woo-hoo! I just finished the last three drawers for the bench my Shoptask machine sits on!
it's only been 9-1/2 years since I build the rest of the bench
an average of 0.000002 fasteners installed per day :)
something like that
they stick out like a sore thumb - so clean compared to all the rest
alt-032 on Windows machines - I'm not sure how you enter specific keycodes on Linux
[01:51:28] <skunkworks> http://cnczone.com/forums/showthread.php?t=57695
I wanna know why firefox keeps crashing on me
quit visiting the same porn site over and over
Toolmotion is the opposite of what is shown in the live plot in axis - re-run the set up and reverse that axis or rewire the steppers themselves - I suspect it doesn't matter does it? Former is easier to be sure.
change the sign of your scale to reverse motion on an axis
err, is that the save as "invert" in stepconf?
probably - I haven't used stepconf
Ok, checked that, will give it a try.
Err, one more question here.
Got a 1/4-20 lead screw and have that pitch set in leadscrew pitch but distances seem off...
I set up the driver (HCNC 4 axis pro) to 1/2 microstep..
(being off by a factor of 2 is likely to be a different problem than being off by a few percent)
wwhich seems to correspond to what movement I get.
the microstep has nothing to do with altering the leadscrew travel, correct?
yes it does
when emc puts out a "step", that is one pulse
If I command a 1" movement, it won't move 1/2" with 1/2 microstep or 1/4" with 1/4 microstep?
it is up to the drive as to whether that pulse moves a full step, 1/2 of a step, 1/4, 1/10, etc
if you set the scaling right, when you command 1", it will move 1"
Ah, ok, I think I get it now - if I am 1/2 microstepping then I have 400 steps per inch rather than 200?
"setting it right" means taking into account the microstepping if any
Ah, thank you!
would be 800 for 1/4 and so on.
That explains my problem :).
if you're using stepconf, you shouldn't need to do any of this math yourself
but you have to give the right answers to stepconf's questions, so you still need to know what your drives are doing
so if I set microstepping to 2 (1/2) then stepconf should altern the motor steps per rev to 400? If so it didn't.
motor steps per revolution is usually 200, if you've set 1/2 microstepping, enter 2.0 for Driver Microstepping
no, it doesn't alter motor steps/rev
it alters the scale value that it writes to your ini file
look at the bottom of the dialog - it will show the axis SCALE
Ok I got a pen in the chuck\
that scale value is calculated from screw turns per inch, gear ratio (if any), motor steps per inch, and microstepping (if any)
do you have any pulleys?
oops, I meant to say "motor steps per rev" not per inch
At 1 IPS it is god awful noisy.
or is it direct drive
No, direct drive.
half-stepping is noisy
20 threads per inch times 200 steps per rev = 4000, times 2 microsteps per full step = 8000
Should I just full step and be done with it - I have no idea why I would actually need microstepping to be real honest.
I'll bet that the default accel of 30 IPS^2 is too high
no, you want 1/4 or 1/8 ideally
(for noise issues anyway)
no, fullstepping will be even noisier
default seems to be 1 IPS
sounds great at 72 IPS :)
I don't believe you ;)
IPM maybe ;)
Ah, yes IPM
happens all the time :)
I are very easily cornfuzed these days :).
screw it, try it at 72 ips
It might be a piece of crap mill, but I am learning a LOT here.
OK lemme save this and ask some dumb questions with my hand on the throttle.
well, it's best to be sure the scale is correct first ...
Gonna jog the pen down and command a 1" X move...
K`zan DONT DO THAT
K`zan "You'll kill yourself!!!"
Thanks, I needed that. Now to go clean out my seat...
better you than me =)
I hate gnome, no launch feedback.
usually a taskbar item shows up that says "starting <whatever the app is>"
at least when launched from the menu or top panel
My eyes don't stretch that far ;-).
I hate gnome eyes ;)
Ok, incredibly dumb question - re:, in this case the Z-Axis
Where should home / 0 be - at extreme retract or with the tool on the table?
it's up to you, but + means move up
OK, got that much righ :).
usually people reference the top of the work to 0, and - moves go into the work
It was right them, I shouldn't have reversed it - now plus feeds down.
I think he's asking about not work zero, machine zero / home
Setting yp homes / zeros to the table at this point
machine Z=0 is typically at the top of travel
work offset Z=0 is typically the top of the workpiece (before it's cut away)
SWPadnos: When yu get a chance, crack open the cover on the artigo, and gentle turne the fan. Does it feel like it's scraping againest something? Also, can you push the fan down my about 1/8" ?
Don't confuse me more than I am at this point :).
Next question is x and y zero / home: I am assuming it is (facing table) roughly front left?
but I think some people put it in the middle of the table
OK, works for me.
I like front left :)
if it's front left, all your numbers are positive
K`zan personal preference, but MY XY/00 is lower left for me.
JymmmEMC, yes to both
if you don't have home switches, its kind of a moot point - it will be wherever you are when you hit the home button
changing the z direction back to what it was.
Once I get this basically set up I'll add home and / or limit switches assuming I can have both.
Looks like they have to be combined (looking at the parallel port setup).
SWPadnos: I loosened the screws in the fan shroud, it seems to help.
Gonna shut it down and rejumper for 1/4 or 1/8 step - I can go to 1/16 with this driver - which should I use?
1/4 it is Sir, thanks!
K`zan DONT DO THAT =)
Thank the ghods I don't have a tool powerful enough to hurt me too much :)
* JymmmEMC grabs the mic.... "Clean up on aisle 5
Driver Microstepping to 4.0 on all axis
What is considered the minimum number of encoder counts needed for a stable dead band area?
how long is a stick?
wrong universe I guess
(depends how big a deadband you can tolerate)
twice the distance from an end to the midpoint
OK next I work out the table area (or movement) for each axis. Wheee!
deadband should be about a step
I was figuring on using 1250 line on a .2" pitch - gives me a .00004 per count
The Z is moving down with the + button, but the live plot shows the tool moving up
Guess I have the stepper coils reversed on the driver board?
K`zan_emc: just negate the SCALE
could be - just change the scale (-)
for that axis
cradek: How does one do that? I didn't see anything for that - other than inverting the direction in the parallel port setup?!?
sure, you could also invert the direction pin on the parport setup
that's just as easy
I just undid that LOL
that way might be easier in stepconf...
ouch MariTool just raised some prices 30%
The Z axis in the plot still moves the wrong way.
That pulls the actual head all the way up, but the plot goes down.
Time for some Yukon Jack - bbl
Yukon Jack is out of my price range at the moment - can't afford those imports...
oh wait - can't afford domestics either
4" commanded == 1" actual movement...
SkullWorks_PGAB: Yes, but a little of the Jack goes a LONG way :).
I can't imagine I am missing 3 of 4 steps, even my little AVR driver did better than that.
use a slow move, not a rapid or high speed jog
Same at 10 IPM as it was at 72
did you change the config after changing the microstepping?
(and restart EMC)
Yes, I did.
with the correct config? :)
So much ATM doesn't work out I am really confuzed
Step config - seemed to make sense
Covers xylotec and another of what seem to be the same kind of drivers.
and you can enter other values if necessary
takes about a minute for stepconfig to come up on this 2GHz box...
is emc running?
wasn't when I did the reconfig
I'd stop it, then you can actually test each axis from within stepconfig
Lets try it your way :)
Mine is apparently a complete bust.
it's also possible that there's a problem with the microstepping jumper settings, like incorrect docs or something
It seems, and I am not positive about this at this point, but it seems that with 1/2 step it was moving half of what it should have and now at 1/4 it seems to 1/4 of what it should.
Pretty sure about that though.
positive about the 1/4 as I just measured it - 4" command gives 1" actual.
so you think it's set for 1/4 stepping, and it's a 200 step/rev motor direct-coupled to a 1/4-20 screw?
Yes, checked the jumpers twice.
Yep, jumpers are set for 1/4
and when you enter 200, 4, and 20 in the appropriate places in stepconf, does it come up with 16000 scale?
and you do the same thing for the other two axes (substituting the correct values if anything is different)
I frigging HATE dialogs that are not copyable, grrrrrrrrrrrr :-(
Haven't gotten to x and y yet.
Gonn a take a break here before I swap my Youkon Jack for a single jack :)
oh. for some reason I assumed you were testing by moving X or Y, since you mentioned using a pen
No, was running Z up and down.
I'm probably headed to bed soon. good luck
phuuuuk, I gotta blow this install away...
don't do it, it's not worth it
mouse randomly starts right clicking on windows. takes 15 seconds to type hello in any doalog box
everything is completely lagging now
<<< that would be me forgetting the EXACT syntax to the user password.
i am a goddamned dork
i just drew up images of how to very accurately align a rotab under a spindle in solidworks
i clearly have no life
Had a friend drop by, coffee break is over, back on my head :-)
Something(s) are just so hosed....
Hummm, Realtime system did not load
time to do a wincrap and reboot I guess. Either that or I hosed the config somehow.
segfaulted on shutdown...
This is going to require some thought. Tomorrow sounds ideal for it :-).
THanks for all the help folks!
toastydeath: stick a probe in the spindle run a macro..
Good morning. Sorry to ask a Q that's probably been asked many times already, but I don't see the answer in the header or on the website. How long until a new heron/8.04 ISO is released?
joeKr, you can try one now
btw are there any preferred parport pci io cards for emc? i could do with some more io
joeKr, test status only http://www.linuxcnc.org/hardy/ubuntu-8.04-desktop-emc2-aj07-i386.iso
this URL might change in the future, reference
[09:24:56] <archivist> http://www.linuxcnc.org/hardy/livecd.errata
for the latest name, and for a
description of known problems).
archivist: Thanks! Any problem areas that need testing? Is feedback wanted? Thanks again, I'm downloading now.
feed back is wanted yes
Where should feedback go? How much longer is 8.04 development expected to take? Will the Summer exhibit/meeting still be held? I've never gone, but would like to.
should find all on the website and wiki
OK. The wiki lists EMC Fest 2008 ( http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?EMC_Fest_2008
). The website only lists 2000 and 2002-2005 ( http://www.linuxcnc.org/content/blogcategory/15/12/lang,en/
). Not sure what happened to 2006 & 2007.
joeKr, here is the ubuntu anouncement http://firstname.lastname@example.org/msg00948.html
archivist: Wow, they don't call you archivist for nothing, do they? Thanks for finding that.
just a google away (and its in my gmail box)
and erm I do play with archives
I was just noticing your oversize ISO file. I found
How can I burn an ISO image larger than 700mb on a 700mb CD? Has anyone re
Oops wasn't done typing yet
I was just noticing your oversize ISO file. I found "
How can I burn an ISO image larger than 700mb on a 700mb CD?"
Gaah! What is it with copy and paste!?
the 700mb limit is an ish/non fixed limit
OK, let's try this again...
So I see. I found "How can I burn an ISO image larger than 700mb on a 700mb CD?" ( http://www.frihost.com/forums/vt-25985.html
) Has anyone reported good or bad results using InfraRecorder (what I use) to burn the ISO? Is there a published MD5sum for the beta ISO?
Or maybe it has that "verify this disc" feature? More opaque, but OK I guess.
the mail i pointed too
Found the MD5sum, thanks. As for InfraRecorder, we'll find out soon, the ISO is at 99%
I'm burning as always, except I checked "Allow Overburning" under the "Advanced" tab. I've never tried to use that.
Hi, can sobody help me with adding e new pci parallelport card?
when i run lspci -v I get [disabled] next to the adresses ...
I soved it myself, i had to disable plug-and-play in the BIOS =)
i also forgot to addf parport.1.write :-D
JanVanGilsen: glad you got it worked out!
is there any usefull cad software for 2d design under linux?
what are anybody personally using?
dimas, your question begs a definition of useful.
This has some good info http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Cam
Some folk bait you by thinking "useful" meaning it works just like the latest version of some commercial program like autocad
There isn't such a thing in the Linux world.
time "we" wrote one
well... for the price of acad - you can get something. can't you? like synergy (is that the name?)?
actually a nice gui and 2d output to brlcad comes to mind
I use qcad because it's free and easily available. But I'd order a copy of easycad/fastcad in a flash if it ran on Linux.
I admit to being spoilt by Solidworks
archivist, Yep. And folk have been talking and trying.
I think brlcad has the base to work on but its a huge job
I think that is the conclusion some have come to.
Certainly it is a capable system for what it was intended.
rayh: I'd want to draw a few parts of my router
Have you tried doing it with qcad?
but I want to draw them in cad software which I could use for a longer time
That package takes a while to get comfortable with but it can do the job.
rayh: what features are you missing in it?
My problems with it are that it works quite different than what I liked about fastcad.
I guess you'd call it less intuitive.
It is quite capable.
You on a version of linux there?
Do you have a computer with Linux running and on line?
If it's debian or ubuntu you can get qcad using apt or synaptic.
I believe the packages are about 30 Meg.
I will try at weekend
just prepairing new box for emc
now only have emc live-cd there
Great. I recommend the 8.04-emc if the box is fairly fast.
That is what I'm using here now.
emc2.2.2 Ubuntu LiveCD - is it it?
cradek_ is now known as cradek
That sounds like 6.04
Nothing wrong with it though.
I can try qcad on another machine
is there 8.04-emc ready?
Alex has prepared a CD.
let me find the link.
[13:18:00] <rayh> http://www.linuxcnc.org/hardy/ubuntu-8.04-desktop-emc2-aj07-i386.iso
I'm sure that it is mirrored in Romania which would be closer to you.
But I don't have that link.
location does not matter
Okay. Good luck.
thanks rayh archivist skunkworks_
Morning guys, i am trying to use pin 5 which is my a axis step for pwm out, but when i try and link the pin emc tells me pin 5 out does not exist?
net spindle-pwm pwmgen.0.pwm => parport.0.pin-5-out
doh i'm an idiot 05
I've done that!
it always helps when you write down your question..
Guest549, how's the weather up there?
It is really easy to overlook stuff like .5 or .o5 when it should be .05.
I've seen "l5" too, I bet from older folks who learned on a manual typewriter (uh, like me, but I've broken those habits long ago)
not bad at all
finally can open the doors :)
I figured it would be close to here (Essex Junction, VT)
montreal,qc still have large piles of snow here & there tho
it snowed a few miles from here a few days ago, but I think it's gone
hi all, is the emc-environment script automatically installed with 2.2.5 ?
for example, I get: $ sudo comp --install idb.comp
sudo: comp: command not found
it's not necessary for installed systems
or it shouldn't be anyway
but you may need to re-login if you just installed, since the environment in running shells won't be updated
The program 'comp' can be found in the following packages:
that's not the right comp ;)
I wonder if you need some other packages, like emc2-dev or something
I got build-dep, but maybe that does not include emc2-dev ?
yep, now it works with emc2-dev installed
so if you type "emc", emc runs, but if you try "comp", you get an error (in the same shell)
is there a pin or parameter in emc that let's me look at the same thing latency-test shows? i.e. the max interval between calls to servo thread?
hmmm. I don't think so
you can see execution time, but I'm not sure there's an easy way to see actual interval
yes you can scope the interval
the ubuntu repositories are pretty slow the last couple days, and they pushed out an openoffice update today...
nope, that's much smaller than servo-thread should be
or last-period for the count in CPU cycles
ok, I would need to create the max function myself then
you could also look at scripts/latency-test and see how it uses the 'timedelta' component
there's a manpage as well but it's very unhelpful
awallin_emc: man minmax
oh but you can't add timedelta to base-thread, it uses floating-point
that's a better idea. I think I will display these on a pyvcp panel while I'm setting up the machine so I know if a following error or other strange behaviour is due to real-time or something else
hrm. these things are parameters, not pins. so not displayable on a pyvcp panel.
who invented parameters anyway?? :)
oh, so you can't use minmax either. darn.
but I guess the consensus is that parameters are going to be pins in the future?
I think that's what jmk wants
by all means just change them if you want the pins now.
I haven't looked at motion code at all...
change the things in emcmot_hal_data, and change the hal_*_float_new calls
then change any site where they're written (probably only in this file)
you'll be done in 5 minutes
ok, thanks. I'll do that if I find I really need the pyvcp display.
On the list there was an opinion that as I drive my servos only based on encoder counts (no velocity/current feedback) a higher servo-thread rate would be better.
What kind of jitter/max interval would you consider acceptable?
with a 1ms thread I get about 1010us or so as the maximum interval. with a 250us thread I get 260us maximum.
so what's the threshold for a realtime-delay?
awallin_emc: have you had any machines that have run a good latency? I see on your site you mention 50us-100us are ok. I rarely see latencies greater than 30us (unless there is a hardware issue)
skunkworks: now it shows 10us Jitter for both a 1ms thread and a 250us thread
I think well-working machines give 25us or less jitter
260us is a sure sign of a problem (sounds like an entire missed period)
so with 10us jitter what would be the minimum acceptable servo-thread period?
with faster servo cycle you get more quantization on the derivates (velocity)
cradek: this is with a 250us thread, so 260us is 10us longer than it should
I'd decrease it by half (e.g., to 500us) and see if it improved anything
I'd experiment with 1,2 and maybe 3 kHz servo cycle
what have you tried so far? do you have bad results?
similar to base-period, it seems like the absolute minimum servo cycle will be something like (time spent in servo) + (jitter) + (time you want left over for responsiveness of non-rt stuff like GUI)
first I tried 1ms and then for a while 250us. then I discovered this 155ms problem so had to change machines.
I bet almost everyone uses the default 1kHz servo cycle
one problem is that ferror is proportional to speed, while in a torque-mode loop the error is proportional to acceleration
rayh_ is now known as rayh
I see my lathe (encoder feedback only) uses 2kHz
so my actual following errors are bigger at the start and end of a move and very small during a move
awallin_emc: are you sure that's not just a tuning problem? ff2 can often fix that right up
or, maybe accel is too high. your pid output isn't saturating is it?
no the pid output is not saturating, but with a PWM drive it's hard to say if I am hitting the current limit of the drive
I found FF2 could reduce the accel error or the decel error but not both: http://www.anderswallin.net/wp-content/2008/04_xaxis/ff2.png
well these are all errors below 0.05 mm at rapid moves so they are not that problematic in reality. it would just nice to have the error-plot pretty pretty
did you ever set I=0 and then retry the search for a proper FF2 value? I still think that's the problem
cradek: yes I think I did, but gotta try more tuning on the weekend
you cannot find FF1,FF2 if I is not 0
after FF1,FF2 keep the error low for a normal move, you can increase I almost with impunity for good/quick response to load conditions
I have an application where the fast (only) thread runs at 10 kHz, and uses floating point
even in the base thread at 20-50 khz, it may be usable - you'll have to experiment
a few hundred CPU cycles saving/restoring the FP registers is likely to be 1 microsecond or less - less than one in/out to a parport
hi PeterW !
Hi peter. Thanks for adding your experties to the cnczone thread. :)
* skunkworks_ = samco
Had a thought on the PCI-8255 problem. Since the trouble goes away when the data buffer is removed, this at least suggests a bus contention problem.
Since the local 8255 bus probably has no more than 100 pF of load capacitance, you might try replacing the data buffer with a 8 resistor dip resistor network package. With a 470 ohm or so value, this would limit bus contention but still work is the strobe times were a few 100 nS. You would have to offset the leads to match the data pins of the buffer...
that's an interesting idea
<cradek> with faster servo cycle you get more quantization on the derivates (velocity)
The HostMot2 config should improve this by using timestamps to allow the driver to do 1/T velocity estimation
cool. we do not yet have a separate velocity input in the pid module but it would be an easy addition.
or use 2 PIDs :)
hmmm. nevermind. I should drink more (coffee)
a good velocity output would put the mesa driver quite ahead of the others, I think
he's not here currently
With a decent velocity estimator and a velocity input into the PID loop, a pretty good replacement of a velocity mode amp is possible
how would you tune the pid with both position and velocity? would it be 2 sets of PID?
(with 1 pid module)
velocity is used internally for position mode anyway
it's just done by differentiating the position feedback, which can be grainy
so getting a nice smooth D term from hardware would eliminate that problem
I was thinking - inner - outer loop where they where separate.
like tuning the velocity drive pid - then the emc position pid.
if that made sense
that would need dual loops
heh: "Linux is user friendly, it's just particular who it chooses for friends."
awallin: Hallo, have you tested your stepper?
gefink: hi, I switched machines, no more realtime problems now
can move the computer to the mill during the weekend
did run some tests earlier: http://video.google.com/videoplay?docid=-6222166294337898920&hl=en
aw: i confound you with one else.
i cant see google-video, greasemonkey dont detect the downloadlink and flash is baned from my box
good night all
regarding velocity FB for PID - I've been thinking along those lines ever since I added a better velocity output to the software encoder
the vel (if present) should be used for the D term
jmkasunich: so you would not enter a D term?
what I'm saying is that a modified PID that used velicity input would use it when computing the D term
yes you'd still tune with D, but you would use velocity feedback, not derivative of position, to calculate it
we thought about writing a modified PID like that when we were at Stuart's shop doing the mixed scale/encoder feedback
but we wound up just using two PIDs and summing the outputs
if you have tachs and adc, you could use that too, it would work great
one had I gain only (scale FB), one had P and D and FFx (encoder)
anonimasu: re: your response to my rotab thing from a long time ago
the process doesn't rely on the accuracy of the center hole on the table
[22:43:52] <skunkworks_> http://imagebin.ca/view/bmhQu9.html
is GL acting up or is that a screenshot artifact?
anyone got a link to the newest live cd??
rotab?? cener of rotation issues??
I tried a few times - but it always split the image.
still sub 10us
i may have a new face in the house soon... sold the fidea teck on EMC2 in 1/2 hour... ; )
and they do offer a linux based version of their control ... on machines
they have a dedicated motorlola cpu ( he wouldn say what) in the pci slot doing the CNC ALL c and assemblyt apperanty then another pc as awired in GUI
dmess: just a center bore of unknown accuracy
map it with a bore comparator??
is it on a machine??
it isn't an issue, i already explained how to line it up without relying on the table's accuracy
lewin2 is now known as lewing__
any one run the new live cd?? i want to show it to some one tomorrow... will it impress???
if it boots on the PC you use, and that PC doesn't suck for RT, then yes, it will impress