k be right back
That's not an option for me as I do not see it. I am going to try some other settings in my kernel and see if I can get it quicker.
I have multi core scheduler support on, 300 Hz timer frequency, and 0x80 based IO delay
at 16us the only option is c1e ...
Neo_The_User: Are you using the SMP build of RTAI and have you given one CPU/Core exclusively to RTAI?
I am using SMP and I don't know about the other question
[00:46:11] <andypugh> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?EMC_With_Custom_Kernel
Right at the bottom, section 7.1
That got my latency down from 23,000 to 2,509
wow thank you so much
do you recommend optimizing rtai itself using the ./configure options?
If you are asking me, I am not the one to ask, I barely understand the question.
like erm.. cd rtai-3.7.1 && ./configure --disable-user-debug
let me try this SMP thing first. if the problem persists, i'll try triming down rtai to its barebone. ;)
i trimed rtai down too along with RTAI smp support. same thing. anyway 16000 isn't bad. thank you for your support. :)
Did isolcpus work at all? You should see one CPU with 0% load
im compiling emc v2.3.4 over 2.4 beta and i see one cpu with 0% load
oh nice. with glxgears running it went from around -30 to -1000 :)
MarkusBec is now known as MarkusBec_away
Do drilling cycles work with a lathe? I get an error about undefined Y, then another error if I try to set it.
yes but there is no Y. if you want to drill in Z, you will need to change to G17. if you want to drill (peck?) in X (???), you will have to change to G19.
Is there a list of reserved numbered variables and what they do? The documentation says it's in the RS-274-NGC manual, but i can't find it.
Alternatively, what's the easiest way to remember the current position (I just need Z) so I can use it later in a GCode program?
Section 3.2.1? http://www.isd.mel.nist.gov/personnel/kramer/pubs/RS274NGC_3.web/RS274NGC_33a.html#999262
looks like I just bought a CNC lathe
gotta make some room quick
going to be an EMC conversion as the controller is dead
he has dropped the link all ready
it is that Hardinge cradek and I were looking at a few weeks back
I kept it in my watchlist for a while
jthornton: happy for you
thanks alex_joni I've been looking for one for a while
MarkusBec_away is now known as MarkusBec
jt-plasma: cool! have you seen it yet?
on the latency test page there is an intel board with the atom processor
does anyone know if those stats are using the onboard video or not?
[13:28:12] <mikegg__> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Latency-Test
it's the Intel D945GCLF2
I'm betting it is with onboard video
that board only has a single PCI slot, so finding a video card that's better than the embedded Intel graphics is difficult
cool. spec-ing out another machine. I think that will work quite nicely
it's a nice board
there's another one that's a bit more expensive, but is fanless
more suited for an HTPC though :) http://www.newegg.com/Product/Product.aspx?Item=N82E16813500030
milkegg__: That board works great. I'm using the onboard video also - NP
I'm getting really low latency numbers similar to the EMC2 Wiki doc - so those are real
it's even better with the experimental smp kernel
SWPadnos: that nvidia onboard graphics work ok? geez $170? more than double
nvidia can be a crapshoot for RT
I love the fact that the intel board has serial/parallel ports
yes, that's a big plus
so many MB's these days don't even include the add-in card thing
I made up a little bracket that holds a PCI card parallel to the motherboard, in a PCI riser card
so you could use Mesa cards
exactly what I was thinking
how did you do it?
yeah, makes it a little easier to mount in a control cabinet
using 1.75" standoffs to hold it above the board
then a 1/8" aluminum plate, with holes for the PCI card and the two motherboard holes along the back (I/O) edge
then some short spacers (1/16" thick, I think), then the board
I'll have to make one on a CNC to be sure the hole placements are right, the one I made by hand seemed a little off
how did you bridge the PCI bus?
....make it turn 90 deg that is
with a riser card
I actually sell a PC that utilizes that board.
[13:42:25] <Dave911> http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&rd=1&item=160372584424&ssPageName=STRK:MESE:IT
I also have one with a board setup similar to what SWPados mentioned but it isn't on my website yet. Same idea...
[13:42:27] <SWPadnos> http://www.mini-box.com/s.nl;jsessionid=0a01074d1f4305615a461e034d04abd893fc8d50899d.e3eTaxaQbxmTe34Pa38Ta38Qa350?it=A&id=289
Yep, same stuff
ok, that looks like a VOOM enclosure
oh no - M100
mikegg__: that was with on-board video.. You can get better latency with the smp kernal...
oh wait - M350 case :)
I put them together for my own use so I decided to see if they would sell. That is a Minibox enclosure...
i was reading about the SMP kernel. definitely going to give that a shot
Yep, I stuck a bunch of fans in it to keep it cool in industrial cabinets
the idea behind my adapter plate is that you don't need an enclosure at all
you just mount the PC to the control cabinet backplane
I haven't had good luck with open boards in enclosures. Some yahoo electrician drops a screw on it and presto - dead boards
So I put everything in boxes..
don't hire yahoo electricians ;)
I didn't hire them, someone else did! ;-)
heh I fried my first MB when a stray BB fell in the case
you'll shoot your eye out....
It takes very litle to pop a MB - brush a wire against it and presto - dead MB
that's why we keep the door closed while machining
Don't underestimate the stupdity of some people... There is a reason why CNC makers put interlocks on cabinet doors so you can't run with the doors open...
I rarely underestimate the stupidity of people
doors...who needs them......BANG
that's what tape is for
OK, a short story about idiots. The local town near me put in a pumping station just down the road. Everyday some yahoo from the city arrives at the pumping station and opens the Nema 4x cabinets to do something ??? These are 480 volt power cabinets. Rain or shine or snow storm the same idiot goes down to the pumping station each day about 7:30 am and opens the doors. The...
...panels are mounted outside with absolutely no cover. So open the door and let the rain in while this yahoo checks whatever. Sooner or later I am going to drive by that pumping station and see that guy laying there dead in front of the panel. It is just a matter of time. I like living here but some of the people around here have absolutely no sense whatsoever.
480 just tingles a little bit... ;)
but wait, it gets worse. once this idiot kills himself, one of his relatives will sue the town for making him open the doors all the time
Hehe.... did I mention that I live just outside of the town limits.... that is purely intentional... ;-)
getting the occasional shock is good for you
I tied into 480 volts once pretty hard and all I remember is things going black. Then opening my eyes and realizing I was on the floor in front of the control panel. I had no recollection of falling backwards at all.
The rest of the day I was useless... That was about 20 years ago. The newer finger safe electrical gear is a big improvement safety wise.
Ive had more 240 volt shocks than I can remember and one at around 9kv in a tv
reminds me of an old Robin Williams skit
but I do follow the rule of only one hand in the device to avoid cross body
the TV preacher saying "I want you to GRAB the back of the TV set and FEEL the power of comedy"
"... as it shocks your ass across the room, someone will have a laugh" :)
i was just browsing around in dealextreme and found these: http://www.dealextreme.com/details.dx/sku.6187 http://www.dealextreme.com/details.dx/sku.4688
i wonder if those can be controlled in the raw way
I believe alex_joni has messed around with those
I think the ports work, but laptops in general are crap for realtime, so it doesn't matter
well, if you find a good laptop (i guess some old ones might be), it'd work?
it might - you'd have to test it and get back to us :)
one comment in the more expensive one says that the manual says it works under dos. i guess that could mean it could be good
there's just one problem - the older laptops have parallel ports already :P
I think that means that it shows up at the normal IO port addresses for LPT ports
that very likely says nothing about timing or anything else
maybe it'd work for some extra i/o for peripheral things, in addition to an another port already on the mb
sure, that's likely (assuming the laptop isn't crappy in the first place)
i don't need one, but if someone needs, he could try it 8)
they also have these quite cheaply, might be useful http://www.dealextreme.com/details.dx/sku.4638
(the comments say it's good)
i can imagine someone putting one to every pci slot they have 8)
have to watch which chip is on the board
try these insteadhttp://www.byterunner.com/byterunner/product_name=PCI-1284-P2/user-id=/password=/exchange=/exact_match=exact
but you'd have to cut some holes for the extra D connectors
that sure looks like a quality one
they're nice, I have one of those
and an 8-port serial card from them
expensive, though, compared to the dx's $10.90 and free shipping
$26.95 for two ports in one slot
yep, depends on whether you need one or two ports
cool! i have finally got my closed loop spindle speed control working, it is excellent!
! a happy bunny
indeed, the spindle encoder seems to be working perfectly
so i just need to learn a bit more about pid loops then tweak it for some better response
Dave911 are you selling the 40w cnc laser? got there from your posted link (tinyurl-ized now) http://tinyurl.com/yju7qzv
I've got a configuration that I set up with stepconf and microstepping set to 2. All worked correctly. When I re-run stepconf and set microstepping to 4, all movements and speeds are doubled. Is this normal behavior or a known bug?
yes - you are changing the steps per inch by doing that.. (doubling it)
if you would change your drivers from 2 microsteps to 4, the movements would stay the same
adjust the drivers at the same time
that is what this setting actually means - it's supposed to say how your drivers are set.
Sorry. I'm missing something. I'm setting the 'Driver microstepping' for each axis. Is there another setting to be changed at the same time?
what are you trying to do?
if your drives themselves are set to 2 microstep (aka half stepping) you cannot change that setting with stepconf. if you lie to stepconf and say they are 4-microstep, you'll get wrong speeds.
so I don't understand what you're trying to do by changing this setting - can you elaborate?
Just trying to tune my machine. I was trying quarter stepping. O.k. I think I understand. If I remember correctly there's a jumper on the hobbycnc controller for changing the microstepping. That must be what I missed.
yes, there you go
LawrenceG: how is it going?>
btw i have a question about pid loops to control spindle speed. Is it possible to either make a button to disable the pid, or some how disable it during spindle syncd motion like rigid tapping etc
why do you want to disable it?
well the loop wobbles a bit when its trying to correct the speed after doing the m3 to m4 transition at the bottom of the thread
i can just live with it, but just wondered if it was possible to do
configure the loop so that it doesn't 8)
hah yes i've been trying to do that ;-)
yeah, maybe it just needs tuning
it's very hard to tune a system that responds slowly
I wonder if the r/z tuning method would work better for that type of loop.
whats the speed controller?
so i'm finding out! is there some automated method of tuning the loops?
the speed controller is the vfd
ah, they have fixed ramps
i have the spindle encoder smoothed out a but with a low pass filter so the speed isnt too jittery
usually a vfd keeps speed on its own pretty well...?
it does, within a few 10;s of RPM
what is the reason for the pid then?
but i thought since i now have the facility of proper spindle speed readout, i'd make it closed loop
the main reason is for me learning it
I wonder if it can be in a tight loop at all
[16:24:05] <skunkworks> http://www.linuxcnc.org/docs/html//motion_pid_theory.html
maybe once you learn it you will decide not to do it :-)
(I've tried doing this too, and decided it worked better without)
* archivist is just thinking of a loop fight
from what I read - it was used to tune hydaulic loops to begin with.
may the best loop win :)
skunkworks, good morning...
LawrenceG: Good morning! Getting cold?
ah ok so its probably not really worth the effort spending hours tweaking the loop since as archivist says its not really a tight loop as it is
anyway it does work pretty well, and settles to speed within a couple of seconds
vfd has its own loop and will be filtering the analogue in
i'm using the ref frequency input pulse train
as opposed to a 0-10vdc
its pretty quick in terms of changing speed too
i finished off the last of the re-wiring today, so now the machine is back to its operational state
I presume its speed locking to it, is the response in the vfd manual
i had normally open limit switches everywhere when I first built the machine so had to change them all to NC
yeah the VFD locks within a couple of RPM to the pulse train
i measured the rpm with a tacometer, and its near enough
but of course the spindle encoder is giving me a decant quality speed readout
its 360ppr and measures fine at 6000rpm which is the fastest the motor will run
now we need some tapping videos at 6000rpm ;)
thats probably a bit fast for my Z
one sec i will try it in fresh air
on 1mm dia taps in to stainless
not surprisingly it craps out with a following error on Z
but at 1500rpm it is fine
well - I guess we can settle for a 1500rpm rigid tap video.... ;)
rigid snap video ?
archivist: nice pictures btw.
he what..the model engineer ones?
30% of attempts to see will end in fail till I migrate this box
tomp: >>Dave911 are you selling the 40w cnc laser? got there from your posted link (tinyurl-ized now) http://tinyurl.com/yju7qzv
Nope. That isn't me.. :-) I was just asking if the other guys laser was similar to the one I linked....
well, I just mailed the check for the Hardinge
JT-Work: how quaint! how far away is it?
I shot him a price for the machine delivered to my door!
and he took it
hope it makes the trip well. that sounds so easy.
the same one we were looking at a few weeks back
did you go see it, or is it a surprise?
I just went by his feedback and the pictures
If I went that far it would be on my trailer :)
yea, let the games begin in a couple of weeks
was it working until recently, or long dead?
I'm not sure
do you know if it has resolvers or encoders?
by the seat of my pants on this one
Now I have to clear a spot for it while I do the EMC conversion :0
that is the fun part
two minutes ! 170394223207
effing big servos
oh - ebay
2off they went for £128 460mm long motors and the drivers
too big for me
here is a rag-tag bunch http://cgi.ebay.com/4-Large-DC-Brush-Servo-Motors-for-CNC-Applications_W0QQitemZ300359818029QQcategoryZ78197QQcmdZViewItemQQ_trksidZp4340.m263QQ_trkparmsZalgo%3DSI%26its%3DI%252BC%26itu%3DUCI%252BUA%252BFICS%26otn%3D10%26ps%3D63
those are tiny compared
and too far away :(
those yellow ones in devon went for £128? probably a fair price
someone here in town is desperately trying to sell a Land Rover discovery. It's turning into "what can I do with a 3.6L engine that doesn't involve road tax?"
would a start motor last long used as a servo?
>those yellow ones in devon went for £128? probably a fair price
I saw those. That was super cheap.
starter motors probably don't last long in anything
starter motors are not continuous duty at 12v
I'm surprised that someone didn't buy those motors for an electric car conversion
how do you know they didn't?
btw, have you guys considered converting your cars to electric?
batteries are not up to much and too heavy
how do you debug emc itself? you can't do it through gdb :S
Clive Sinclair was on telly the other week. That'll set a few people going
celeron55_: I have an eye out for the perfect electric conversion car ... one of these days I'll do it.
you can get lithium batteries for a car for 8000 euros or something (which is about 150km)
I've been prosecuted for riding an electric bicycle
the only problem is the BMS for the batteries
practically there isn't any on the market
Well I don't but 128 pounds is not far off scrap value for those things. Obviously the demand for those things in the UK is not great. If those were in the US they would have gone for a lot more than that.
they were sitting there for a £10 when archivist mentioned them. maybe he shouldn't :)
>>I've been prosecuted for riding an electric bicycle
Is that a big no-no in the UK??
probably concidered a motorized vehicle and needs to be licenced as such
it was at the time
is there a way to output everything from emc over ssh or something?
yeah metal chain, no pedals
in finland we have a law that allows electric bicycles which go 25km/h at max, only power the motor when the driver turns the pedals and the motor is <250W
Neo_The_User: what are you trying to do?
well my ultimate goal is to solve the overruns
139000 jitter isnt good
oh that's a hardware or bios problem, not an emc problem
the distinction in UK is the drive method. any friction drive (up to and including V belt) is "electrically assisted" not "mechanically propelled"
often video card or video driver
so basically unless i have a PCI video card, i cant do anything... right?
In the US - Indiana, you have have a Moped up to 50cc and a max of 35 mph and not need a license. Go too fast and you can get a ticket. I don't think they would even pay attention to an electric bike since it should be pretty quiet
no, that is not a correct summary
what is your video card and what driver is it using?
I am currently trying to get TTM working with the openchrome driver but I need to fix the source to libwsbm
Dave911: we have that too, though a licence is needed
ok those are three things I don't know about
i am on a CN700 VX700
onboard video? what chipset?
>> friction drive distinction - seems very odd - so you could have a 10 hp engine with a V belt drive and still be legal in the UK?
oh I see VX700 is maybe the driver name
I don't have experience with that one.
[19:28:43] <Neo_The_User> http://pastebin.ca/1645514
dunno about power limits, but they probably exist too
it's 12kW for a "learner motorbike" under 125cc
Neo_The_User: have you checked power save settings in the bios?
Neo_The_User: if it's video related often you can see the latency jump if you move or scroll a window, etc
only when i close firefox for the 30th time or so
ill check the power saving options in the bios. should they be on or off?
turn off everything you can
oh thats easy enough. thanks :)
does your hard disk spin down? that might happen after a long time.
usually once every hour
I've seen machines that freeze while it spins back up
you get one latency jump per hour?? yuck.
i dont pay attention to the hard drive noises. ill pay attention. and yes
I've also had a laptop that would pause when the cpu fan turned on or off
If everyone still uses matrox cards then I've got a 2d accel driver that works in DOS
alright ill be right back
new kernel and bios settings :)
how many threads will EMC use at most when running?
emc is not multithreaded, but uses several processes
wellll I guess AXIS uses threads
there are always at least three processes plus the realtime code loaded into the kernel directly
I'll have to ask the FreeDOS people if they've got SMP support yet
all these latency problems are the same thing the audio software industry went through when they moved from DOS to Windows. It was painful and still not 100% reliable today
not quite what I was looking for. EMC the server company
we're much cooler
EMC the open source cnc software
I see that :)
cradek: servers can be cool too..
we carve servers out of solid granite here (or MDF)
but you need lots of kW in refrigerators :)
I just had a vision of a PC with an oil sump, and a coolant hose spraying the cpu
bah. American books from the 60s cheerfully telling you to buy neat peroxide and saltpetre from the local hardware store
even when I was a kid I couldn't get basic stuff like that
but back then I /was/ making bombs >:)
my dad had a stash when I was a kid
Has anybody compiled rtai with -funroll-loops? latency seems to have dropped down a lot
what is a 'lot'?
it's a VIA C7 cn700 vt8235r, with all the power management turned off, it has onboard graphics but heavy loading doesn't effect latency
a lot meaning more negative max latency numbers than positive since i set that really dangerous flag
was that the only change you made?
-funroll-loops shouldn't be dangerous
we even disabled the network controller with no effect, even hot-unplugging the HD had no effect on the latency on the once per hour or so 120uS jump
it's only going to be used with the EPP and 7i43 with 2 servos for positioning
what kind of servo rates are you planning on?
it acts the same with two different c7 + cn700 + vt8237r boards
120us only works out to ~12 microns of travel at ~100mm/sec
SWPadnos: some other flags too along with some kernel hacks
ok - I'd be curious to see what was done and what the effects were
also, if you see negative numbers, I'd look at the difference between the highest positive and the lowest negative number
I basically just did some cherry picking from the zen source as far as the kernel goes
L84Supper: have you tried using both boards? motor control on a headless, UI on the other
will try digging up a PCI graphics card to see if it's the Chrome-Pro onboard acting up
I believe there's some tuning used to make the user code execute at the right time, so they pre-fire the timer interrupt to take the housekeeping into account
MattyMatt : that's another thing we're going to try
negative numbers (may) mean that the housekeeping took less time, and if you would re-calibrate, all the numbers would go up, since the pre-trigger would be reduced
run it in text mode
and no network
for the latency test?
just monitor, keyboard, and maybe mouse connected
that should eliminate graphics as the problem
will do after I see it jitter again (if ever)
you can use the RTAI latency test
anyone know if the pc2500 boards mentioned on the wiki used the onboard graphics? it's the same chipset combination as these boards
and/or make a HAL file/script that just prints the values every so often, like the RTAI test does
L84Supper: you are at the mercy of the drivers and xorg and gtk and gnome, although using LTS versions of Ubuntu should help prevent that causing problems
we're not tied to these boards , but I'd like to use up the last few left over from other projects
MattyMatt : we're seeing the same issues with ARCHLinux
they may use the same drivers as LTS
do what SWPadnos said, and if that fixes it try it again with network enabled, and if that still works add the 2nd board for the UI if you need it
but it's the same drivers that would be used with the pc2500 used successfully as mentioned on the wiki
heh, we already have a second board running XP for a XP only app that needs to run something else :)
have you tried that app in Wine?
Wine isn't good with low level printer controllers that are not standard
Wine is getting really quite good now. most stuff works
it's odd that it only happens around every hour when run for a several hour period of latency testing
someone could make a port of the UI side of emc to windows 8)
unfortunately doesn't work with Wine since it needs USB, we even worked with SUN to tweak virtualbox, close but still not there
should be easy, gtk is on Windows
i guess it would be useful for a lot of people
we're in the middle of moving everything off of XP, but it won't be ready in time for this project
it would be useful to me. I'm embedding my emc machine and it would be nice to use my XP lappy as the UI
20GB drive not enough to dual boot
it's more of a "what the hell is this once per hour anomaly" before we move to a different board
use it headless, or in text mode, and you've got a permanent solution
:) I love cli
I dont think emc itself though can run headless.
MattyMatt: "if" it's X or GPU causing the glitch
will make sure DMA is off for the tests with ARCH
try running the latency test without X?
SWPandos : 1mS will be the servo rate
ok, in that case 120 uS is OK, but just barely so
Can anyone (Cradek?) confirm if the G33 patch (http://timeguy.com/cradek-files/emc/0001-Improve-initial-threading-synchronization.patch)
is in the 2.3.4 release? The release notes mention "improved initial synchronisation on threading moves" but I thought that last night he said it wasn't in.
SWPandos: if the servo rate is set higher than our 120uS anomaly will it even be an issue?
what do you mean?
setting it to 50us, for example?
L84Supper: You would normally have a base thread much faster than the servo thread
the change in 2.3.4 is not the same as the patch you have there
git master has a totally new algorithm
celeron55 : yes, I understand that if the servo rate is less than the latency there will be problems
Ah. I expect I would be a good guinea-pig for the new algorithm as I need to compile from source anyway to work with my SMP kernel?
depending on what you're doing, having too much hjitter in the servo period can be a problem (it adds phase noise to PID, for example)
my understanding is that the pulses will deform 120us at max if that's the latency, so they would be 1000us-1200us or so then
celeron55_, he's not going to be running a base thread at all
he's using a 7i43
Are velocity calculations etc based on assumed thread rates or measured thread rates?
when using the 7i45 what is actually controlling the the +-10V to the servo amp based on encoder signals?
I though that the DSP/controller in the fpga will do all the fine detail work
andypugh: assumed, as far as I know
sp though/ thought
I'm still looking over he EMC architecture
all we're doing with this is having the servos repeat a move from 0 to 100k and then back again to 0 with a set accel and decel
you could do that with bare hal
the moves by the servos from position A to B will only be for 6 seconds max.
x is ~ 700mm , y is 300mm
cradek : can you make changes to HAL without restarting EMC ?
you can run/stop/modify hal without ever running emc at all
ok, will look this over, just getting this this point :) thanks
daft question, but what is git master called?
ie I guess I want to git checkout <some version name>
cradek : so I may just be able to use the 7i45 with just using HAL?
andypugh: git master is named simply 'master'
git checkout master
At the risk of proving that I am not fit to be let loose with git, is that the 2.3.4 version or 2.4 one?
master is 2.4
Right, lets see how it goes then.
[21:05:16] <micges> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Git
That was the page I was on, but it doesn't seem to tell you how to work out what branches etc are available to checkout.
try: git branch
with git, a checkout includes everything
err, a pull does anyway
then you can check out any branch, which is a local operation
you don't have to choose one or the other when you pull
I seem to have done a pull and rebase, but ./configure is now not working...
andypugh: try ./autogen.sh first
That fails with errors.
line 6: autonconf: command not found
(Except it makes fewer typos)
[21:15:49] <micges> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Installing_EMC2#On_Ubuntu_with_EMC2_package_already_installed
try second apt line
ok, ran in text mode only with GDM, HAL, Fan, dbus disabled and still get the ~120uS glitch
Wierd thing is, it worked last time...
when is 'last time'?
also, what does "configure is now not working" mean?
what does it do?
A couple of weeks ago when I built 2.3.3
what error does configure give you?
installing autoconf seems to have been a complete solution
master require autoconf from about two months
and this is the same chipset and cpu and Jetway J7F2WE1G
L84Supper, any rhyme or reason to it, timing pattern, network activity ... ?
now churning through make
that's is what is wierd, network is off, running without X and it's just random, avg 1 per hour
I sem to recall eliminating SMI as a suspect?
have you tried the latency measurement tool that Michael Buensch (?) posted to the list recently?
is there a way to disable SMI from kernel command line?
no SMI in the factory BIOS, I could run coreboot and see as well
SWPadnos: what measurment tool do yo speak of?
[21:22:48] <SWPadnos> http://marc.info/?l=linux-kernel&m=124469647405168&q=raw
Neo_The_User: Not that I know of, you need to install a kernel module
[21:22:56] <andypugh> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?FixingSMIIssues
that only works on certain Intel chips
probably not on a via chip
(but who knows)
I did say we had eliminated it, I was answering a different question...
Neo_The_User, that baiscally turns off software execution (by halting all the CPUs), and checks the TSC for latency that the OS has no control over
wonder what is different on the two Jetway and the one VIA board tested on the wiki? http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?action=browse&diff=1&id=Latency-Test
there are a thousand configuration registers on those chipsets - who knows what matters
wow. I just got 3 more overuns in the same minute :(
Neo_The_User, in a follow-up email, he mentions that that tester should not be run on an RTAI kernel - use a stock kernel with that patch
yeah, it's just odd that those three boards are fine and the two different here with different BIOS act up
sure - the BIOS configures the NB/SB chips
there may be settings related to video refresh priority, for example
we have another cn700 + vt8237r from ECS, might try next to just see
Looking at that latency page, the 2500nS latency I get with my old dual Xeon server (with SMP, with the SMI patch) looks jolly respectable?
regarding C1E issue, idle=poll kernel option should disable C1 with C1E, so latency should drop.
* Neo_The_User tries it
well without the odd jump we get, the jitter has been reduced to <2uS
so the optimizations are helping with everything else
Pah! Another partial success!
ohh wow. 3 more again in a row
328853 is the highest i ever seen it
EMC2 - 2.4.0~pre ....... insmod: error inserting '/home/andypugh/emc2-dev/rtlib/hal_parport.ko': -1 Unknown symbol in module
I have seen that before
you need to run sudo make modules_install in the kernel and the reboot
Actually, just rebooting probably makes sense. Ooops!
anyone know if the Jetway mini-itx boards were tested with PCI graphics cards?
andypugh: i think your actually missing kernel modules
It was working 30 minutes ago!
Well, I have pulled and compiled a new version since then. You might have a point.
well if you only update vmlinuz, it probes for the legacy modules
unless you have sub version moudle checking thingy on, it wont work on the fly IIRC ;)
As I have never previously heard of a vmlinuz I doubt I did anything to it.
2.4 run-in-place is working on a sim config now.
Whilst fully aware that watching me compile emc is akin to watching a train crash, any suggestions what to do next? I would like to install the 2.4-pre that I just built. Can I do this by ./configuring it without the run-in-place flag, or is the process to debuild the version I have just created? (Or both?)
I am (probably naively) hoping that an installed version will be able to find the kernel modules, but then my previous 2.3.3 build had no problems in that regard
(The proposal to 'make modules_instal' failed with an error related to installing modules with a run-in-place build)
woops, left AGP Fast Write enabled in BIOS, seems to have been the culprit so far
1.4uS, with a jump only to 42uS
probably onboard video, have to dig up a PCI graphics card
OK, I am somewhat puzzled. insmod is failing with hal_parport.ko, but that module is newly compiled along with everything else.
[22:29:38] <andypugh> http://pastebin.ca/1645763
It is trying to load the right module, and it was all built just now on the kernel I am still running. However I really have very little idea what I am doing.
Perhaps it is worth deleting all of emc2-dev and starting again?
thats windows defeatism :)
andypugh: sounds like maybe parport_pc isn't loaded
check if it's there using lsmod
You are right, it isn't there
I am sure it was there before...
Hi, can anybody help me get an axis running in EMC2 with a ULN2003 chip?
andypugh: if you had a regular emc2 install, then it's prevented from loading
I have the circut all set up, I'm just not sure what isn't working right. The motor twitches from now and then...
I would really appreciate the help
Guest998: You have the logic right? high voltage grounds the outputs?
Yeah, I *just* ran the circuit on an Arduino, and have transferred the logic pins to the parallel port
How do I expunge my regular emc install?
Which pins? Some p-port pins are hardware-inverted
Pins 2 and 3 for the x axis.
what does hardware inverted mean?
Guest998 is now known as Hunter890
The actual pin output is the opposite of what you ask for
But 2 and 3 aren't (1 is)
the voltage is the opposite? or the location of the pin
voltage is the opposite
andypugh: you need to check in /etc/modprobe.d/ for a file called emc2 I think
or simply apt-get remove emc2
that will remove the complete installed package
I meausured the voltage coming out of the parallel port on pins 2 and 3, 2 read at about .4 volts and 3 read at 5.0 or so
might as well remove it, it is broken (2.3.4 on an SMP kernel is a non-starter)
get the 2.3.4 sources, then build a new package :P
andypugh: So is there anything else I should be troubleshooting for?
I have built 2.4-pre
Hunter890: Try the voltages with the ULN2003 connected?
I assume this is a dc motor, not a stepper?
No, it's a stepper motor
I checked the voltage with the chip and everything else connected after I inverted pin 2's signal and the motor is now twitching...
So that helped somewhat
Don't you need 4 pins to drive a stepper?
Can you turn the motor by hand?
not with the power hooked up
it's locked up tight
It's working then
I guess I just need to play with the fine tuning settings?
But I think you are just hopping from one step to the next and then back again
How is the motor wired?
what do you mean?
Oops, our logic supply voltage was at 1.2v
I'll fix that and check back
Gosh! apt-get remove is thorough, isn't it? It removed all the dev files too....
Don't you just love linux :)
You say you are using 2 parallel port pins. I don't think you can drive a stepper motor with just two pins and a buffer, unless there is hardware you haven't mentioned.
How many wires on the motor?
I have the parallel port breakout board, a ULN2003 chip, a crapload of wires
There are 5 wires on the motor, unipolare
So, power to the common wire?
I had the same configuration running off of another microcontroller, so it is possible
yes, power to the common
And each of the other legs to a paralle port pin through the ULN2003?
no, the uln chip takes 2 inputs (step, direction) and power in for the motor and converts it to the pulses needed for the stepper
You only mentioned 2 pins, you see
Let me check the datasheet again.
uln is a power driver it does NOT do that
Hunter890, you are mistaken on the uln2003
[23:15:38] <andypugh> http://www.elecfree.com/electronic/uln2003-control-stepper-motor-by-parallel-port/
You can do it, but you need 4 pins and a stepgen running in EMC2
4 pins used which is the only way
Okay, what is a stepgen?
the conversion to phases that a stepper needs
is that something I do in a command line or the EMC2 GUI?
A bit of both. You need to configure it in a HAL file related to your EMC config.
MarkusBec is now known as MarkusBec_away
Easiest would be to pretend you had a step/direction driver and use the stepconf utility to create a config (which you probably already have) then edit the whatevernameyouchose.hal file to use a step type 5, and then wire up the extra pins using the existing HAL lines as a clue
im looking at that file right now
so modify the line stepgen step_type=0,0,0 ?
Yes, to 5,5,5 for three type-5 axes
However, if you have 3 axes you might have a problem, as there are a maximum of 12 output pins on a parallel port
There's gratitude for ye!
I was about to go on at some length about what a bad idea it was..
multiple parallel ports like the dual pci card this morning
yes, bad idea, get him some l298's
gota be dinky motors run off a sprague driver
True, but still no microstepping, unipolar only (so half the torque he could have) etc etc
l298 297 can do a decent job
toshiba like the ST chip has some microstepping (1/16th?) http://www.atnet1000.com/pdf/TA8435H.pdf
Yeah, he could buy a 3-axis set of those on a board from eBay for $40
[23:34:31] <jt-plasma> http://www.linuxcnc.org/docview/html//common_Stepper_Diagnostics.html#r1_3
one 4axis drive with those was included in a micro gantry from taiwan, i never fired it up tho
jt-plasma: i think a of of guys dont have indicators, if that page talked about how to use a 'gauge block', then more people would be able to use it ( substitute 'nice block o hard shiny stuff' for gauge block)
the emc2 file in modprobe.d says "install parport_pc /bin/true" is that good or bad?
zero on top of block , run test, put block under after test
re gage blox: the use sees if is it tighter or looser?
the user sees...
LOL ok tomp
tomp: I used a tape measure on my plasma table :)
[23:40:58] <Neo_The_User> http://pastebin.ca/1645834
any idea how to fix that?
You need to install some modules I think
i copied /bin/lsmod to /sbin/lsmod and it fixed the /sbin/lsmod errors :)
Is that the standard latency test or that clever one mentioned earlier?
its the one that ships with emc
Does the one in stepconf wor?
First page of stepconf
ohh! :) i didnt havent pyxml installed!
is there an ldd for python ( see what a python app requires )
tomp: i think its mostly guessing
Can I install parport_pc by hand?
i think its in the kernel source @ andypugh
It used to be there, I think. Unless I need to recompile my kernel?
ill check give me a sec
(And I have never managed to make a working RTAI kernel)
ill be back in a bit. run make menuconfig and hit / and search for PARPORT_PC Cheers! :)