tomp is now known as tomp3
if emc can be pushed to update the servos at 5khz... well, a few of the 'high speed/hard milling' controls are claiming 0.5mS servo update periods!
thats pretty cool for emc, even at the 'nominal' 1khz, thats 1mS emc vs 0.5mS big time control.
I bet 5kHz is easy with the pci cards
dunno if you can do it easily and still have linux running though
I'm pretty sure I was doing 1khz with an ISA card
its not so much the card update that concerns me, its running a task on the host at that rate and still having it actually happen somewhere near the time its sposed to rather than going in fits and starts
1Khz is already pretty fast for an interrupt type thing in an X86 type machine
Valen: guess you haven't tried it then? it works great.
you tried it at 5Khz?
If they did the thing I said and put the servo loop in the FPGA you could have 20Khz without a problem ;->
many emc2 users generate step pulses at 50-75kHz interrupt period
by many, I mean almost all
5kHz is nothing at all
are those steps generated by the PC or is it hardware though?
so the CPU is flipping bits?
turn step line on, turn step line off => one step
doing the servo loop is a little prettier
I am downloading emc at the moment
ETA 53 minutes
and I am getting some linear scales here in an hour or 2
the big cnc guys have been using proprietary hardware so it's not very fast
You don't need to be paticularly fast doing it in hardware
eric_unterhausen: it's true that EMC was doing 1kHz servo updates on ISA hardware 10 years ago
pretty sure they aren't using hardware
In hardware you could have it run at the encoder rate
yes but that's not necessarily a win
the closer you get to the encoder rate, the more quantization you get
cheaper to use a pc
"did not get an encoder count this time" => "so how fast are we going then?" => "zero, I guess"
encoders are asynchronous, probably want to run at 5khz or something like that
I was thinking run it in reverse
"oh we just got an encoder pulse"
I'm sure the ideal servo cycle rate is "it depends" but clearly faster is not always better
how long was it since we got the last one?
I was thinking that control theory suggests a constant sample rate :)
ahh that means our velocity is X and our position is U
our target X and Y are A and B, so adjust something
eric_unterhausen: that's sure the normal way, but maybe you could do it Valen's way too
pid is more complex (well, everything is more complex)
you still run a PID loop, you just base it on a different input
there's have to be a benefit to offset the new complexity - not sure what it would be
faster loop control i spose
yeah, researchers went nuts with that sort of control in the past
Its actually not any more complex, in terms of hardware implementation its actually easier in many respects
you would then have to sample a clock
measuring and resetting a counter is really easy
much easier than meshing synchronous and non synchronous events in a standard setup
pid doesn't work anymore
you just need to keep a history of past "events"
you still have a constant axis, its just position rather than time
f=ma doesn't work anymore :)
it does, you just need to thin a little more to work out what a is
t is a variable rather than a constant
I did, you are solving a problem that is uninteresting
i mean the controller needs to do more work
well I think it's interesting to think about, but it is definitely a solution looking for a problem
If your software is properly written for PC you need to account for the fact that your time samples arent perfect anyway
they are close enough
the latency and jitter in latency means that your 1Khz isnt actually 1ms from pulse to pulse
I control a fatigue machine with a pc, it's open loop unstable
but it's self correcting
if its close enough then whats the benifit to running at 5khz?
cradek, in the sense that you crush the specimen?
he means the 1ms is correcting
its always going to be around 1ms
a pid loop with imperfect but close time slices is self-correcting
(on average though its going to be 1.00 something ms)
the "best" way would be to integrate a variable time into the PID variables
the pc makers seem to finally have gotten their act together, this pc has 6 microseconds latency
what is it?
I forget :)
dual core amd something something
i want some nice machines for my mill and lathe conversion
the P3s I use almost always have < 10us
I'm going to try some via min-itx boards i have laying about
Valen: yeah just try various machines from your junk pile, some will work well, others not so well
or should i use the old dual Xeon i have sitting around ;->
I haven't tested all my p3s, but the one I have attached to my mill has almost 40microseconds
why not try it.
is there a SMP kernel for EMC2 floating about?
an early P4 is going to suck, intel never should have released that dog
only down side to the xeon is it needs a really good 600W supply to run it (and it sounds like a jumbo jet)
it pops cheap 400W supplies after about a week
Valen: you can build one. several people have but they're less likely to run on any old machine. they're more custom than the uniprocessor one we use.
I wish someone would make me feel better about the atom
I could use the space
whats bad about it?
I don't trust it
I am going to be sticking the liveCD into an atom board today as well
seems like it should stink at real-time
cheap, integrated video
There is a SMP build as of the last couple of days, for the atom.
oh hey know any sources for old DC servo motors?
they seem to all be gearmotors atm :-<
can't unbolt the gearbox?
probably not if you want some torque
wrt the atom, the onboard video is intel, so open source drivers, that is the main thing
that is good news
should give good performance really, they are one of the best supported drivers available
[03:55:58] <geo01005> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?EMC_With_Custom_Kernel
somebody else was running a SMP one and said they had "ok" if not awesome results on an atom
JMK set it up
an atom board is ~$100 anyway so just try it
He ran emc2 on an atom? 270? 280? or one of the new 5xx series?
worst case you got a nice media player front end ;->
I think 330.
I have seen the 330 around for like $65
$100 Au ;->
My son bought me nettop based on the Atom which I'm using now. I just installed Ubuntu 9.04 on it
Well, better $100 Au than $100 Pb, or $100 Hg
Nothing wrong with gold :)
Au meaning australian ;-P
330 is 81$ from newegg
that's the one JMK has
Valen : I know. I'm teasing you.
I want to know how well it'll work on an eee 701 with the USB controller
they are ~$300 Au
can you even get the newer atoms?
Is there any reason why I can't use Ubuntu 9.04 with emc2, if I take pains to go through the torturous process of applying the rtai patches?
no reason at all
except for library issues, if any
Look at me, two weeks ago, I didn't know a linux kernel from the end of my ass... now I've learned how to compile the kernel and apply patches. Imagine what I'll learn next week.
writing kernel modules?
Now if I can figure out how to make my videos upload to YouTube, I'll be set!
the 330 ITX motherboard in in database on the wiki
It has good Latency numbers
eric_unterhausen : Only if kernel modules can be written in python, basic, COBOL or forth.
I know, but I don't believe those numbers :)
eric_unterhausen : Or C
And of course, they are basically C code .. but I'm 54, and frankly I have no interest in writing kernel drivers.
I'd rather be maintaining the machines, darnit.
How, in Linux, can I tell what processor a machine has?
uname -a would be a good start
Is there something in the gui like "system" in Windows?
eric_unterhausen : I assume at the console?
more than you ever wanted to know about your CPU
Martyn, is that an emc install?
model name: Intel(R) Core(TM)2 CPU 4400 @ 2.00GHz
is what it says about my current desktop
Ah, this machine has an atom 330 with 1M of cache
that is a nice processor, I think.
lightweight but pretty decent
pretty large cache
my son installed Leopard on it, and it works well
yeah they did that to help with power consumption i believe
it feels peppier than a Mac Mini
shouldn't do i don't think?
how do they determine the bogomips?
unless this one is a solid state disk
BogoMips (from "bogus" and MIPS) is an unscientific measurement of CPU speed made by the Linux kernel when it boots, to calibrate an internal busy-loop. An oft-quoted definition of the term is "the number of million times per second a processor can do absolutely nothing."
[04:10:43] <Valen> http://en.wikipedia.org/wiki/BogoMips#How_are_BogoMIPS_computed.3F
The Dell that's powering the tormach is a Dell Precision T7500
Two 2.5Ghz xeon processors, but I need to install more RAM on that machine
I was wondering what a bogomip was...
we always install matlab and run bench to see which computer is faster
whats the latency like on that dell?
Valen : Almost nonexistent
which translates to a number of?
Valen : I got something a couple times when trying to run the latency test .. a "segmentation fault"
just so i can compate
i was talking about servo update rates of 5khs, no steps, and by servo update i mean wth am i, wth should i be, get the new Voltahe to the amp ( 5000 times a second )
my son had to do something to slow down the computer
and i was told emc could do it
tho not by a dev
lol a divide by 0 might cause a seg fault
tomp3, I don't think there is any problem
can you spec a different update rate easily?
try it and see what happens
eric. me neither
max jitter 9248
Hmm i know you can have issues with harmonics and normal servo systems, I wonder if my "no fixed time" system would suffer from that
max interval 31446
i spose an encoder update rate might hit that harmonic frequency still :-<
the other two are max interval 1.0ms 1003151
and 1.0ms jitter 7170
that looks pretty good i thinks
compared to http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Latency-Test
none of the processors on that page are Xeons, are they?
can you move the hamonics on a machine? like add mass? and how to do harmonics tests? i remember some graphs and code that varied a sinus fed to drive, from slow to fast while it watched for the harmonic area.
a freind has a mill in australia, that gets a problem mid sped range ( artifacts in mold surface being hard milled )
sine sweep should do it
yeh, but does 'bode plot' ring a bell?
there are sine sweeps that have constant amplitude, I'm forgetting the name
and the bode plot told yo a characteristic about your machine, in general, for emd, you wanted the harmonic to be above 200hz hopefully above 400hz ( meant your machine was stiff )
yes constant amplitude, frquency varied
this is a system id technique
I suppose I could go read my thesis, but what's the fun in that
looks like this http://www.danahermotion.com/website/common/download/document/Controls_Toolkit_single_pgs.pdf
that's an obfuscated plot if I've ever seen one
The BodeTool is designed to help the machine builder analyze and compensate for
ok, the nyquist plot identifies it as a pole
bode tells you if it's a pole too
depends on what the phase is doing
wah! freebie !http://www.ece.utah.edu/~bodson/acscr/
book on adaptive controls
can you run steppers with encoders?
I love control theory, but it was kinda killed as a research topic
Valen, there is a thread about that at cnczone
probably in the wiki
You can run steppers with encoders, basically with FOC
needs high resolution encoder and high (say 30 KHz) sample rate
(running them as 2 phase AC Servo)
we are putting linear scales on the axis of the mill
and using a "0" backlash setup with acme screws
steppers seem more available 2nd hand in the right sort of size
linear scales don't match too well to the resolution of motors
so i'm wanting to drive them like you would a DC servo
i think you need to check the logs, i seem to recall, yes you can use steppers and encoders, but also, no you cant correct , just fault
yeah thanks thats what i thaught
did you check cnczone?
We've driven Steppers as full servo, its just a 2 phase AC servo
did it cog at low speeds?
No, smooth as glass
then it no longer steps
( i come from very very very low speed arena, edm, like rpm means revs prer month
No its Field Oriented Control
No real advantage over 3 phase PM AC servo
did you steal that idea from Marriss?
except maybe more torque at low speeds do the the
large number of poles
supposed to be a smiley, sorry
Our SoftDMC firmware as a closed loop stepper option
peter c? like microstepping is replaced by true sinus?
Well as good as the table/DAC resolution,,,
i see, any dac would be steppy to some small degree, just get the bits up and you wont notice the jaggies
You can also run 3 phase PM motors as microstepped step motors
never seen one, but i can understand the drive technique
haha that last pdf originated at Copley ( motor drivers )
I have a german one made to run from 320V B+
about 5" diamter by 6" long
We are going to try and drive it from our 8I20
ever run a moog/pegasus valve? got some hydraulics here , its really a 20ma device, but analog
Its pretty easy to generate good sine wave currents at slow speeds with step motors
it all goes to hell when you get faster though (Were messing with a stepmotor controller ATM)
Ive heard of those, is that for running a double acting cylinder?
maybe you dont need the bt rez at higher speeds ( gear shift 'down' to less bits)
yes it is an amplifier itself, a 'spool' valve moves and lets high presure flow into left or right side of dbl cylinder
you get loads of torque, cheap ( usually messy )
but its lke a voice coil
make that gear shift up, as i imagine the lsb's are chucked
Just need to keep electrical current phase aligned with rotor,
resolution will determine how many steps of drive change you will get
with 1/4 step (90 electrical degrees)
What are frequency response does the valve have?
cool, sorry, back to orcad, board designer screwed me again, got a db15 wired in mirror
Oops.. One of our first PC/104 cards we made, we reversed one of the bus connectors
sold some of the reversed cards to a customer tha made the same mistake on their motherboard :-)
tomp3, hey.. moog valves... i designed a servo positioning board that used temposonic linear transducers and mooog valves for position bandmills in sawmills
they were nice, but $2500 each for the valve 20 years ago
we had some high speed positioners that moved slabs of wood sideways at 60ips.... very exciting
LawrenceG: cool, (looking up temposonic scales ) . i used moog on elox's and pegasus on euro brand edm's
the sensors mounted inside the rod of the hydraulic cylinder and used a sonar like technology to read the position of a ring magnet
cool, looking up freq response of moogs ( less response of hyd cyl ;)
opps.... not sure what dcc chat was supposed to do?
i got booted from dcc chat by local sys
so went back here
we used a 1ms servo cycle...
SE3 servo cartridge 170Hz http://www.moog.com/home/search/-/q_servo%20valve%20electronics/start_20/
most of the systems had 1000's of pounds of mass, so valve response wasnt really an issue
right, these were big casting too, but had to be quick, not a high top end, but quick
seems to me we used a 50mhz clock on the sensor and got about 0.001" resolution in 1ms
argh, get back to orcad! soorry bye
pcw_home, you cant use softdmc with emc though can you?
DCC char is direct client to client
it bypasses the IRC network
hmmmmm... i now have an dso on the Xstep and Xdir lines
seems the pulses are negative logic with 3.4 usec length and with a space of 84.6 used inbetween
after a direction change there are 30 usec delay before the next pulse
hm.. i can get it low as 27.1usec
and in this one case i could properly capture 30msec delay after a pulse to a dirchange
yeah.. it moves
syl bf20 with emc 2.3.0
anybody used encoders on a parallel port?
i wish.. only steppers and flying blind
just looking at a DRO type application first up
Valen, parport only has a limited speed
some are using it though
what sort of speed can it work at?
think it'd work to read the linear scale without dropping counts?
it cant handle counts to is limited to the period emc is running at
just got my linear encoders and checked them on the scope, they are quadrature encoded so thats nice
Valen: you can probably expect 20kHz count rate or so
just tried to boot the livecd on my atom, no joy :-<
Valen: what's the error?
if you get some error messages like these: "ata1.00: revalidation failed (errno=-5)"
then try adding 'irqpoll' to the boot parameters
steve_stallings is now known as steves_logging
pcw_home_: just chaging my status from work to home ;)
see you later
Joyhandle question. Is there a reason why it was not included in 2.3, or in fact does not appear to be in trunk either?
tomp is now known as tomp3
hmm.. what kind of gcode iss ngc?
Nist G-Code isnt it?
RS274NGC is the standard its based on
[13:35:18] <archivist> http://www.linuxcnc.org/handbook/RS274NGC_3/RS274NGC_3.html
archivist are there converters which take what comes out of CAMExpert (the cam module for qcad)
[13:41:21] <archivist> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Cam
is our main collection of routines etc
if CAMExpert is as good as qcad, it should be good
i have had nothing but success with qcad
any ideas how to use step, iges, dxf or so? heekscad (3d cad) does that
i've found most of the free/oss CAM conversion stuff to be pretty bad
roh, see http://www.ribbonsoft.com/camexpert/manual_reference/chapter05.html#1
to set a suitable setting for emc
ngc is Next Generation Controller . was this before or after the startrek series of similar name ;)
proctor didnt act like a trekky
don't be silly, that's NCC
steves_logging is now known as steve_stallings
ok, i had to look that up, apparently that was the letters in the ships name
i meant start trek the next generation
Star Trek TNG
You guys would suck at being a Trekkie =)
cradek: What DONT you like about the thing you use to edit/write your website?
I had to dink with it a lot to get it to display correctly on old IEs. fortunately that is in the past and now even IE does CSS.
OK, whats the name of the creature that kept multiplying so much that it basically took up ever space on the ship?
other than that disturbing memory, it's fine
cradek: Anything else? Feature?
sorry, I don't understand. are you asking if it is missing a feature I want?
cradek: Sure, or anything I wish it did/didnt do this or that
I know many people think they want to allow blog comments, and it doesn't do that as-written. I don't want that, though, so it's not a missing feature for me.
cradek: LOL, same here =)
so it's just fine... simple to use - I spend virtually zero time maintaining it.
cradek: What about lets say a thumbnail previewer for inserting pics?
yeah it does not have any kind of photo album browsing thing.
cradek: Simple wysiwyg editor?
ha ha ha
(if I may speak for cradek, that is a "do not want or need" feature)
JymmmEMC: not important to me. it uses markup which is the natural way for me to write text for a website.
for an odd web interface http://www.stevenson-engineers.co.uk/files/index.php
cradek: It doesn't index and provide site-navigation, does it?
no, there's no kind of index - only search
it is a very basic system
it sounds like you want something a lot less basic
cradek: ok, thanks. If you think of any basic functionality it should have, let me know.
JymmmEMC: image processing, wysiwyg, and so on
are you looking for a project to improve, or trying to choose software for a new site?
jepler: No, I'm writing a straight forward intuitive one. Including the basic functiality to bring up any website online and "just add content" in a matter of minutes.
oh ok, I see what you're asking then.
Knowledge of html implied
what I want probably doesn't match what most people want...
I've looked at a dozen CMS'es thus far, and they all go WAY WAY overboard and you just can't get anything done.
yeah, aether isn't from that world.
(probably why nobody uses it)
never use an available wheel when you can invent a new one
cradek: Well, it is in part. Instead of file based, db based.
you'd better be prepared to learn a lot about wheels if the one you invent is to be any good
steve_stallings is now known as steves_logging
I invented my own wheel for archive type data :)
jepler: I've been thinking about this for 5+ years, now it's just time to DO IT!
hide features and allow clicky additions in a control panel
JymmmEMC: it might be too late for simple CMSs. The new way is to give your content for free to a few big popular sites who use it to make money for themselves.
they do this in exchange for the simplicity they present to users who don't want to figure out how to set up a CMS.
Im just getting payback on my archive site, it can be done
cradek: This isn't for revenue generating, just site management.
maybe I wasn't clear - I meant that I think very few people want a simple CMS now.
cradek: All the kiddies just scrap content from other sites.
cradek: Ah, well this isn't for them =)
it better have no drupal tendencies
either way, I want to see what you come up with - sounds like you have some of the same goals I do.
we get far to many sql questions from that
My own REAL issue is the design for use with CSS layout - I hate mucking with css crap across all browser for layout and such.
I made a "cms"
its reeeal simple
theres 2 php pages and a database
database != simple
lol its really simple its already there as part of the LAMP stack
sure it is
i'm hungry for lunches
www.vapourforge.com for an example
I made my "search" fast enough that it can be used for links
I'm on my 4th machine running the live cd
best one i have found has a jitter of 7000 or so and about the same added as the max values
JymmmEMC, learn inverted indexes for that trick (sphinx etc)
although this latest one with an nvidia card has about 6000 with jumps when it accesses the cd rom
I wonder if there are any benifits to be had running EMC from a ram disk
mainly to negate pesky disk accesses when your doing something important
it's not significant due to rtai
it is if its half way through a disk access when it wants to do some RT stuff
's why there is jitter and latency in the first place
have you found that disk IO makes a large difference in latency on your machine?
on the one i just tested it did
running glxgears with the open source driver it was sitting at about 7000 jitter
md5sum /dev/scd0 jumped it up to 2400
do you mean 24000?
wonder if disk does that too, or if it's just the cd.
that machine has no physical disk
RTAI prolly cant interrupt a disk access
stuff in the hardware device driver layer
Network access also does it. Latency tests done via a VNC connection will be much larger.
Network latency may be due to PCI bus blockage by bus master (DMA) transfers
from Ethernet/WIFI (a 1K packet block transfer would block for ~10 uSec)
Systems with PCIE Ethernet should fare better
you could almost say that the only reason for rtai is to help with disk access latencies
on a standard pc, they can add up to 500ms
you will see some small latency spike with disk access with rtai, but not on the order of a half second
SMP should help with all those things
not with DMA causing latencies
latency-test shouldn't be bothered by DMA should it?
still only one bus!
that is pretty much outside the CPU's hands
kinda the point
no, that's one of the sources of latency
the other is interrupt latency
the point i was making is on a non DMA machine i was seeing large spikes caused by disk access
what is a non-DMA machine?
cpu cant access the code/memory during dma
sorry non SMP
have you built an smp kernel?
not yet but it would be interesting to test it out on the installed system
I was benchmarking all the spare PC's i have to find the best one to use
doing a md5sum of the CD caused large spikes
I've never really done much testing on a SMP rt kernel
CPU intensive tasks alone seemed to have little impact
SWP keeps saying that the best thing to do is put the second processor in a busy loop
cpu intensive doesn't mess with the interrupts
interrupts should be per CPU i would imagine
the real problem is allowing an interrupt handler to disable interrupts
so the second core could do other stuff while the first is otherwise occupied
you'd think so, doesn't seem to be the case in general
but I haven't tested a rt kernel
find out when i get an on disk installed
is there something missing from the generic ubuntu RT kernel?
there is no smp
kernel, it's single processor
jaunty is SMP RT
Valen: I wasn't following the conversation but saw your reference to smp. FYI, I just built EMC for smp to run on the Intel Atom 330. The real time tasks run on the second processor, everything else on the first. I am getting latency test results of 3617 and 1976 for servo and base threads respectively.
damn thats awesome
did you have to do anything to get it to boot?, the emc live cd barfed on mine
you are running 8.04 on it?
I had to build a custom kernel, then rebuild emc. In boot/grub/menu.lst you must add isolcpus=1 to the end of the "kernel" line.
i meant the install
Valen, get the mailing lists / me was just looking for the mention when ehj said
or did you start from an 8.04.2 base then install EMC
the livecd is giving me udev errors and dropping to a busybox shell
There is a full tutorial, just a sec and I will post the link.
I didn't look into it much more though
Valen, sounds like a drive problem
its the same drive i used in 4 other computers
[17:56:43] <Jon_geo01005> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?EMC_With_Custom_Kernel
then it's a cable problem :)
beat you :)
slow fingers. :)
and the system itself boots 8.10 fine
you broke it while removing it from the other comp :)
that was the 2nd one i did
its a known problem between 8.04 and atom
8.04.2 solved it
emc doesnt ship an 8.04.2 based CD though
henc me askin ehj how he did his install
did he do some magic to get the cd to work? or use a later install then put emc onto it
well, did he?????
i dunno he has told me how to rebuild a kernel
It was built from source, then generated packages. Move the packages to the target machine and install.
i am *not* talking about the kernel
ehj, what distribution did you start with?
I have downloaded the emc-8.04.iso burnt it to disk and stuck it in a drive
works in 4 other PC's
not in the atom
Oh, there is another tutorial on that, just a sec.
The kernel compile is based on this tutorial:
[18:01:00] <ehj> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Debian_Lenny_Compile_RTAI
ehj: which distribution did you start with?
For the kernel, Lenny.
The packages plus the config file for the kernel build an be found here:
[18:02:33] <ehj> http://www.linuxcnc.org/experimental/smp/
FYI, per the tutorial, I used three partitions to do the whole thing. Partition 1 - Debian Lenny to build the kernel. Partition 2 - Ubuntu 8.04 to build EMC. Partition 3 - Xubuntu 8.04 for installing and running EMC.
The config file then gets copied to ~/ubuntu-hardy/debian/binary-custom.d/ for building EMC, etc. on partition 2.
I don't see why you would use lenny to build a kernel for hardy
Because there was a tutorial for it. :)
Actually as I recall, I started on Ubuntu and ran into some problem, then found the Lenny tutorial, and other than changing the version numbers of the current versions, it worked exactly as written.
Hi all, my first message in the EMC IRC.... wheeeeeee
Oh.. guest... sorry... Rainer it is...
you can do a /nick whatyournameis
unless someone the said nick already
Anyways... got a RTAI message that my kernel is tainted... huh
Massive unexpected realtime errors... like 3456985 are there docs I can browse?
why do you with say steppers for example need a realtime kernel ?
as you issue each step from the computer anyways, it shouldnt mater if it comes second 1 or second 6 ... you ? :)
because its a mechanical system!
and things are moving and must happen at time T
but a step is kinda instantanious anyways ... even tho moving two motors to make a diagonal line they still would make something looking like a stair right? :)
as long as the steps comes in the right order it shouldnt do anything ?
no its not instantaneous
ok then a slong as you do not step faster then the steps can move :P
at speed its a smooth motion
stepper are worse in that there is an acceleration curve to be followed
eg tyou cannot go from 0 to full speed in 0 secs
not on failblog but good enough http://news.bbc.co.uk/1/hi/england/north_yorkshire/8017535.stm
Are there any hardware interfaces that aren't ISA or parport based? Is there anything PCI based that can be used while still closing the loop in software?
sure look at the mesanet cards http://www.mesanet.com/
I thought they only did PC/104.
alex_joni: thanks. I missed the wiki.
a bit easier to have it gathered in one place, than search the web for each of those
hard to find page alex_joni was just searching the wiki for it search fails for mesa, M5120 two terms it should find
archivist: it's linked on the first page
under getting started
search should still not fail though
archivist: the mesa board is actually called '5i20'
and it comes up if you search for that
same for mesa
I found a page with M5120 and that page did not get returned in the search
m5i20 is an older driver for the 5i20 board
archivist: otoh, the wiki is a community maintained place, so feel free to adjust it how you think would be better
Can the 5120 handle 3 fully interpolated servo axes, quadrature encoders, and limits?
probably 8 or 10
3 surely ;)
Not bad for $200.
Pretty damn good actually.
you might still want some extra hardware
like isolation for the I/Os
and some conversion from the PWM to analog
(depending of the drives you have)
I got a 7i31 (24 LEDs for debugging only)
a 7i33 - 4 axis analog servo interface
and a 7i37 (16 in + 8 out - isolated I/O)
if you get those, there is a T version now with terminals on them, which saves you some trouble
I think a 5i20+7i33T+7i37T would be fit for your specs
(at around 340$ though)
I wish I had seen Mesa's PCI boards when I had some extra cash laying around. I hate this economy.
I've been wanting to dig into some VHDL or Verilog but haven't had the chance.
Are their code blocks open source?
GPL, BSD or the like?
[21:30:16] <alex_joni> http://cvs.linuxcnc.org/cvs/emc2/src/hal/drivers/mesa-hostmot2/firmware/src/
all the firmware for the mesa cards is in our CVS and is released under the GPL just like the rest of EMC2
-- * GNU General Public License (GPL), version 2.0 or later
-- * 3-clause BSD License
I think I'm going to have to get myself one.
voxadam: there are other options out there too
we're not trying to sell you one, btw :)
(even if I think it's a very good solution together with emc2)
alex_joni your recommending using mesa???? looking for insight from the elite
coden4life: well, I try to stay neutral
I find using just the parallel port for now is fine for me but I am thinking more robust machines and I/O is the key
is usb the answer
usb certainly not ;)
Not exactly deterministic.
voxadam: there are some "gimmicks" that do some motion control over USB
1394 maybe but USB...
but they add so much limitations, that it isn't worth the trouble
I'm sure they're very gimicky.
rt-eth maybe :)
pc based cnc with more than 4 axis
voxadam: seen EtherCAT?
rt-eth is another posibility.
the I/O is my hangup
coden4life: get more parports
that's probably one of the cheapest solutions
the emc2 hal driver supports 8 ports iirc
I seem to remember rtLinux supporting rt-eternet
voxadam: RTAI does too
[21:36:40] <jepler> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Emc2HardwareDesign
but the problem is the device on the other end
I'd take parport over usb any day.
I understand parallel that would be the route I would want but is it time I get into the new age lol
2 ports / PCI board seems ok (for about 10-15$)
parallel is still available via PCI addin.
It's a more stable option than USB.
although I don't know of anyone who actually tried one with emc2
What kind of PC hardware would EMC need to close the loop in software for four axis and handle the UI locally? Pentium? Atom 330?
I think some users used atom PCs lately
for servo thread only, you don't need much of a PC
a P3 should do it
the userspace stuff like the GUI is much heavier than the realtime stuff
depends on the GUI used
So a dual core hyperthreaded atom should do okay.
consider AXIS, which likes to make tens of thousands (if not more!) of opengl primitives for every redraw..
but if you're using AXIS, then probably it depends on how big the files you use
I would start with a cast off machine (p3 like alex_joni says) and then evaluate if a faster machine is required, rather than buying a very new and possibly expensive machine
voxadam: hyperthreading is usually evil for RT
anyway, bbl -- the weekend begins
alex_joni: I hadn't thought of that. I'm sure you're right.
voxadam: also, consider grabbing the LiveCD and taking it to the PC store
to try out a couple PCs :)
Or I could just spend $80 and buy an Atom based board.
[21:43:38] <voxadam> http://www.newegg.com/Product/Product.aspx?Item=N82E16813121360
voxadam: although I think you can switch off HT in some cases from the BIOS
That shouldn't be too much of a problem.
voxadam: hang on, looking for a link
[21:44:53] <alex_joni> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Latency-Test
there's a D945GCLF2 in there
with Atom 330
The "noht" option on the boot prompt will disable HT.
voxadam: nice chatting with you, I'll head to bed now
good night all
Thanks for the help.
I'm running EMC 2.2.8, and Mesa's 7I43, 7I33, and one of their digital I/O cards. Trying to get X,Y,and Z to work for right now. Y was no problem, X was backwards (is now fixed), but Z is an opportunity.
It goes in the right direction, homes in the right direction, but stops short of homing, or give me a "home sw inactive before start of backoff move" error.
I'm not sure what that means. If I reset and try to move after that, I get a hm2/hm2-7i43.0.encoder.00 dT <=0, how can this be?" error - lots of them.
I give up, how can this be?
halscope the Z home switch input -- is it noisy?
if it is, maybe you need software debounce
I have watched it and didn't see any transitions while it was moving.
set halscope to trigger on the rising edge of the home signal, then initiate homing
Can I wiki and find the debounce?
[22:28:07] <jepler> http://linuxcnc.org/docs/html/man/man9/debounce.9.html
OK. THanks, I'll check it out.
this is a config file (not for mesa, though) which uses debounce on some switch inputs: http://git.unpythonic.net/view?p=zenbot.git;a=blob_plain;f=pluto.hal;hb=master
just search through it for 'debounce', you'll get the hang of it
but first it'd be nice to know whether it is bounce, or whether it's a red herring
do you need me to talk you through using halscope to view that home switch input?
No, I've done enough of that to understand it.
I'll make another trek out to the barn tonight with the added knowledge, and see if I can see anything.
looks like the default debounce delay isn't documented. apparently it's 5 (or 5ms if you haven't changed the default servo period)
is it going past the switch before it can stop? the switch should stay triggered all the way to the end of the axis's travel
what kind of switch is it and what actuates it?
bizarre, I thought the "how can this be?" error was fixed in 2.2.8. are you sure that's the version you are using?
jsr__: he said 2.2.8
you're "loosing" it
It's on a Bridgeport BTC-1. It's a big, old "microswitch" brand switch.
a cam pokes it?
some switches has a lot of bounce
true, they do
It's driven directly by a ramped dog on the spindle.
What's the "how can this be?:" error?
I saw that error once when trying to do a fast move on the tormach... but it went away on the second take.
Martyn: it's a debugging message reported by some versions of the hostmot2 driver. I think it says "hm2/hm2-7i43.0.encoder.00 dT <=0, how can this be?"
I am going by what the dmesg said when I started up to get the version number.
jepler : The tormach doens't use the hotmot2 driver though... right?
it's a bug in the velocity code of the mesa driver. It is fixed now but I'm obviously not sure in which releases exactly
jsr__: does that homing dog keep the home switch actuated all the way to the end of axis travel (limit switch)?
Martyn: I suspect you saw some other message, then
Yes, the dog keeps that switch actuated all the way up past the limit switch, and up to the physical limit.
ok, good, just checking - some machines can go past the home switch, which is bad
I agree you should look for a loose wire and then check for bounce with halscope
cradek: the message was removed between 2.2.8 and 2.3.0.
Author: Sebastian Kuzminsky <email@example.com>
Date: Sat Jan 31 06:10:35 2009 +0000
I think this fixes the encoder velocity bug
jepler: yeah I was just going to say that too. darn.
OK. I was sort of avoiding changing versions while I am in the middle of a retrofit. I'll just live with that one until I get all the axis running.
jsr__: I don't think this specific problem will affect your system
optos and the servo-loop speed polling input should make you fairly immune to regular bounce. maybe some wiring is actually loose, or the switch is pretty bad.
I got bounce big enough to see in the servo loop, but that's with cheap microswitches
oh really, ok
either way, halscope will tell the true story
I checked on both ends, but mice could have gotten hungry in the middle somewhere.
I have a feeling these were of the most rugged switches in their day.
rated for a million operations or more
I'm pretty sure at least.
Maybe this explains why the LSI-11 on it kept crashing, too.
jepler: 4 debounce instances were loaded in your example, but it appears only the first one was added to a thread?
Thanks, everyone. Bye.