anyone take a look at why M66 fails to respond to hardware inputs?
BigJohnT: all I can say is that it works for me..
are the inputs long enough?
hmm, how long do they need to be?
the input is on for 10 seconds
so I must have something wrong
do you have a snippet of code where it is used for an example?
hmm.. not really
but it should be something like:
I had M66 P0
perhaps that is the issue
P0 should be fine too
good morning all
finally! I am loading software on the G&L.
I think all peripheral systems will be completed today.
There are some new pictures on www.mpm1.com:8080. take a look
alex_joni: me brain == jello (been up all night)
stustev: is that you in the last pic?
stustev: gonna run emc2 on it? looks quite nice :)
stustev: what is 'W' on that machine?
JymmmEMC: sorry to hear that
Yes, it will be running EMC2. The W axis is the table movement in the Z axis. The machine has 4 axes - all linear. XW on the table. Y is the column up and down. Z is the quill.
yes that is me in the last pic
stustev: hi. do you think you could run 'convert' on those images to make them a little smaller?
fenn: sure - is convert a command line application?
for i in `ls` do; convert $i -geometry 1024x768 small_$i; done
it's the command to use the imagemagick package
I will do it
maybe -sample will look better than -geometry
i wonder if art will make mach open source
fenn: I doubt that
someone else is taking over for him
does he have a cadre of loyalists to take up the mantle?
what's to be gained from that -- a linux port?
dunno.. what's to be gained from keeping it secret?
fenn: he does.. he set up a company, and employed 2-3 guys
they will be taking over
fenn: he keep selling it?
er.. i guess i misunderstood the meaning of 'retired'
stustev: how long does it take you to tune a axis? and what is your favorite way of doing it?
fenn: he stops developing it.. but I think that's about it
Just this once .. here's the log: http://www.linuxcnc.org/irc/irc.freenode.net:6667/emc/2008-01-11.txt
I have only tuned the 3 axes on one machine with EMC2. It didn't take very long. Chris Radek was here when we tuned the axes the final time. It was surprising how quick they tuned. We added P until the axis started to oscillate, backed P off a little bit, added some D until the halscope showed the improvement was slowing then added some I. This was very easy and quick and very accurate. Chris may remember is a little differently but this is what
Neat - thanks
stustev: you got cut off at "differently but this is what
Chris may remember this a little differently but this is what I remember. The G&L will be our second machine with EMC2
stustev: great stuff :)
are these your machines? or do they go out to customers?
(does chris get one? :P)
* archivist finally finds the pic with the cast iron, a serious lump!
the G & L
[14:46:42] <archivist> http://www.mpm1.com:8080/machines/GandL/S1030182.JPG
* skunkworks_ loves those tumbstones.
stustev: it looks great. has it moved yet?
hasn't moved yet
the pictures on the website are smaller now
stustev: are these your machines? or do they go out to customers?
these are our machines - the hitachi may be for sale when we get it up
the hitachi looks like a neat little machine
stustev: how long did the retrofit take for the G&L?
does/will it have a pendant? I ask because I don't see a jogwheel on the control head
months and months and months - the actual work I will know later - if we devoted the time to it you are looking at about a week's work for two people
it will have a jog wheel on a cord/tether
cradek: the center knob of the three black knobs under the display is a jog wheel, the left knob is feed override, the right knob is spindle override
* alex_joni is happy someone actually uses spindle-override
spindle override rocks, its the tool saver
stustev: aha, I couldn't read the labels in the photo
alex_joni: I use spindle override on the lathe
hmm, I wonder how it works with CSS
BigJohnT, cradek had the right answer for you
(on the tool offset question anyway)
dave! how is the tuning going?
skunkworks ... slow
I need to get the system set up on a real machine
I think I can get a step response off halscope and a battery box.
PowerPath on Oracle Unbreakable Linux "Starting PowerPath: This package requires RedHat RHEL5. failed"
but then I need to know how to use the information I get
SWPadnos: I was away from my computer
and sadly, the logger was gone when he explained it to me :)
but basically you need to use G43 instead of G41/G42
G43 is "tool length offset", which with a lathe-type tool table offsets in X and Z, rather than just Z
< making notes to try when I get back to my emc computer
I did figure out the G92 Coord Sys Offsets after I logged off
I was missing the point that it makes the "current" point to the coord set...
ah - good that you figured it out
The G43 makes sense that it would offset in the X as well as the Z, I'll give that a try when I get home
Thanks for remembering
I see you survived the trip to Costco
I've got the coupons lined up for the next trip :)
So assuming I can get a step response off my axis how do I get from there to PID?
first, use HAL to generate the step, don't just hook the motor to a DC supply :)
also, the step response thing is just one way of doing it
have you read the Gecko G320/G340 manual?
OK, slow square wave so we get full response
Not really but I have one
ok, take a look at the scope traces in that manual, they show a good representation of underdamped and overdamped response
I was hoping for something that takes less art and is more quantitative
incidentally, this is the "scientific" method of tuning, it may not be the most practical :)
then use Ziegler-Nichols
I've usually done just the first part of ZN
[16:45:54] <SWPadnos> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Tuning_EMC2/HAL_PID_Loops
take P to osc then P = 0.6*osc value
yep, then use the osc. period to calculate I and soemtimes D
You mean in 60+ years we don't have anything better?
heh - dunno, I'm not that old ;)
ZN was conceived for tuning hydraulic actuators - I think it is definitely better now
hmmm. I could have sworn there was another wiki page with PID tuning info, but I can't find it
the wiki stuff is good as far as it goes but expanding it could be most useful
your mission, should you choose to accept it ...
a bit more nuts and bolts on the hal config would help
conceptually I kinda know what needs to be done but ....
oh right, the "multiple ways to do it" and "more steps to follow" sections could use a little more meat
I don't think it's necessary to tune a step response
I use small incremental moves which are a lot like steps, but you can easily program them in gcode
yep - that's probably more practical, since EMC won't output real steps - only small acceleration-limited ones
I like to make a long program with small rapids back and forth with delays between
do you use an insanely high accel at first, to get it close?
step response should give you a lot of information in one shot
then you can set up the scope trigger so it constantly shows the latest moves
I used to have the paper by the NIST intern but it needed a bit more fleshing out also
driving with a sine wave is a lot easier on the mechanics
One should be able to do a Bode plot ... that just gives you freq response ... not values for PID
but you need to sweep the sine wave, and measure amplitude over the full range
I'm not sure that's really useful for PID tuning
and it would take a long long time to get a plot
Not really ... just enough cycles at each freq to equilibrate
well, it can take seconds for RF plots, with much lower frequencies involved, it should take much longer to get the data
It may depend on how accurate you need the data
sure, plus you don't need a 6-decade frequency sweep
0.1 to 20 or 40 Hz should give you slope
60 Hz+, in case there are line-related issues
(not that there should be with a DC system, but still)
Yeah .. Jon says his system has a resonance at about 250 Hz
actually, since the control bandwidth is 500 Hz (based on a 1KHz servo cycle), you should go to 500 Hz
that would be 1/2 of the control bandwidth, which could be explained by the discrete-time nature of the system coupled with feedback lag
response should get pretty bad at high freqs ...
or I should say "possibly" :)
the system runs at 1000 Hz, but there's a delay in response - this isn't a continuous system
so although the timing resolution is 1 KHz, there may be a component that's at a lower rate (and 1/2 sounds as good as any)
Jon tried his system at 5 KHz ... and said the problems due to quantatization of D decreased dramatically
but it pushed his 600MHz system rather hard
I can try that on a 1.8 (?) GHz system
if it has a parport. hmmm
That had better make a difference
not much actually
well, not in motor control anyway
the GUI should certainly get faster :)
it may depend on the chipset
I was surprised - I have some embedded PCs here, with Intel chipsets
PCI cycles are nearly as slow as parport cycles
and their realtime is good
at least I'm getting 32 bits at a time
the chipset doesn't do any read or write combining, so each individual read/write is a full PCI setup and single word transfer
well, the RT response is good, but the PCI speed is crap
I guess it depends on what you need
with a core 2 duo (with one CPU doing bogus nothing), I have RT latencies in the 1 uS or less range
that should keep you alive
it was ~200ns max at one point - I don't remember exactly what made that happen though
yeah. but it is funny that loading one core down makes the RT core so much better
I've seen neg figures also but wonder if they are rollover
haven't figured that one out
I don't suppose that running the GUI on another machine will speed things up much since most of the work is probably done before
it gets sent over the wire ????
well, one thing I had to do on those machines was not start X
I stripped down the init
(actually made runlevel 2 what it's supposed to be - multiuser but no X)
so you really don't have a emc machine at that point
it runs a HAL application at startup
it could run EMC, but I'd have to do it with keystick or something that runs in a terminal (or at the console)
SWPadnos: did you try using the no-hlt kernel boot option to see if it had the same or similar effect to running a busy-loop?
I don't have it set for auto-login though, so that would likely be a little sticky
jepler, no, I didn't. I can try that
SWPadnos: just something that crossed my mind as a possible explanation
hlt causing some kind of long to exit state like power save mode
sure - keeping the CPU alive was something I had considered
ok - I'll try that. didn't know there was a no-hlt boot option :)
is that using isolcpu?
hmmm. it'll have to wait a little - I need to pull the Mesa card to get at the IDE connector so I can use a flash drive that's configured correctly ...
it's of no great importance
don't break a working machine just to do that
"never touch a running system" unless you're bored
I was just using a normal HD during development - flash is pretty slow
it's not a running system, it's my development system for this product :)
It depends on which running system .... some people are rich enough to have an extra "experimental" machine around
oh, I should post the photos of the power supply - I figured you guys might geta kick out of seeing something like that controlled by HAL
daveengvall__: i've got piles of "experiments" laying around - does that count?
some people have customers who provide the equipment ;)
anyone using the C7 mini-itx boards besides Ray?
SWPadnos: so, they have insurance or something? who pays when it explodes?
so what does hal do for you PS
fenn, yes, they or their customer will pay if it explodes :)
daveengvall__, well, it's kinda silly, but HAL is talking to a PLC and generating pulse waveforms
an dnot doing much else
so ... if it does what you want
oh, it does what I want. it also does some error management and scaling as current drive modules fail, that sort of stuff
at 10 KHz
pretty fast for a plc isn't it
no, the PC/HAL runs at 10 KHz
s can be fast
the PLC is slower, but it's a very expensive PLC
(something like $17k in the configuration they have)
allen bradley ?
with about 15 I/O modules, and an extra backplane
to hold them
gotta run ... still don't have the machine ready to go to the shop but I'm getting close
lot's of analog
see you Dave
not so much analog, though there are about a dozen analog connections to the PC
once I get the shop stuff running I can play with tuning ... maybe by different methods.
plus a dozen digital connections, plus modbus :)
SWPadnos: that sounds like a fun one
yep, for strange values of "fun"
the more "fun" a project the more we charge LOL
heh - this sure fits that category. I just sent out the $50k+ invoice :)
that's what I'm talking about
it'll be better when the check arrives
that's the part I hate the waiting for the check
we have direct deposit with one customer
one customer we do a little work for takes 3-4 months to pay
their a pain
I know the type
and it's for piddley stuff like call out's
you know the call, "my allen bradley quit can you program it"
and someone cut a prox wire or something dumb
"yes, but to go to your site, we'll need acreit card number" :)
we don't take credit cards so it's cash or check
our first customer 6 years ago wrote us two 6k checks that bounced
that's tough to swallow
we finally got our money 1 1/2 years later when we put him in jail
hmm the mill is quiet gotta go change parts
so.. i add myself to a group "usermod -a -G fuse fenn" then "chgrp fuse filename; chmod g+w filename; touch filename" and i get permission denied - what?
you have to log in again
check 'id' and you will see that you're not (yet) in the group
you're right - but id says i'm in fuse
if you ran 'id' in another shell (xterm?) that may have been a new login shell
oops yes it was a different terminal
cradek: saw pics o my mill ?
you put them up on wiki?
alex did it for me
[18:19:22] <micges> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Micges
why do you need bellows for a laser?
what is bellows ?
covering the x axis I bet
the black rubber thing
this is table mill
laser is next to build by us :)
I thought it was a laser
ok, I added the link on the Case Studies page too
micges: is that a router?
is it for cutting wood or metal?
desktop PC is there
by router I mean a 30,000 rpm cutter for wood not a network device sorry for not being clear
micges: how much Z travel does it have?
it has 24000 rpm spindle
from 5000 to 24000
phew. I didn't kill the embedded PC by plugging in the flash disk
(and moving the 5i22 around)
glad to hear that
z has 60 mm
made for cutting sheet material?
to make very precise (0.01) matrices in thin (1..2) mm materials
ok, now I understand
how it is say when machine leveling material
I dont know the work
prepare material to be same height on all sheet
do do a face cut is to cut the top of the material flat
jepler, no-hlt didn't help
but it may not be spelled right - I'll look into it
my dic have no that word
latencies are ~17 us without a do-nothing (with no-hlt), and 600ns or less with a do-nothing process
hmmm. so it may have helped the loaded case, but it's not noticeable in the unloaded case
it does have a peak at ~15us every once in a while, but otherwise, it's well under 1us
some of the max_lat numbers are negative :)
err - lat max
ok my point is machine is stiff enught to do facing herself table with 10mm tool :)
negative is not good either
next time dave is in - someone should tell him to try the at_pid. It does the Zn method.
they're still <1us absolute value
it got me close but my period was outside of the zn specs.. (just the motor)
wonder if 0xfed8 is a valid ISA address
to all: tool height compensation have any known issues ?
I plugged in a cardbus parport
it shows up with I/O range: FED8-FEDF, FED4-FED7, FEE0-FEFF
and memory range: FFEFF000-FFEFFFFF
well, see if it works! :)
it's not a Netmos.. it's an oxford OX12PCI840
SWPadnos: got nothing to test it with atm..
at home? nope
not even a meter..
I could bend a paperclip, and check if an output goes back to an input
got a spare speaker?
a paperclip is all you ever need
and chewing gum wrappers - can't forget those
[18:53:13] <alex_joni> http://superyay.com/wp-content/uploads/2007/01/macgyverclip.jpg
now.. the clip is easy.. where do I get an emc2 now :D
wonder if vmware can replicate it
[18:54:35] <SWPadnos> http://en.wikipedia.org/wiki/List_of_problems_solved_by_MacGyver
in case you missed that yesterday :)
doesn't show in lspci :/
in vmware or a real Linux boot?
if it's Windows and you're running Linux in VMWare, then I'd imagine you need the Windows driver loaded before VMWare will see it
it is loaded.. that's how I got the I/O range
but vmware doesn't replicate all stuff
only USB and cdrom
do you see any parallel ports in lspci?
micges: nice work, what does the air cylinder do?
hold and vacuum out
has connected vacuum removing of chips
micges: is your shop in the forest? I see lots of trees outside of your window...
haha.. I found a BDI4.08 cd
its no forest
< playing Eric Johnson, Cliffs of Dover very loud...
we generate lot of noice
trees are to reduce it to neigbors :P
BigJohnT: what you are doing?
working and playing music
helps me focus
this is me http://i47.photobucket.com/albums/f163/johnplctech/hunting/HPIM0281.jpg
me to :)
ok :) what you do ?
design and build automation equipment for factories
bbl got to go to town
the deer was tasty btw...
How can I build control of deep of tool in material ? We have one solution but I hate it :)
with a mill or a laser/plasma?
mill and laser
I don't know for the laser - you probably have to control the laser output power, and keep it proportional to motion speed
let focus to mill
for a mill, add a Z axis :)
and if I have material height 10 mm on left and 15 on right and want to make row 2mm deep ?
no more or less
you need some way of measuring the height, like probing
once you have that, you can set the measured height to 0 (using G92 offsets maybe), and run the G-code
yes I have feet with connected resistance something
oh - are you talking about something that is not level, and you want to move from the tall part (15mm) to the short part (10mm), maintaining depth?
like engraving on a computer mouse, for example :)
hmm.. anyone knows where I can find cheapish linear guides?
precise tall part have 1mm, short part has 0.7 mm and deep of row let have 0.5
yes sth like that
that isn't possible to do unless you have a way of continuously measuring the part height
if you know the profile, just use your CAM program to do the right thing :)
but if you need to measure, that's a harder problem
maybe you can set up something like the THC for plasma
all drilling in material is controlled by our hardware and its ok but EMC have no access/control of it
only if you have a means of continuously measuring depth of cut
no preview/stopping/restarting of it
(or part height, which amounts to the same thing)
analog resistance signal we have
its converted and by PID correct the Z pos
ok - that's the kind of setup I was imagining
I think there isn't much else you can do, except maybe change the sensor (to a laser interferometer or something)
food time - bbl
crap.. doesn't work
any idea what I could try?
lspci shows it at 6000, 6020 and 6028
but it also says [disabled]
this is what mine looks like, and it works for me
mine says lines like:
try loadrt probe_parport
Region 0: I/O ports at 60000 [disabled] [size=32]
fenn: it's loaded
probe_parport probably doesn't do anything with PCI ports
only ISA PNP ports
would the board need initialisation by the 'parport' driver first?
e.g. linux parport or parport_pc ?
there is a lot more 'detect & enable' type code in the parport_pc driver that I did not put in emc's probe_parport
so -- worth a try I guess
modprobe parport_pc doesn't seem to do anything
rmmod parport_pc reports module does not exist in /proc/modules
maybe its that line in modules.conf
use insmod directly if loading is disabled
or 'man modprobe' for the flag to ignore modules.conf stuff
-a or -i I think
the animated picture is not quite right
-i causes a segfault
well that's interesting
but thats great and simple idea !
micges: if your table is really sloped by 5mm, you should cut it flat in place
if the sheet material is just warped, you can use double sided tape to hold it down
no, 5mm was to imagine problem
scale is 0.5..1.5 mm :)
this is not single sheet problem but technology of production
will try the parport_cs driver after the next reboot
yay.. it works :D
I had to load parport, parport_pc and parport_cs
after unloading them, I loaded hal_parport cfg="0x6020"
and now I can toggle input pins with a wire an toggling an output pin
good night all
er, what is the peak AC voltage in the US?
at my house it's more like 120
but what does "120V" mean? AC amplitude? or RMS voltage?
AC voltages are measured as RMS
so yeah peak-to-peak is *sqrt2
wait a sec -- isnt peak2peak twice the amplitude?
cradek: don't you live in finland or something?
maddash anonimasu lives in sweden
[21:30:15] <cradek> http://www.ee.unb.ca/tervo/ee2791/vrms.htm
oops disregard what I said about peak-to-peak
"peak" is *sqrt2
[21:41:15] <maddash> http://incredimazing.com/page/Hi_Im_Vista
i burst out laughing everytime I lay eyes on the Unix avatar
i really think this is more appropriate http://mytsoftware.com/misc/linux2.jpg
[21:47:55] <SWPadnos> http://incredimazing.com/page/Science_vs_Faith_a_flowchart
fenn: that's so untrue
[23:26:01] <LawrenceG> http://www.youtube.com/watch?v=RJEXbEXn0CQ
emc2 servo project
what do the blinkenlights mean?
they are on the output of the servo drive.... + and - current drive
ah - ok
they actually are showing the voltage across the motor... since the drive controls current, they show a little jitter as the drive command jumps around
ok - they didn't quite seem to match up with movement - more with accel
I am not sure how well sinked the videos are