas: No, it is missing. Look for the [HAL] section and add the line HALUI = halui
That adds in the extra user-interface pins
andypugh: but, here the contents of custom_postgui.hal : http://pastebin.com/m25d998c0
That's custom hal, ye need a change in the .ini
andypugh: you are right :-)
andypugh: EMC started and looks correct, see here a screenshot: http://www.flickr.com/photos/43558168@N00/4053130825/sizes/l/
Yes, that looks pretty hopeful
andypugh: I will document on RepRap wiki this information, so others can install/use EMC ;-) -- many thanks!
cas: this must be you? http://www.cnczone.com/forums/showthread.php?t=92032
what version of emc is that?
andypugh: that wiki howto is a bit old, don't be too disappointed if it doesn't work
skunkworks yes, right ;-)
frallzor: see here: http://objects.reprap.org/wiki/Builders/EMCRepStrap
OK, I seem to be falling at the first hurdle. Which of the files at http://www.linuxcnc.org/hardy/dists/hardy/base/source/
is the actual kernel source?
OK Can I re-write the entire emc with custom kernel guide?
That is the "wiki way"
how about you use a vanilla kernel and patch with rtai 3.7.1 from www.rtai.org like I did?
you can still make an ubuntu package so you wont be bypassing anything
you can use checkinstall to build rtai 3.7.1, you can use make-kpkg to build the kernel, and emc can also be built with checkinstall. things are cleaner that way too
me thinks that would be great to have EMC2 running on Ubuntu 9.10 ;-)
* cas thinks that would be great to have EMC2 running on Ubuntu 9.10 ;-)
i could do that :)
all using ubuntu packages too so you wont be doing any sudo make installs :)
just sudo dpkg -i commands :)
I have tried the vanilla approach with approximately zero success in the past though. I never got RTAI to work
Presuming that the instructions on that page made the kernel I am now using (which works fine with 2.4-pre unless I actually want to use a parallel port) that seems like a likely route to success
ah finally got a good enought quotation for steel!
What are you buying and how much?
420kgs of profiles
and a beam =)
[20:30:30] <cradek> http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&item=180425248497
that's no boring bar...
I don't know, it's not dreadfully exciting
thread mill :)
* frallzor is happy like a kid at christmas
hell - buy it for the thread mill.
* skunkworks_ doesn't know why he was worried about the 40 taper
no BIN - I'll probably bid on it
go for it.
wonder what those inserts cost though...
I actually have a machine that can make round holes - I could use these now
hm I should get a 1/4-20 thread mill to play with. I use that size a lot.
you really should get an encoder mounted...
yeah I know
maybe I'll work on that tonight. I found some inkjet motors and they have that HP module and a metal wheel in them.
504 line, iirc
That would be cool, constant-contact thread milling...
I love stuff not described properly
cradek: with index?
no, but index is no use on mine anyway, since there's a belt
I'll just hook one of the channels also to index so it looks like constant indexes
for tapping, who cares
not described properly = opportunity!!! I bought a 1" air wrench once. It was on ebay as a drill. :-)
hook up an index sensor to the spindle
but leave the encoder on the motor, since that's easier mounting (marginally)
it's fun to have all the threads on a part point the same way, but other than that, it really doesn't matter
SWPadnos: it's true maybe I could do that
got an 8 chan HP digital storage scope that way
THIS TV DOES'NT WORK
loon described it as a logic analyser
There's a difference?
its an HP 16500A with scope plugins instead of logic analyser plugins
andypugh im writing a new guide for you that is actually up to date and works. what is the name of the kernel config in /boot? exact name please
the non smp kernel config
non-smp is config-2.6.24-16-rtai
thanks. guide will be ready in just a second. will post link.
[21:09:36] <Neo_The_User> http://neo-technical.wikispaces.com/emc2-ubuntu
sudo apt-get install build-essential and so on. sorry. not sudo apt-get build-essential. im going out for a bit. ill look over the guide more later and fix problems when i get back if it doesnt work. Cheers!
OK, let's see how I get on.
Speaking of .....
I'm trying to load all of the files onto this pc so I can recompile EMC2 from scratch, I already have cloned the source files and I was installing the other needed files and I got to this ..
sudo apt-get install build-essential autoconf
At this point I get a 404 error - basically website not found... Uh, what do I do now?
Try again? It worked for me yesterday
(well, autoconf anyway)
I think I have tried it 5 or 6 times now. Usually I screw something up but so far I'm stumped ...
I got through this one ok.... sudo apt-get build-dep emc2
Perhaps I should wait for 1/2 hr or so .... maybe some DNS update thing is going on..
Making new Gibbs for my lathe... http://picasaweb.google.com/bodgesoc/Gibbs#
Hmm, has Neo gone?
I find myself thinking that I should use the smp .config and add parport_pc rather than the normal config and add smp...
Did the old gibbs wear out?
The old gibbs were bits of cast iron plate, and the adjustment was how tightly the allen screws were done up.
No jack screws, no lock-nuts.
Just three not-quite-tight allen screws
anyone that is using heekscnc?
cas a few in #cam
andypugh: looks nice. bronze? did you scrape it in or just use it as-machined?
but its a bit raw at the moment
archivist_emc: thanks ;-)
It was a big improvement in that parting now works (3.5" stainless even)
Ideally I suppose I should adjust the mounting surface on the saddle to be exactly flat to the ways, but it seems to work fine already
Well, I am grateful to Neo for trying, but blindly copy-pasting his command strings didn't work.
fakeroot make-kpkg --initrd --append-to-version=-rtai kernel-image kernel-headers
Oops, wrong paste
I got that link to work properly, things are compiling now. Must have been a DNS glitch
andypugh: are thouse vanilla or ubuntu sources?
andypugh: also it looks like you don't gave any permissions there, so better run it with sudo instead of fakeroot
I have no idea, it was suggested I as too dumb to think, so I am just blindly following instructions
or give yourself write permission to /usr/src
Aye, I tried sudo instead, and it went further then gave up
I am trying again with no thought at all this time (last time I decided to try using the smp .config and adding parport_pc. This time I am using the standard .config and trying to add smp
The instructions look like they will create a kernel with the same name as a good, working one though, which worries me
andypugh, it won't since the version string is different
and you should use sudo
sorry I haven't used ubuntu in i don't know how long
It got his far last time though
permissions problem ok run sudo make clean
I got everything to compile and I did a run in place....
If I uninstall the existing 2.3.0 so I can compile the newest sw it will not take out my hal folder and the short cuts on the desktop right??
I mean install instead or run in place..
then do: sudo make && sudo make modules_install && sudo make firmware && sudo make firmware_install && sudo make headers_check && sudo make headers_install
No, but it will remove your emc2-dev....
it bypasses the package manager but it works :P
Uh, that's not good
Well, that was what sudo apt-get remove emc2 dod tome yesterday
(did to me)
OK, how do you get around that..? copy emc2-dev somewhere else - not under emc
The details are fuzzy, but I just recompiled so it wasn't a huge issue. The configs and hal files survive, which is the main thing
OK, how about just copy it somewhere else, then make a directory called emc - put the source code in there and do a compile and run in place...
I'll just save stuff off and see what happens..
OK, this compile is running a lot longer
I should have booted without the dedicated RTAI processor option, for more compile speed, I suspect
oh and be sure to run sudo cp arch/x86/boot/bzImage /boot/vmlinuz-rtai when you are done and edit menu.lst or grub.cfg accordingly after its finished
Is that extra to your other instructions?
I can't copy and paste from the IRC client to the CNC machine, by the way, as that is in the garage being VNC-ed to
ill just edit the guide and you can refresh the page
Thanks, that would be clearer
ok now... that should fix everything
Bother. I am 20 minutes in to the old 8a...
Stop it and start again?
just wait until its done and run sudo cp arch/x86/boot/bzImage /boot/vmlinuz-rtai and edit your boot loader config file
your hard drive modules and stuff is compiling INTO (Y) the kernel, yes?
unless you changed that, you need to start all over again, as said in the guide, your hard drive drivers need to be inside the kernel
Though I see a munch of IDE drivers compiling right now
you must change those settings to your hardware though
wait... nevermind. your using the ubuntu config
I was wondering...
heh bit out of it today
I am puzzled why it is better to take the non-smp config and add smp rather than take the smp config and add parallel port
Except that I tried the second option first and it didn't compile very far
CC [M] drivers/media/radio.... I am sure I don't need 90% of this stuff
you don't. its the default kernel
you can hit control C at any time and re configure the kernel. but you will need to restart the compliation process and run make clean
And the chances are high of me deciding I don't need something, and being wrong.
i had to compile about 700 kernels from source to learn the ins and outs of the majority of the options
* Neo_The_User spends too much time recompiling kernels from git
i guess, to build a bootable kernel, you just basically have to be sure there is a driver in the kernel (the Y option) for the ATA controller and for the filesystems used for booting
for the base kernel, yes
but to make it somewhat usual, you need to know what networking options or leave on or off, native language support options, sound options, etc,
i usually just check that and then go through most of the pages in menuconfig and select/deselect whatever seems reasonable
you can always add things as modules after you've got it running :P
the problem with modules though it constantly running modprobe (depending on if you have module auto loading off) and sometimes udev takes longer to load
which is why i go for kernels under 900K or so. fast as lightning
CC [M] drivers/telephony/ ....
i guess some of that stuff is quite old and takes ridiculously little space
motherboards with built in graphics card etc is no problem for running emc with right ?
Hmm, nice to know that the Phidget drivers are a default install
MrSunshine: It depends. Try it and see. Most seem fine
well, its about buying a new one, cant afforda a full ATX motherboard that supports my AWSOME 140W cpu :P
hell even a mATX costs like hell for that cpu
my ms-7267 was pretty cheap
well i have a cpu, and memory and everything, problem is that lightning strook and cut my mobo in half
just don't use the closed source graphics drivers and you probably will be fine
so need to buy a new one, dont wanna waste the cpu :)
although not always
too bad it doesnt have a small PCI E 1X slot for the fatal1ty champion pro sound card
Are you sure everything else survived?
andypugh, tried everything except the cpu realy
RAM, graphicscard etc everything is fine
lightening is likely to have killed everything
so the cpu should be also i hope :P
fans etc starts up, but no picture on screen
I would be astonished if the CPU survived a lightning strike
but im trying to find someone to test the damn cpu, but now finding out that i need special mobo for a 140W cpu i dont know if i want to put it in someone elses mobo :/
andypugh, well why did everything else survive? . .and it was not a direct hit .. power spike
no beep? try no kbd
I would never plug that into a good board
andypugh, AMD Phenom X4
if beep then rom is run
no beeps=no cpu
tom3p, or the mobo can be dead and do not give power to anything? :)
MrSunshine, I can test it but I can't promise I'll return it if it works :P
cpu dead ps alive
How did a power spike cut your mother board in half? Or were you exagerrating?
andypugh, exagerrating ... it just died =)
nothing shows on the mobo
no damage anywhere
what kind of mobo?
pull kbd, no beep = dead cpu
Well, then I change my opinion. But you can't blame me for taking your statement at face value
Neo_The_User, dont remember some crappy stuff .. that does for 1 not support 140W cpu :P
tom3p, even if voltage regulators are fried from the power spike ?
cause i guess the other components also need power to be able to beep
vregs ran fan
It's not like a car you know, maximum CPU power consumption isn't a great feature.
tom3p, well lets say it like this then, i guess the vregs that run the cpu isnt the same that runs the fan .. as its so high powered cpus today
so, if the vreg to the cpu is dead == no cpu also :P
but ye, a voltage meter i guess i could find out with if i find the vregs =)
one alive and others dead is unlikely, a v meter is simple to use, ears simple too
andypugh, what CPU do you have?
tom3p, well depends on where the voltage spike went doesnt it ? .. if it went out everywhere or if it just took one path .. say 12V cable :P
Not sure :-)
just have to find that damn voltage meter
lost both that and the display for my AVR :/
yes your thinking is absolutely correct, determine path with vmeter
aye gonna take that as a project tomorrow if i find the voltage meter .. thanks for the help didnt think of that realy :)
I was being obuse, the CNC machine is a pair of olde worlde Xeons, and the indoor one is an iMac with a Core Duo
not ever gonna buy a new computer anymore
when i can get a descent one for pennies :)
when someone changes their old :P
bought this one to play games .. $800 or so down the drain as i played for only a month
I can see me buying a brand new 27" iMac in the not-too-distant..
then it hasve just been standing there
tom3p, me ?
in sweden but trying to make a currency thingie that people can understand =)
Gosh! Token ring drivers being compiled now, how useful...
anyone from the UK around?
The good news is I just saw the parallel port drivers whistle past
Yeah, and archivist, MattyMatt...
[23:19:38] <frallzor> http://www.ukmail.com/index.html
ever used these guys?
Cant find any eta on deliverytimes on that site, thought some uk-guy might have used them =)
The deliver most of the snail-mail spam I get...
But I hardly ever post letters
well you cant mail packages :P
before some smart-ass replies :P
I have always used royal mail parcels.
no idea where to find eta-times then? :P
hmmm i thought international post used a unified system... like a single site when i track post from thailand/sg/taiwan/aus
post as in post office
sorry old emails dont get me an url
must be wrong, i found i used EMS but thats not worldwide, just far east post
hmm, all the old mobo says is "M3A <GREEN>"
cant find any info on that :/
about the board
hold it up so a strong light behind it
do menu.lst entries require an initrd?
tom3p, still stuck in the chassi :/
OK, lets see what we get then...
not rqd, but uou tell it where root is
and its half past midnight here so i guess i have to do it tomorrow :)
andypugh, only when you use mkinitrd or something like that and generate an initial ram disk but it boots up slower
so your best off just having a base kernel
Of course, I won't know what it is doing, as it is in the garage with the screen off
sata drivers = sXX pata = hdXX
with an inited though, it only uses sdXX
Bah! no VNC conection, best wander out and see what it says..
try editing /etc/fstab to mount the hard drive first
ah ok more details please
not synching something about VFS, mention of sda0,0
I probably should have taken a pen and paper
yeah then i could help you fix it
can you still modify files on the HDD?
By booting back into the old kernel, yes
You mean there was a risk I wouldn't be able to do that?
No i just wasn't aware that rebooting still keeps the network alive
but you would've known that if that were true
No, I wandered out to the garage
Let me get the exact panic message
[ 0.880113] Kernel Panic - not synching VFS: Unable to mount root fs on unknown-block(0,0)
please tell me there is more
Not on screen, no.
ok well.. recompile kernel
first boot into the working kernel
then im going to make a very small config for you after i get some info from you
it will compile much faster ;)
I made two changes to the .config. Presumably that was at least 1 too many.
I turned on SMP, and I changed the processor type from "Pentium Classic" to "Older Xeon"