ack! Anyone around that can offer advice on excessive following error alarms?
I'm adding these axes one-by-one, running siggen to tune the PID settings. Yet when I run a part program in EMC or try to jog, I am getting following errors at random during the motion...
Primarily during accel
what is your following error set to?
FERROR = 1 MINFERROR = 0.005
and how does the following error look in halscope?
the position is tracking a squarewave fairly well
set halscope to trigger on axis.N.ferrored
I increased P until oscillation, then backed it off
is that .005 inch or mm?
then brought up I to get closer to command velocity
1 inch is way way too big then
(unless you are tuning)
okay, but won't decreasing it aggravate the alarms?
yes, I am tuning, this is a new machine
well, I believe the PID settings are quite close
if you are still tuning, increase min_ferror
I thought I was finished tuning; I had tried to run a part program
if you get a plot of commanded position (and ideally also velocity) as well as following error maybe someone can offer advise
10% feed rate works fine, but get above 75% feed rate, and it faults
well these errors mean that your axis is not tracking as well as you require with your FERROR settings
alright, I get a screen shot
so, you are not done :-)
I suppose not :)
GNieport: you will probably be adjusting p,i,d and maybe ff2.. (don't ask me how to do it though)
(not just p)
what type of stimulus is preferred for the scope screen shot
square, sine, triangle...?
hey guys, I started a general machining chatroom, and posted some links on the various forums
if you are interested, the channel is #machine on freenode (here).
GNieport1: I just make a short program that repeats short G0 moves back and forth with G4 pauses in between
then you can trigger repeatedly and see the changes as you tweak the numbers
cradek: do you tweek the numbers within hal - or do you use the axis config page?
I've done both
I type too slow to do it in hal I think ;)
that seems like a cool way to do it. keep adjusting the numbers as the program is running back and forth. (jmk might not like it) /troll
I always tune in HAL with setp
yeah you can quickly narrow in on good settings if you have a new plot every few seconds
yes - and you do it with just hal.
history helps a lot - you can up-arrow to "setp pid.0.Pgain 100" and just change the number
I am using HAL :)
that is nice. Playing with the at_pid - i did that
I only ran a part program to check the progress
[00:43:37] <GNieport1> http://imagebin.ca/view/8I-w_rk.html
lots of true - false settings
that is Halscope, P (8) I (7) D (0.1) siggen.0.square freq 0.5 amp 0.01
I could not determin how to plot velocity
it doesn't settle on target
More P = oscillation
more I = doesn't settle
More D = VIOLENT
D = 0.5 did a good impression of a telephone ringer
D is definitely touchy
Physically, the machine has about 0.0005 inch backlash. Straight coupling from motor to ballscrew
D can stop the oscillation of a higher P
encoder is on the screw or the motor?
I suggest you use emc jogs/rapids so you don't have this step response
then you can twiddle FF1,FF2,D to get very good tracking (possibly turn off I while you do this)
once you have excellent tracking you can increase I a lot to make sure it settles correctly and stays there
but, if your tracking is bad and you increase I, it will wind up and make everything worse
well, I wrote a gcode sequence to shuffle to and fro 0.5"
you suggest to run that and scope the ferror
so zero out I
increase P as much as possible, using D to prevent oscillation
Do i set P jsut below oscillation and leave it there?
okay, be back later :)
you can often raise it further
this may take a while
after you do that, use FF1,FF2 to track well (they are VERY touchy)
i.e. 0.01 increments touchy?
once you have that, increase I a lot so it fights you if you try to push it off position
yes the numbers will be very small, you'll see the influence in the plot
great, I'll get started
if you get it, please write a wiki page :-)
and beware this is how I do it - lots of ways I'm sure
just to recap, load up a kinder stimulus; increase p, increase d to cancel oscillation, repeat a while; try adding FF1 FF2 in small increments; bring up I to finish
yes that's what I've had good luck with
u guys are talkin about pid and didnt even let me know
okay, got to document so I can write the wiki - lol
if it tracks well, you can add I with impunity
guess whos got 40 dollars in sushi all to them selves! yay
if not, I will totally screw it up
good luck with that
haha "I will"
y is nobody opped in here?
we don't need it
what if someone flooded teh chan?
then I'd kick them
but we don't ever have any trouble
thats cool. cradek i search through logs from like years ago and i still see ur name way back then lol thats awsome that uve been helpin out for so long.
thanks I appreciate it
it's a good project
u and everyone else deserve it. it is. not too many people know about it though
I suspect everyone who looks for open source machine control software finds it
that's what counts
it was certainly the first hit I got
welll... not too many poeple think to look for it... thats the problem
think of how many people are using mach3
no skin off our backs, there are lots of answers to the different problems people have
my answer needed to be open source, so I joined the project and helped make it into what I thought was best
same for others
u can just tell how much more this is advanced and worked out than any other ubuntu distro.
its not just a tool kit
ok, little help please, what does this mean: core_servo.hal:42: Ini variable not found.
Roguish: check on or around line 42 of that file
this PID tuning is BS
trial and error with 6 variables is rediculous
it definitely takes practice
I am not making any progress, the system is totally unstable. Any upset and it goes into oscillation and shuts down the amplifier
you need to divide and conqueror
turn off all the FF gains (set to zero)
turn off D and I
I outlined a procedure earlier - are you stuck at one particular step?
Chris, I got P all the way up to 36 and D to 0.4. Any change in amplitude and it falls apart
sorry I don't understand what that means
when P is high enough it barely oscillates, you can damp it with a bit of D
at that point you should have some kind of stable response
it won't track very well yet, but you should be able to run your rapid-back-and-forth program
I am using a sine function to stimulate. Frequency 4 and amplitude 0.01. When P starts to osc., I backed off 0.5 and added 0.1 to D
I quit the program, the moves were not aggressive enough to expose any weakness
I have no idea what sine response should look like
went back to HAL
I woudn't use sine
Well, about the same as g01 0.5; g01 -0.5
you want step, or something a little milder but still steppish
no it's not the same
gnieport: that's steppish innit?
the idea is to have a change, then a steady command, so you can observe how it approaches and holds the steady value
I think you want the same motion profile emc will use
yes, a g0 move has an accel, cruise, and decel
sine wave never stops, so you can't observe the settling
you can tweak the accel/decel with FF2, you can tweak the cruise with FF1
you can tweak the final end position with I,D
a sine doesn't tell you anything about how emc will work
hmm FF2 from 0 to 0.1 the the motor tried to throw the screw through the side of the machine :(
way too high I bet
jmkasunich: what is steppish but milder within HAL
square wave thru a limit2 block
or an EMC move, like cradek uses
probably the latter is easier
what is a good way to allow a short settling time after a g00 or g01 move?
18:29:46 < cradek> GNieport1: I just make a short program that repeats short
G0 moves back and forth with G4 pauses in between
^^ 3 hours ago
you could also use M0 pauses, if you want to be able to tweak something, hit a key to make one move, tweak some more, hit a key to make one move, etc
you gotta look at the scope pictures
as cradek said
for me, tuning is kind of a zen thing - you stare at it, and change things, and stare at it, and change things, and after a while you start to "get it"
I typically allow several hours to get an axis tuned, and just try to get in the zone
It's bound to hit me soon, I've been staring a halscope for 12 hours in the last two days
it's not magic though.
there are direct cause-and-effect relationships between each of the gains, and things that happen
there's all this mysticism around it. ignore that and learn what each of the numbers do
but you gotta be willing to really think about it
the "mysticism" is just because it can sometimes be hard to get all of the effects in your brain at the same time
i know i said this before but i'll spam it again: new general machinist chat, #machine
you really should try to get your output scaling "right"
It's all too vague. If there was a rule, this would be much easier. "Increase P until you get three decay oscillations. Then subtract 5%. Increase D by a factor of 0.58"
Too bad i don't need a dissertation on control theory
there are rules like that, but they are only applicable in certain circumstances, and they aren't necessarily "optimal", they're just "functional"
some things can't be put into recipies that can be followed without understanding
The guys nhelped me set up scaling yesterday. It's spot on.
input scaling AND output scaling?
input scaling means that if you move the machine 1" by hand, emc things it moved 1"
Yes, I did that two years ago :)
SWPadnos hey can i call u tomorow maby if ur not busy? lol
output scaling means that if you disconnect the PID loop and force "1.0" onto the output signal, the machine moves at 1 inch per second
1600 rpm max speed on the screw. 5 turns per inch. Output scale 1.875
Gamma-X, I don't know that it would do much good really - it's not much better than IRC (especially when I don't know your machine at all)
10V to the amps = 1600 RPM = 320 ipm = 5.333 inches/second?
so if you command 1" per second, you want 1.875V, so scale = 1.875 - gotcha - that sounds perfect
that means that the ideal FF1 gain is 1.0
It got really ticked when I tried .2
if the command is changing by 1" per second, an FF1 of 1.0 will result in an output from the PID of 1.0, which will make the DAC produce 1.875V and the machine move at 1.0" per second
if you apply a step (square wave, etc) to it, you will get nasty results, because a position step = infinite velocity for a very short time
and FF1 looks at velocity
cradek: thanks for the g4 suggestion 3 hours ago that I missed
thats why you want a softer input - like a G0 or G1 move
doesn't that mean that tuned for a step response = best possible tuning?
maybe, but emc will never ever step the position
should my gcode use g00 or g01 moves for the 'step'
it is physically impossible to follow a true position step
and do I want to use my maximum travel speed for the feed?
moving from here to there in zero time.....
G0 produces a ramp instead of a step
G1 produces a slower ramp
use whatever speed you like - you can control it with feed override too
I'd tune with G0, but check with G1 at a few different F values every once in a while, as a sanity check
yes you'll want to try different speeds
I want to be able to jog the machine at maximum without a following error.
yes of course
then I will at some point be using that feed rate in my tuning gcode, yes?
same as G0
btw thanks all for explaining this just one more time :)
What's the best way to determine what my INI max acceleration should be?
one thing you could do is watch the pid/dac output during the acceleration phase
if it's near the max output you know you are pushing it
personal choice is part of it
cradek's idea is a good way to see if you are asking more than the motors can do
which you MUST avoid
I think that's the only quantitative thing you have to go on
the rest is probably just how it sounds/feels
you might decide that even though the motors can do X, you only want to do X/2, because doing X makes the machine jump around, or just seems abusive to the screws and ways
accel also relates to speed
you want the machine to be able to slow down from top speed, or accel to stop speed, in a reasonable time
LOL tuning PID seems abusive to the screws/ways
if it takes two seconds to get to speed, that will be kind of annoying, and may limit some things like rigid tapping
going to full speed in 0.1 seconds on the other hand, might be a bit faster than you need or want
380 IPM in 0.1 s; not going to happen with a 1HP motor :)
Wanted to mention that both infinitely fast moves and controlled moves are valuable for PID tuning, they give you different information...
PeterW: I got a new 7i33 this week, apparently pulling ENA0 low is fatal =-O
bbl, putting on PID war paint
Pulling it low should be OK , its just a HC CMOS output, but connecting it to some external voltage would likely be fatal (as fatal as a bad 74AC08 is)
Well, I had all of the ENAX tied together, through a relay contact to GND
fatal enough for me, I don't have hot air rework :)
bah - all you need is a 1/64 tip
Maybe shorting 4/6 of the outputs to ground at once killed it
and fine cutters
Or perhaps shorting 4/6 of the outputs to ground many dozen times
1/64 tip? I've never seen one :)
hmm, next I'll be shopping for a microscope
You can also make a heat bridge with some #18 solid wire and pull the chip...
bah - all you need are really really good eyes ')
oh, pulling is ewasy - cut all the pins with the fine cutters, then scrape the remaining legs off with a hot iron
Unfortunately by eyes are getting worse at the same time as parts are getting smaller...
lucky you! ;)
hmm, perhaps I'll give it to my buddy that plays with Arduinos all the time
a stereo microscope is a wonderfull thing
I'm still OK with anything >= 0402
but those aren't easy, especially when they're the 8-element resistor packs with 0402 spacing
If I use 402's i always bring about 20 when I need 2, the rest get lost...
a stereo microscope with large working height and adjustable power is even more wonderful
thats what we have at work
and one that's impervious to resin smoke
yes I solder under it sometimes
about 4" from the lens to the focal plane
nice zoom range
PeterW, yeah - it's really important to control your breathing around the small parts
and never ever sneeze
the only trick to not losing parts is to never pick up any part more than an inch off the work surface
heh - my ex-wife is a bench jeweler - they refer to small diamonds as "sneezers"
that helps, but coughing or sneezing knows no bounds for parts loss
just coat your bench in flypaper
its not like you ever take anything off the bench once it's been put on
hmmm. while that has some advantages, I'm not sure it makes the overall assembly process easier
I'm running a tuning gcode program. I am not having any luck finding something to use for the scope trigger
are you doing moves like last night? (+/- 1/2" in X)
trigger on position
Like last night. In fact, tuning the same axis.
yep - like cradek said, trigger on position (axis.0.motor-pos-cmd I think)
It's staying flatline
check the scaling of that trace, and also make sure the scope is triggering (put it in "run" and "auto" mode for that)
do you have dwells/pauses after each move?
dwells; auto did the trick
ok, that's not going to be good for tuning, just to set the scaling and trigger level correctly
it will not retrigger in normal
For those with employeer benefits, are you paying anything for them or 100% by employeer?
GNieport1, are you seeing the position command change between the high and low values?
How would I go about setting this up to use my realworld scope? AARGH
you wouldn't, the PID signals are all in software
have you selected the trigger channel?
I have two unused analog outs on this 7i33 :)
ok, and what's the level?
does it show up between the high and low position?
I see the rising arrow and trigger line on the scope window. It crosses the waveform
what's the total time shown on the scope?
it should be 4, 8, or 16 seconds
it captures for 8 seconds, I'd sure like to reduce that :)
turn the number of channels up
It was set to 16
hmmm - that sounds like a non-standard number of samples
This computer is posessed. i'm getting the can of gasoline.
got the fire out?
okay, fresh reboot
nope, i'm still fired up 8-)
I was talking about the gasoline ;)
lol, I decided to save that option for later
so I want axis.x.motor-pos-cmd pid.x.command and pid.x.feedback?
axis.x.motor-pos-cmd is probably the same as pix.x.command
they are connected in HAL I would expect
you want command, feedback, output, and error
all can be gotten from the pid block
what to trigger with
probably error is easiest - it will be near zero when stationary, and will deviate when you start your move
so set the trigger just a hair above or below zero
is there a way to trigger on crossing, not either/or
if you set the level at 0.001 and the edge as rising, it will trigger when it crosses 0.001 going up
you can switch to falling edge, but you can't do "any edge"
[05:13:08] <GNieport1> http://imagebin.ca/view/Q8UZED5.html
now, the traces match almost perfectly. Is the PID in tune?
what are your accel and velocity limits?
(that doesn't seem like a very agressive move)
do you have a channel showing the pid.x.error?
I see four colored buttons at the bottom, but only three traces
the error is nil
nill isn't a number
0.01"? 0.001"? 0.0001"?
I increase the gain on the error channel quite a bit - so I can see individual counts from the encoder
oops, my fault, gain was low
typically 1m or so per division
well, if I keep increasing gain, teh trace goes offscreen
then its not nil ;-)
then it's not centered, or it's not close to 0
[05:24:43] <GNieport1> http://imagebin.ca/view/PXxDW98b.html
I see what to adjust! :)
EMC will not allow the g0 move to be performed any faster
I suppose this is close to worst-case if I'm not going to tune with a square-wave
my max_accel is three times my max_velocity
that means it takes 1/3 second to get to top speed
that may be a bit slow
(maybe not though)
I'll have to check the DAC to see if it's pegged :)
4.8 out of 5.0; it's pegged
5.0 on the DAC is 100% PWM duty cycle
you have a scale of 2?
for the 5i20
I think that's supposed to be +/- 10
the 5i20 puts out a 5 volt PWM
maybe i'm confused, today has been too long
my output scale is 1.875
are you looking at the FPGA to 7i33 connection, or on the output of the 7i33?
I'm looking at m5i20.0.dac-00-value
are you limiting that to 5 in the ini (MAX_OUTPUT)?
max_output is set to 10, because the docs say that is real-world volts.
yes, that should be the real-world output
the 7i33 does output +/- 10V
(paragraph 1, page 1 - pg4 in the PDF)
so dac-00-value has a scale of =/- 10 also?
I think that should go from -10 to +10, yes
sorry, a range of +/- 10
yes, I believe the "value" parameter should go "full swing"
* GNieport1 hugs Halscope for saving the window settings.
Do I want pid.0.error to follow closely to pid.0.output, or to go to zero?
the error won't stay at 0
pid.0.output should follow pid.0.command pretty closely
what is my tuning goal for Proportional?
err - that's not what I meant to say actually
pid.0.command should closely match pid.0.feedback
one of the methods is to increase P until you get ringing, then back off some
[05:46:21] <SWPadnos> http://www.linuxcnc.org/docview/2.2/html//motion_pid_theory.html
[05:46:27] <SWPadnos> http://www.linuxcnc.org/handbook/part2/pidtune.html
lots of reading there, if you haven't already
would you be so kind to tell me what method works for this machine? ;)
I wouldn't know actually :)
Yes, I read it all, and have been using loops for a while
SWPadnos: You know, you never used to be so synical before you started hanging around me =)
I'm handicapped by industrial products that have an autotune button
once you back P off ( to something like 2/3 the oscillating value), you can add in bits of I, D, and FF0/1 to make things "better"
Jymm, thank you very much. now go back to the garage! :)
you could try the autotuning PID component
SWPadnos: I'm in the garage!
SWPadnos: but you knew that already!
it's called at_pid rather than pid
oh hell, I've come this far, why use the 'easy' button
If you contribute (let's say $10,000) into a 401k, can you loose what you've contributed?
especially because we're not positive it works correctly
Jymm, it depends on the investment
Jymm it depends on how the vesting works. And if it is in company stock (Enron)
if the dollar fails you're screwed anyway
I think 40i(k) accounts can be invested in the stock market, for instance, in which case they could lose value
well, shit... and I dont see anything specificyling details of the plan in this 3lb binder they shipped me
ok, I lied... found the 401k stuff
taking apart hard drives for those finger pinching magnets is FUN
[06:04:40] <GNieport1> http://imagebin.ca/view/gTZfKup.html
Does the fuzzy pid.0.ouput trace indicate that I have too much P or D (the only nonzero variables)
GNieport1: i'd guess D since it's more sensitive to encoder quantization
wow grizzly belt/disc sanders are expensive
$600 for 6"x48" belt 12" disc
might as well go up to the 4HP model for $700
fenn: Be VERY leary of it being true or even somewhat straight
fenn: Also, see if it has a cast iron dis and not an aluminum one
eh i'm not buying one, i've already got a 1950's 12" disc sander
thinking about making a belt attachment for the shaft that sticks out the back side
fenn: got any pics of what you are wanting to build?
example pics that is
this pretty much http://grizzly.com/products/Combination-Sander-6-x-48-Belt-12-Disc-3450-RPM/G1183
this is cool though http://www.grizzly.com/products/G0489
Jymm is now known as Jymmm
Hi all, I am having problems configuring emc2.2.2 with --enable-simulator flag enabled
on my laptop. I get the error "configure: error: GNU PTH library is required"
emc2.1.7 configured with just --enable-run-in-place flag runs perfectly.
fsdafsd is now known as Unit41
[14:56:29] <jepler> http://www.netdisaster.com/go.php?mode=manif&url=http://linuxcnc.org/
netdisaster ... cool idea
running out of cap screws on a sunday is not fair /me needs 12 4mm x 20mm long
stainless or 12.9
drop over I have some
jepler: send the bot if that crowd gets unrulely
BigJohnT, possibly a large pond in the way for "drop over"
ok I'll e-mail them to you
I have a large supply left over from radio control helicopters from 2.5 to 6
heh, you do need spares in stock for r/c helicopters
archivist: yes you need plenty of spares
BigJohnT, I had one for a short period, I only damages it once
usually the damage is not a nick in the rotor but something quite amazing...
actually the project fell apart so I sold it, was for a remote camera, and was a project to add gyros to stabilize more than normal
I went to see a friend the other day and he had bought a new one, and a spare
what kind is it?
my oldest one is a 70's Kavan Jet Ranger from Germany
never been flown
finally a glade/python tutorial that works!
[16:59:30] <BigJohnT> http://www.arson-network.org/index.php?class=tutorial&subargs=430
now if I can get a glade gnome python to work...
[17:08:09] <tomp> http://www.manning.com/grayson/
excellent pay for pdf book on python & tkinter http://www.pythonware.com/library/an-introduction-to-tkinter.htm
excellent free book http://infohost.nmt.edu/tcc/help/lang/python/tkinter.html
great tech ref for each call http://docs.python.org/tut/
guido ( creator of python ) his own tutor for python ( not tkinter ) & many more to be googled
i use those because they cover the python libs used in emc2
A Comprehensive Introduction to Python Programming and GUI Design Using Tkinter by Bruno Defour ( also in my desktop library )
libraries want feeding all the time
tomp: do any of those cover glade with python
dunno, i dont use glade proper, but did get very confused by the many variants of python to glade libraries. i'm not even sure GL is needed by emc (as in not sure )
GL is needed for the AXIS preview
what is GL?
err - openGL, the 3D library
i see import gtk but no text containing gl, maybe gtk includes it?
the reason I got interested in glade is because the stepconf uses it
Looks like GLX is the linux link for openGL
stepconf.py sez "import gtk.glade" ... i >>think< gtk.glade allows the xml formated guis to be parsed and generated, and gtk.glade is in ... dunno
and glade itself is a drag&drop interface builder ( saves whatyousee as an xml file )
yes kinda when you run it it resizes to fit LOL
glade makes gui design easy
so , you might not need to know much about glade just to use an xml description of a user interface or to use emc or to use stepconf
glade makes gui design simple, not gui use ;-) ( whatcha gonna do with that thing now? )
use it with python which is kinda easy to program
the biggest hurdle was sorting out the "signals" between the glade gui and python
the biggest pain in python is generating the gui, glade makes that part fast and easy
yeck. 2 different circuits go to 1 electrical box.
(this was done in the 90's)
hello users of emc2 ..
someone is here ? i would like to know what is your solution to milling some PCB with EMC2 ?
i generate gerber file from DXP..
I use Eagle's user language scripting to generate gcode directly
I don't know how gerber is used, or how to get gcode from it
hello cradek, everytime here ;) happy new year att all of course ;)
an import solution from gerber in emc2 is maybe in development ?
all my search give eagle solution .. i dont use eagle .
if it's possible to translate from gerber to gcode, you could do that in an AXIS input filter
I do not know if anyone has written something like that.
what is axis input filter ?
a program that generates gcode from some other file
how i try it ?
[19:09:38] <cradek> http://linuxcnc.org/docs/html/gui_axis.html#r1_11_1
ty i will read it
if i right understand, i need to provide a translator script ..?
:( ... everybody use eagle here ?
cradek, do you have a link to see wich tools you use to milling pcb ?
some information on timeguy.com
one of the problems with a gerber input filter is that most gerber files are positives - they tell you how to draw the copper
when machining, you want to "draw" everything that is NOT copper
yes i allready look at here, but my english is not perfect, i prefer look a picture to be sure ..
gerber->png->"some prog generate gcode to cut the black/white"
yes jmkasunich, need a great tanslator .. i try KCAM, i'm not sur that he do a ggod job
translating from positive to negative is NOT a trivial problem at all
no Gcode from Altium DXP ( Protel ) .. :(
imagine a 0.050" diameter round pad, with three 0.015" wide traces leading to it from three different directions
4 lines in gerber - flash the pad, draw three straight lines
but to machine the profile, you have to do both sides of each trace, and multiple arcs between them
and you have to know about all three traces before you can generate any of the code
yes i agree jmka
unlike the gerber code, which can draw the lines and flash the pad in any order
the product job of KCAM seem realy not optimized ..
weel, my little girl wait me to eat .. need to go, thanks all, will try next week end ;)
have a nice week all
So, if I donwloaded some gerber files from the web, I couldn't run them thru the script that jepler created to create the gcode files?
pcb milling seems to be a common need
I must have a look at it one day
for single side boards it's great
I should get outside - it's 45 degrees
i'm back for some minutes ;), yes Archivist, it's a common need ;)
I have a pcb to do soon
what is soon for you ? next week ? lol ;)
is it possible to add import gerber under emc2 ?
I spent many a year designing pcb's and getting someone elsta to make them, but that was at the previous job
I see two ways a quick and dirty, and a better way
i maybe help you to design your next pcb ..?
gerber to bitmap then use a bitmap to gcode program
that would be quick and dirty
the proper way is way more involved
i think it's not a good way
i look at Youtube, there are some good video job
the proper way would be to create a vector representation of the outlines - which isn't simple because of the various line widths and flash shapes (pads and whatnot), then create the toolpaths from that outline
SWPadnos: and that's usually easier in the program which created the gerber in the first place
should be, but it depends on whether a script writer has access to that geometry ;)
there is maybe some good algorithm on google ?...
* archivist needs to read his graphics gems books for an algorithm
follow outline, merge paths , clean or ignore unused area
If "isolation" routing is OK, this is pretty easy. Convert to a bitmap, then trace the outline of each area, converting to gcode on the fly.
some soft like Galaad do that job ..
yes and under windows ..
time to write something
loool YES !.. ;)
what is the best way to install eagle ? i have download the rpm file .. just click on the file ?
I had downloaded it using synaptic package manager.
i search on synaptic, but the only source find is something about MODEM ...
but i have an error when i run synaptic
many lines like :
W: Impossible de localiser la liste des paquets sources http://fr.archive.ubuntu.com
breezy/main Packages (/var/lib/apt/lists/fr.archive.ubuntu.com_ubuntu_dists_breezy_main_binary-i386_Packages) - stat (2 Aucun fichier ou répertoire de ce type)
So, if I downloaded some gerber files from the web, I couldn't run them thru the script that jepler created to generate the gcode files?
not all gerbers are the same
jymmm: I haven't written any programs for gerber
and the flash setups could be different
RS264D and RS274X are pretty standard, but there are different specified formats that not everyone supports
err - RS274D
phil_: the "breezy" version of ubuntu is no longer supported by ubuntu or by us. http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Breezy_Upgrading
-- you can upgrade http://emergent.unpy.net/01186407274
-- or change some configuration files to get the last version of all packages for breezy
Jymmm: eagle does not load gerber files
phil_: or do a reinstall with the current ubuntu dapper live cd http://www.linuxcnc.org/content/view/21/4/lang,en/
jepler: excucuse the symantics, but didn't you or cradek write something to generate gcode for milling PCB's?
i know jepler, buti'm newbie under linux and i'm scare about to change anything .. my EMC2 station work, i don't want broke something !
that's a .ulp file, which is a script that runs inside eagle to generate G-code
it is an eagle script - generates gcode from within eagle
it doesn't take gerber files and "translate" them for machining
what SWPadnos said
Ah, I didn't relaize the output from eagel wasn't gerber format.
it is, but the script runs within eagle, and has access to eagle internal data structures
eagle files are not G-code, but it can export G-code for a photoplotter
(at least I hope it can - it would be pretty useless if it couldn't)
SWPadnos: eagle has various export possibilities, I am not familiar with them (have never used gerber format)
it's basically the only way to get a board produced (by a board house), so I'm 99.9 sure it can do it
"export for fabrication" or similar
output CAM files ...
A lot of board houses will take an eagle file directly... makes for a lot less mistakes
interesting - I haven't seen any
oh- maybe Olimex or one of those
i confirm, most pcb maker use Gerber .. i send all time gerber file and the maker send back the PCB .. www.eurocircuits.com
[21:34:14] <LawrenceG> http://www.cadsoftusa.com/Boardhouses/
i have used http://www.custompcb.com/
i have found : "gerber_to_gcode_0.1.6.tar.gz" .. the sources files maybe includ under AXIS ?...
where wot? google cant see
[21:40:40] <phil_> http://sourceforge.net/projects/gerbertogcode
I just found it
I would like to be able to generate gcode from pcb patterns scanned from magazines.... any bmp_to_gcode?
some goods links on google with "gerber 2 gcode" search
Question... In this cog making video, he changes to 1/8" bit (it looks like). Now, when he cuts out the cog from the stock (not shown in video), it's too short to go thru the 1/2" or 3/4" MDF. How would you typically accomplish that? I've never seen tooling of that nature.
[21:49:34] <Jymmm> http://video.google.com/videoplay?docid=-3133898738886800969&q=cnc+router&total=712&start=0&num=100&so=0&type=search&plindex=1
Archivist, you look at gerber2gcode ?
LawrenceG: Inkscape can trace a bitmap to vector format
i try to install Eagle .. but unlucky !.. i read that : - If you are installing the downloaded TGZ archive, enter the commands
tar xzf eagle-lin-eng-4.16r2.tgz
and wait until they have ended.
that would help, but the cutter realy needs to be offset from the edges of the copper.... the tricky part
- Enter the command
to invoke the product registration.
i'm root, but not possible to mkdir opt/eagle !..
LawrenceG: Once you have a vector format, you can use eagle I believe
Gerber to Eagle to G-Code
Jymmm: check out http://imagebin.ca/view/Zau2sfg.html
... It was cold here last night
i thought those were decorative window blinds for a second
more than a seconds
some cool icr crytals on the outside of the glass
yea... looks like that cheap plastic film you glue on bathroom windows!
no i meant like, it looks like someone painted tress/waves on some hanging window blinds
i didn't realize it was glass
[22:14:38] <LawrenceG> http://imagebin.ca/view/5Wloibu.html
front window... wipers at bottom
all melted now.... very close to freezing still
a few minutes in the sun and it was ruined
why there is a PCB picture on the top of tha AXIS web page ?...
well .. can't creat eagle directory in /opt .. don't know why ( i'm superuser ! )
i have try with "sudo" instead of "su" and now it's work ..
LawrenceG: looks very cool though =)
SWPadnos: Hey, guess what... not everything will physically fit into that enclosure!
well everybody, go to sleep . .. hope a guru of Axis make a good "import gerber" option .. ;)
Juanjo is now known as juanjo
Has anyone tried any of the mini-atx boards for RT latency?
Or have a suggestion to a mobo that will fit in tight spaces?
Jymmm: VIA EPia is reported to run fairly well
alex_joni: any specific model?
[23:09:59] <Jymmm> http://www.via.com.tw/en/products/mainboards/motherboards.jsp?motherboard_id=530
[23:10:23] <alex_joni> http://www.trilobyte.ch/pix/index.php?spgmGal=fraese/Beckhoff
VIA EPIA M10000
jmk ... you around
alex_joni: who's that by? And when did the latency test go gui on me?
oh, the gui is 2-3 releases old
some guy in germany tested that board
alex_joni: base thread looks bad
I'm used to cli, so I might be reading it wrong
[23:13:08] <Jymmm> http://www.trilobyte.ch/pix/index.php?spgmGal=fraese/Beckhoff&spgmPic=7#spgmPicture
it's supposed to be 25000 every time
the max value recorded was 32067, which gives a max latency of 7794nsec
which is *very* good imo
alex_joni: ah, I *AM* reading it wrong.
and if you get negative numbers????
dave_1: that's not good either
it is wrapping?
that means it ran a bit faster than it was supposed to
you need to use the max jitter as = abs(max-negative-value) + max-positive-value
anyone feel like helping with homing?
juanjo is now known as Juanjo
encoder index is always TRUE
if so.. it's broken
as far as I can tell.
it might be active low..
even on the channels that are not connected
so it only goes low for a very small amount of time
hmm.. what HW?
do the encoders work?
counting I mean..
when I tested on the bench ... with a scope I could see the encoder pulse going high
yes, counts just fine
and now it's stuck at HIGH?
different Koyo encoder
I would first try to see (with a scope) if it's all how it should be
after that I'd try to get a halscope plot
(trigger on index low.. or something)
guess I have more work to do... ugh.
I'll grit my teeth and go play
dave_1: you still here?
I don't know any more about the motenc than anybody else
including even what the "index" hal pin means
index-enable is the one that is used for homing, etc
what are your symptoms again?
normally Xindex gets set by the index then it sets index enable the way I read the hal
well, I honestly don't know if the motenc board follows the canonical interface or not
with halmeter the index pin is always T and the index-enable is always F
hal pin <whatever>.index is meaningless on a hardware encoder counter. because the reason you are using hardware is that counts and index pulses may be too short for software to see them
hence, having a HAL pin that is simply a copy of the physical index input is kinda dumb
demo_mazak should tell me ... basically the same board as Galesburg but 8 chan
but that isn't going to tell you anything about the .index pin
because I'll bet you its not used in that config
but it should tell me what you did
index enable is what matters
is there an rpm version of emc or something easier to install ?
cxix: there is a deb
what distro are you interested in ?
im using mandriva
so motenc.0.index.0 or what ever should drive the enable directly?
i tried installing the ubuntu live cd, but after fiddling for a day straight, it still wouldn't recognize my keyboard.
dave_1: index-enable is a bidirectional signa;
i could not type to log in to ubuntu
not a bit?
dave_1: it is a bit
I mean.. tried plugging in an USB keyboard?
bit is the data type, bidirectional is the direction of data flow
so low | high T | F
it's an odd error report though..
most pins are either in or out
i went through the config about 20 times with different keyboards, nothing worked
no, i didnt try a usb keyboard
cxix: I bet it's not the keyboard which is problematic
installing on mandriva *will* be a much bigger pain
as you need to compile/patch with RTAI a kernel on your own
guess it is time for me to go see what you did at Galesburg and to scope the encoder
only after that you need to compile emc2, etc
so, now i instaled mandriva, and downloaded emc, well you have to install rtai, to do that you need some other crap, and to instal that you need other crap... i've been at this since 9 this morning
dave_1: I'm waiting for the other conversation to die down
cxix: I'd try the stock Ubuntu 6.06.1 Live CD
cxix: you do NOT want to use mandriva
cxix: installing emc2 on mandriva will take a week at least
i think you're all right about that
using a distro for which there is no RT kernel will take a lot of tricky work
ok, so what could the issue be if it doesn't recognize the keyboard?
when exactly doesn
when exactly doesn't the keyboard work?
cxix: we need more data to say
does it never work?
does it work during the install, but not later?
jmkasunich: if I understood it correctly the LiveCD works..
it works on the live cd, it works at the terminal, but when i try to boot into the ubuntu gui, it will not let me log in. no lights on the keyboard, no input possible
cxix: that sounds like it locks up
here's part of the problem...
can you still switch to console? (Ctrl-Alt-F1..F6)
if that doesn't work, then the most likely is that X crashes your machine
try a different driver for X then (vesa or fb)
does the cursor blink normally?
something's weird with my computer i'm going to use as the controller, it's a bit slow - no operating system will instal on that computer - they all freeze... so i take the hard drive out and put it in my server computer, install the OS, and put it back into the controller computer. well, the graphics cards are different, so i had to run through the configuration (sudo dpkg-reconfigure xserver-xorg), and the try to login and it doesnt work
i can not switch to the console, no inputs work from it.
try running a memory test from the boot up screen
fenn: it blinks here
the cursor blinks fine, the mouse works and everything too
if no OS will install on that computer, maybe its not a good choice?
it's just a weird problem, memory shortage i would imagine.
the computer runs xp and mach3 fine
bad memory i think
maybe some busted memory..
run memtest overnight
i will go ahead and re-install ubuntu real quick
and take the computer apart and blow all the dirt and dust out
hardware is so plentiful and cheap it's silly to waste time on bad hardware
could be a heat problem (fan)
cxix: you just said "no operating system will instal on that computer - they all freeze", and then two minutes later you said "the computer runs xp and mach3 fine" - so which is it?
jmkasunich: simple, XP isn't an operating system :)
jmkasunich: installed on a different PC, then moved the HDD over
the hard drive is fine.
what alex said, yeah.
although that's quite bad for XP usually
well, i use to be more of a fan of linux, until i try to use it for something
trying more operating systems is not the answer here - fixing or getting new hardware is
cradek is right
cradek - so, why does the keyboard work off the live cd, and not when i install the os?
starting a flame/troll war is also not the answer
cxix: sorry, I decline to magically remotely troubleshoot your hardware
good call, very helpful.
if i modify 5axisgui.py, do i have to re-compile?
not that it will help but do you have a second video card?
cxix: first, you are going off on a tangent when you say "the keyboard doesn't work" - from what we know, it sounds like the entire computer is locking up
really, try memtest if you can boot the live cd that far (it's on the menu)
Roguish: yes, "make" copies the script to the bin/ directory. otherwise, your modifications won't be used.
how does the machine do on latency?
Roguish: you can also just change the 5axisui from bin/ .. but a subsequent make will overwrite it
(and you don't get the nice spellchecking like with make)
the computer does not lock up, i can move the mouse, the cursor blinks, i can choose those options like "language" with the mouse, it just will not accept keystrokes
is your video card on board?
i don't mind recompiling. i just did not know.
dave - me?
so why didn't you tell us that back when alex said: "cxix: that sounds like it locks up"
yes it is
i want to convert 5axisgui down to a small inch system.
jm - i was typing, i apologize, i missed that
dave - i have another video card i could use
cxix ... I just went thru that with a M6VLR ... nothing but problems using on board video....
cxix: understood, it happens - but it tends to lead us down the wrong path
dave - the keyboard issue as well? i can get the monitor working by reconfiguring it, but they keyboard has issues
dinnertime here, bbl
so, when i run (sudo dpkg-reconfigure xserver-xorg) it gives me options for all these keyboards, the one it uses automatically is pc104 - should it be something different?
cxix: maybe you need to look at /etc/X11/Xorg.conf (especially the Input sections regarding keyboard)
cxix ... yes I had keyboard issues ... and mouse issues ... finally I just gave up and swapped MB's as I couldn't find a video card that would install with 6.06
alex- i'll give that a shot, it'll be about another 5-10 mins before it finishes installing
thats one plus for linux - it installs much faster than windows.
and 6.06 is a bit old.. the newer ones should be a bit better :)
maybe I have the number wrong ... what ever the latest Live used
if nothing else, if i can get the keyboard and monitor to both work on the server, i'll just swap the hard drives and use the slower computer for my server
and upgrade later
I'm going to go play with hardware ... bbl
dave_1: 6.06 is the latest we have with emc2 on it
ok, so the keyboard works using the server