oh -- I see that alien hal style is seb's, not goslowjimbo's
I was about to criticise it
net the_expected_position_to_which_the_motor.0.should-move axis.0.pos-cmd pid.0.command
steves_logging is now known as steve_stallings
cradek: I was able to run the 5axis simulator. How do gat it to control my machine?
what specifically do you mean? how far are you?
cradek: I am adding two axes to my mill, and I am trying to create a config for it.
Looking to use a parallel port interface.
ok I see. unless you have the same configuration of rotaries as an existing demo, you will need to write your kinematics code.
I am good at code, but kinematics? no.
do you have a picture of the machine configuration? maybe we can brainstorm.
I thought I just specify machine joint distances etc
don't be too scared - kinematics is usually just a little bit of simple trig once you get the idea
well, it depends whether the kins are already written.
there are many configurations of 5 axis machines.
Doing something like your frist XYZBC machine. Using a Taig with B rotating around the Y axis miving up and down on the Z column
C is just a rotary table sitting on the base.
then you can use maxkins directly
when you build it, you should arrange B so it rotates around where the tip of a typical-length tool is.
that way, you don't waste a huge amount of X motion to keep the tool tip in the same place when B rotates
I think that is the only tricky part of that machine configuration
That's tough. I will need a very long arm of the rotating axis
do you understand the problem I described with wasted X motion?
Yes, but I am not sure how much it wastes. If the tool tim is 3 inches past the center of rotation of B. how much x is wasted?
depends how far B turns
if it turns +- 90, clearly 6" of X is used
I see. That slows it down also. I have 11.5 inch travel on the X so that should leave me a good 4+ inches for the size of the part rhght?
Icould accept that if I could keep the arms short, ant the machine stiff.
yes I guess so
what is the diameter of the C rotary? remember your workspace is pretty much round in XY because of that
I will move it up at some poing to speed things up and remove the extra x travel.
yes. I plan to have my part round in the XY plane
it's surprising how small the working volume gets when you start thinking about this
then you stick a drill in a drill chuck, and you might not be able to reach anymore
Yep. The Y and Z travel are small. Too small for the toolchanger. I try using collets directly on the spindle and short arbors.
No drill chucks for me.
under emc/configs I found max. Is that the one?
max5 I think
no, it's max5, must not be in cvs. one minute.
[03:35:58] <cradek> http://timeguy.com/cradek-files/emc/max5.tar.gz
Got it. Thanks.
I see maxkins is in cvs, so it may just run as-is
but it might be stale - I haven't used it for a while.
Loaded max5 into 2.4.0~pre. Any obvious veriables I need to change?
I have no idea :-)
you should look through it and then test.
Cool. Thanks Chris!
welcome, hope you get it all going.
With the amount of time I am willing to spend, It will work. Eventually :-)
steve_stallings is now known as steves_logging
alex_joni: are you on and alive this morning (5:20 here)
I checked my 'atom' installation last night and it definitely does not run with the emc2 rtai kernal thing.
please refer to ubuntu bug 239602 and ubuntu foum 869967. much better explanation than i can provide.
after the ubuntu 8.04.2 install, i did the apt-get ....emc2-install.sh and everything went well as far as i could tell.
The atom board?
rebooted, picked the proper build and error.
I have installed the live cd mine here - seems to work great.
well - 17us latency - but never higher.
mine will not go unless it's 8.04.1 or newer.
put in 8.04.2 (what i could find to download) and it installs sweet.
[12:26:45] <skunkworks> http://imagebin.ca/img/z0vrLSPy.jpg
was chatting with alex_joni yesterday morning.
yeah, that's what i am doing also. wonderful. mesa board and all.
I guess all I have tested on it is the emc built livecd's - even dapper
did you update the bios?
ssd drive? or hdd?
I have not - right out of the box.
i am going for low cost, low noise, low power, ......
80g hdd now, but want ssd
intell D945GCLF board. 69.99 Fry's
is that the single core?
don't knoe for sure.
I think that is the single core
This is what I bought http://www.newegg.com/Product/Product.aspx?Item=N82E16813121359
cool. i see promise in these as they are really inexpensive.
yes - I am acutally using one as a media pc. working good.
(using the s-video out to a tv)
is there possible to write keyboard driver for hal ?
you mean like hal_input, which treats a keyboard like a bunch of switches?
yes, it's possible to write that :)
but more realtime than hal_input
well, it may be possible to do that with a PS/2 port, but not USB
situation: loaded axis with 700.000 lines of gcode, keyboard delays are in seconds :| (tested)
I don't know how you could tell the OS/BIOS to ignore a PS/2 keyboard though
I don't think that's the fault of the keyboard driver
but delay can be removed when keyb stuff will be in rt
you can plug in a second keyboard (make sure it's a different brand), then load the hal_input driver for that keyboard
oh I didn't know that
so hal_input is generic input driver ?
put a + sign before the name, and hal_input will prevent the OS from seeing anything on the device
it supports most any device that the Linux input layer supports
(joysticks, keyboards, joypads, mice ...)
there are some issues with some devices though, so test first
plug in a SpaceNavigator for a really good time :)
[13:03:56] <SWPadnos> http://www.3dconnexion.com/3dmouse/spacenavigator.php
heh, I have few of simmilar stuff :)
if I have time I'll test it :)
There isn't any access to the led's on the mesa 5120 board - is there? (I don't see any link)\
hmmm. there used to be
the m5i20 driver had a single byte that would set the LED pattern
which may have been changed to a set of bits at some point
hey skunkworks, you have a couple of the D945GCLF2 boards, right?
ok. I may have some questions for you as the day goes on
I think I'll be ordering 15 or so today :)
[13:48:12] <skunkworks> http://pastebin.ca/1413253
I should design a small case that lets you use a single PCI card and a mini ITX motherboard, but which isn't the size of a medium pizza box
wow - 15?
I have xp on one - emc/hardy on another - and one is in the box.
do you have access to any of them over the course of the day?
I have access to the hardy one for sure. The one in the box could be hooked up also
ok, cool. I may ask you to run some benchmarks or power checks or something, if that's OK
not a problem
oh - actually there is one thing I'd like to see if you have a minute :)
you may have to talk very slow..
(it starts already ;) )
there are 4 USB ports on the back and two additional headers, right?
if you have a USB2 device, like a camera or external hard drive, I'd love to know where those ports connect internally
the cipset has only one USB2 controller and 4 USB 1.1 controllers, which are internally connected through a crossbar switch
I have a both.
So you want me to connect the camera to each port and give you some sort of feedback?
the way to test this is to run lspci -v once before plugging anything in, and then run lspci (no -v) when the device is plugged into each socket
actually, the lspci -v may not be necessary
err. of course I meant to say lsusb every time I said lspci :)
[13:56:31] <skunkhardy> http://pastebin.ca/1413258
nothing plugged in
did you need the -v?
no, but I should have been saying lsusb instead of lspci :)
heh - hold on
I have a benchmark that several people ran before - I'll get some instructions together if you'd like to run it (image processing - easy stuff :) )
[13:58:55] <skunkhardy> http://pastebin.ca/1413261
jmk did this on his board, but I can't seem to find the numbers
oh right, this is the RTAI kernel, isn't it
yes - I could download the livecd from ubuntu
or install the generic kernel and reboot :)
if you like. (I am here all day)
also - I don't have any headers plugged into the motherboard at the moment.
sure - is that easy?
jepler: how do you get networking to the beagle?
[14:01:14] <SWPadnos> http://pastebin.ca/1413264
lsusb output on jaunty, with all the information you could ever need :)
I think there may be a PCI ID database that you're missing (or something like that)
in any case, lsusb -v should tell us which of those root hubs (BUS numbers) is the 2.0 controller
and once that's known, a simple lsusb will tell which hub the device is plugged into
(the bus number of the device will be the hub it's connected to)
[14:03:53] <skunkhardy> http://pastebin.ca/1413268
.ca is almost an rss feed now
do I need to sudo?
bus 5 is the USB2 one
you can see that on line 5: bcdUSB 2.00
the others are all 1.10
upper left rear
[14:06:17] <skunkhardy> http://pastebin.ca/1413271
cradek: I bought a HDMI-HDMI cable, but that cable would work if you had a monitor with DVI input. I think dit has them for less (6' about $9)
[14:06:55] <jepler> http://ditcorp.com/itemdesc.asp?ic=CBLGENHDMDVI%2D06&eq=&Tp= http://ditcorp.com/itemdesc.asp?ic=CBLGENHDMI%2D06&eq=&Tp=
hmmm. is that quickcam a USB2 device?
cradek: you're thinking about your media player project?
my thumbdrive is usb 2 - I will use that.
I'm a little worried about getting the old software to build on it. The alternative is to start from scratch I guess.
maybe do lsusb -v, just for the heck of it
cradek: what software does it need?
cradek: festival, mpg123?
hacked festival, hacked mpg123
how much can they have changed in the meantime? both projects must be just about finish
do you have source trees? I can try building them for you
I'll have to really dig around for them... or maybe firing up the old machine, if it'll boot, would be easier.
or connect the hard drive to another machine
looks like my interface is currently 5 input bits
are there 5 bits somewhere that I can read directly with an inb?
oh, I didn't think it had one
cradek: there are 18(?) GPIOs on the expansion header but they are 1.8V so you have to build level converting hardware
oh - sorry. forgot you were talking about the beagleboard :)
and here I was looking up cross-compilation information for it
I haven't tested the board I made for that purpose, but here's an 8-line equivalent to the part I used: http://focus.ti.com/docs/prod/folders/print/txb0108.html
did you want this? http://pastebin.ca/1413274
you hook the low voltage signals and low vcc to one side, the high voltage signals and high vcc to the other, and you're done
seems like I could just use five zeners
skunkworks, yeah - that's the kind of thins I'm looking for
you could find an LED with a ~1.6V forward voltage, then you get indicators for free
SWPadnos: I didn't look - is the camera 2.0?
it says 1.10, but I don't know if that's because of the camera or the bus
it's better to check with a known USB2 device
it does look depressing though - if the back panel connectors are connected to USB1 hubs, that's a problem for me
according to the datasheet, acceptable high-level inputs are 1.17v-2.0v for vdds=1.8v (vdds*.65 to vdds+.3)
heh - this is plugged intot he same port
[14:19:47] <skunkhardy> http://pastebin.ca/index.php
I wonder if the crossbar is automatic
I thought it was set up by hardware pins/BIOS
do you want me to run my cruser around to each port and give you a lsusb -v?
hmm, it has rs232...
and the motherboard ones, if you have a suitable adapter for them
RI, DCD, CTS, DSR, RD = 5
not sure if RD can be read directly though
RD can't be read directly
however, the serial port is probably also GPIO, if you set it up right
cradek: it only has RX and TX
not much of a cheese shop then, is it
here's an example use of GPIO pins: http://linuxjunk.blogspot.com/2009/01/beagleboard-gpio-input-driverless.html
since the OMAP is a big microcontroller, you should be able to use most any pin for GPIO (assuming it's brought to a connector of some sort, with minimal hardware in between)
(maybe not the best example!)
SWPadnos: lower left http://pastebin.ca/1413286
is this going to help - or do I need to plug in 8 devices? ;)
if you have 8 device, that would also be interesting :)
another example of gpio access, through /sys/class/gpio: http://blog.makezine.com/archive/2009/02/blinking_leds_with_the_beagle_board.html
Give me a few - I need to find motherboard usb plugs.
if I plugged in a usb parport cable and let the kernel do its normal thing, could I read the five inputs with inb?
skunkworks, thanks a bunch
cradek: no, not with inb
this device would let you get some 3.3v or 5v I/Os on USB, but again you can't read them with inb(): http://www.sparkfun.com/commerce/product_info.php?products_id=718
some kind of ioctls, I think
for a parport cable, see the ioctls in /usr/include/linux/ppdev.h
unsigned char bits; ioctl(/dev/parport0, PPRDATA, &bits)
something like that, there's probably more setup to do
ah, that's simple
as long as I can do it from C without too much bizarreity
I have a USB parallel cable if you want one
I'm 95% sure those ioctls will work on usb parports
that would let me just plug it in. (the less I have to mess with the avr interface board, the better)
do you (any of you) want some parts from DigiKey that aren't worth spending the $5 handling fee on?
(since you're talking about IO level shifting and stuff ...)
that you'd bring to workshop?
I'm placing an order today or tomorrow, and will probably place another one early next week
let me see how much I'd be spending to get the parts to make proper ribbon cables for my project
how many conductors?
I need a 28-28, a 26-26, and a 50-50, one each
all relatively short, I imagine
so you could take a couple of those old useless IDE or floppy cables and use them for the 26 and 28
oh, 0.1 spacing or 2mm?
the one of the 26.01
argh, what was that
one of the 26p connectors is shrouded, so that one has to be "right"
initially I'll just use http://www.sparkfun.com/commerce/product_info.php?products_id=8430
for the 26 and 28p connectors, and a 50p cable that's just too long
there are only 7 and 6 lines used on the 28 and 26 connectors
anyone know if is it planned to add stepgen velocity mode for hostmot2 ?
I can shorten the 50 for you if you don't feel up to it.
hmmm. cool. I may need some of those :)
oh I don't have a lot of spares
I'll try to remember to toss them in the stuff I bring though, if you think they'd be handy
they're expensive at DigiKey (I think)
oh, I was thinking of getting 20 or 30 to have around
SWPadnos: seb lives right by sparkfun, and I think he's even offered to bring stuff from there to fest
SWPadnos: if you want some, maybe you should contact him
oh, excellent. thanks
and tell him I need a https://www.sparkfun.com/commerce/product_info.php?products_id=8548
heh. new phone? :)
SWPadnos: newish phone, yes
not useful for communication, but it beats bringing a wall wart as a charger
(I have one at the office and last time I went on a trip I forgot to pack it.. packed a mini b by mistake)
I replaced the battery in my phone, so I get to keep using a mini-USB cable for that
actually that was an interesting experiment. both my phone and my wifes had maybe 10 minutes talk time, most of a day standby (unless you talk - you get more like 1-2 minutes at the end of the day), trouble making calls, dropped calls, etc
all those problems were fixed with new batteries
even a problem where you'd dial, and the phone would sit there for about 30 seconds, then say "call failed"
and the total cost was $30 or so, including shipping (not Motorola brand, of course)
I left my phone on for most of a week, and talked for about 1.5-2 hours on it on a single charge
(with the new battery, of course)
huh, a 28p idc connector is $10, while a 26p idc connector is <$2. I guess volume does make a difference.
interesting - they don't even carry 28-conductor ribbon cable
which may explain why the connectors are so expensive (they also appear to be individual crimp style - the photos are not like "normal" ribbon connectors)
oh I didn't look at that
where does the 28-pin cable go?
SWPadnos: from beagleboard to level-converter board
no shrouds, so I'll probably use a floppy cable (34p) there
order me 4 of D89126-0131HK
that gets me one cable plus two mistakes, which is probably about par for the course
damn. DigiKey is pretty expensive for cable assemblies: http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=SAM8235-ND
it's only $1.24 more for hte 6" model
SWPadnos: all 4 in back filled
[15:16:40] <skunkhardy> http://pastebin.ca/1413339
got a kill-a-watt meter handy? :)
(or another way of measuring power draw)
heh - no. I could measure current at the power supply though. I could borrow the fluke from the shop.
are you using a normal AT supply, or some DC-DC thing?
normal cheap at power supply
hmmm. ok - nevermind the power calcs :)
did you install a generic (SMP) kernel?
whatever is on the live cd
actually, I think Eric Johnson (?) made a SMP RTAI kernel for 8.04 - you could give that a try at some point
hmmm. I think I need to put the latest one on linuxcnc.org, come to think of it
ok. Hmm - either the plug that I have to hook into the motherboard isn't standard or I am having problems with it seeing devices plugged into the motherboard headers
no biggie. the 4 external ports are all capable of USB2, which is the main info I was looking for
ok - good :)
so, install dcraw (you may have to enable "universe" to see it), download http://www.thothsystems.com/images/dsc_0291.nef
and tell me the output of "time dcraw -q 3 dsc_0291.nef" :)
before lunch! ;)
thanks again. I owe you a Red Bull or something ;)
not a problem. Least I can do.
SWPadnos: what's a good dcraw score?
hmmm. good question
[15:30:01] <skunkhardy> http://pastebin.ca/1413349
ok, that's about the same as my A64 X2 4850e system
phenom 9600, so you should be able to do 4 in 5 seconds
cool picture of the 7i31 ;)
ok, those numbers are very close to jmk's numbers (I finally found them)
btw - dcraw was already installed.
if you do it > /dev/null, it may be a little faster, but the user number will be about the same
heh - good. At least I am not screwing things up ;)
nope, not yet ;)
I did a multicore test by doing "time (dcraw ... & dcraw ... & wait)
add more dcraws for more cores
ends up about the same wall time, which is nice (and expected)
is this for your camera project?
I sure hope so - I have to get it working for a demo in Dallas by the end of next week :)
he was also using the single core kernal - right? would smp help?
so we can use it for an event at the end of the month
SMP helps in that you can process two images at once
luckily these computers will only be connected to 4 cameras each, so the total time should be relatively quick
which interface does your primary hard drive use? (IDE or SATA)
It is hooked into sata
ok. could you time a large-ish file transfer from a USB-connected hard disk?
just cat $BIG_USB_FILE > /dev/null
err - time cat ... :)
SWPadnos: how do I do it if the file name is ESTIDATA.MDB?
cat: .MDB: No such file or directory
time cat /path/to/USB/drive/ESTIDATA.MDB > /dev/null
use tab completion - it helps :)
how big is that file?
close to 1gb
I will get the info when it is done
nice. thanks again
should be a minute or two I guess
[15:45:28] <skunkhardy> http://pastebin.ca/1413353
hmmm. no, theoretical 17 seconds, but could be limited by the actual read speed of the drive
ok, not bad - 50% fill factor or thereabouts
recompiling dcraw with -mtune=native trims a little bit of time off (consistently just under 5s wall, instead of just over)
huh - does native check the CPU and set things automatically?
the actual application will be imagemacick actually, so I'll have to see what can be done there
4 in parallel take about 5.5s wall
cool. is that to /dev/null?
no, to the same default output file
could you try it to /dev/null?
it makes less of a difference than run-to-run differences
oh. cache effects I guess
interesting - mtune=native will optimize for the native CPU, but will not cause architecture-specific instructions to be used
you need to add -march for that
so -mtune=native -march=native could give a better result
using sched_setaffinity to make each instance run on a different core also doesn't make enough difference to measure
oh. thanks for checking
-march=native also doesn't make a difference that can be measured
you could always try intel's icc too
I hear there's a version that's free to use
hmmm - haven't seen that (the free-ish one)
I'll leave those optmizations until after I have the computers, the networking, the control software, and the storage systems figured out :)
SWPadnos: thanks for looking for preassembled cables, but a 2" cable is too small anyway -- to avoid putting a twist in the cable, I end up having to pass it under the led board.
yeah - you probably need 6-8" or so
hmmm. are you referring to the cable that's installed in the photo?
strange. it shouldn't need a twist
mounted that way, it doesn't
I originally had the 7i31 rotated 180 degrees
yah, you';d still need a 6-8" cable if you wanted to put a twist in it :)
the other option is to swap the 7i43 and 7i31 so that the cable passes over the 7i43 which I don't need to see
yeah, but you may need the second connector some day
BJT-Work is now known as BJT-Gone-Fishing
this sounds like an AXIS bug: http://www.linuxcnc.org/component/option,com_kunena/Itemid,20/func,view/catid,10/id,208/lang,en/#212
well, probably both
or all three
I thought incremental/continous is a GUI thing, not task atm
well there's an incremental jog nml message. if getting it in teleop mode makes it go forever, that's not a gui issue (except the gui shouldn't have sent it either)
* alex_joni agrees on all three :D
otoh, I don't plan on working on this, except if we take on designing it from the beginning
I think at fest jmkasunich is going to work on the free planner vs teleop.
he has a new design, and part of it is even done (isolation of the free planner)
it would be great if you can come up with a design for jogging (both joint and world) at fest
afaik, the design is done
I'm ok to do the SMOP, but haven't seen the design :)
and I started without on joints_axes, and got mudded in
I think: have a free mode planner including jog wheel for each joint and each axis
so you get continuous, incremental, and wheel jogs in both joint and world
jmkasunich would remember the details - he even has a drawing of it I think
jog wheel for axis will probably be a bit harder to do
but it's probably similar to incremental jogging
conceptually it doesn't seem hard: send the new axis position through kins, set the new joint commanded positions accordingly
hm, I am now realizing that I don't know the design
I hope he can find our notes from a year (?) ago.
well, if you come up with anything, I'd be glad to help coding
will you be around during fest time?
when is it again?
but yeah, most likely I'm around
21-28 I think
jepler: carrier no; hardflow no
cradek: in cu?
I think there's no way to tell cu except by using the uucp config files
this doesn't seem to be documented in the cu manpage..
anyway, qq was quick to write
I think maybe it's a mistake to think cu can be separated from the rest of the uucp package
although I understand why they did it.
for users like me who want a simple terminal program, I imagine
like you said, the choices are minicom (crappy qmodem clone) and cu
it sounds like the only thing missing from cu is a bit of documentation, and I'd have been happy
or maybe the proper signal loopbacks in my serial adapter
it's in the manpages, but they are in the uucp package
err, maybe in the info
BJT-Gone-Fishing is now known as BJT-Work
EMC: 03jepler 07TRUNK * 10emc2/src/hal/components/joyhandle.comp: documentation markup fixes
EMC: 03jepler 07TRUNK * 10emc2/docs/src/lyxtree.py: when building lyx documentation into html, convert to the version that lyxtree knows how to handle
heh, that's fun to read when you're tired: '/usr/share/lyx/lyx2lyx/lyx2lyx'
EMC: 03jepler 07TRUNK * 10emc2/docs/src/lyxtree.py: apply the conversion to lyx files included from the main lyx file
huh, that's a good trick -- whenever I exit lyx it kills my window manager (icewm)
does it restart at least?
wonder how it manages to do that.. (kill icewm)
huh, they changed the design for SF again
EMC: 03jepler 07TRUNK * 10emc2/src/ (configure configure.in): switch to lyx 1.5.x for documentation
EMC: 03jepler 07TRUNK * 10emc2/docs/src/lyxdep.py: switch to lyx 1.5.x for documentation
EMC: 03jepler 07TRUNK * 10emc2/docs/src/README: switch to lyx 1.5.x for documentation
This is cool http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&ssPageName=STRK:MEWAX:IT&item=110256821264
I have seen pictures of it in computer books explaining cnc
heh. fennville - fenn ought to move there :)
let's move this over here
$ inivar -var Y -sec X -ini badly.ini | od -a
0000000 Z cr Y = W nl
yeah, a similar problem exists for inifiles
(I'm not sure what dos2unix will do with X^MY though ...)
you can do things like sed s/^M/\r/g can't you?
EMC: 03jepler 07TRUNK * 10emc2/src/hal/utils/halcmd.c:
EMC: detect errant carriage return codes and signal the user about it
EMC: "goslowjimbo" presented us with a hal file which had a sequence like
EMC: net ....^Mnet ...
EMC: which was intended to be two lines. This case is now an error. DOS-style line
EMC: endings (^M^J) are a warning, but won't stop executing the HAL file.
EMC: backport candidate
durr. of course ^M is "an M at the beginning of the line" :)
I shudder to think about getting an inexperienced user to type the right thing at the commandline
dos2unix isn't on a base system?
it's part of "tofrodos"
skunkworks, is it just me, or does that machine look like it's got a giant handwheel on top of it
SWPadnos: giant jog wheel
that's what it looks like to me :)
[23:44:21] <skunkworks405> http://www.nit.ac.jp/center/museum/tenji/width450/kosaku/14.jpg
looks like something from star wars
yeah, the little lightsaber defense practice droid
on a machining center
that would fit in the corner of the garage... ;P