Just this once .. here's the log: http://www.linuxcnc.org/irc/irc.freenode.net:6667/emc/2008-11-17.txt
i've thinkin of Network multi-player Game, like counterstrike or other RPG game... they SENDING Data with 'REALTIME' alike
cos i 'almost' never find any jitter on them
except if there is 'to many' People on-line that day
it would require: Modem / LAN
you are being fooled by the game software
you as a human wont see the errors in jitter that a machine sees
(yes it is require data-pocketing that IS far from realtime, i understand)
how about PC --Data-pocketing-- Buffer -- uC/FPGA --- Machine
delay loops cannot work with delays
my last idea:... unless we bring G-code-executer being external as uC-programmed...
ARM processor / minimum PC
G-code executer contain PID+feedback processor on them
if PC -usb data-pocketing- MinimalistPC-WithPort -to the machine (servo amplifier)
"minimalistPC-withPort" would do the RTime...
just an idea, would i like to shere
use a card which has already has an fpga set up for the job eg Mesa Electronics Anything I/O boards
:) mesa ...that would lot cheapper and less in difficulties...yes still
it is just idea for today laptop problem, i guess
since laptop do not has it any more.
:) suggestion only
laptops have a number of problems, just not designed to be in a machine room environment
i gtg... have a nicest day
oops, looks like I just bought an olpc
* archivist expects EMC on it in a week
I hear that laptops are bad for realtime :-P
no excuses for you
do they have any ports?
no useful ones
* archivist wants to reduce the hate of usb one day
just need another few hours reading the spec and a logic analyser and reatime usbdriver and an fpga and month or two of time
oh I fully expect that someone competent with enough dedicated time could come up with something .. I'm just not sure it could ever be emc2
for instance, "send 1ms time sampled positions out on usb and hope for the best"
I think we can get enough packets both ways as there is a synchronous mode
archivist btw i've not forgotten that linear rail, i may be in warwick next week, possibly...
all it takes is, as I said, lots of effort by someone competent and dedicated
that person's sure not me
pjm_ Im bidding on some bits in brum which are pickup. may manage both on same day
archivist good idea!
seller wants pickup before 26th
ok well if I go it'll probably be next week
jepler: how did you run emc2 with valgrind ?
modify emc script and command line failed
micges: three main ways to run parts of emc under valgrind: * modify the runscript to invoke valgrind at the desired time * put it in the inifile: [TASK]TASK = valgrind milltask * if it makes sense, run it at the commandline manually: valgrind userpace-component
micges: depending what you want to valgrind, you may need to use a different method
emc shutdown doesn't work proplerly when valgrind milltask, because the process name is not 'milltask'; you have to kill it manually after each run
1 leak in nml lib
did you update from CVS in the last day or two?
4 days ago
update again. that leak may have been plugged :)
I'm not sure of it, but seb did a few commits correcting problems found with Coverity, and a couple were in the NML library
I see his commits, I'll check it
I think it would be accidental if the leak is fixed, but it doesn't hurt to check
is there way to restart RT without restart system ?
yes. stop RT then start it again
if your EMC runscript doesn't stop it, you could try halrun -U
or /etc/init.d/realtime stop
is hal_lib.so exists ?
sorry error show that hal_lib.so is exists
with halrun -U ?
what is valgrind?
a memory usage checker
ERROR: Module hal_lib is in use by timedelay,debounce,wcomp,abs,mult2,mux2,ddt,flipflop,conv_float_bitmap,laserctrl,vctrl,hctrl,pid,and2,or2,dio,and10,motmod,threads,trivkins
ah - ok
hmmm. that seemed like userspace. oh well
<jepler> emc shutdown doesn't work proplerly when valgrind milltask, because the process name is not 'milltask'; you have to kill it manually after each run
SWPadnos: is that where you are?
Essex Junction is 8 miles away
I suppose I should go get healthier by getting some breakfast. bbl
SWPadnos leaves and everyone shuts up. What's up with that? :)
we defer to the master....
So I see
har har har
I mean, uh, cool!
which standard .h defines true
I don't know that it is defined for C. it's in C++ though
doing my engraving "patch" for ttt.c
archivist: everybody and their dog defines it, you should too!
* archivist shocked, compiles and doesnt crash still
* skunkworks_ only has cats
* archivist needs a new pussy....cat
#define true (1==1)
#define FALSE (0)
#define TRUE (!FALSE)
#ifndef TRUE, of course
#define TRUE FALSE
#define TRUE MAYBE
#define TRUE SOMETIMES(TRUE)
if ((!true)&&(!false)) printf(maybe\n");
#define TRUE 911
should be 999 like the UK
well last coffee arrived, and code walks around outer of the character glyph :)
and doesnt crash!
good job boss was out most of the day :)
is source forge having a problem today?
stustev: what kind of?
the webpage opens (albeit slow)
I have received almost no list mails today. I submitted one about an hour ago. I don't see anything.
I saw one from you
maybe it is my problem
archivist: was it about the cincikins.c?
2 hrs ago cinci problem with oscillation
this must be my problem - I will look into it at this end
stustev, I haven't gotten any since yours either
that was the last for me
same here (last from stustev )
one before was dac problem
or tuning problem :)
specially with the reply
heh, yeah. "What's tuning?" :)
archivist, good day... what exciting new additions to TTT are you working on?
how is it going
just started, got an algorithm idea, started coding
have a third executable
I would be very happy to help test... I have been trying to figure out how to do that for a long time
has anyone looked at the cincikins.c?
I have it open in an editor, but I need more coffee :)
so I guess I'll go grind that up. brb
* archivist had coffee and now its Monday play time see ya tomorrow
stustev1: with those types of kins.. YOu would not have to level/tram machines that don't have a physical connection between say Y and X/Z. Cool Just bolt the machine down and make a few measurments. Then adjust the kins :)
kind of the idea but adjusting the last few thousandths out of the machine is tedious
the machine sits on pad and is not bolted to the floor
stustev, is there a diagram somewhere that shows what the various compensations are?
to adjust it completely I need to pull down on a corner
I have not created one
(ie, where does ZcompX fit geometrically)
ZcompX is supposed to move the X axis while the Z axis moves. This will allow the center of the tool to stay above a point on the table
oh hey, I could actually read your whole email and find that out :)
you could do a ball bar sphere test
fenn: i am not trying to determine what the error is - I know what the error is - I want to compensate for the error
I hate (not really) technology. I had this machine running with MDSI's opencnc for years. Did the adjustments with angle plates. Made a lot of good parts.
but - emc is cooler.
Now, with EMC2 and a laser tracker I see the capability to comp much closer.
paralysis by analysis
emc is not just cooler - it is much better ( I don't know the current state of OpenCNC but it is windows based and I don't want to know)
hmmm. I think you'll have a problem compensating Z the way you want anyway (even once the kins works)
if the spindle isn't perpendicular to XY, there's no way to make a perpendicular hole
even if you move the tool so the endpoint is in the right place, the other points further up along the tool won't e on the correct XY location
SWPadnos: the spindle is perpendicular to the XY plane. I can adjust the A and B axes to have that
(the compensation only applies to one point on a non-perpendicular tool)
so your spndle moves would be in W?
SWPadnos: good point on a 3 axis machine. (what I was thinking)
err - "vertical moves"
ok, so this is compensating for the whole head moving vertically, not the spindle itself?
is there a quill too?
are the values in joints absolute machine coordinates or coordinates in the current system?
(should be absolute)
the W axis is a coordinated XYZ move along the spindle axis - at whatever orientation the spindle is in
I am using joints in my program. I think that is the way to do it.
if that's correct, then you have some far-from-zero number in joints, which then gets multiplied by the correction factor
you think you're at 0,0,0 - but you're not
theoretically, changing that comp number should cause the machine to bounce to a new position and then sit there
what's the Z travel?
14 inches total
hmmm. that's not much
(well, mathematically ;) )
I was at Z home
the joint display showed zero
I suspect the limit error is a NaN problem somewhere, but I don't see it yet
I put 'setp cincikins.ZcompX .0001' in the hal window
the machine sat there
(or something like atan2(1,$really_tiny_number)
I moved Z and the machine jumped as described in my email
when the machine was not on Z and I set the parameter the machine jumped immediately
ok, that's what I expected
that behavior is not what I expected :(
not what I want either :)
that's one reason why you can't go to coordinated motion until you've homed
14 inches is about normal around here :) (for a machine like this) :)
there's no way to know if the machine is in a place where there's a kins solution unless you've homed
I have always homed the machine
this kind of compensation changes the joint positions (as you know) - there's not much you can do if you turn it on away from your reference point
it's like changing a scale on the fly (away from zero) - does't quite work the way you want ;)
I expect the jump if I am not at zero - I don't expect the oscillation and errors
I have the parameter so I can set them in the hal file
they will be working all the time
oh, right. the jump is what I was expecting - not the oscillation ;)
my question is why did it oscillate?
oscillation is not good for finish or tool life
I'm not sure. just to confirm, did you set the parameter at 0,0,0 then move away, or did you set the parameter when the machine was not at 0,0,0 ?
when I start the machine the parameters are set to zero in the hal file
I have issued the setp at Z zero and with Z not zero
when Z was at zero the machine did not move until I tried to move Z
when Z was not at zero the machine moved when the parameter was se
moved as in oscillated
I was expecting a very slow X motion as I moved Z
have you scoped the motor-pos-cmd outputs?
about .0001 X move per inch of Z move
yep, that's what I'd expect. please scope the command output, PID output, and PID error
does the cincikins.c look as if it will do what I want?
I think so, but I'll be able to tell better when I load it up on a machine where I can compile it :)
what would I be looking for?
running the code in my head is a bit less accurate (and much slower) than trying it in sim :)
I want to see which place the oscillation starts
I will post a picture in a little while - thanks
oh, also stick motor-pos-feedback (or whatever it's called) on the halscope
might be worth to convert the joints and pos-> to hal pins and scope them
I was thinking that also - just output everything for debugging
are the HAL files and ini in the same directory?
stustev: it feels funny that you have the same compensation calls for forward and inverse
* alex_joni didn't completely think about it, but a gut feeling is something to consider sometimes
that did seem odd to me as well, though it makes sense for this part of it (it's linear compensation)
alex_joni: the signs are reversed but maybe that is the problem
the rotational part may not make sense
the rotational part is for the geometric compensation - the error between the centerline of the spindle and the centerline of rotation of the A and B axes
bbl - must go make pictures
stustev: you might be right that it's the same transformation, but it feels odd that it's in teh same place in both cases
what bothers me about those calcs is that there seems to be no offsetting to an origin point
that may be OK, since it could be defined that machine zero is the origin for the calcs (which makes sense)
yeah, you always need that to be 0,0,0..
can't change that for kins ;)
I guess I just wonder if the translation and rotation origins are actually at the same place :)
but I guess it's trivial to make the translation origin match the rotation origin, so no matter
the picture is on imagebin - it is not when I moved Z - it is when I changed to MDI
[20:20:59] <alex_joni> http://imagebin.org/31409
the first time I setp I was not at Z zero - I must have been in MDI
the second time I was at Z zero - I changed to MDI to move Z and the jump happened
the first time the X axis oscillated 6 or 8 times before stopping
K`zan, good afternoon
sunny in Seattle? nice day here once the fog cleared
Today it is, yuk :-). I *like* rainy days :-) :-) :-).
Has a nice thick fog this morning here too.
CNC stuff is kind of floating while I sort stuff out. My little pipedream mill I think I can write off as an educational experience. Time to move on to the HF uMill, NOT going to be cheap :-/, which comes as no surprise.
I suppose its time to go rake some leaves.... it doesnt take long for them to turn into soggy goo
Learned that my drive (HobbyCNC) runs unipolar which derates the steppers torque a LOT :-/.
Yep, true, but makes for a nice slide, it you are into that kind of thrills.
cool.... I have a rotary table on my bench that I am trying to figure out if its worth putting a drive on
it would be nice to be able to make Al timing pulleys
IIRC, (was on my first vacation last week in over 10 years) but the steppers need to be increased by ?1.4? to make up for the unpolar driver.
Ah, rotary table and another axis :).
yes.... I took some of the table apart and it had hardened grease everywhere... very poor quality mechanical parts, but it does go around!
yay, it seems like my package is in sweden now
Lucky you, that is a ways down the road for me at this point. Between steppers and mounts, it is going to be about $400-500 to do this.
That would be almost the direct other side of the earth from me, hope you are closer to it anonimasu :-).
stustev, still around?
stustev, just for kicks, try doing your comp experiments using a different GUI, like tkemc. (I don't have any specific thing in mind, but it could help zoom in on the problem)
stustev, what mode did you switch from (when you went to MDI)?
K`zan: oh, I have all my stuff, just that making mounts for the mill is hard
K`zan: I'm waiting for micrometer blue and a scraper -_-
howdy yall, and thanks for the help the other day, I was not able to get the functions inline but oh well
scrapping is fun
hehe scraping you mean?
you need some blue and a scraper?
I have them comming in the mail
I dont know how much I'll be scraping :p
you have a good flat?
yeah a really good one
(at work that is)
or are you just going to match some bearings? ah awsome
[22:15:54] <anonimasu> http://www.powells.com/cgi-bin/biblio?show=HARDCOVER%3ANEW%3A9781114266612%3A92.95#
<- I've got that comming too
you looking at scrapping for fitting or you looking at fancy?
I've been pondering making a small lathe :)
and I figured scraping is a good place to start at
dont scape the upsides of the ways
despite how cool its going to look :)
anything that will leave a hole for dirt to sit on
oh, it'll be box ways
I just think they are easier to make ;)
practice a lot with your bit's tip, you really want to be good at keeping it sharp
well, I'm far off from making anything like a lathe.
you could with a good scraper widdle down a massive hunk of iron :)
I want a 500mm bed machine with a toolchanger
my plan is to rough it out on the big mill of mine
cast iron base or bold on?
I didnt decide yet :)
not aluminum though right?
hot rolled construction steel in worst case..
if your going to rough it on your mill, your really not going to need to scrape that much
if any at all
unless your talking about frosting and flaking
the pretty hook looking stuff
just to get it super super flat..
you shooting at sub .0001?
about there would be sweet
cause that would be a fine work of art for sure
less then 0.01mm is what I'd like
you can do it, scraping is a blast
how do you make a box way?
either in a T or inverted L
tell it it's fat?
...in those jeans
is that a fat joke?
so it is not really a box shape...
gezar: If the bon bon fits...
I think they are easier to machine and get really nice
never climbed in to look at one
seems like all new macnines go for box ways
new machines are using rails for the most part :(
not the heavier ones
but big heavy lathes have a fancy v set up though
less stuff to mess with in terms of fitting
[22:26:22] <anonimasu> http://www.machinerywholesalersne.com/images/CNC_Bed_Mills/cnc-2616.jpg
anonimasu : you going to use a shim on each side?
there you have box ways
I've been thinking of making a small XZ table to strap on my manual lathe...
just machine it very tight and scrape
I just want some free time to piddle once again
oh, I made a 90 on my most recent cal 3 test :)
gezar: well, it's a time/work tradeoff
I have a southbend I want to restore, I milled the bed, and sorta screwed that up pretty good, its goign to be a lot of work to get it fixed
well in this case work==money, its hard to find a big machine to work on
gezar: I want ot machine as much as I can as that's easier to find time for then scraping
but my mini cnc machine is doable, I have zero time
anonimasu : since youll be rough milling, and using square ways, your going to be at .001 right off the bat
just indicate everything in really good, so that when you make the matching cuts they are || with each other
yeah, that's what I figure too
and you know, if you get .0005" off the bat, call it good, and buy some trucite, and scrape that stuff
ah the big mill of mine's a nc mill so getting them paralell should be easy
you can get enough trucite to cover your moving surfaces for like 20 bucks
and linear scales so getting even closer should be doable
then scrap that stuff in and add your oil groves and you will end up with a uber machine(they use uber on max pad comercials to describe the absorency so I can officially use it now )
anonimasu : how does that sound versus using metal on metal?
hehe, im in north sweden getting micrometer blue is hard :p
oil paints man, oil paints
or wait wait wait, persian blue, or red, um, engine repair companies?
yeah, but it's hard to find here :)
engine rebuilders use the red stuff and its better to work with, you really dont want to use the oil based stuff
I can squeeze you some into a container?
that sounds really really bad, but I can do it
thanks for the offer! though not yet :)
[22:34:02] <anonimasu> http://www.toolfastdirect.co.uk/acatalog/stuarts_micrometer_marking_blue_large.JPG
that's the stuff
let me know, ive got a 3 " tube
oh wow, let me see if I can find a link of what ive got
the goal is to make a machine rigid enough to bore out some bushings and stuff
yeah, your going to want to use trucite
i finaly got in contact with rogern. he will soon be sending me all the info from the anilam control he converted to emc.
but im like 500 miles away from my machine.
and then i have to figure out how to hook up the vfd
SWPadnos: there are only two modes to choose in Axis - manual and mdi - the same action in TKEMC when changing from manual to either auto or mdi
stustev, right - I realized that you must have been in manual mode to home
Axis automatically switches to auto when you run a program
* JymmmEMC just got partial order from deal extreme... I'm happy so far =)