Not with snow up to your waist. You need this: http://www.youtube.com/watch?v=zBjlSJf4274&feature=fvw
screw driven, cool
It does seem effective
theorb is now known as theorbtwo
i think i'd opt for one of these though
[00:05:36] <DaViruz> http://www.youtube.com/watch?v=K9RWc2M6AvY&feature=related
Typically, the Russians made something bigger and tougher. http://www.youtube.com/watch?v=afJ18eJeNgU&feature=related
pretty cool that it can also drive sideways
I am not clear if they are trying to...
you're right, they might actually be trying to rotate :)
This might cope with your snow, and carry your machine shop: http://www.youtube.com/watch?v=uHaj5bDSgRA&NR=1&feature=fvwp
They used to show Truck Trialling on Eurosport, but not for a few years now.
i could have used one when i moved my vmc this summer
(Competition off-road trucking): http://www.youtube.com/watch?v=dVj22ybQ-y0&NR=1
I have to say my little GasGas 200 would work a lot better on that terrain.
not if you had to move 10 metric tonnes over it ;)
but for pure personal transportation i imagine walking would be viable
I could do it 50kg at a time and stil be faster and use less fuel :)
At the truck trials, when it all goes wrong they bring in a tank to pull them out. I don't know what you use if your tank gets stuck.
random tank link (P1000) http://en.wikipedia.org/wiki/Landkreuzer_P._1000_Ratte
[00:25:06] <DaViruz> http://www.k-makris.gr/Gallery_images/sky_crane.JPG
one of those maybe
[00:25:53] <DaViruz> http://upload.wikimedia.org/wikipedia/commons/6/67/CH54B_and_M551_Sheridan.jpg
Or, bigger still: http://en.wikipedia.org/wiki/Landkreuzer_P._1500_Monster
800mm cannon :D
Just a little bit silly.
The P1000 with ordinary naval guns was a better bet
most nautical artillery is much smaller..
naval, that's the word
9 tonne payload on the skycrane. You would need several in formation.
andypugh: what kind of resolution are you getting out of your arduino resolver interface?
i just remember i have a resolver lying around, and they are kind of absolute, are they not?
I don't know. I have never actually used it.
However, it is quite likely that after 6 months on the back-boiler I will be setting it up tomorrow.
The Arduino has a 12-bit A-D but I am using it at 8 bits.
(Because I want speed)
i'm looking at 1rpm
1 revolution per day would actually be sufficient for celestial tracking
Working very slowly at 12 bits with 100x oversampling I reckon that your resolution would be effectively unlimited.
hmm, a little faster is needed, a sidereal day is shorter than a solar day
i guess it boils down to imperfections in the resolver then?
Yes. There is no inherent limit to the accuracy of a resolver
You might eventually hit issues with the regularity of the windings, but then you could compensate for that in software.
if i have something to calibrate against
More practically, you could mount a laser on the telescope and measure how high it reaches in a tall building.
probably a bit overkill anyway, i'll settle for 0.1 degrees
Actually, that might be imprecise. How powerful is your most powerful laser? Could you use Lunokhod 1?
i have a 1W infrared lying around
not sure what lunokhod 1 is
It's a lunar rover with a retroflector
I'm starting to think i have a burnt out PCI slot on this machine or something
AR_ I think that'd be beyond my capabilities
well check it out I'm testing the parallel port and the on board one works with my test
I can set and reset it
now I have an add on port card I have it all configured in the OS it shows up in /proc the works
I do the same test and nothing
I'm starting to feel beat on this whole deal
How are you testing?
I have the cable in the D connector to my break out board and a meter on it and I hae a program that can toggle the port
and it works with the on board just not my add on card
so the test method is valid I'd say
Paragon went through an exhaustive list of combinations last night and finally found that the combination which worked was with ioport_hi below ioport.
I am not familiar with ioport.
What base address is your program using?
ah I got them from hwinfo
they're also in /proc
though in decimal form
I might be misremembering, but it is a setting in the hm2_7i43 config string.
How many adresses do you see, and what width are they?
well i am just using the kernel module now to sort out which is who
parport1: PC-style at 0xdc00 (0xd880), irq 9 [PCSPP,TRISTATE]
What motherboard? The D510MO BIOS lies
that is what hwinfo said
this is an old gateway
(though probably only about the onboard port)
I think it has an IBM in it
yeah the on board all works
what does lspci say?
I'd just prefer to use my add on card for reasons of safety
(lspci -v IIRC)
yeah i have it in my file as well it has like 5 lines of course two match what hwinfo says
I/O ports at dc00 [size=8]
matching line to the other i pasted
oh wait it isn't
I/O ports at d880 [size=8]
that one matches
What really matters is if EMC2 can find it, I suppose.
yeah but I'd like to know which one to use
I'd actually prefer to use the ribbon one
Try this: http://wiki.linuxcnc.org/emcinfo.pl?Parallel_Port_Tester
case i burn it out i can work my way back you know?
If that works, then EMC2 will work
OK let me give it a whirl thanks
I run in place so I need to put these with halrun right?
That's what it says
mmm not exactly
To run this place both files in the same directory and open a terminal window and use the command
I think as long as you are in the same directory as the files when you start halrun, all should be well
You might need to run the emc-environment script
Also note that the first line of the HAL file needs to be edited to suit your port address
so how do I know what port this is working on? I am going to move to the on board where i suspect it is working
You might as well run through every port address you can find.
OK what exactly is this hal file?
yeah it was configed with the n board let me move the wire to it just for a control test
That uses all the EMC2 drivers and infrastructure, so is a better test
time for port guess #1
# is a comment in this file right?
either this card is baked or the slot is
Try every address you can find. It seems that p-port is now so "legacy" that OS support is getting flaky
What card is it?
There are problems with Netmos not actually doing EPP, though that shoudn't matter for bit-twiddling.
hey i got it!
the freaking LED is flashing when I hit the button!
man I couldn't get the kernel driver to work for love or money over here
It's a nice tool that, and a really good example of the flexibility of EMC2 / HAL
although it seems inverted to me
I guess red is low
Some pins as hardware inverted anyway
yeah I don't know maybe my BOB is inverted?
But yes, I can never figure out what colours are high and low in halmeter etc.
who cares it works!
Aye, every p-port oin has "in" and "in-not" so it matters not a jot.
if it really bugs me i can go in the drawer and get a green LED
then say see it even matches the screen!
thanks man oh i was going nuts
At a guess your button is shorting a pin to ground? So pressed would be low?
I've the anode of the LED to pin 3 then the cathode of course to ground with a dropping resistor
and when I hit the button I get LED on
and I have a volt meter on the pin as well and it goes up
just one i happened to have a wire stuffed in
was either that ot two
Your switch is broken :-)
well I hit pin 3 button in PortTest and LED on pin 3 lights
3 is an output?
says it is
now i sort of want to move the LED around and check them all out
Ah, right, sorry, I thought you had a switch into an input
nothing so sophisticated over here yet
But you might as well test them all
I just wanted to find the danged port!
I got just the one i wanted now too
the ribbon one
EMC2 seems to have more "forgiving" port detection than some other systems
yeah I'm going to try 2
I'm telling you I unloaded the kernel modules fed it this: modprobe parport_pc io=0x378,0xd800,0xdc00 irq=7,9,9
should have worked
I'm still not seeing anything wrong with that
Don't ask me, PC hardware and Linux are mysteries to me
the on board was happy with it
the other data is just as good
I''m an Apple fanboi, it works or it doesn't :-)
yeah i mean I'd really like someone with a 190 IQ to walk in and explain to me in terms i can understand just what is going on
but hey I'll take it just works too
just ran through all my outputs the kid can still solder
I just found a web site that will let me take an IQ test for $9,99 that hints I can score 190....
my whole homebrewed BOB works
I reckon I can offer 200+ IQ tests for $5 and become rich!
andypugh if you're dumb enough to fork over $9.99 for an online OQ test i can guarantee you you're on the wrong side of the bell curve :)
OQ? Obesity Quotient?
I've taken real IQ tests and they're pretty involved and subject to the poinion of the doctor
Yeah, me too, I come out as "fairly bright"
the further i get in them the funnier looks i get :)
ha this is all I wanted to do today and i did it
I rather expect that the same is true for anyone on this channel, it's not a game for dullards
why didn't the other stuff work?
I donno you can plunk down 6 grand and walk away with a turn key
No idea. I guess conservative drivers giving up too easily
not a good one but it'll do something
well if the on board hadn't worked I'd have been less sure of anything
but with it working I knew my test method was good
turnkey solutions don't send you to the emc2 forums.
depending on how bad they really are
They send you to Maas support, or Mach
hey sherline ships with emc2
or how allergic you are to M$
I can't believe anyone gets that stuff to work at all
It seems that the Sherline config Just Works. I don't think I have ever seen one here. Perhaps the problems end up a Sherline?
skinnypup earlier I hit a page on cnczone where someone was crying the blues about an add on port card and mach and it made me feel better :>
I like Visual Basic and Excel. I use what works.
skinnypup they ended up buying another card
cant figure it out? just spend more $$$ lol
skinnypup thats how I looked at it
I foresee linuxmint debian going on the up and up. :D
was more like Windows was a brick wall and they just got tired of running up against it
i've been given pc's by windows users b/c that one kept getting viruses
There are some p-port cards that EMC2 can't do EPP on. Buying a new card there is the only solution
andypugh I read the docs first got a NetMos Technology PCI 9815
for me buying the right card off the bat seemed the best solution
its so cool hit the button on the screen and the needle jumps on the meter
The right card might have been a 7i43 on the onboard P-Port, or a 5i23.
one of those cost more than my whole project
7i43 is $80
OK twice as much
I suspect your project will grow.
yeah so far I've had no luck scrounging up some materials to make the machine frame out of
well a little but not everything I would like to use
but i got my BOB my motors and my drivers
oh and lead screws
I am at the point where I wish I had spent more earlier. I sold the micro lathe bought a mini lathe/mill, sink £1000 into that (on top of the purchase price) and I am now converting a proper mill, made properly by Yorkshire folk.
and what do you make on your machine?
all i want to do is route wooden signs
I'm not looking for a whole lot
hey it routes wooden signs I'm a happy boy
2 years in, and the thing I wanted to make has become secondary to making a really nice CNC setup
well I've looked online and drawer slide machines seem capable of doing what I want to do
though I'm going high class with right angle skate bearings over here :)
Yeah, MattyMatt seems to be doing OK with drawer slides
hey I wouldn't look down on a good set of drawer slides
He has arranged them in pairs at 90 degrees which is fairly cunning
I could see that
I rescued some very nice 3-layer ball-bearing ones from work. I think they might be part of a toolchanger
lots of this hardware is still out of sight price wise
ive been doing some wood routing on my mill recently, poplar mills nicely with little loose grain needing sanding
skinnypup i snag used pallets for wood and there is one kind of wood some of them are made out of that routes so nicely
skinnypup its not popular I know that stuff it is this really soft close grained whitish wood
I have a problem with wood, I am an oak-snob. Anything else seems like a compromise (with the possible exception of pitch pine)
really pale in color stains like crap but man it routes nicely
andypugh I've terribly bad news for you
andypugh oak is poor man's wood :)
i did some red oak as well, few places the grain would rise
yeah oak has a habit of rising
it is also very open grained and prone to splitting
white pine wasn't worth turning the spindle on for
granted quarter sawn is pretty
Oh, oak is a pig to work, especially english oak.
But it's strong and lasts long enough.
but its next to casement woods
cherry maple mahogony are all better
I had a piece of oak go bad on me here
All my own work (http://www.bodgesoc.org/Slaithwaite2/end_stairs.JPG)
(actually, the lowest beam in the picture has probably been there 400 years)
you made the beams?
No, trees made the beams
no they really don't
This is the 500 cubic feet we started with: http://www.bodgesoc.org/wood.jpg
and its a lot of work to get it to that!
I count two original beams in this photo: http://www.bodgesoc.org/Slaithwaite2/Hall3.JPG
Hopefully it is hard to tell which two.
andypugh what are you a hobbit?
that looks like Bilbo Baggins house!
It's my parents' house, built around 1450.
There are lots of houses that old. In someways the web page I wrote in 1995 is rarer: http://www.bodgesoc.org/slaithwaite.html
now that is place in serious need of some roofing!
Back in the days when people had 13k6 modems. The full-size photos are thumbnails today
is this you? http://www.bodgesoc.org/truss3.jpg
Aye, we roofed it in 40 tons of sandstone slabs
14.4, 33.6 56
56K only worked on long distance
local it was 53.3
Jymmm: Quite right. I had a 28.8 at the time
[02:28:38] <skinnypup> http://imagebin.us/images/2k3ia1kuwgh5rbwuo50.jpg
skinnypup like a nordic knot
pcncx: No, that's my dad. I am the one at the edge of the roof in http://www.bodgesoc.org/roof2.jpg
(where yiou get a feel for how big some of the roofing slates are)
well least you didn't have to get them very high up there
'tis a challenging job, each one needs to be the right length for the course, the right thickness to work, and the right width to stagger the joints.
you guys have a wet saw?
not sure if its gordian, just took some "tatoo" art and worked it up in gimp
skinnypup yeah I'm not sure either its cool
I just picked up a little PC router I may use on my machine
first piece of wood i ever put on the mill
skinnypup how high RPM can your spindle do?
pcncx: No, saws give sharp edges, which look wrong. You can try with a chisel and hammer, but generally you just have to find the right one. It's dry stone walling with one more degree of freedom
those were done with 1/8 endmill
routers are 20K RPM and run 1/4" or bigger
might want to try biger to up your SFPM
But then you lose detail.
I know when I try to mill wood on my machine its a big marginal
but mine only goes to 2,400 RPM
yeah working on getting together a gantry design
which is even worse
well you can use a side spindle
like mount a router to your mill spindle somehow
still limited to 12x24 on the mill though
or like i said just use bigger bits
Does wood mind too-slow? Too-fast burns it, but you don't get the same work-hardening and strain-softening as metal
yeah too slow it comes out hairy
just don't get the crisp cut
or they'd make routers a lot slower
Hand tools are very slow, and work...
half the art of hand tools is lost today
if hand tools are wicked sharp it can be amazing
its a whole different ballgame you art the wood with hand tools
true hand tool wood craftsmen are like magicians
you watch them see the results and are like how'd he do that?
its so clean and crisp and smooth
I really enjoy working oak with hand tools. I hate pine and jelutong. I can imagine that many woods I have not tried are even nicer, but green oak cuts very crisply with a big enough hammer.
I like hand tools got a bunch but I like power tools too
[02:40:01] <skinnypup> http://imagebin.us/images/8u8xrcsrcyho6naogct.jpg
after a lil stain and laquer
skinnypup do you seal before you stain?
like use a shellac wash?
popular can be blotchy
i have some sanding sealer to try out next time around
your dark came out very even is it a water gel?
water based though?
no just regular stain
get a can of shellac stuff is handy anyways
then cut it down to nothing and do a seal coat
like to half pound or so
yeah i have some playing to do on finish products
well shellac is some magic
get ammonia and alcohol when you get it
and you'll have the whole kit
last one cut was a 12 featured knot to make a clock from
I like the dark middle one the best
what do you make your g-code with?
wow works well
I've fooled with that some
everyone wants to take the dark one home with them
just on the software side never cut anything with it yet no machine
yeah its neat looks undercut to me
[02:46:35] <skinnypup> http://catalog.ebay.com/Museum-Early-American-Tools-Eric-Sloane-1991-Paperback-Reissue-/164895?_pcatid=4&_trksid=p3907.m570.l1313&_rdc=1
terrific book for handtooling
skinnypup: The middle carving I like. It's great.
thanks guys !
<laughs at "early american">
[02:47:53] <pcncx> http://img46.imageshack.us/img46/8965/hornplane.jpg
yeah didnt say the title was accurate
andypugh hey we got a house from like 1620 around by me
andypugh that you guys tried to shell
well did shell
called the cannonball house
The other two look too rounded and routed though. Nice in their way (somewhat Gaudi-esque)
because there is a cannonball sticking out of the wall
andypugh they fooled the british though they hung lanterns in the trees a mile up the river
pcncx: Pretty plane. Have you tried it? Some of those old planes are lovely, and some basically don't work.
andypugh its a two cherries blade and yes it works nicely
I got some other wooden planes
[02:50:55] <pcncx> http://img72.imageshack.us/img72/6417/pict0822h.jpg
the sandusky works too
They do seem very variable (which is not too surprising)
oh yeah you got to smack them just right in order to get them to work
You know how to adjust them?
pain but they work different than the iron ones
oh yes you hit them with a hammer
I use a urethane one
I see frm what you said as I was typing that you do :-)
I made the handle for the big one out of mahogony
when i got it it was missing
msde the tote for the coffin plane in the back of the drawer too
you can sort of see its the same wood
check out this mad hand carving: http://img840.imageshack.us/img840/6763/sawh.jpg
I took some of that never hardening clay and made a model like I liked then traced it onto the wood and had at it
cool, looks like that would feel good in the hand
skinnypup oh i kept holding it then hitting it with a rotary file or sand paper
skinnypup til it fit me custom
I make like little wooden boxes a lot: http://img197.imageshack.us/img197/8479/sb006.jpg
but I'd like to do some CNC carving on them to jazz them up some more
thanks yeah it's relaxing
I was looking for something else, but wow! http://blog.craftzine.com/davidesterly.jpg
andypugh that doesn't look very relaxing
no not at all
imagine if you broke a leaf
ahhh then keel over!
I met the #2 bird carver and man his stuff was nuts
the pins inside the feathers
he was like yup I'm only #2
gack! how do you clean up after hal ( hal file gone amok ) ? halrun -U didnt get rid of 'ERROR: Module hm2_pci is in use'
killall -9 halrun ?
I was actually looking for Grinling Gibbons when I found that, it seems that that chap is making repro stuff to replace damaged Gibbons stuff. So perhaps the skills are not lost, just as expensive and rare as they always were
tom3p: If you find out, let me know. I just restart the machine
check out my primative rig I used to grind my lead screw ends: http://img199.imageshack.us/img199/6379/lead2b.jpg
should waste it
then init 2 to go back to regular
its almost a reboot though
yah, may be cleaner
or just ctrl+D it at the prompt should bring you back up
pcncx: Ingenious. I used a CBN bit and maximum spindle speed: http://www.youtube.com/watch?v=wIMEPkK-Doc
andypugh I used a right angle grinder while it was going then a little filing
I'm going to slip rubber hoses between it and the motor shafts
not the best but it should work
OK, back on topic. What the heck is happening here? http://www.youtube.com/watch?v=Yri_BELrXjE
looks like they're threading to me
andypugh that carving isn't one piece it is glued up
its still wild but its not like it is solid
hey I carved a duck decoy all solid then found out all the heads are glued on
even the real ones. Trust me, it's true
I was in a shop that sold collectable decoys and I'm looking at one and the person in the shop is like yes its lovely isn't it?
turns out the thing was $30,000
it wasn't *THAT* lovely!
I got the hell out of that place
haha! dont try to work on pyvcp screens that invoke the mesa pci 5i20 driver... when you're on a laptop :0
and that pic of the leaves and vines was incredible
tom3p I'll try to remember that
yeah there's always someone more psychotic out there
I'll never forget this one carving I saw at the met museum when I was a kid
this little 3" diameter ball that opened up and inside it was the whole scene of christ getting crucified
I swear there were thousands standing on the hills!
tom3p: Why does that driver cause an issue with pyvcp?
inside this little tiny wooden ball
i like the temple poles in Taiwan, all dragons and phoenix, like lace
tom3p thing thing made me feel like I'd just fallen down the rabbit hole
andypugh, insmod: error inserting '/home/tomp/emc2-dev/rtlib/hm2_pci.ko': -1 No such device , then...
yeah some of those asian soapstone carvings are crazy too
or the 7 nested ball things
andypugh, then ^D, then run again ( not noticing the impossibility) and you're stuck with "tomp/m5i20-23oct2010/tjp-m5i20testPanel.hal:38: child did not exit normally"
I think I've seen the worlds record nested ball carving and its pretty kooky
andypugh, and need something strong to allow halrun again
epic tiny-ness http://www.youtube.com/watch?v=CD7eagLl5c4&feature=player_embedded
oh I've seen this
can you do a dmesg after the first error?
andypugh you seen the boorklyn space balloon yet?
I think you'd like it
just search for brooklyn balloon it should come up
yeah top result looks like it to me
people are nuts
Yes, we are.
(perspective here, you are on IRC chat about an open-source CNC package. Look at the Venn diagram, and see where you are.
That's really cool. I am persuaded against the Flat-earth theory.
andy, nada in dmesg, but messages had http://pastebin.ca/1971463
hard to explain if you wanted to get warranty on your iphone though
andypugh, nm, its a problem that can be avoided easily
tom3p: Sorry, nothing in that is telling me anything (and I have spent the last week watching Hostmot2 drivers crash)
I should perhaps point out that I mean half-baked drivers of my own creation. Seb did a fine job of the others.
heres some fun, what to do with 16 rc servos http://www.youtube.com/watch?v=ejROvUC-gWU
[03:31:49] <andypugh> http://www.youtube.com/watch?v=yYguYgCNoY0&NR=1
Turns out it was a fellow Yorkshireman who did that weather balloon trick :-)
[03:32:27] <andypugh> http://www.youtube.com/watch?v=yYguYgCNoY0&NR=1
I've no sound :(
lol that was cute
nice use of duck tape!
now i met a guy who claims to have put a rocket on a balloon that light off when the balloon popped and kept on going up
ah, it's space.1337arts.com ( hard to read that screen on the video )
he claimed to have gotten 3 orbits out of it before it landed in Italy
but he might have just been BSing I don't know
he had a pic of himself with an awfully alrge rocket though
I just noticed that it is 0439am. Not that I have much planned for (errr) today, but still...
good morning and thanks again
fragalot is now known as Guest39863
Guest39863 is now known as fragalot
is there some docs on the 3-phase pwm generation in fpga?
I found this http://www.linuxcnc.org/docs/devel/html/man/man9/bldc_sine.9.html
but is there a write-up, images, videos, tests of someone building a power stage and using the fpga-based ac-servodrive?
he used a commercial drive iirc
Lets say I have a Tool with 1.6mm diameter and I have G-Code without compensation for cutting a rectangle from (0/0) -> (0/10) -> (10/10) -> (10,0) -> (0,0)
Now I want to use G41 for left compensation
what else do I have to specify apart from G41?
It seems I have to insert a lead-in move, but can't find more information about it
on my dx32 control (DOS 6.22 based) I have one file that has a space in the name. From the control I can't delete or rename it as I get an error about bad parameter. Would dropping to dos and using del file nam.txt have any issues?
"name with space"
like "file 123.txt"
that should find it
with double quotes around the name?
archivist: I get parameter format not correct... I wonder now about using a wildcard filler
the file name is "dis shaf.txt" would del dis*.txt work
yea,using a wildcard worked
oh the lovely days of DOS such a fond memory
I still use the command prompt in windaz
so do I for some things
I still use dos
Jymmm: is there anything else I need to do to back up my hard drive in DOS 6.22 besides a DOS back up?
Well, I got rid of the portable rack <sniff>
JT-Hardinge: the MBR
Hi Jymmm I'm over here :)
jthornton: you all look alike... the mbr is is where that partion info is stored.
it has a solid state hard drive
jthornton: What has DOS that you backing up?
my BP VMC
I just want to be able to restore the hard drive when it dies
jthornton: Ok, hang on. I need to look for something.
jthornton: Is that a modern machine? Like is has USB 2.0?
no, it is a 386 motherboard
17 years old
ok, still checking something.
jthornton: How do you plan on backing up? To another hdd or ????
well I did a back up to floppies using DOS backup command and it took a whopping 3 disks
main thing I'm interested in is being able to restore the system when the hard drive fails
jthornton: Right, I understand.
jthornton: Is this flash drive IDE ?
yes, I think so... it just plugs into the IDE slot and is quite small
jthornton: Yeah, it's just a Disk-on-a-chip. I have a 32MB one
jthornton: What you could do is install a 2nd hdd on the ide cable, then backup to it, or pull it out, then plug into a modern machien and backup/restore it that way
jthornton: you have a windows machine? XP?
when I backed up to the floppies did I miss something using DOS backup?
also the ss disk plugs directly into the IDE slot
jthornton: It doesn't backup the MBR (partition info) just the regular, hidden, and system files.
to restore it, you'd need a bootable dos floppy with format/fdisk/sys, and a copy of all the files that are on your hard disk
cradek: (assuming they didn't use any hidden partitions)
I have never seen a 4MB disk on a chip before, the smallest being 8MB. So Im not sure why he was able to backup to only 3 fdd
disks aren't always full
I'm not sure what dos backup even does...
cradek: Sure, but I've seen weird things with hideen partitons, etc.
yeah if they made it weird on purpose it might be very hard to restore to the same state
Yeah, I'm working on something so he can backup without having to deal with that.
ok, I'm back
I can format with system a disk... have to look up the syntax
jthornton: patients grasshopper, I'm almost done.
Jymmm: there is not much on the hard drive
that is why it took only three floppies
jthornton: Can you seen the hidden files?
I don't know what do I look for
just hang on, I'm working on somethign for you jthornton
looks like dir /ash should show them
I'm going to wander back out to the shop now, I'll be on the Hardinge computer
jthornton: JT-Hardinge Where are you?
I created a dos boot disk with sytem
JT-Hardinge: other window
moopy is now known as mungkie
what are the gear reductions found in robot elbows called?
i have the name on the tip of my tongue..
that's it! thanks
[16:27:52] <Jymmm> http://www.dealextreme.com/details.dx/sku.37373
[18:03:00] <skunkworks> http://www.youtube.com/watch?v=9xDPqFXo_5w
skunkworks: coke is bad for your health
yes it is
:) - Hi alex
only a couple more bits & pieces left, right?
and it's still free :)
anyways, good job (to you too, but mostly to your dad ;)
(in and out working in the yard)
Is that lowering when the pallet is in place part of the locking mechanism?
boot|sleep is now known as bootnecklad
anyone worked with OpenMP? I'm trying to spawn off two tasks, and each task has a parallel for-loop in it. It's not going very well...
my friend has been busy making moulds again this weekend! http://imagebin.ca/view/I3BwDwIV.html
awallin: sailboat parts? nice work!
yeah its a mold for casting a lead weight. the design weight is around 2.5kg. The weight sits at the end of a carbon fin keel.
awallin: what's problem with omp?
i was working a bit with it this spring...
jthornton: here are casting-pics from 2003: http://www.anderswallin.net/2003/05/making-a-bulb/
psha: with 'task' I do seem to get two function-calls executing in parallel. But the functions themselves have their own 'omp parallel for' sections, and those don't seem to execute in parallel
psha: if I call funcA() and funcB() in normal serial code then their internal 'omp parallel for' sections do execute in many threads and I see speedup
psha: but if I put funcA() as one task, and then funcB() as another task, the parallel for-loop inside them doesn't work anymore
I think OpenMP isn't abstract enough :) I'd like to just describe the parallellism in the code and not have to worry about number of threads and scheduling etc
wait a bit, i'll try to recall magic omp pragmas for such cases
awallin: it's impossible :)
this is nice: http://supertech.csail.mit.edu/cilk/
nice but i've not seen any real-world tasks with it
mpi and openmp rarely
yep, and I don't know how much compiler support there is for cilk. only intel compilers? gcc?
that's interesting, have you tried cilk with gcc on linux? compared it to openmp?
only 'theoreticaly', i can not remember but we've desided not to consider it later
pcw_home: yes - the pallet gets set on 4 tapered pins
hydraulic pressure clamps the pallet down on the pins
psha: something like this http://docs.sun.com/source/819-0501/2_nested.html
somethink like this
i was looking through my spring project for pragmas for this :)
we were optimizing/parallelizing nuclear plant simulator :)
awallin: i think i've created minimal nested example
cool, pastebin or email?
[19:43:59] <psha> http://psha.org.ru/p/nested-omp.c
run with OMP_NUM_THREADS=4 ./a.out
for me it's using all 4 threads for for loops
endfor is not needed :) it's fortran rudiment
uploaded new file
or instead of omp_set_nested(1) you may export OMP_NESTED=TRUE environ variable
I'll try your code soon. mine is still not working right :(
g'day all...looking for some help in getting Axis to run on this machine...This is an older (2002) Dell Optiplex GX240 puter, Ati rage 128 video, all other (TKEMC, MINI, XEMC) apps start and run without problem...Axis locks on startup
Have you run it from terminal to see what the error is ?
jthornton: JT-Hardinge ping
no, I haven't
have only used the gui
awallin: it's ok with parallel programs to work incorrect :)
launch emc from term and see if some errors get dumped there
will do, if it locks I will have to reboot and come back in
emc > error.log
will put the terminal errors in a log file for ya since lockup may be an issue
Hello all, anyone with experience with Celerra?
hm.. I think I am in the wrong channel...
sorry, later all..
machine locked up without providing any visible errors...only mouse worked but mouse cursor appears in black
did you run with
psha: yep, your minimal example seems to work over here. thanks.
no, I did not...
use less or nano to read the log file afterwards
will attempt again
awallin: np :) this question is from area where i'm a bit confident since distributed computing is my main job :)
and thanks in advance for your time and input
i need to work more at the mill
took the wrong freakin dimension of endmill, totaly crapped up the dimensions of the raw material ...
btw, why doesnt emc retract to home or something when tool change is issued?
for me atm it stops like 10mm over the material :/
EMC2 - 2.4.5
Machine configuration directory is '/home/mike/emc2/configs/sim'
Machine configuration file is 'axis.ini'
very disappointing...that is the extent of the error log...
kb8wmc any errors shown at the output of dmesg command?
will check now
pjm: not a thing showing as error with dmesg...
try doing dmesg -c
then dmesg again
at first I thought maybe there was something in the sim/axis.ini that was not correct, but it looks ok...
will do, be back in a bit
psha: still there? the minimal example works fine, but my larger code doesn't :( in my real code omp_num_threads() = 4 in the func() calls, but I still see load only on two processors (corresponding to the two sections)
psha: but maybe it's something else. the two func() calls may have the same private(variable) and they are competing for access or something...
maybe private variable
one way is to drop all 'private' statements to see if they really compete for one var
dmesg does not show any error
if event without private they still don't parallelize you have to dig futher :)
psha: I think my task objects both have a const-pointer to the same object. and they are calling that shared object a lot
is it really needed to be declared 'private'?
if pointer is const then variable is read-only
it's not declared openmp private
yes but even if it's read-only it's maybe not possible to call in parallel?
so it would not block anything
jthornton: JT-Hardinge ping
no :) with such hard requirments there would be no parallel programs at all :)
psha: maybe I should try this in my minimal example...
dmesg does not show any errors
psha: hmm, the minimal example still works fine when the tasks share a const-pointer
snow your parallel for pragma
#pragma omp parallel for schedule(dynamic) shared(calls, fiberr) private(n,i,tris,it,it_end)
it's in miniam or in prog?
this is the real program. calls and fiberr are modified by both tasks... maybe that is the bottleneck
how many cicles in for?
it doesn't want to split that??
dynamic will work good for it
i've tried your pragma with for(100) and it's not splitting
i've forgot to export OMP_NESTED
may you print something like omp_get_thread_num()/omp_get_num_threads() in cycles
like it's in minimal example
it'll print number of threads in group
also there may be some internal barrier
could it be that I am trying to get more threads than I have cpu cores? I have a quad-core machine, but I would really want 8 or more threads for this
well omp_set_num_threads(8); does not help... still I get load only on two cpu-cores when launching either two tasks or two sections
i have dual core and running 4 or 8 threads
i'm setting it with OMP_NUM_THREADS environ var
well it seems there is no performance advantage in nested parallelism if I am creating more threads than cpus.
the serial tasks TaskA and then TaskB is faster
there are some cases with advantage when creating more threads then cores (+1 or +2) but in general you are right
about 1000-fold ovepriced?
isa vid lvds, that's kind of rare
Sorry to be dense, but with reference to the wiki article http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Axis_Embed_Video
, where would I find .axisrc ? And am I looking for a file or a folder? (I'll update the wiki when I know)
Jymmm: The same seller has got a 10-BaseT ISA card for almost $100: http://search.ebay.com/200528956930
maybe i've forgot to upload .axisrc file
KimK_: run 'wget -O- http://wiki.linuxcnc.org/uploads/axisrc-dynamic-tabs
>> ~/.axisrc' and you'll get it
KimK_: Maybe I should sell my 3c809 cards for $100 then =)
uh ok, 3com.com redirects to hp.com
i have like a dozen 3c509, with various port combinations
psha: There's an uploaded file in the article for .axisrc, but I don't know what to do with it, still looking for an existing .axisrc
psha: Not sure whether to add it to exiting text or poke it into an existing file
run 'ls ~/.axisrc' in console
and you'll see if it's exists or not
No such file or directory
so crate new one
A file or a folder?
or better download file and rename it to .axisrc
OK, that was my next Q, whether .axisrc has some boilerplate at the top, apparently the answer is no. Thanks!
it's standard header i'm writing to all pythonfiles
it may be safely droped
but useful for vim users :)
psha: OK, I added it, but it doesn't seem to do anything to Axis (no "camera" tab), shouldn't it at least say "no camera found" or something?
oh, this script only allows adding custom tabs into axis
Yes, that's what I wanted to see
for camera you may use plain 'mplayer' with 'tv://0' flags (it's mentioned in wiki)
or something else
i have a bit hacked camview program from camunits project
Oh, OK, I have to add to the config file too, hold on.
[22:04:47] <andypugh> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Axis_Embed_Video
Seems odd the way it need to take two bites to do a change.
But looks cool.
yes - it is odd
I suppose that made it work in space provided - distance it had to go
KimK_: if you'll have questions about video tabs - catch me tommorrow
psha: OK, thanks!
skunkworks: Nice, Sam. I like the yellow safety corner, you'll have to paint that yellow and black safety diagonal striping around it.
Or just stright yellow, easy to see and easier to mask!
What's left on your K&T agenda?
My Arduino resolver to quadrature converter is being odd.
How's that, Andy?
KimK: b axis - then a lot of tweeking.
It works OK (but slowly) if the debugging serial port stuff is on.
But if I turn that off to get a better following speed, the encoder that is following it goes one way, then bounces back.
It first I thought there was a problem with my wrap-round at zero degrees, but it doesn't seem to be that.
andypugh: That does seem odd. Is the simulated encoder following via a PID loop?
I put in a 50uS delay in the main loop, and it is working OK now, but can't follow fast rotation.
[22:20:59] <andypugh> http://www.pastebin.ca/1972225
How long does it take to digitize/sample/measure/whatever the sine and cosine components?
It runs a PWM sine wave, through an active filter, then samples the sin/cos voltages at fixed points in the waveform
From that it calculates an arctan, and that is the internal value of the angle.
I have a counter which basically "chases" that angle, outputting quadrature pulses as it goes.
And all/much of this trig is lookup tables?
Yes, the arctan is done with a LUT
I think that it is 14uS per arcan
Does the "chasing" need a "queue" to dump into, or can it keep up?
It can keep up if I don't have the delay.
But if I don't have the delay, the encoder module acts peculiar
<voice>"You've got volts"</voice> , LOL
What's your audio drive frequency (PWM, active filter, etc.) to the resolver?
PWM is 32 kHz creating a 1kHz sine wave
btw - on the video - you can see what the last thing that was machined on it before the old control died. a 9 light switch cover.
The sampling is synched to the sine wave, and the calculation and quadrature counting happens asynchronously to that (hence the "You have new volts" flag
I am wondering if, on passes through the loop when there is no new data to work with, the quadrature pulses get too short for the encoder to register.
Perhaps I should look at making the 50us delay calculated on the basis of how long the loop has taken so far.
That seems like trouble to me. Can you sync the sampling, calc and posting a new value to the output buffer all synchronously, and let the buffer to encoder pulse output happen asynch?
Where "That" is "The sampling is synched to..., etc."
It nearly is. The sine generator kicks off an A/D measurement at three fixed points in the cycle and when the A/D finishes it calls an interrupt that sets the "new measurement" flag.
I can't do the calcs in an interrupt thread, the PWM function needs to be able to do several updates while the calculation is running.
Well, OK. So what would be the fastest rotation (Nyquist?) that you would expect to keep up with, 1/2 rotation per 1ms?
One feature of the "chasing" encoder is that it automatically wraps the 0-360 degrees resolver position into an open-ended value
That's the limit due to the resolver and the waveform.
One issue is if the "chaser" falls too far behind, it will start running backwards.
Which I think is what I see at high speeds.
But that doesn't explain the issue that I get when I speed up the "chaser" loop.
Yeah, plenty fast, if it works. One-half rotation per 1ms or 500 RPS (x 60 secs/1 min) = 30,000 RPM.
KimK_ what are you making a Dremel?
A bigger issue is 1 quadrature pulse per 5uS with 1024 pulses per rev.
Maybe make the chaser run at some multiple of the DAQ rate (thats effectively what ours does)
(is it possible your chaser violates quadrature sequence or speed if some delay is missing)
Can you build an output simulation and have it ramp up in speed (with no resolver) just to see?
Ah, PCW. How long does the Hostmot2 encoder need to see a pulse for before it believes it?
pcncx: Welcome. It's andypugh's arduino resolver-to-quadrature converter we're talking about
KimK_: I think my first change will be to start off with a 50uS delay in a variable, and knock 15uS off of it every time an arctan is computed. Then delay for what is left.
Depends on the filter setting (also the TTL inputs have a RC filter in the 100s of ns region)
If filter is on its 15 counts of 50 MHz = 300 nS long to be recognized
OK, there is no way that I am running the loop in 300nS.
andypugh eat your Wheaties
Valen: Hi, I see you've just joined us too.
figment of your imagination
with this font I'm running i can read pretty much whatever I want to it is so illegible
[22:59:29] <Valen> http://www.youtube.com/watch?v=zGZIztjoHb8
wonder if they are running it with EMC
I wonder if I can code my resolver thing a bit differently. Perhaps the angle and encoder should be allowed to count up a few thousand turns, but lose their top bits any time they get back in synch (to avoid overflow)
Though I have just noticed that Q is an unsigned int. So if it tries to count backwards, it all goes wrong. That's not right, and might be part of the problem.
It sounds like a eureka moment. We'll look forward to your next report.
I, however, do not have mplayer showing on my third tab in Axis, so clearly I am still doing something wrong. (No, I don't want want to watch DVDs on Axis, I just don't have a camera.)
KimK_: just pr0n
Did you look at the Wiki link?
Yes, that one :-)
Yes, but I still haven't gotten it to work (in test mode with mplayer, no camera here yet.)
andypugh: what issue are you having?
Nothign wrong with that. If it wasn't for pr0n most of us wouldn't even exist!
andypugh: Waht do you mean doens't work? Does it sit around on IRC all day???
in any paticular way?
Jymmm: Yeah, smoking and drinking. And then it gets drunk.
I'll bet Andy is re-coding right now and will soon be back with a good report for us.
And the beat goes on....
It seems a bit better,
I think I need to get the motor actually turning under power to really test it now.
And the beat goes on.....
And Valen, the problem is a bit hard to describe.
If I don't have a delay in the main loop, the encoder which is watching the output wraps round, the axis position wobbles around when the motor is rotated, rather than moving. If I put a delay in the main loop (which really ought to only have the effect of limiting the maximum following speed, which is bad) then it all behaves a lot better.
But is doesn't look like missed pulses, as the limits of the "wobble" are consistent.
I wonder if you are toggling output faster than EMC can read it
what are you reading the encoder with?
It is being read by a Mesa card
what about line drivers and cable lengths?
I just took the delay back out again, and with the Q as a _signed_ int it seems to all work.
I have no idea how a delay (initially it was serial debugging comms) masked the problem though.Very odd indeed.
it just wanted to fuck with you
Right, that's enough for today. Tomorrow I will figure out what I did to break the 8i20 driver (it was working, now it crashes EMC2, which is not what is normally considered to be "progress")
breaking things in new ways is progress in the field of programming
Then I will have a motor, driver and encoder, and can run it as a proper servo loop.
(the video I posted of the 8i20 driving the motor was basically driving it as a 6-pole stepper with 1000x picostepping.
We used the 8I20 to drive a 3 phase 320V step motor, worked OK
1000x would technically be millistepping
8x is microstepping...
i'm speaking literally
8x i'd call octostepping
In the pursuit of pedantry, I am going to start calling 8x "centistepping"
(working to orders of magnitude)
now you're just beeing mean
mungkie, that would by 10x
deci sounds better but its not base 8
Yes, decistepping. Whops.
ahh there ya go
deci steping, its just base 8
There are 10 types of people in the world, those that understand binary, and those that don't.
base 8 is oct
so it would be 1000
yes that is 8 in binary
duodecimal would be 1100
Anyhow, I have uploaded the working versions of the software to the Wiki. If anyone else fancies a go at a homebrew resolver to quadrature convertor: http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?ResolverToQuadratureConverter
But if you have any sense, then you would buy the Mesa or Pico versions.
what is buy things?