I also think we should goo-gleisy our friend
Did you see my post about multiphase PWM, I'm planning to add such a module to HM2
no, I didn't, on here?
I think I just missed you
Do you need deadtime between PWM and /PWM?
would be nice, but not absolutely required
Also how do you drive a 4 phase motor? 4 1/2 bridges?
pain in my behind
Well its a pain for my multiphase PWMgen as well because I was planning to have 3x10 bit PWMgens in
one 32 bit register so they get updated atomically
But I suppose if you dont mind the write overhead, it could be identical to the existing PWM gen
but with logic so that the update is only done when the last of (2,3,4) PWM gens is updated
I guess those odd-balls with 4 phase motors could do with 2x 3 phase PWM gens
could do that
I just thought of this today
Eric: are you planning current control or just voltage mode with current limit?
I'm thinking current control
OK, how will you read the currents?
although the only feedback is halls
that's a good question, maybe I should look into limits
position feedback or current feedback?
Is this a spindle or servo?
the lack of a/d is a problem
How is the motor wired (never seen a 4 phase PMSM)
wye has three legs, not four
X maybe ;-)
I'd suspect two independent phases
pretty sure that's right
there are 4 wires
two independent windings would have 4 wires
I did the back emf to verify that there really were 4 phases
4 wires can be 4 windings in an X, two independent windings, or three and a ground (or wye point)
was not happy when I realized there was
so all four are connected to each other (according to ohmmeter)?
yeah, and there are 4 sine waves generated when you spin it
0, 90, 180, and 270 degrees?
hmm, measured with respect to what
three sine waves measured with the 4th wire grounded
how do you get 4?
you can also just ignore ground
if you are hoping to prove that there are 3 phases and 4 wires, you don't want to have one grounded
if you know that there are 3 phases and just want to know the voltages, that works
I was just wondering how you made the measurements
dSpace and a hand drill
scope ground hooked to the (inaccessible) X point, and four channels looking at the four leads wrt the ground, would give four nice sines at 90 deg increments
but with one grounded and scope (or dspace) looking at the other three, the phase relationships would be more complex
we do back emf testing on the 3 phase motors, you really can ignore ground without any problem, just hook up all three leads and spin it. The voltages will vary around ground.
Did you measure the amplitudes? With 4 phase you should get V or V*sqrt(2)
so you hook scope leads (or dspace probes) to the winding terminals and use no ground at all?
depending on which pair you read (I think)
PCW: i did measure the amplitudes and I would have to look at the files
PCW: that sounds right
jmk-robot: don't forget I'm a mechanical engineer :)
adjacent terminals would have 70% voltage, terminals across from each other would have 100%
since I'm taking all 4 voltages at once, the lack of ground has no effect, I can ground to an arbitrary point for noise purposes
you are a mech E
there is no such thing as "a voltage"
there is only the voltage between two points
generally accepted shorthand
generally accepted because the implied 2nd point is ground
but you just said you can ground any point
call the motor terminals A, B, C, D
but if you are measuring the voltage across 4 connected coils, you don't need an absolute ground
you are measuring A with respect to ground, B with respect to ground, etc
the dspace is ground referenced
then you are NOT measuring the voltage across 4 connected coils
because the far end of each coil is inaccessible
the X point is buried down inside the motor
the voltage across the first coil is A-X, 2nd coil B-X, etc, where X is the point you can't access
you are measuring A-G, B-G, C-G, etc, where G is ground
X != G
if you make differential measurements (either in analog, or by subtracting afterwards), you can do (A-G)-(B-G) = A-B, and get the line-to-line voltages
there are 6 of those - A-B, B-C, C-D, D-A, A-C, and B-D
there is no reason to expose the virtual ground though
and it wouldn't give me 4 out of phase sine waves
I'm just trying to understand what you measured
my experience with 3 phase PM synchronous machines says that I can ground one phase lead and measure the voltages from the other two, or measure the floating voltage from all three and get the same answer
when you say "measure", do you mean with a meter, or with a scope?
do you mean with a meter, or with a scope? ;-)
it's an a/d
and you display it as a plot vs. time?
then its a scope
if you calc the rms or mean and display a number, then its a meter
if you insist
I like to call that a time series
scope then ;-)
you're really hurting my head with this floating stuff though
yeah, meters aren't very satisfying
lets talk about the 3 phase case
I know floating is bad
if you ground A, then you can measure B and C wrt ground (also wrt A)
that is two measurements, two time series
if you do a "floating measurement", how many series do you wind up with?
ok, what are they?
in the first case, we had B wrt A, and C wrt A
you'd have to know what the preamp was like in the dspace
is it a differential input?
a couple mins ago, you said that you can gnd one and measure the other two, or do a floating measure on all three, and "get the same answer"
heh - it kind of depends on the excitation now, doesn't it?
phase to phase voltage
if the first case gives 2 time series, and the second gives 3 time series, in what strange world are those the same
I meant with a suitable transformation
you can only do the transformation because you know the coil configuration
otherwise that's an unknown
sure - if you take the B series (B wrt ground) and subtract the A series (A wrt ground), you will get the A wrt B series
4-phase sounds like a strange motor, which is why I'd be questioning you on your measurements
I've only heard of steppers being 4 phase, but that's just me
let me assure you, I was quite disappointed
there are 4 and 5 phase that are common
I think if you let the common point float, it doesn't float at the same frequency
google seems to only know about 2-phase AC or 4-phase steppers
SWPadnos: there are 5 phase steppers
rare these days
sure, I searched for 4 phase ac motor
I know about 5 phase also
4 and 5 phase AC would be rare
4 phase brushless
actually, Vexta still makes a large line of them, and they're fairly common in China
there is some company that is going to save the world with 36 phase brushless
yeah, rare is relative
the more phases you have, the faster you can go
plus harmonics :)
once you get over 3 phases, yer a pervert as far as I'm concerned
but otoh, these are some really nice spindles
the first heartbreak was when the resistance between the 4 leads was identical
How many HP?
they are dremel sized
for pcb routing/engraving
zero, without appropriate drivers
[02:09:08] <SWPLinux> http://www.patentstorm.us/patents/5043641/description.html
I am going to start with the dspace
I don't think I have my data files here
they were afraid to put their name on it
That patent only has 4 switches
believe it or not, I did build a 3 phase pm drive from scratch
yeah, I looked at that patent when I figured out I had a 4 phase motor, I wasn't too impressed
Well it must be pretty low current, a couple of small IGBT modules could drive it,
but it looks like "A Project" ...
looks like it
I need to go understand how trapazoidal 3 phase controls work a little better
If its just a high speed routing spindle why not just commutate it with its
Hall signals and treat it like a BLDC?
it's a minor extension
I'm sure that's what the original drive did
do you have a reference for that?
Torque ripple isn't going to be a problem at 50K RPM
No But you could experiment pretty easy with some 12V or so low power 1/2 bridges
I'm talking at you from my Atom board
but I haven't gotten around to reproducing your steps
k, how is it working?
and unfortunately probably won't for a while
so far, not bad
its been a long and winding road
I started the install sunday evening
detours for everything from making up a cable for that 4-pin 12V connector, to stealing a CD drive from another puter, to finding out that I couldn't burn a CD anyway cause my media is ancient
then attempted USB boot, but the 1G usb stick is dead
yea, been there done that. I must have done 20-30 installs playing with that problem.
put a netboot installer on a 128M usb stick, but had problems fetching packages from the net
finally put the iso on my local webserver and had the installer fetch from there
still have grub/mbr issues - won't boot unless I have the USB stick in
I seem to recall telling you about getting the adapter for the 4 wire 12 V cable. :)
I think it was $2 to $3.
the 6-24V power supply I intend to use on the robot has that
the older ATX supply I'm using now doesn't (and I haven't gotten the other supply yet)
I got the 12-32V PS. It did not have the adapter so it was good I ordered it.
anyway, I just installed the experimental SMP RT kernel
I think I'm gonna ignore the MBR problem for the moment, see how the SMP kernel does
ehj, where did you get your ps?
Just a sec, I intended to order it from mini-box, but they did not have the 4 wire adapter, so I ordered from someone else. Give me a sec and I will look it up.
looks like JMK bought the last atom from newegg, they don't have a listing for the one he bought
The dual core atom at mini-box were listed as out of stock last I checked too.
Looks like they are back in stock at mini-box, see: http://www.mini-box.com/Intel-D945GCLF2-Mini-ITX-Motherboard?sc=8&category=100
seemed like everybody and his brother had the atom when I was looking
everybody decided to build an htpc
or a robot
When I was looking it up to post the information on the boards on the wiki, I noticed that it was out of stock at mini-box.
I like the picopsu, they have an AVR running them
looks like the Goal went back down in price
ok, who was reporting the eth0:avahi problem on the atom?
it was Guest835 I think
oh, I thought it was one of the erics
I have that problem now
using the experimental 18.104.22.168-magma SMP RT kernel
No not me.
what is avahi?
I was OK with the mainstream linuxcnc kernel, 2.6.24-16-rtai
ifconfig shows eth0 and eth0:avahi as two different interfaces, neither one works
I dunno what avahi means, probably an ethernet chipset or something
avahi is zeroconf
[02:54:30] <eric_unterhausen> http://avahi.org/
a 'meta' interface for http://en.wikipedia.org/wiki/Avahi_(software)
so you're saying that it showing up the same time as my network breaks is just a coincidence?
it might actually have to do with it.
that's weird that it shows up in ifconfig
nah; it makes an alias interface to bind to
ps says I have _two_ instances of the avahi daemon, and _two_ of avahi_autoipd
dunno if that's 'usual'
I have a normal DHCP server on my network, why do I need this avahi?
can you see if it's really two processes, or is it perhaps a thread ?
jmk: A google search gets quite a few hits for eth0:avahi and listing a number of problems, especially with DHCP. I don't see exactly what you are describing so far.
just reading the wiki, it doesn't seem like I need it
you don't *need* it.
ehj: yeah, I did the search too, was starting to read
apt-get remove avahi* ;-)
apt-get doesn't have a list command?
justa: my ps-foo is weak, I can tell you that they have consecutive process numbers
eric_unterhausen: 'dpkg -l' lists all installed packages
apt-get install fetches them from configured repositories and 'internally' uses 'dpkg -i' to install them.
jmkasunich: I have 2 avagu-daemon
to search for available packages, use 'apt-cache search somestring'
but no interface for avahi
I see that i have two avahid's running too.. One 'running', the other is a 'chroot helper'; I'm guessing it's a priviledge seperation scheme
well, they're gone now (uninstalled and rebooted)
but my network still doesn't work
worth a try
anything in dmesg about 'etho' ?
did you replace a/the card with a new one perhaps ?
how are you configuring the network ?.. via /etc/network/interfaces ? or via the gnome/KDE network-management-thingy ?
background: this is a new Atom mobo
I installed hardy, the net worked
installed the standard linuxcnc RT kernel, the net worked
installed xubuntu, the net worked
installed an older, "experimental" RT kernel with SMP support, the net doesn't work
I saved dmesg to a file and am rebooting that box back to a working kernel so I can stop switching back and forth
and here I am
jmkasunich@robot:~$ grep eth0 bootlog.smp
[ 3.103942] eth0: RTL8168b/8111b at 0xe001a000, 00:1c:c0:b4:c1:95, IRQ 20
[ 39.678186] r8169: eth0: link down
[ 39.678234] ADDRCONF(NETDEV_UP): eth0: link is not ready
jmkasunich@robot:~$ grep eth0 bootlog.up
[ 19.700767] eth0: RTL8168c/8111c at 0xe000e000, 00:1c:c0:b4:c1:95, XID 3c4000c0 IRQ 11
[ 58.355377] r8169: eth0: link up
[ 58.355396] r8169: eth0: link up
[ 76.573703] eth0: no IPv6 routers present
similar but not identical drivers are being loaded
the uniprocessor kernel is 2.6.24-16-rtai, the non-working SMP one is 22.214.171.124-magma
maybe the .20 has an out-of-date driver for this nic?
hmm.. possible; can't say
RTL's are always a pain, anyhow
yeah, thats why I'm using a kernel that cradek built well over a year ago instead of trying to do my own
but; at least it's not a 'mac-address changing' issue with the /etc/udev/rules.d/70-persistent-netdevices file shifting device-labels around
too much stuff goes on behind the curtains anymore
its nice when it "just works"
but when it doesn't, I have very few clues
*nod*. I'm a linux sys-admin; we do Ubuntu and Debian so it's not too hard to stay ontop of things.
but when it's just 'your tool', then it's harder. Chances are, however, that in this case it might be a driver-release that just doesnt support this chipset (properly)
so if I want to use both cores (and realtime) on this project, I need to build an SMP kernel, using a more up-to-date driver
From what I saw so far; i can't see anything else 'obvious' that it might be.
except perhaps RT-breakage in this driver
something that prevents it working with RT-extensions at all
that seems unlikely, the RT extensions aren't really even loaded yet, just the bare minimum patches in the kernel itself
although anything is possible
copying the working driver from /lib/modules/working to /lib/modules/not-working is a recipe for instant crash, right?
well, instance on reboot
more from dmesg:
jmkasunich@robot:~$ grep r8169 bootlog.smp
[ 3.103069] r8169 Gigabit Ethernet driver 2.2LK loaded
[ 39.678186] r8169: eth0: link down
jmkasunich@robot:~$ grep r8169 bootlog.up
[ 19.700114] r8169 Gigabit Ethernet driver 2.2LK loaded
[ 19.700394] r8169 0000:01:00.0: no MSI. Back to INTx.
[ 58.355377] r8169: eth0: link up
[ 58.355396] r8169: eth0: link up
looks like that driver had problems in the 2.6.20 kernels, supposedly fixed in 2.6.22 and later
that could then be it
decisions decisions - do I try to build a kernel, or do I do this project with only one core?
I'd have to check the changelogs for it.. but; ah well.
Well, why not start with a kernel that works ?.. Compiling a kernel for it later isnt too much of a bother.
if it'll work with one CPU of horse-power.. it should fly with two
the project is an autonomous robot using vision to navigate a maze
two cores = twice as many images, half as much time to move off course between images
approx twice anyway, I'm sure it won't thread perfectly
but even if the 2nd core just let me keep the RT code and the image processing from stepping on each others toes it would help
I'll use the single core RT kernel for now
on the subject of stepper-drivers... any suggestions on good approaches ?.. the 297,298-design ?.. the 6203 ?.. the Allegro a3977 ?.. Discrete FET's ?
depends on your needs
Why not stick in a dirt-cheap Intel card into the PCI slot ?
Intel cards just ... work
the slot is likely to be obstructed
bought this card because it is small. 6.75" square by about 1" thick
Well; i'm getting 3 x 4.1V, 1.2A steppers. Aim to make a small XY table with it; first with DVD-laser head; later with small spindle
the vehicle won't be much bigger than card + battery + laptop hard disk
are those nema23 steppers?
buy geckos and be done with it
geckos are certainly the best performance, but at $120ish a drive, might be overkill for that size motor
jmkasunich: how big is this robot?
jmkasunich: they're likely from an old copier or printer
i have yet to procure them; but the sale-agreement (if you can call it that) has been made.
jmkasunich: IS the hdd IDE or SATA?
40-44 pin conv
jmkasunich: A CF card is bootable and direct pin-to-pin compatable with IDE http://www.dealextreme.com/details.dx/sku.10309
JymmmEMC: I know, that is on the "if I have time" list
CF->IDE is yum
jmkasunich: Since you are not using it long term that is.
jmkasunich: free shipping on that adapter too.
JymmmEMC: to be honest, I'll be pushing to get this working by the deadline
jmkasunich: no EU sales.
if I have some spare time, I'll spend it trying to make an SMP kernel to get better performance, not reinstalling the OS on a flash card
wtf, that isn't very nice
jmkasunich: Not install, image.
i find dealing with US shops as a .EU'er rather tedious at best; impossible at worst
jmkasunich: It takes 12 minutes for 6GB
the xylotex drives are a nice half-way point - less power and performance than geckos, but much less cost
And since I'm capable of etching my own boards (or have them be made for me); i'm looking for a nice design to build.
since they are bipolar chopper drives, they do a lot better than most homebrew drives
There's the boards at 'hobbycnc.com' that look promising, though
if you are good with electronics, you might want to look for the allegro chips that xylotex uses
there is very little on their board besides that chip and some bypass caps, etc
I could probably dig up the number
jmkasunich: yeah; i was looking at the Allegro A3977.. Microstepping driver for 2.5A, 35volt max,
do you know how your motors are wired? unipolar, bipolar, or universal?
thats the one
justa, did you look at pminmo.com?
The A3977's would cost me about $5 a piece
which sounds like a rather great deal .
eric_unterhausen: looking now
heh, if you are equipped (test equipment, knowledge, etc) to use them, should be great
[03:58:44] <jmkasunich> http://pminmo.com/ss3977/ss3977.htm
or you and 9999 friends can do a group buy g540 for $120
got scope, meters, parts; just not quite sure about motor-driving specifics.
oh and got some stabilized lab-supplies and all that.
if you don't have power electronics experience, budget a few chips for smoke/education
interestingly, motor controls may work better on an unregulated dc supply
justa: I use a board based on A3977 (xylotex.com), but I am far from the voltage and current limits. It works well for me.
eric_unterhausen: I wasnt planning to dedicate a lab-supply to the in-system board; but for testing,.. hey.
1.2A steppers would be far from the current limits
geckos are more in yurp though
jepler: *nod* I aim to stay well under what the board can do too.. Just to be sure that I'm not running into weird issues on that side
yeah, 1.2A is comparable to my system. My DC supply is 27V (switching regulator)
4.1V nominal and a 35V max supply is a bit under 10x overdrive, it will work, but not be the fastest possible drive
jmkasunich: flank rise times will be big ?
drive as in motor torque at high speed
using rated voltage on a stepper gives you rated torque at zero speed only, speed drops off very fast because of motor inductance
chopper drives let you apply much higher than rated voltage for better torque at speed
but that's motor-related; not board/driver related though ?
10 to 25x rated is considered good
you will have about 7x or 8x
I think you said your motor voltage is 4.1V?
so you pulse 10-25x more than 'what the motor is supposed to take in sustained mode', and keep the duty-cycle low; right ?
yes, but "you" is the chip
something like the allegro 3977 does that automagically
I'm getting them cheap; second-hand.. etc.
Yes; i figured that. I've been looking through the datasheets of a number of the drivers and some of the 'standard application notes' for a couple of them
I haven't looked at the pminmo site before - under the horrible site layout, I think he tells you everything you need to build your own board, or you can buy bareboards from him
the boards he sells are nice
Yeah; the info looks interesting; the.. design of the site had me .. stunned
but he may not want to ship
justa: where in .eu are you?
since I wouldnt buy it anyhow; 8 euro for a blank PCB that the plans are posted for is something I can easily do myself
especially since it's single-side
and I noticed that the 3977 comes in TSSOP and PLCC ... and i'd really prefer it in a PLCC
it is either very late or very early there
getting late here (11pm)
I need to walk the dog and get some sleep
thanks for the help with the driver issue
jmkasunich: no prob. Hope it was useful
kept me from chasing that avahi thing all over the place
it might yet escape and eat holes in your fence
kg is now known as karlo
ewig nicht mehr in nem channel gewese
mh, 100 Leute angemeldet und keiner da?
sorry, maybe this is an english channel?
karlo: sometimes we understand german too ;)
but usually more people are around on US times
ok, no problem
karlo: just ask if you have questions
karlo: are you from cncecke?
if english is not ok for you then /join #emc-de
btw.. this is really cool: http://www.chippc.com/thin-clients/jack-pc/
where is the info what M62 M63 do ?
what mean synced with motion ?
micges : in the manual, there's a little at chapter 12.10
sorry, custome was here
mh, how works "answe to"?
alex: yes, cncecke
micges: it's supposed to work like this
with the ones synced to motion, they should happen just before the next motion line (so at the beginning)
the ones not synced with motion will be turned on/off just as they reach motion
(you know that motion lines get put on the TP queue, and get executed when they are due..)
karlo: /join #emc-de - dann koennen wir deutsch reden, geht etwas schneller/besser hoffentlich
alex_joni: thanks, this info can be in the user manual
there is too little about M62 63
I think BigJohnT can take care of this
micges_emc: I'm not saying it works like that currently
it should work that way.. if it doesn't then it's probably buggy
the reason is that I fixed it once to work ok, but that was before the TP rewrite (waaaay long ago)
I don't have idea how to test correctness of this
I'm testing now table mill
I'm sure that M64 M65 works ok :)
cradek_ is now known as cradek
since there's more people awake now, I'll ask this again: Anyone here have experience with 'cheap chinese laser engravers' and the type of interface they use ?
justa: cheap chinese in my language means: don't buy it !:)
micges: this is where the problem lies: It has been bought ;)
I have one chinese machine here, after few month only chinese part left is steel table :D
micges: don't worry.. It wasnt me that spent any money on it ; but i'd be liking to use it with less weird software
whats wrong with software now ?
micges: it's called 'Newlydraw'.. Windows (xp) only, USB-dongle... half-chinese
heh I had simmilar one (in english even)
we easy put EMC on it but more problem is how to create gcode for it
you must know many thing about it if you want to put EMC on it
have any documentation ?
justa, you had about the best answer yesterday, use a logic analyser to find what the pins do
I just noticed that in version 2.2.8 the kinematic modules have hal_parms while the newest version in cvs has hal_pins, so would a module compiled with pins work in version 2.2.8?
and the other way around
the whole idea is that we encourage less parameters, and more pins to be used
(that way someone can connect things to them.. you can only setp params)
for kins it's unlikely you want to connect something to the pins definind how the kins work.. but it may be possible
(think a pin for setting the thermal expansion of a screw, then you use a temperature gauge, and connect the temperature value to the kins pin)
ok thank you :)
archivist: true; just wondering if anyone else here had any real experience and/or knowledge about these typical devices.
justa: no worries about asking, but I doubt you'll get a reply from people here
(I mean the reply you're after..)
Perhaps I should ask in some chinese forum ;)
well.. good luck with that
another question: if I try to load a module in halrun and it fails to load so that I cant unload it anymore, neither with unloadrt or modprobe, is there away of geting rid of it without rebooting the pc?
there was an Eric guy doing emc2 & laser around here
Mark_FAPS: fails to load should mean it's not loaded
if I type show it says as state: initialising, and I cant getr rid of it
justa, most of us in here would just rip out the specials and drive with EMC2
Mark_FAPS: did you configure for simulator?
alex_joni: where would I do that? sorry I'm still not shure I quite understood the whole compiling process
archivist: i would (and will , eventually) too.. if it ere my machine. I'm sure that the owner will end up frustrated with the system as it is..But , ah well.
err.. maybe tell me how you do it now?
alex_joni: with sudo comp --install
i thought there was discussion of cheap laser engravers on cnczone
eric_unterhause1: there is; just no real resolution/conclusion. No info about feasibility of leaving the controller intact
and also much wailing and gnashing of teeth
the biggest issue i'd be having is how to interface the laser-PSU to something I can use.
what is the interface?
skunkworks: it's paralel; but it doesnt seem to be 'passive'. There's a good bunch of controllers (40-pin DIP chips) and an EPROM with a firmware-revision label on it.
(on the controller board).
ah - so your going to have to do some digging.
skunkworks: indeed. Just was hoping somebody had any experience with these things. Seems that there's a whole series of 'brands' all selling the same boxes with the same specs, software, etc. The thing that 'worries me' is that they took the trouble of cutting off 5 of the pins off of one 40-pin DIP, aswell as go over the markings with sandpaper..
So.. i think there might be a reason why 'nobody has a clue what these things do'.
[14:13:39] <skunkworks> http://www.cnczone.com/forums/showthread.php?t=72439
skunkworks, how's the servo h-bridge going?
Good - have not had a chance to really test the latest revision other than knowing the logic works and it spins the motor :)
oh dear, well i'm off to bed
oh there you are
was there errors in the previous logic?
i let out some smoke on a new servo controller today :(
I had to add a diode across the input to the comparator for negative voltages.
well it should give me some use for my new smd reflowing solder station
anyway, time for bed
I notice your ip says faps.uni-erlangen.de
I was wondering what you are planning with emc2 at FAPS ;)
I'm a student and I'm supposed to try and get a kinematik for a tripod working in emc2
first as a sim and then maybe in real
tripod should be fairly easy
did they point you towards emc2? or was it something you decided?
yes they said look here a linux cnc programm go and look at it^^
the kinematiks are quite complex though a bit like this maschine : http://www.drives.co.uk/images/news/ABB%20Flexpicker%20first%20gen.jpg
I also got a working matlab kinemtiks which I'm now trying to implement in c for the emc
oh, so not a simple tripod then
I know the flexpicker..
scary fast though :D
ohh yes theres one in the same hall as me, its very fascinating to watch
[16:08:50] <alex_joni> http://www.youtube.com/watch?v=ADLMXMxLRy8
that's 4 dof's .. right?
degrees of freedom
DOF = degree of freedom (freiheitsgrade)
its got 3 motors (joints) so I think that makes it 3 dof, but I'm not too sure, I'm rather happy I don't have to figure out al the maths, but I'm really having a fight with al the matrixes here
I think it has a 4th motor in the middle
that rotates the gripped part
ah yes in the video thats true, our one is used to wind wires for motors so it doesnt need a fourth dof
Mark_FAPS: you can join #emc-de .. da spricht man deutsch
(it's pretty empty though :)
I will help fill it then :)
just have to figure out how this is the first time I use IRC
Mark_FAPS: type /join #emc-de
alex_joni: Mark's english was perfectly fine - why a special channel?
cradek: not necessarely for Mark
he left anyways..
Ive noticed over the years on IRC that the germans have no trouble with english
archivist: only slower in typing
but depends on the language literation level :D
uups, back in emc ???
alex: can I use default from menuconfig rtai?
it should be fine
building rtai is .. maybe 5 minutes
so if something is still missing, we can always do it again
configure: error: HAL patch not applied/enabled in /usr/src/linux
hmm.. is your kernel source in /usr/src/linux?
and /usr/src/linux/.config has IPIPE mentioned in it.. ?
mine says" CONFIG_IPIPE=y
then something is not right
the kernel source you patch must have that
whish category in menuconfig is it?
how did you compile?
menuconfig make .....
did you use only menuconfig? or did you build debian packages?
at the end I build debian package
did you build a headers package?
following the steps I found
Not shure, how I can check?
[16:54:04] <alex_joni> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Debian_Lenny_Compile_RTAI#Installing_kernel
a headers package basicly contains files that are different then the standard kernel source
I have linux-headers-2.6.24-rtai_r1_i386.deb but apt-get will not eat
sudo dpkg -i linux-headers-2.6.24-rtai_r1_i386.deb
now cernel new make?
once it's installed, you can go back to the rtai menuconfig, and point it to /usr/src/linux-headers-2.6.24 (or whatever the folder is called)
no, kernel is ok.. we only need to compile/install rtai against the headers
I dont find the poin to say "there are the headers...
configure: error: HAL patch not applied/enabled in /usr/src/linux
schau nach wo linux-headers-* installiert wurde
dpkg -L linux-headers-2.6.24-rtai
it should be /usr/src/linux-headers-2.6.24-rtai/*
Im new at debian, long years a SuSE freak
yes, it is
I switched maybe 4 years ago..
ok, now you need to tell rtai (in menuconfig) that the linux kernel is in /usr/src/linux-headers-2.6.24-rtai
not in /usr/src/linux
I was long time beta-tester, but since Novell......
yast was something I missed initially when I switched to debian..
now you need to tell rtai (in menuconfig) this is what im searching
now you need to tell rtai (in menuconfig) this is what Im searching
you're on your own.. don't remember where.. but there's an option for kernel source
isn't it called "Adeos" or "Interrupt Pipeline" in the kernel config?
SWPadnos: in menuconfig yes, in the config file it's CONFIG_IPIPE
but he's doing the "rtai" menuconfig now
different one :)
oh, *that* menuconfig :)
sorry - just having my first sips of coffee now
donald83: the rtai menuconfig has lots of levels..
but I'm pretty sure the kernel source was the first option
menuconfig ask for linux-source tree and installation directory, no mor chance to put in a path
where it asks for linux-source tree you need to put "/usr/src/linux-headers..."
"Part is an aluminum frame 3.5x1.4x.25. Profiled out of .375 stock, plan was to turn over and face off the back and have the part fall out."
uh, yeah, that'll work
configure: error: No Linux kernel tree in /usr/src/linux-headers
you need the full path /usr/src/linux-headers-2.6.24-rtai
configure: error: HAL patch not applied/enabled in /usr/src/linux-headers-2.6.24-rtai
donald83: sorry.. you probably missed something..
[17:14:52] <alex_joni> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Debian_Lenny_Compile_RTAI
that instruction is very good.. try to follow it exactly
ok, se you later, thanks
see you later :)
you think, my kernel is ok? or I start wit it?
if test -r $LINUX_KERNELSRC/include/linux/adeos.h; then
donald83: it should be ok
elif test -r $LINUX_KERNELSRC/include/linux/ipipe.h; then
but do as jepler says
look for /usr/src/linux-headers-2.6.24-rtai/include/linux/ipipe.h
it is checking for one of these files inside the linux-headers directory. If these don't exist, then the problem was probably in building the kernel
building or packaging
probably the second one, ipipe.h
the kernel he installed and booted is 2.6.24-rtai (in uname -a)
so that part looks ok
(patch and all).. but the headers deb might be wrong..
find / -name adeos.h didnt found it
it's probably ipipe.h
also not found
adeos was the older name.. still used up to 2.6.1x or so
then there's something completely foobar
did you look in /usr/src ?
if you don't have that file anywhere, including in the location where you built the kernel, then you built the kernel incorrectly
I try find /
donald83: then probably the kernel is not ok either
that file should have been created within the kernel source when you applied the patch from rtai to the kernel
[17:26:20] <jepler> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Debian_Lenny_Compile_RTAI#Applying_RTAI_patches
arghh..... now I run in the same problem with etch..... there is no kernel-source 2.6.22 and no rtai for kernel 2.6.24....
sure there is for 2.6.24
3.6.2 or 3.6-cv works ok for 2.6.24
the patch is in the x86 folder though
not in i386
-rw-r--r-- 1 root root 154680 2008-04-11 22:47 hal-linux-126.96.36.199-i386-1.3-04.patch
-rw-r--r-- 1 root root 154680 2008-04-11 22:47 hal-linux-2.4.36-i386-1.3-04.patch
-rw-r--r-- 1 root root 227117 2007-02-10 23:42 hal-linux-2.6.19-i386-1.7-01.patch
-rw-r--r-- 1 root root 249524 2008-02-13 14:18 hal-linux-188.8.131.52-i386-1.12-03.patch
-rw-r--r-- 1 root root 248825 2008-01-05 16:30 hal-linux-2.6.22-i386-1.10-12.patch
-rw-r--r-- 1 root root 252799 2008-02-13 14:18 hal-linux-2.6.23-i386-1.12-03.patch
starting with 2.6.24 or so i386 and x86_64 have been merged into x86
there is only one patch for both platforms
then wiki is fault
so you want rtai-3.6-cv/base/arch/x86/patches/
patch -p1 < ../modules/rtai/base/arch/i386/patches/hal-linux-2.6.22-i386-1.10-12.patch
donald83: it's a wiki.. if it's wrong, it should get fixed
the wiki describes using 2.6.22
if you use 2.6.24 then the instructions will be different
"NOTE: Debian way is not applicable for now because linux-source-2.6.22 was deleted from the repository and linux-source-2.6.26 which is in the repository, does not work with RTAI 3.6. For now, use the Debian way with the snapshot repository or the Classic way to get kernel source."
he got the source from snapshot.debian.net
live CD rules
archivist: in the early days, everyone had to do this individually to run emc...
I can imagine
donald83: I'd use 07
heh. why make 07 if 05 works perfectly? :)
maybe 05 doesn't work perfectly then?
huh. could be
I'd use 07 :)
then the question is .. why package 05 in the tar.gz
maybe th 07 has issues aswell :D
I'd use 09
07 was the last one
[18:19:41] <alex_joni> http://download.gna.org/adeos/patches/v2.6/x86/older/
then I'll wait
then you've been waiting since 22-May-2008
I'm not holding my breath, just waiting :)
SWPadnos, see http://pastebin.com/d20f9770b
the screensaver is probably set to lock the session after some time, and then you need to enter your password again to unlock
I'd disable any screensavers or auto-logout functionality
but that's good that removing xrandr fixed the vesa problem :)
ok it also shuts down evry thing thats running
not if you had an EMC2 error window there
we can shut down the screen savoer
I'm hoping that you didn't try to run both types of test at the same time ...
ok - phew :)
I remmeber why the xorg.conf shows a viewsonic VA721 Thats because that is the monitor I was using when I installed Ubuntu and emc. I later procured a viewsonic VE510s and began using that. Should I now do sudo dpkg-reconfigure -phigh xserver-xorg?
To straighten out the config for the different monitor?
if it works, leave it alone :)
the name doesn't matter and there are no mode lines in your config, so the available modes should be auto-detected
Checked and under Systems,Preferences, Sreen saver, I find that it is not activated. So something else must be knockin g it back to the log on screen
The default 1024 x 768 screen resolution seems to be fine
Should I be concerned that t I get knocked out of that latency test?
I've never had that happen before
Under System, Preferences, Power Management it is already set to never (sleep) and do nothing.
evening all, are there any pci parport experts here? I'm having real troubles with a new pci dual parport card that I just bougth of ebay
pjm__ is now known as pjm
pjm: if you ask your question, maybe you'll get an answer if someone knows it. the problem with asking for an expert is that one doesn't know whether they are an expert, since they don't know if they know the answer, since they don't know the question. I think that's why often, nobody says anything when faced with a question like that.
thanks i will ask
first: a PCI parport works perfectly under windows, when booting linux and lookign in lspci, the mem addresses are showing as 'disabled' , insmod'ing doesnt seem to help. http://pjm.dyndns.org/parport.txt
are you trying to use these with emc? is emc installed (and therefor parport_pc kernel modules is disabled)?
emc is installed yeah, i tried these base addresses directly with EMC but of course it errors on start up
since it cannot find the ports
set the BIOS to "non-plug+play OS"
did you loadrt probe_parport?
the bios is set to non-plug and play, and i did the 'reset config data' so it assignes relevant io space to the pci cards etc
loadrt hal_parport cfg="0x378 0xb040 in"
setp parport.0.reset-time 500
is what I have
that PCI ID should have been supported since sometime in 2000 (according to this: http://cyberelk.net/tim/parport/archive/current/3496.html)
ahh ok thanks
so it should work.. hmm
does probe_parport print anything in dmesg?
ok it shows nothing in dmesg when I start emc
it starts and works with the onboard parport which controls the steppers etc
it for some reason cannot see the PCI parports
probe_parport configures isa pnp ports; it does nothing for pci ports
* cradek <- obviously not an expert
can't you run lspci -v and get the actual port address?
but, I think probe_parport does always print something in dmesg
* skunkworks should read back..
reading the source code, I'd expect it to print either "parport: PnPBIOS parport detected" or "PROBE_PARPORT: no PnPBIOS parports were detected"
loading parport_pc may enable the card
then unload parport_pc
ok let me give that a whirl
take a look at lspci after loading parport_pc, and again after unloading it
ok it reports [ 1619.121010] parport 0xb048: You gave this address, but there is probably no parallel port there! for each of the base addresses
and yet Region 0: I/O ports at a040 [disabled] [size=8] <- still showing as disabled, this is what I think is the problem
the addresses btw ignore, as I've since run the 'reset config data' so they have moved
you shouldnt have to specify any port addresses with parport_pc
hmm, this isnt the EMC is suspected when looking for mass storage, right?
THANKS FOLKS!! I knew u guys would know the answer!! its working
I had to modprobe parport_pc which found the card / parports then rmmod ppdev, lp, parport_pc
ok so now that is working I will add the next parport card
ra3vat is now known as dimas_
no power here since 10pm yesterday :( good thing I have a generator and gas heat
sleet then ice then sleet then ice then snow!
I can't even get my 4w drive truck up my driveway!
sounds about like here BigJohnT
we got 3 inches of snow, half inch of ice, then about 8 more inches of snow
our total is about 6-8 inches
I'm in Poplar Bluff MO here
.. i gotta make a motor coupler and get my machine running again, i got a tool pot grinder i'm going to try and convert into a new spinle, for real cutting!
(it will at least be better than my dremel, HA!)
tool post grinder*
you doing pcb or milling metal?
i did a buncha wood carvings and a tiny bit of engraving on aluminum/chrome
**UNTIL.... my Z axis motor coupler crapped out!
I gotta do a wood badge sign for my partner on my mill... I did one with my plasma but after all it is a wood badge so he wants one from wood
I hope they get the power back soon
at least with a tool post grinder you can get some good rpm I bet
is the emc2 ch?
whoa, back to the future
my system's been off for nine months (moved) and the buffer has my last... mcFlew
ok, tmalsh, powered up and ran synaptic update, now i only get rtai exceptions?
update from what to what?
ok then you have your base period set too low and now you are being told about it