dmess: i really havent used it enough to teach, but read the manual, JMK did an excellent explanation ( i could work thru the examples and see it work, so can you)
just go thru it slow
i'll try it.. i NEED it so i gotto
see if i can get a lesson on a regular scope tomorrow
a scope is just a way to visualize electrical stuff. you can see if its bigger smaller or faster slower. you get 2 rulers to measure with, up down is bigger smaller (volts usually) and left right is time. so you can see what changed & when.
its just like a vernier or micrometer or thermomenter
but with lots more buttons
dmess: did you ever get one of those cnc4pc boards?
you jumped to another thought, no never used/had one
maybe back later, traveling home now, try the halscope tutorial, it's actually fun.
dareposte1 is now known as dareposte
i have the speed controler and spindle index pulse cards in transit... SEE what I get
hopefully an early x-mas
anyone ever use vpnc? I can't get it to read a .conf file
I've used VNP, and VPN, never heard of vpnc
ok, which one now? =)
trying to get past a firewall using vpn
you using NAT on your network?
penn state firewalls most useful computers
have to log into their vpn to get there
you using NAT on your network?
then enable vpn passthru on it
man page for vpnc says you can enter "vpnc filename" and vpnc will look for "filename.conf" but it doesn't
eric_u1: Do you have a router?
then enable vpn passthru on your router
my problem is more along the lines of "vi somefilename.txt" and it never finds somefilename.txt no matter what
but it appears that vpn passthrough is enabled
maybe vpmc /path/to/filename
well, it works if I rename my file default.conf, but man page says I don't have to do such an ugly thing
well if it works, do it
good news is it worked, bad news is it wouldn't connect to the subnet I needed
so I'm still firewalled out of my computer :(
didn't even notice that I changed names
it's possible that my school might donate the pc for my machine
awesome o.O (still pending on that suggestion tho)
a supplier of frame extrusion http://www.minitecframing.com/
also in uk Im still looking for a better supplier I saw at a show
bah he wented again!!
hmm bug in gcode parser "G0 X#<curve1_xstart> #<curve1_ystart> z.5 (move to cut start)" error message says missing = where the real error is missing Y
archivist_emc: bug lerman_ :)
took me a while to spot the error as the message was sending me off!
archivist_emc: the proper way to deal with this is to submit a bug report at SF
I know but the lurkers in here may say not a bug, expected modal behaviour
well.. you can't expect it to say Y
but you can expect that it says modal operator or whatever missing
I think it might even be right
you have: G0 X<var> <var>
it would expect a <var>=value
I think it's right as it is..
I wonder if <var>=etc should not be after/in the middle of a move to stop ambiguity like that
not sure if it's allowed in the same block, but I think it might be
just found a real parser bug
damned silly thing wont allow a copy from a msg popup can only get a screen grab
I think they also go to stderr
cradek_ is now known as cradek
* archivist_emc types the words "near line 52 of x.ngc Named parameter not defined" er what named parameter!
as the parser had to lookup the parameter, it knows the name it failed to find so should be in the error message
I just had a lil' idea for the USB version of this,.. Just make the controller handle some of the feedback things if it can, such as the home or limit switches, so it can take action before the PC is able to receive that info
nah just get the usb comms deterministic and fast as thats required for the servo loops
have the USB also tell the controller how /FAR/ it wishes to move, (eg. instead of "step.. step", just say 15 steps."
archivist_ub: problem is just the massive latency USB has
esp. with communication from the machine to the host,.. always waits for the host to REQUEST the data :/
standard usb has latency yes
most problems could be solved by making the controller smarter i'd say
hence the req is in the realtime part :)
that's why i'm suggesting to like,.. not make the PC itself realtime, just have the controller take action :p
granted, it's a challenge
REAL challenge ;)
but, in the computer world, nothing is impossible
see what parts get moved to the mesa fpga and do similar
the home & limit switches can easilly be done on the fpga, so can the homing process..
the problem would be making it do complex curves as USB won't let you move the axis independantly at different rates as easilly as a parport does
unless if you program the movement into the FPGA's memory module first.. :/
fragalot: if you want to spend energy .. better use a RT-ethernet connection
alex_joni: but how will you waste your time on youtube while the machine is running if you do it that way ;)
hmm that would be nicer cabling
escape key doesnt work when a debug window is waiting for a click! potential damage
archivist_ub: that's to boost the economy
* archivist_emc begs for debug and msg to go to a scrolled window in axis so focus remains safe
aah, vmware won't boot completely, it's taking ages
fragalot: second eth?
archivist_emc: that's already implemented in pre-2.3
er what Im 2.2.6
alex_joni: that'd work on mobo's that have that, my issue is that i can't add the hardware I need to my boxes, would have the same issue with eth :p
fragalot: USB-eth ?
not for the RT link though, but should be fine for youtube
archivist_emc: in the development version of emc, here's what axis does with errors and operator messages: http://emergent.unpy.net/files/sandbox/notifications.png
jepler I thought a scrolling box would suit best
hm, this little plug-in power supply with a USB jack measures about 5.6V, and that's with over 30mA of load. maybe it's not so great a way to power a homebrew USB device for testing without endangering a computer's USB port
5.6 is somewhat over!
do you connect both the power supply and the computer at the same time?
SWPadnos: no -- this is just for smoke testing, it can't communicate with the computer this way
ah - ok. I had misunderstood your concers there
may get a valid smoke test
(but still 5.6V is within the absolute maximum ratings of the at90usb162, so it shouldn't cause damage)
5.6V shouldn't hurt anything on most microcontrollers
if you see smoke, get a different power supply ;)
still leaves me scratching my head about why this board is flaky
what's it (not) doing?
on monday it wouldn't even show its power LEDs. today it does (without me changing anything) except one of the 4 LEDs on it seems to be damaged (it doesn't light).
oh well, I should head to work instead of playing with my hobby projects
bring the hobby project to work ;)
hobby is more fun
archivist_ub: believe me, I know it
same here :)
Im playing with EMC as a slight work excuse
would be faster to get a hacksaw and file out
[12:45:47] <jepler> http://emergent.unpy.net/files/sandbox/tqfpboard-schematic.png
<-- here's my board -- USB, microcontroller, LEDs, and some I/O on headers. no specific use in mind, just to experiment with this new-to-me microcontroller family at90usb
did you pay for the usb developers software? iirc it's necessary to use the usb functions
fenn: no. there's USB software available under the MIT license from here: http://www.fourwalledcubicle.com/MyUSB.php
to make a "real product", I think that you may need to pay for an allocation of a vendor/product number. this doesn't matter much to me
for a bit more on that: http://www.fourwalledcubicle.com/files/MyUSB/Doc/1.5.2/html/page_vidpid.html
thanks for the links /me looks
I haven't actually uploaded any firmware with a usb function to the device yet (aside from the bootloader, which is preprogrammed)
hmm interesting /me contemplates usb IEE488 controller
it comes with a usb bootloader? (from digikey?)
fenn: the at90usb162 (and I think other parts in that family) are preprogrammed with a "DFU bootloader" which can transfer a new firmware over USB -- no need for the 6- or 10-pin programming header and a special cable
fresh from the factory it automatically enters the bootloader. after that, if you haven't modified the fuse settings, you can enter it by having a specific pin brought low when reset goes high; I think that you can also include a jump to a specific address to enter the bootloader from your own firmware
the downside is that the bootloader takes away 4k (?) of program space
is there a place in emc where the gcode is parsed into structs? (eg lnum gcode x y z u v w a b c i j k mcode... )
in the interpreter ;)
* archivist_ub is beginning to hate the center format arc with a vengeance
why is that archivist_ub?
hand coding end points and debugging
You don't use the g code wizard for arcs?
Arc Buddy here http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Simple_EMC_G-Code_Generators
my arcs are connected not on nice angles/coordinates
archivist_emc: I usually end up drawing them in cad ... I agree it's a big pain to hand generate them
they don't have to be connected ...
er but its the shape :)
cradek: so would I but when in EMC I cant get at windaz for cad
archivist_emc what kind of shape?
just a mo
[16:28:39] <archivist_ub> http://www.archivist.info/cnc/p1010246.jpg
hour detent (pic on its side)
the drawing has a large amount of guess work!
and let me just say that I don't really understand the drawing at all (but probably very clear with the part in front of you)
yes it sure does
just remaking parts that are missing or made so badly
this one is missing
well the one in the clock was nothing like (worked the wrong side of the pins)
archivist_ub is that a progressive arc like a french curve?
Im treating as simple arcs
BigJohnT: your running trunk on your plasma that blends arcs (like it already does with lines?)
ok - did you see the post on cnczone?
does that work faster?
why not just use R format arcs
[16:53:43] <skunkworks> http://www.cnczone.com/forums/showthread.php?t=65515
erfm scrollback underflow error
fenn: because they are less precise..
* anonimasu loves the heidenhain rnd command
underconstrained, not less precise
but usually it seems to do the right thing
I like doing L x10 RND 5 L y20 and get a round corner
the Arc Buddy kinda does the conversion for you from R to center format...
in soviet russia, arcs format _you_
and in the UK atm
blending... of the people!
skunkworks: I thought the arc line arc blending was on 2.2.6?
do not confuse blending with the naive-cam detector
lines and arcs have always blended with one another (no stopping)
I stay confused some time
the new code eliminates unnecessary tiny tiny arcs that some cam systems generate (why?) if they do not affect the path according to the given tolerance
was that in trunk only?
this is really not going to be an issue with archivist_emc's hand-generated code
BigJohnT: probably, but I'd have to look to be sure
no I was answering the post that skunkworks posted on cnczone
cradek: is it possible to make a script that will run EMC simulator on some G-code, and when it fails on Rtool>Rcorner, it will make an arc dith R=Rtool of that corner in g-code, and so on and in the end it will output new G-code for review and report : "I've rounded 20 corners, please check"?
dushantch: anything is possible, including that
that skunkworks pointed me too not posted
cradek: ok, but is it a good solution, as it doesn't change interpreter, and it's easily checked in simulator, and done only once?
with simulator I meant in some other sw
ah, I sensed that "is it possible" wasn't the question you really wanted me to answer :-)
personally I don't think it's a good fix for the problem
well I had to make an intro of some kind :)
for one thing, parsing gcode is hard
but parsing it exactly the same way emc's interpreter does it is easy - you just use emc's interpreter
which points us to the better solutions - which are some variations of "fix the interpreter to do what you want"
that's why I thought this would be easy, as it uses current interpreter
sorry, I don't see any advantage to your solution, only unnecessary complexity
isnt it better to write proper g-code in the first place?
yes parsing depends on the gcode 'variety'
but the problem is that machine makes decisions then, and how do we now that they're correct?
by programming it to make the correct decisions
how can we check it without running the real machining?
did someone call my name?
anonimasu: writing proper g-code is a lot more (unnecessary) work in this case
lerman_: duck and cover
cradek: you idealist :)
Not a bug. The parser can't read your mind.
should be able too by now
fenn: uh..yeah tool radius > corner radius is alot of work.
if someone wants emc to accept concave corners for compensation, he/she should continue work on the concave_ccomp branch I started. The work is partly done. IMO, other hackery to work around it is wasted effort.
Murphy says that when you program something to make correct decisions., it'll do the oposite, but only when you're not looking/when it costs a lot :)
(Although I have to admit, that it didn't have a very good guess).
anonimasu: it is, you have to add an arc for every concave corner
where is some description of the concave_ccomp branch?
I have very little patience for continuing to argue about whether it's good/bad/necessary/unnecessary
I dont even know what a concave corner is.
when that happens I just go do something else...
and I propose that adding of arc to be a new g-code which can be reviewed, and run on other machines with same results
anonimasu: g1 g1 where angle is < 180
ah.. I see..
cradek: I don't want to argue, I just ask
fenn: not right. arcs-line and arc-arc make concave corners too.
'inside/outside' corners depends on the direction you follow the path (more common term)
all of line-arc arc-line arc-arc line-line can make a concave
does that break the interp?
if you use tool comp?
anonimasu: depends on what you call "break"
no, your program doesn't run if you have an inside corner when using cutter comp
they describe a shape that cannot be cut, so you get an error (currently)
break I mean give you a error message about it and stop cutting..
that is by design
If the interp is asked to do something that is impossible, it stops.
so what with concave lines made of lot of small lines?
I would like a different design, and have already written most of it
ugh so I need to rewrite with hand compensation
archivist_emc: not if you describe a path that can be actually cut
Like making a mould from output of some CAM sw?
where is some description of the concave_ccomp branch? wiki has the word concave just twice.
cradek: sorry if I bug you :)
The CAM software should be smart enough that it isn't a problem.
whats for lunch cradek ?
lerman_: afaik not in real life , and not with some stupid people writing postprocessors
for use cheapskates emc is the nearest to cam we can get
the solution is to add a rounding g-code like on heidenhain controls.. so you can round corners, with a parameter ;) rnd tool radius
* anonimasu hides
anonimasu: and I propose that it should be done with a script which check all the places where interpreter bails out
No, the solution is to find someone with appropriate skill who cares enough about this to implement a change.
i think this is called g02 or g03
yeah and then you need to work out where the arc and line intersects..
(notice I said "change" not "fix")
anonimasu: actually that does seem quite handy
how to browse cvs?
I think throwing in a RND command between the two lines to round the corner is quire neat.. no calculations or anything requires..
see the pulldown below the file list with branch names?
not sure how to get diffs between branches though
adding RND is a non-trivial task on several levels
not least of which is that the interpreter only deals with single-letter codes at the moment
I've written gcode subroutines that: given three points defining non-colinear line segments and given a radius, converts them to two line segments with an arc connecting them.
when you selest a file you have a option to diff it
so that would look like an R with no parameter, an N with no parameter and not at the beginning of a line, and a D with no parameter - all of which are at least one error
but what if arc is so large that the second line would be <0
unless you change the interp to use spaces as delimiters, which would have other implications
It gives an error.
dushantch: then it errors
or if you dont define a tool it'll cut it
like tool tip on path..
I would rather add a -R parameter to g1.
round to the next line
I know, it would be nice
the heidenhain control's has lots of neat stuff.. for doing hand programming
fenn: sorry i see no pulldown below file list with branch names . theres a File and a General Options area.
tomp: there's: "Show only files with tag:"
above general options
please, where? (i've even text searched for 'files' and 'branches' ) http://imagebin.org/27776
eh sorry you have to click on emc2 first
thx. got to concave_comp now
i think this is what was changed: http://cvs.linuxcnc.org/cgi-bin/cvsweb.cgi/emc2/src/emc/rs274ngc/interp_convert.cc.diff?r1=1.106;r2=184.108.40.206;only_with_tag=concave_comp
my agenda is now officially full.. I had to put post-it's in, and put post-it's ontop of the other the post-it's
real computing, using a stack
*g* the hammer drops on 19 october
then the school board decides if i should get a free PC from them or not
hmm does quicksort work on post-it's
archivist_ub: not even bubble sort works
archivist_ub: simplesort works
but by the time you finished sorted, the post-it's probably won't stick anymore
gets stuck to the fingers
add btree index to the post-its
or avl.. should be good enough
Just try & filter out the middle one if you nolonger need it
>.< then realize you forgot what date it was stuck to
methinks doubly linked list
that's some fancy post-it's
messylist is the function I used.
this page is really cool: http://www.sortieralgorithmen.de/bubblesort/index.html
I've never pictured sorting algorithms in such a visual way
(bubblesort is really bubbly :D)
v3 looks nice
MSsort is quite fast :)
I never knew there were so many sorting algos :p
well.. it's a simple problem you need to solve :)
there are still a couple missing :D
I think some have multiple names tapesort->filesort
it's nice to see how they work on reverse ordered
lol, shearsort is the nicest to look at
I solved the Problem with image to gcode
it were, as you thought missing libarys
thanks for pushing me in the right direction
heh the stoogesort (nyuk nyuk nyuk) http://cg.scs.carleton.ca/~morin/misc/sortalg/
a-l-p-h-a is now known as a-l-p-h-a1
good night all