03rayhenry * 10emc2/tcl/tkemc.tcl: disabled calibration and testing menu items until scripts are gone or fixed.
Jacky^ is now known as Jacky^afk
03cradek * 10emc2/src/hal/drivers/hal_ppmc.c: extend the counter to 32 bits, the emc1 way
new here whats up?
anyone here need a good deal on a VFD?
to use as a phase converter?
what's up cradek?
my mill moves again
I found a really great surplus power supply
I scored the deal of the century today
a pallet of vfd's?
new in the box, with manuals, VFDs
wow, holy crap
all three phase input, so derate accordingly for single phase
they're "Benshaw", which I had never heard of, but has a web site with good info
the 10HP has card slots for PLC interfaces and such
the 7.5's and up were $25 each
the rest were $15 each
no, the 2HP's were $10 each
anyone need any?
you probably shouldn't have told us how much you got them for
I'll pass along my deal to folks who will *use* them
folks working on EMC, anyway
that's a terrific gesture
I'd rather you didn't just resell them, but if anyone needs one...
some of us are working on small (or nonexistent) budgets
yeah, many open source hackers are
believe me, I've met quite a few
I'm just waiting for someone to send me a complete minimill
I'm sure it'll happen any day now
that might be more of a curse than a blessing
jepler: what are you going to do with it?
CNC it up?
djb_rh: with a cnc minimill I'd probably mill circuit boards mostly
I got one recently
and have the CNC kit from Stirling Steele on order for it
I dunno, though, I might get back to this dxf-to-ngcode converter cradek started two years ago
and will likely use it for that purpose
that would be nice to have
(the dxf converter, I mean)
how do you know which side of the part is the inside with dxf format?
dxf polygons or dxf lines?
there's an implied surface normal from the direction of the triangles/quads
cross product of the vectors from point 1,2 to point 1,3
the dxf file should represent a closed surface, dividing space into two regions. The "inside" is the region that's finite.
ha, that's true too I guess, but maybe not so easy to figure
fenn: with the scheme I described, the surface normals point "out"
cradek: so it depends on the order in which the vertices are listed?
btw know of any .stl viewers for linux?
looks like I wrote a stl-to-dxf convertor, and toolpath (and maybe lots of other things) display dxf
[03:47:21] <cradek> http://timeguy.com/cradek/stl-to-dxf
seems like a dxf to stl converter might be more useful
what produces or consumes stl?
that's just as easy to write
i mean, somewhere you are going to have to figure out the surface normal direction
I told you how ... it's trivial
what app is .STL ?
just about any 3D CAD app
.DOC is M$ Word, .STL is ___________________
.STL is the industry-standard format for stereolithography files
it's like DXF
but there's actually a spec
(only actually pretty standard)
dxf is just "whatever loads into autocad"
DXF is AutoCad
DXF is "Data eXchange Format", not AutoCad
though they may have been the originators of the format
Jymmm: it's like .html, but for 3d graphics
3d solids, actually
no texture information or anything
it's basically a surface with a mesh of triangles representing some CAD model to some accuracy
would it be hard to hack AXIS to work in standalone mode? as a general purpose g-code viewer/verifier
fenn: probably, since AXIS doesn't actually interpret the G-code, it leaves that to emc
though it's probably possible to make a standalone G-code and canonical command interpreter
SWPadnos_ is now known as SWP_Away
(tricky - wasn't that? :) )
no time for rebuttal
of course, cradek and jepler are the ones to ask, but AFAIK, AXIS basically sends teh program to emc, and pipes the output of the interpreter into axis instead of the TP (or something similar to that)
03paul_c * 10emc2-auto/wiki/ (12 files in 9 dirs): "Auto update wiki from a cron job. Thu Dec 15 05:30:01 GMT 2005 "
03petev * 10emc2/configs/m5i20/ (m5i20.ini m5i20.var m5i20.tbl): Initial revision.
03petev * 10emc2/configs/m5i20/ (m5i20_io.hal m5i20_motion.hal): Updated to latest pin names, etc..
03petev * 10emc2/configs/motenc/ (motenc.var motenc.ini motenc.tbl): Initial revision.
03petev * 10emc2/configs/motenc/ (motenc_io.hal motenc_motion.hal): Updated to latest pin names, etc..
Jacky^afk is now known as Jacky^
morning operose :)
how's it going today?
im reading the news.. :)
I'm a bit late for work
just got here
oh, you know tonight the moon will be at it's highest point for the next 18 years
have to share that with everyone
can you observe it ?
after this week, once I'm back to regular working hours (rather than 60 a week)
but I'm gonna put together an emc box
and get some motors
don't have any tools to control yet
but I wanna play around
you can watch the motors spin on the table.. it's more fun than you would expect
that's the plan
id like to stress it in avery way :P
too see how it work
there's some cool gcode file test in emc
if I remember well 3D test
may written by rayh , cant remeber now..
operose already tried the backplot feature ?
its another nice thing :)
I haven't had a chance to play with it really at all
been reading documentation
while at work
but lately it's been wake up, go to work, come home, go to sleep, wake up, go to work.. etc
is there any chance this would work with freebsd or netbsd?
maybe with linux emulation or something
I think all is possible ..
has it been attempted?
but how many time and work could it required ?
note sure about the advantages youll take
im interesting in embedded linux system
advantages for me would be that I already have a bunch of netbsd/freebsd machines I wouldn't have to reinstall on
ive a small handheld pc ..
yeah I'm rather interested in playing with embedded netbsd
I found my handheld required a lot of hacking to work with emc
patch on kernel, cross compiling, drivers ..
hw adapters and more
i done the idea for now
I confess I wrote 3D test
I needed something to see if I had the axes reversed in the display.
cool code ;-)
operose: You anywhere near antwerpen?
antwerp, I am close to
not sure what antwerpen is
I saw the .be on your addy
there was a company giving away .be domains a while ago
dunno if they still are
oh I see.
grabbing tarball now
Long ago there was an attempt to run emc with freebsd
who are the mages that tackled such a beast and where must I adventure to to find them?
the emc2 stuff sifted out anything but linux.
First task would be to find adeos and rtai for bsd.
if it's too much work, honestly, I'll use linux
interested in getting it working sooner, rather than later
it takes me so long to finish a project, the last thing I need to do is port emc to freebsd
operose: I briefly looked into realtime freebsd but it looks like there isn't one and there's very little interest.
did you ever get your ppmc working again?
it's a wonder I still have power here...ice storm in progress
good morning johnboy
hey les: math question
if you represent a set of boundaries as a space, does it make sense for a rectangular space to have rotational boundaries?
for instance i have a 6d vector describing the position and orientation of an object, and i dont want it to go further than <x> units in such and such axis
news fro EU: http://www.europarl.eu.int/news/expert/infopress_page/019-3536-348-12-50-902-20051206IPR03225-14-12-2005-2005--false/default_en.htm
only for “specified forms” of serious criminal offences
i bet that will last precisely 0.2 seconds
we can send away logger_aj
haha they dont even pay for it either
sorry fenn was away
may vanish any second...ice storm
fenn, sounds like possibly a job for quaternions
They are unique in that they require only 4 numbers, and they don't suffer from gimbal lock.
rayh is now known as rayh_away
crappity crap.. ij=k, but ji=-k
haha yeah non commutative
you can create a space from quaternions?
transforn them...I'm no expert though. I generally use vectors, but they DO gimbal lock
lots of stuff on the web
you read it and tell me and then we'll both know
i dunno.. complex matrices.. not my cup of tea
I did em in school a lot...
ahem 25 years ago
how does a vector's gimbal lock manifest itself?
not enough resolution?
division by zero?
well, certain orientations lose a degree of fredom
let me give an example...
latitude and longitude
say you were traveling near the north pole
and want to know lat long
as you get close to the pole any small movement makes big changes in longitude
more and more sensitive the closer you get
right at the pole longitude basically has no meaning
makes math blow up
that is classic gimbal lock
Till was talking about applying quaternions to the emc a couple years back.
they are used a lot in 3d graphics I guess
really.. i thought people used transformation matrices
but that lack of gimbal lock is the main reason for using them instead od matrices
OpenGL certainly uses transformation matrices everywhere, no quaternions in sight
People usually think of the movie apollo 13 when gimbal lock is mentioned , huh
Apollo used a 3 gimbal gyro to save weight
so certain poses were not allowed
they are both the same type of singularity.. it makes sense to call it the same name
if they wanted to pitch they had to roll first
I'd have to read up on it some though....I know openGL does not use them
jepler: you can represent a quaternion with a complex matrix
or a 4x4 matrix
another dumb question: inertia is described by the mass term in f=ma right?
but rotational inertia is described by J in t=ja
so how do you convert rotational inertia (N*m^2) to kg?
j is polar moment of inertia
torque= j* angular acceleraTION
shift key sticks.
still there les?
We (machine designers) usually use mechanical impedance for such things
It's a complex number...in MKS units are newton *sec/meter
right.. i'm trying to subtract the "apparent mass" caused by the inertia of the motor
er, add not subtract
laplace transform of force/velocity
* Jacky^ confused :/
motioneering does that automatically saves a lot of calcs...
for example in my machine x axis motor rotor inertia has the same apparent mass as the 500 kg gantry!
i believe it
les_w: thats very cool app
yes very handy
but how do i get that number if i'm calculating by hand?
id like to see in the next future some similar app open source :P
oh gosh let's see
you need reduction ratios and stuff
would be very nice
so let's say a 5mm pitch ballscrew
and a 1:4 gear ratio
ok. first establish the relationship between angular position and linear position
I jsut drank a cup really really fast
so x=( angle/2*PI) *5mm/4
x = .25*rev*5mm damn
i was trying to balance units.. but angle has no units
you guys are crazy smart
well, radians is dimensionless is the better way to say it
nah...but spent a few years in engineering school
no thanks to the american education system here
this channel is an engineering school in itself ;)
anyway you can start taking derivatives...
I've got my mind completely made up to go get a 2 year mech. engineering degree
to get velocity?
to start learn ? O_O
to get a degree
don't need college to learn
just for the paper
I liked school.
i think mechanical engineering would be much more fun than computer science, thats for sure
les_w: right, all depend on teachers
i've always considered the idea
dx/dt = .25*5mm*(da/dt)
time for "brunch" as my boss puts it
hmm been a decade or so since calculus
let me get pencil and paper
les_w: the wierd thing, I was reading today in the it news, only 25% of students who finished theyr school dream to do the work for what they have studied
[15:11:16] <Jacky^> http://www.repubblica.it/2005/j/sezioni/scuola_e_universita/servizi/giovanilavoro/almalaurea/almalaurea.html
yeah because school beats the life out of you
sorry .. is Ialian
so, the school sometime fail :(
in 50 % of cases
another strange thing
ok how about this: dx/dt = .25*5mm*(da^0/dt) == 1.25mm/dt
ok, now you have a relationship between angular accel and linear accel after 2 derivatives
in some UK school are experimenting to pay students to learn
they say are getting good results
dx/dt^2 = 1.25mm^0/dt^2 == 1/dt^2 ??
you can plug into f=M*d^2x/dt^2 ant T=J*d^theta/dt^2
fenn: ugh ! whats that ? -> les_w> you can plug into f=M*d^2x/dt^2 ant
Jacky^: that's the shortcut they tell you after you do the hard work
* fenn looks scornfully at his units program.
Im really thinking at open source ver. of mootionengineer ..
or similar app
the formules should be free
the problem could be motors data
Jacky^: it should be easy enough then without the database.. get writin' code!
really, les_w could help in calculations functions
it could be a nice app for emc
well, with the units command there's not a whole lot left to do but this particular bit of calculus
unless you're doing a robot arm or something
robot arm move in 3 dimensional space, it should not hard to implement routines for that
just a lot of math
a guy some day ago pasted an interesting url about 3d space..
yeah, right, tell me when you're done writing a robot arm trajectory planner
can you wait 3-4k years ?
oh I'll get right on that
ah, operose was you , right
I need to get readin' the wiki
thanks operose, i knew it was easy after all
yeah it was really easy
only like 5 lines
what's the best supported controller right now with EMC?
I think it's called a controller
servotogo xylotex or pico systems it looks like by the wiki?
eh, depends how much money you want to spend
what if I don't want to spend a lot? :D
power is bumping...had phone call about trees falling from ice
fenn let me get you a link with all the calcs
easier than deriving them here
also irc text is real bad for doing math
operose: get a hobbycnc board.. it's like $75 and has the amps built in
les_w: actually i find it to be easier to read than standard math notation
cool the xylotex is pretty cheap
hobbycnc you say
you can spend upwards of a zillion dollars on "the best" driver electronics
no point in me doing that though
just playing for now
hobbycnc is probably the minimal, which is more than enough for your first system
you do have to solder it together though
yeah that's fine
operose: what are you intersting in ? milling , drilling , wood, steel ?
oh and you still have to get a power supply
well, honestly, I don't have any machine tools whatsoever right now
but I'd probably put together a little mini-mill if I could get the CNC stuff working
got some random projects in mind
[15:39:46] <les_w> http://www.techno-isel.com/Tic/H834/HTML/H834P009.html
some calcs here...
oh, I dont think you need to spend a lot of money then
what are the tolerances this will work to?
I'd like +/- 0.002" at least?
you can take a lot of advantages learning to hack a cheap machine before
I guess that depends on the machine though
les_w: go on :)
most calcs we want are on that link I think
and the other pages of the site
* Jacky^ bookmarking
fenn: sorry for the break, go on
so i should end up with something like.. (kg*m^2)/((m/s)/(2rev/s))^2 = kg
les_w: great website
would it be an alright idea to get a package from hobbycnc with controller and steppers for $235 or could I get the steppers cheaper somewhere else like ebay?
calcs are there too fenn
nevermind that question
always looking for the easy answer
I'll search around
just make sure your steppers are compatible with the driver
and that they still work
demagnetized motor looks fine on the outside but won't do anything (or do it poorly)
so in other words, maybe shy away from motors on ebay unless there is a good guarantee and good feedback etc.. like usual
well, if the person doesn't know what it is, they probably don't know where it came from either
out of interest, has anyone here managed to build adeos patches into a recent kernel version?
delacroix2: I think most people are using the latest BDI version which uses kernel 2.6.12
did you get the equation for equvalent mass fenn?
woops sloppy notation
paren around 2* PI
you can solve for m as well
still not sure how the 2pi got in there
plug in j and lead and you will get the equivalent mass
delacroix2: if I read the wiki right, the kerner 2.6.12 is using magma, 2.6.10 used adeos.
magma is a verson of rtai, adeos is the kernel patch to make rtai work right
am i getting confused here?
lead is movement per 2PI radians that's why
delacroix2: see http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl/emcinfo.pl?BDI-4_Compile_EMC2
have read that
ah so j has a radians in it also
have followed it, and couldnt get the patches to apply right, im guessing i was using a wrong combo of kernel and patch version
delacroix2: I'm in the dark just as much as you
delacroix2: the rtai documentation is not good.
delacroix2: few here build their own RT kernel, and the ones that do tend to use older stuff (I use rtlinux on a 2.4 kernel)
delacroix: you were having problems with applying the patch or with building it after you patched it?
i can apply the patch, but i dont get a tickbox in make xconfig for adeos, also the kernel built doesnt have the adeos hooks
although the patch appears to apply fine
ah, ok, I'm learning here. Sent the ref so you could see where I got it from.
i have tried to get rtai working before, and the wiki is wrong in a couple of places, my real problem is that i cant find any release notes for adeos, anyone know if they have a webpage?
delacroix2: if you figure this mess out, please let us know (maybe add a page to the emc wiki)
cradek: will do
might try the rtai mailing lists
delacroix: try a make distclean? (wild guess)
fenn: oh have to put a square of reduction ratio in there too
or just dive lead by reduction ratio
so equvalent mass is 16 times bigger than direct drive with your 4:1 reduction...
(armature mass that is)
that ends up as 1600 kg apparent mass
yeah...gotta watch those grear reductions...square term...gets big fast
but fear not...
for your reading pleasure
SWP_Away is now known as SWPadnos_
i dont really know what inertia matching means.. i was just going to see how quickly it could accelerate a mass to top speed
i'll read it though, so i'll know
you don't have to precisely match motor and load inertia
but if motor/load is huge like in your example...
it swamps everything else
so the motor J is the limiting factor in how fast you can accel
steppers are not known for good torque to inertia ratios
otherwise known as angular acceleration
if you know the rotor inertia, you can figure out the torque curve pretty easily right?
if you know the torque and inertia, you can figure out the accel curve ;)
but angular accels of 10-20,000 radins/sec^2 are common in servos
some are as high as 100,000
talk about winding up fast...
i dont want the accel curve, i want the torque curve! *wah*
more precisely i want the current to torque ratio
ok just a motor spec
so that when i command a particular current i know how much torque is coming out
that should be given by the manufacturer
and torque curve is a constant with most
well, what if i dont have a spec?
hence the term "torque constant"
no spec? well you are SOL ;)
apparently, you can calculate the torque constant from the voltage constant
you can measure it with a weight or fish scale easily
but I'm skeptical, and I also don't have the conversion constant
les_w: duh why didn i think of that
too easy :)
yeah, torque from voltage constant needs some assumptions
string, weight, power supply
yeah - like max voltage == voltage at max speed
which isn't true for my motors, for example
little matter of IR
fenn do you have a power supply?
capable of constant current?
if not you can simulate it with a voltage source and series resistor
resistor value at least 10 times winding resistance
(if you have dc motors)
this is great.. you can get force feedback with just an encoder and a current loop
that depends on the speed, no?
why? torque is constant with motor rpm (you said)
er, les said
you can get torque with just a current loop
yes but not feedback
so you dont even need the encoder to get force
that's in the drive electronics
i'm using a very very simple current loop
comparator and a dac
sure - the drive monitors current, so it "knows" force
just create a known torque with a pulley, mass and string
torque= pulley radius* mass*g
cank in just enough current to lift it
question - could you get the torque constant by letting the weight fall with an ammeter across the motor leads?
no external power supply?
I don't think so - the current would depend on the speed and the external load
you can measure the speed with an encoder though
yes you could. You would have to put a load on the motor and time the drop
a little sensitive to friction though
you'd have to measure pretty accurately how much slower the weight drops, so you'd need a good timer rather than a power supply
jtr_away: your servo driver should work as a constant power supply
car battery and 100 ohm resistor is cheaper...
126 MA current source! (kinda)
and you don't even have to take it out of the car! ;)
no machine yet, no drivers - just hanging around trying to get an education and some mental exercise.
only need a 2 watt resistor or so...
jtr_away: just start asking about motor drive electronics.. plenty of mental exercise there
especially microcontroller programming.. that's a good one
Been doing embedded programming for years, just not with a normal language.
Done some assembly, c, RATFOR. Built a test executive to test a lamp controller I designed.
few years ago, built an analog speed controller for the motor on a cassette deck
had to recover a tape made with dying batteries.
follow the bais ...
yeah bias actually records