geckodrive seems to be dead
Mariss said he was having someone working on his website about a week ago.
wanted to check something out :P
wanted to read up on there servo drivers
ty googles cache
now how would i do that? never tried it before
You don't have permission to access /index.html on this server.
Additionally, a 404 Not Found error was encountered while trying to use an ErrorDocument to handle the request.
Apache/1.3.19 Server at goofy.link4pc.com Port 80
[00:21:22] <Jymmm> http://www.google.com/search?num=100&hl=en&lr=&safe=off&q=servo+site%3Ageckodrive.com&btnG=Search
nope... she's dead
works for me
[Global Notice] Hi all. We're currently testing a patch which should resolve some kline removal issues. At some point this weekend, we're probably going to want to do a network-wide reup. We'll keep you posted.
PhydBleep_ is now known as Phydbleep
cleaning up some of the semaphore code a little more...
did you get any further with the mode change bug ?
I didn't try further
let jmk take care of it
I only isolated it
paul_c: are you nice enough to pass me that wet kipper?
right out of kippers
dang.. then the wall should do it
* alex_joni smacks his head against the wall repeatedly
while (1) do smack_head();
my init function returned 0 that everything is ok
but furtheron I expected for it to return the base address
and because I didn't have a STG at address 0 it didn't work :(
I didnt sleep too well
had been sleeping too much during the day
* anonimasu feels like he's had a close encounter with a truck
I just found what I borked
oh well, could be worse..
* alex_joni suggests some coke
or any other form of caffeine
I've had a cup of tea
earl grey with some honey
should be plenty of caffine in it, but maybe I'll make another cup
wanna hear smthg funny?
[13:31] <alex_joni> my init function returned 0 that everything is ok
[13:32] <alex_joni> but furtheron I expected for it to return the base address
[13:32] <alex_joni> and because I didn't have a STG at address 0 it didn't work :(
great stuff :)
I only wasted a few hours to track it :(
I am thinking of starting to play some with HAL
I really dont have anything better to do right now
*waits for another cup of tea*
it seems to be working ;)
getting inputs to HAL
I need to buy mor honey..
outputs working also
anonimasu: did jmk get anywhere last night?
I mean.. any luck with that mode change?
alex_joni: no he it stopped with the fact that he's X sessions wouldnt work
now why didn't he set up a remote GUI....
an0n's firewall might have prevented that :D
I've poked a hole larger then norway in it..
hell, you could probably smuggle a truckload of drugs past it.
iptables -A INPUT -s druglord/32 -t drugs -j ACCEPT
the tea is pitch black this time.
gimme ten minutes, and we will try to set up a remote session on this box.
nice thing this interne
03alex_joni * 10emc2/src/hal/drivers/hal_stg.c: various bugfixes. moved some functions around for better reading. DIO is tested and works, rest needs further testing
OK alex_joni - You wanna give remote GUIs a go ?
why not :)
started emcserver on an0n's box?
gonna send you an nml config
I got one.. lemme try
You may need to forward port 5005 through your firewall
Let's try without first...
what port does emcsrv run on?
Can you copy that nml config to emc2/configs
ok.. had my remote nml set to connect to port 1234
emcsvr is set for port 5005 in the NML config
I connected succesfully
but X&Y are red
and Z is yellow
lemme shutdown & do a fresh compile.
does seem to work
All gone green here.
ok.. probably my old emcsh.exe then ;)
you're not running Doze are you ??
oops, you caught me ;)
* paul_c slaps alex_joni with a very large wet cod
heh.. new fishes ?
lemme try a newer emcsh.exe =))
did you forward port 5005 ?
why should I?
I need to connect to 5005
Just wanted to check...
all green now
* alex_joni tries to open a file
seems to run
the one you opened
* anonimasu is slaps alex with a mop
ngc file names is still one area that needs looking at.
I am having a cleaning day
* anonimasu mops the channel floor until it's nice and shiny while at it
the $filename is assembled by the HMI and it should be task that does it..
anonimasu: you can come over here and mop mine too
paul_c: too tired to think about that right now
it would presume using my memory
and that's still cluttered with stg stuff ;)
I just pushed abort
did it stop?
ok.. seems to be working ok
do you want to try the other way too?
right, now to adjust period
can you connect to my ip?
emcsrv running on port 1234
5005 is taken
oooh .. new people
how was europe?
I remodeled an old farm house for my kids.
Good and drink were awesome.
took some pics?
yea.. I figured
* alex_joni and paul_c are playing: I control your machine, you control mine
Just a few. My old sony digital went tits up before the trip
turn DEBUG down, & the error can be triggered sooner.
and tkemc :P
paul_c: tried to connect?
Joe M. has been trying to contact you...
I got a couple of messages when I got home.
'k.. I also need to discuss some of the ini params when you get a chance. accel in particular
I was looking through SF. Lot of changes.
been tracking a bug with the GUI locking up at intermitant intervals.
INI anytime you ae ready.
Tkemc or all of them?
might be emcsh related
stress out xemc/yemc and they also lock up.
Wot it just quits connecting?
it just sits there
paul_c: tried to connect?
nope. None of the buttons respond, can't close the window, and redraw dosn't work
Turns out it is not a tcl/tk bug, nor is it anything I've done
hello rayh :)
True if it happens with x..yemc as well
Figured it out - It's a futex bug in the pthreads library
Can you connect to the running emc using another instance of a gui.
So it is just the one instance that dies.
It doesn't die. It freezes
What conditions cause it?
I don't know.. Ran loads of traces on emcsh, and it was something at a very low level
I'll have an updated library on the next build.
I've seen a few setups with very small PERIOD values where I can lock up stuff but I assumed that it was priority inversion in rt.
so it's bdi-4 related?
I mean.. pthreads lib used on the bdi
Fedora may also get hit by the same bug
and any other system that uses it
Moving on.... Sherline ini
You had increased the vel & accel in TRAJ....
sorry.. a question for rayh :
Yes. I did that when we added individual axis max for these
rayh: is it possible to run a program from an arbitrary line ?
i'm using mini, emc1
is it too difficult to do ?
There is a restart set of buttons that pop in to the right of the program display.
Unfortunately I didn't allow for a type in number but the blue line
in the program is the restart line.
At least it is supposed to be.
rayh: cool :) thanks
i will try
Let me know what happens.
You can show the restart buttons under the view menu as well.
Do we have problems with sherline accel?
lost steps on Z
Ah. We're back to that. What do the lost steps look like, random?
This is much more repeatable... Turning down accel in TRAJ cures it
Turning down accel there slows the accel of the rotary.
the vel & accel settings in each axis section only affect jogging
I didn't think that was true when I set up the first rotary.
The max Vel (per axis) may be honoured by some calcs at the task level
The only time the rotary did not honor the individual accel was during linear/rotary coordinated motion.
but the accels are not when doing a coordinated move
So... We can either reduce accel in the TRAJ section, or do some additional coding in tp.c to limit the max accelerations.
If you turn down the accel in traj, you get 180 degrees of accel and then 180 degrees of decel.
I spent days testing that setup and did not see any lost steps.
Do they only happen with za motion or with any coordinated?
any coordinated move
* paul_c has to go for lunch.
* paul_c is away: Dinner time.
I don't have a system set up here now but will do that
rayh: I have progress on the STG driver ;)
dio is tested and works
Great. I saw your additions.
You got the STG card.
yeah.. a while ago
thanks to Matt for that
You got servos running?
nah.. just metering the outputs
no servos around .. unfortunately
Well that works.
hook it up to the robot
anonimasu: heh.. I wish ;)
And stand back!
no idea what the interface is though
the new ones run on CAN
with 1kW ac drives
btw.. got that turn/tilt table on friday
8 tons ;)
Wah! 8 tons.
smthg like this: http://www.weldplus.com/printit.asp?t=RP210003P&i=prod_images/RP210003P.JPG
yet a bit bigger
Wow big a part can you place with that?
1910 mm turning radius
I'll have a hard time to fix it properly
That's almost the size of the kitchen I built for the kids!
mounting plans say: use a 40mm sheet fixed into the concrete with 400 mm long sticks
and weld the base plate on that
We need kinematics for this type of device.
yeah: smthg like... take it slooow
With actual tool tip feedrate.
the new controller on the robots handles up to 18 axes coordinated
typing is not my best
Beautiful country you live in.
Imperator_: Try using tab (for autocompletion)
works on the nicks
maybe because it is raining the hole summer now
* alex_joni goes hunting for some food
* alex_joni is off to the sea next week ;)
See you in a bit alex_joni
This was my year to see a couple of botanical gardens. Great in the cool, wet weather.
Mostly worked on the house though.
you have visited germany ?
Several times. I have a son/daughter-in-law/enkelin there
Think I may have spelled that wrong. Grand daughter.
Enkelin is right, i think
I know I pronounce it wrong almost every time.
sorry i mean do you have visited germany this summer ?
How are things with you Imperator_
no problem on irc .-)
Where are you located now?
still in germany, same place
but i have to move a bit closer to Stuttgart the next time
Oh. I thought your job ran out.
have a new job
Alex is a fast food guy
nah.. I just prepared it
now I'm about to eat it
* rayh need more coffee.
prepared as in heated up ;)
is it possible that halscope will not be compiled at the moment
the other guy who is working with me with the mesa cards cant find it
halscope and halgui only get compiled if gtk is installed
i have a emc2 that is some weeks old, there i have halscope
you probably have gtk on your machine
tell him to install gtk-1.2
how's the mesa stuff going?
I#m still drawing on Eagle, and he is fighting with the pin config and so on
of the FPGA
rayh: where in germany have you been?
whereever he was, he was in the rain
this summer is terrible
that bad over there?
it's nice over here
last week we had about 10-20�C
and 7� at night
we got 26 right now
i was close to switch on the heating again
Imperator_: this is where I'm going
[14:10:34] <alex_joni> http://www.tzgrovinj.hr/eng/home.asp
Not far from Dortmund.
Flew into Amsterdam
That country is a bit flat for my taste.
* rayh reads back.
What Mesa stuff are you working on.
rayh: how was the rain in DOrtmund
do you know the Mesa FPGA card?
Always rain. never a day without.
I do know the PCI version of the Mesa
we are making here some breakout boards for this card
Oh. I've used both the 7I33 for analog servo
i seams that we make about four
And the little 100 watt servo drive.
What sort of breakout are you making.
brb gotta put the breakfast rolls in the oven.
we make one board with a quad 16-Bit DAC and one for the digital io needed for the amps
Imperator_: use a STG *g*
and another one for counter input and one for the limit- and referenceswitch input
alex_joni: :-) toooo expensive
now seriously.. what are you using for counter input?
only some RS-422 drivers, all is done by the FPGA
RS-422 recievers, of cause
So you are writing your own fpga code.
* paul_c is back
at the moment is only Jan (the other guy in north germany) writing some stuff for the FPGA, but i will learn that also
Okay. We got Pete at Mesa to write a version for the mesa
stuff that we put in SF.
Pete is a employe at Mesa ?
PeteV is not. PeteM is.
PeteM=Pete Mesa ;)
rayh: Did you ever get a chance to order a 4134 board from Mesa ?
According to PeteV Mesa made some mistakes on the PCI bus.
I did and they were out at the time. I'll try again tomorrow and see if I can get one direct to you.
Thanks for the memory jog.
4134-2 would be best if the funds cover it..
They should more than.
alex_joni: the ways i take with the bike are normaly nowhere to find on cards
I agree ;)
rayh: what do you mean with mistakes on the PCI ?
Oh. PeteV tried to explain it to me but I'm limited.
It seems you can only address stuff at the 16 bit wide port because there is some sort of overlap issue.
Imperator_: smthg like: it's actually using timings from the AGP bus, but if you stick it backwards in a PCI64 it might work under low stressload
And if you use more than one card there are problems with the bridge and auto loading of the fpga.
You probably should get the problems from PeteV himself.
is that a software or hardware problem ?
I don't really understand it all but I think a small hardware change has to be made.
drill some holes into the fpga ;)
We did get three systems running okay though.
do you get some samples from Mesa ?
using the 16 bit definitions
Yes there is a bunch of stuff in their software package.
how do you mean ?
The SF stuff will install a complete system using HAL for
four analog outs using the 7I33 interface
jep, have sean that
It will also read up to 6 or 8 encoders so you can run a handwheel and 4 axes
i have only bought the PCI card itself
Okay. Sounds like the interfaces you are thinking of will make a very nice addition to the system.
because i think that PWM for driving a real amp is no good idea
that is only usefull with the boars from mesa, i think
but i want to use real amps
I liked what I saw of it with Jon Elson's stuff but that all comes from the same engineer.
have some Siemens Simodrive and some smal amps from galil
with smal DC-Servos that might work nicely
Do they take +-10 volt signals?
with tacho feedback
They will make a good start then.
brb time to taste the rolls.
you mean a comercial start ?
What kind of machine are you thinking of running with the Mesa?
rayh: iirc a bigger portal machine
with 2 slave axes on X
there are two machines, my own one and one that i want to realise tith some friends
err.. master/slave axes for x
my own one would be a normal machine, but High speed with a polymer concret frame
and the big one is a gantry machine withmaybe a granit stone as table
we found a chaep stone but now we are searching for a place for the machine
thats the bigger problem at the moment
What is the thinking about axis slaving? HAL or above?
HAL is ok.. but homing will be a problem
Yes. I wonder if it would be possible to just freewheel one ballscrew while homing on the other.
don't think that is a good idea
depends on the screws lead I guess, and could lead to some twisting
i think i will do it the way Alex suggested
I know that one company tried making both work together during startup and often ripped the balls out of one nut.
hm, absolute encoders perhaps..
to export some slave axis
have only short absolute encoders
for the big one we have normal linear encoders
the big monarch gantrys dont use screws, but those insanely expensive zero backlash rack and pinions
OK... Who wants to try remote networking of emc2 ?
paul_c: I can open the hole if you would like to play at the box :)
anonimasu: Already have it set up & running here
just wanted to see if anyone was wanting to try tkemc in client mode.
but it does require a recent emc2 compile
everybody is quiet :-)
Just switched the Z axis over to current feedback (torque mode)....
Got an initial tune on it, and it seems quite good....
Have to do more tuning, its doing abut +/- .0005 .....
paul_c: why a recent emc2 compile?
alex_joni: 'pends on how much has been altered in emc.hh
not much so far, afaik
it works ok over here (with an emc.hh from emc1 a while back)
SWP_Away is now known as SWPadnos
long time no C
no c++ either
as I can tell from the logs
yeah - mostly PIC and AVR assembly lately
thasa point... Someone was asking about GNU compilers for Z80 ?
I'll get with you on the truck business.
Look at SDCC - Got support for 8051, Z80, AVR,and PIC
interesting - do you know if it supports special intrinsics?
(probably a bunch of others too)
I'd suggest gcc for AVR
tried keil for 8051.. works ok
I usually need much more efficient code than a compiler can give me on a microcontroller
Keil is OK, Tasking is good if you need floating point
at least on the '51
bummer - sdcc says AVR is no longer maintained
well.. if you've got gcc.. why would anyone bother with sdcc
good question :)
gcc doesn't support microcontrollers
there's an AVRGCC that does
though it's probably not "real" GCC
[15:16:05] <alex_joni> http://www.avrfreaks.net/index.php?module=FreaksTools&func=viewItem&item_id=145
paul_c: gcc certainly does support avr
[15:16:57] <alex_joni> http://gcc.gnu.org/
! avr-gcc -v
Reading specs from /usr/lib/gcc-lib/avr/3.3.2/specs
gcc version 3.3.2
there's a good free libc with FP, trig, etc
lemme put it another way... gcc does not support the majority of 8 bit embedded microcontrollers.
darn, sdcc doesn't do 6800
Only problem with this is after an upward motion, the motor is stalled with 7.5 volts across it....As it can t get all the way to 0 following error due to friction....
this is strange...
anyone ever met this?
Default Trap Handler: vector 14: Suspend RT task xxxxxxx
Yes - Seen those before
any idea what causes that?
You used a NULL pointer in the RT code
nothing pops up
nothing at all ?
a long list of the boot log
then the last messages from /var/log/messages
and it ends with the message above
nothing between the adeos init message and the suspend ?
ok.. seems it's my driver doing smthg wrong
only normal messages
did you use a sleep in your code ?
ok.. isolated the function which is responsible.. will ask if I need aditional help ;)
and which function was that ?
the adc read function
in the stg driver
alex_joni, it looks like the stg_adc_read function could busy-wait for nearly a millisecond
yep - I think the the inb takes roughly a microsecond
*1000 = a millisecond
you mean the for at the beginning..
right it should be done already
but I probably didn't check tghe first time it runs
I also noticed that the stg_read_adcs doesn't check arg for null before it's dereferenced
sorry - stg_adcs_read
(which could lead to null pointer problems :) )
will fix that later
jmkasunich: any luck last night with that bug?
no - never got ssh -X to work
I saw paul talking about other ways to approach that this morning
you could set up tkemc to connect to an0n's pc
I have limited time today, leaving in about 90 mins
Hi. Sounds like you are making progress on the machine.
You might try a little bit of ff0 for that friction problem on z
Ok...about how much ?
is typical ?
Not a clue. FredP is the one who tunes this way.
what does the FF values do?
Last time I tried any FF0 there were some oscilation problems...
Try 1 and see if it makes a difference.
I m runing FF1 at .25 and FF2 at .03 right now, with about 6800 P, 3000 I, and 12 D ...
anonimasu: ff is feed forward
These will interact a lot the way that you are running it now.
anonimasu: it's supposed to insert some apriori knowledge about moves that are going to happen
before they actually do
Ouch. That is far different than I'd run but sounds more like Les's way of tuning.
*dosent understand at all*
The FF1 and 2 help to tune the following error flat and rise times..Which should both be zero...kinda like tuning an osciloscope probe...
the FF parameters basically put the commands into the PID loop
normally, PID only responds to errors in position
I tend to tune p less than 1200, I and D less than 10.
I seem to get allot better rest following errors with the high P s...
with FF, you get outputs that "know" what the velocity commands are, before any errors occur (and have to be corrected by PID)
ff1 somewhere in the range of 8. ff2 0 and ff0 just enough so you don't get movement while none is commanded.
FF0 is position, FF1 is velocity and FF2 is acceleration feed forwared...
If ff0 is to large, you will see an offset from commanded position.
I think I had FF0 s down like .005 and it was still making things unstable before I gave up on it ...
ff0 should not cause instability by itself.
I'd say you need to turn down P a bunch when you see that instability.
The HUGE I value is holding P from oscillation so both could come down when you see instability from ff0
What kind of drive amps do you run?
I turned up the I to try to get the rest value to settle close...There was not much effect from 1 to 3 K , though....
No there wouldn't be.
Let me add here a big IMO 'cause my way of tuning is not the only one out there.
I use the Gallil PWM amps, the Z axis was just changed over to current mode, from voltage mode ... Helped allot...For some reason a .2 inch (!) following
error had crept into the system in the voltage mode....
Now every thing is down intot he tenths...
Do the amps have their own tuning pots?
and were they factory mated to the motors?
The amps have offsett, input gain, loop gain and current limit pots...
I set those up with a stack of resistors for the load....
Okay. These will all interact with the EMC tuning parameters.
No...the motors are DC brush servo motors....
I have all the parameters for the motors, and put them into the .INI file...
wb9mjn: the params in the ini are for simulating a DC motor
not your case
Ok...thought they were used in the control loop...
I believe what I'd try is putting back the FF0 of .005 or .01 and try reducing the P and I by half.
Ok...will give that a try....
Hi Dave. wb9mjn needs tuning thoughts.
can you open the ports to your emc box again?
hmmm so what is happening
Dave can help alot with looking at accel and following errors.
Well, got rid of the big problem with the Z axis by going to current (torque mode) feedback...instead of voltage feedback...
The following errors are good, but the settling error is about +.0006 -.0001 ...On an upward travel, it stops with 8 volts across the stalled motor...
Misses the target short by .0006 ...
I'm going to log for awhile.. going to church, then maybe swimming with kids - Ray: I'll call you later & catch up with you - we might want to plan what to do next with Dave's lathe, your lathes, & Roland's Mazak...
is there a way to calculate load, baised on "accual position" vs "following error", where say the tool should be at 1" but because of the load of the cut, its accually at .989, with a .011 following error?
Probably friction related...30 pound milling head on 1/2 inch dovetail ways....
I was beginning to wonder what the plan was for the Mazak
keep me informed
The plan is: Finish it ;)
thats a goal, not a plan
plans involve who, when, etc
Need to get out for the walk ...back after lunch....
Will leave this up....
gezr...that is a lot of following error...can you raise the gain?
Seriously, I think some folks (probably me, Ray, Tom, etc. will go out & finish the hardware, then you might join us either in cyberspace or meatspace to help with software
dave-e : its not an accual thing on a machine, I dont have a machine yet, I was thinking in terms of a load meter type issue
I'm with Matt. Will plan on a return to Cardinal after Labor day.
ill finally have the money to finish the bike then work strictly on machining stuff :)
In the mean time, I'm working here with emc2 and the Motenc-lite card.
I want to come out there (in realspace)
Okay John that sounds very good to me.
hopefully when the weather is just a bit cooler than it was in June
but not too cool
week of Labor day I'm visiting family
I talked with TomP yesterday. He's ready whenever I am.
does the weather cool off that much in early Sept?
I'm guessing that the re-wire will take most of a week to land all the devices.
hopefully the trip can be a couple weeks later, my boss won't be thrilled if I want to take two weeks vacation at once, we will probably be deep in the throes of testing a new drive by then
Weather should cool some.
I think it would be nice if the wiring crew has some diag drivers to test individual tool change components
I don't see any problem with mid to late Sept for your work.
I'll get there a bit earlier and make certain that all the devices are in place and work properly.
how many tons is that Mazak?
No sticking power relays this time.
I can't lift that :)
I've used a pinch bar to move em a few mm.
they roll nicely on 1.5" bar stock
starting to sound like a plan
yeah - it's the first 1.5 inches that's the trouble :)
We had three guys on an extension bar.
gotta go -back later
mshaver is now known as mshaver_away
I need to go very soon too, back in 7-8 hours
have fun... john and matt
Good we can firm up a plan in the next few days.
jmkasunich is now known as jmk_away
* alex_joni just dropped his keyboard over the test box
hit the HDD.. but it still runs
Lucky. How come when that happens to me I have to replace the motherboard!
paul_c: You about?
heh.. I searched a while till I got this system
thought ISA was still around ;)
there are still a few boards
I got a couple of new ISA MB's a couple months ago.
Now that PCI Express is here, it may be a good idea to get PCI cards all figured out, before they also become (relatively) obsolete :)
one day, I'll have the time to install my Mesa card
rayh: Yes ?
The tp.c and individual axis max accel/vel.
Have to do the walk in the afternoon...hi...Tried playing with FF0...
I was wrong on my numbers ...In an upward movement it misses .0003 short, in downward .0001 long....
with FF0 set to 0
A weight issue.
Setting FF0 to .22 results in a downward movement settling on 0...and an upward movement .0004 short....
Yep...the milling head is about 30 pounds....
Stability was not effected....I thought I remembered that from several months ago, but apparently that is wrong, or maybe its an issue with Voltage mode...
In that case, reduce I by half and see if that will allow P to drive the axis closer.
I see the opposite Ray...Reducing P by half about increases the resting error to .0006 in the upward moves (from .0003)...
That s how I ended up with the high P, kept going up, and tweaking I and D for stabilty, and each increase reduced the settling error...
The drive screw is a 5/8 s double nuted ball screw....
The I value works like a drag on P. If reducing I doesn't do much but make it unstable then might add a bit of d
Yes...The D had the major effect...
any chance of using a small air cylinder on the spindle slide (supporting) like what people do for the knee mills?
Oh I missed that in your original note.
Don....is your encoder on the ball screw or on the drive motor/
I m thinking about that Dave-e ....
D=12. IMO this axis is way overtuned.
Its on the motor...Have another encoder, and one of the improvements would be to direct mount it...
wb9mjn: What happens if you cut P,I, and D by half?
The error when the axis comes tor rest will double, I think...I ll go try....
I tend to tune as simply as possible ...P first then ff1 then a bit of ff1 and a very small amount of ff2 then maybe a bit of D like 0.2
but then I've not done well in tuning the motenc card so maybe I can't say much
Ok...with FF0 back at 0, P at 3500, I at 1500 and D at 6, the upward movement stops .0006 short, and the downward movement stops .0002 short..
So a move from 0 to 3 stops at 2.9994 and a move from 3 to 0 stops at .0002 ...
Try a big reduction in I. Like I=5.
But keep one foot on the estop.
Did not seem to effect stability...The downard movement had no apparent change...The upward movement now stops short by .0008 ...
at 2.9992 ...
Could you list here all six values as they are?
P is 3500, I at 5, D at 6, FF0 0, FF1 .25 , FF2 .03 ...
A FF0 of .25 will zero out the downward movement rest error ...
What does D=10 do to it?
Will give it a try...
no difference apparent...
If the D is set to high, it will oscillate...have done that on occaision...
My thinking on the large I was that as things slow to a stop, the I term will integrate up the small errors into big control voltages, and shouild help push
it to zero...
When you set the drives pots, were you able to get full voltage out when you put 10 volts in?
ok .. back to basics...with all other values at zero ... what P will just give you an undamped oscillation?
i'm sure you did this at some point
I have a 48 volt supply, for 3 motors...I set the amps with a gain of 3.6, to get a max of 36 volts to the motors, which is the motor rating.....The 3.6 is put
into the .ini file too...
When i put in 10 volts, the amps puts out a PWM signal, with an average value of 36 volts...
or 10 v in gives max current in torque mode
Yes, I did Dave...As I remember, it was around 600, I think....
hummm....with ZN 60% of that would be your starting P
and then you add ff1 etc to refine the control
of course every system is different...my mazak tunes much easier than my Cincinatti
but encoders on the ball screws seem to make any system easier.
Yep...need to get the encoder on the ball screw...especially for this axis, I think, due to the friction...
I have encoders on the screws for the x and y ...
And those ones are 5K encoders, versus the 200 line encoder on the Z ....
ok .. so how many counts on the z for the error you are seeing/
aaarrrggg... geckodrive is still down :<
Its a 2 to 1 drive, on a .2 inch pitch screw....At 1600 quad counts per turn, its a resolution of .000125...so, the worse case with the best tune would
be 3 or 4 counts...
Gotta run. Hope I can get back later.
seems to me a better encoder here would help
Witht he 5 K direct coupled encoders, again .2 in pitch, that s .00001 in per count ....and I see about .0001 error either way at rest on the x and y, or about 10 counts...
that is in the right range... :-)
Ok..I have a 5K encoder for the Z which needs mounting yet...guess that is the next project...
not that more counts will cure all the problems but it can't hurt
sound good to me
do you have enough speed so you could change the gear ratio on Z
Yep, with the weight of the Z , I m sure the belt is stretching some...
At 240 inch per min up, the motor is getting 30 some volts right now.....Which is about right...With a bigger ratio, the motor would need to go
overvolt to stay at 240 inch per min...
any one have a spec sheet of the geckodrives?
* alex_joni goes away
Is the geckodrives site down Yuga ?
well it is myside
but unless one is doing something pretty soft 240 ipm is not a machining speed....just a fast way to get someplace
any one have a spec sheet of the geckodrives?
tool change, traverse between drill points, etc.
The 240 is the fast retract speed...Its pretty agressive when it throws the milling head up 7 inches at that speed...
But, with the axis tuned, there is not allot of shock , comes to a smooth stop...
yes that can be pretty impresive. My machine is spec'd for 400 ipm but I limit at 225 just to not drive it so hard.
have you tried the new tp
I hear les liked it but have not talked to him.
Yuga, try to search for the drive you want on Google, then bring up the cached page...
tp ? what s that ...
Oh...Yes...that s what started this alll...
"bring up the cached page..."
I had been loafing with the 100 degree THI down here for a few months...
paul made a small but important change in the traj planner...only makes a difference at high speeds
Then they mentioned that fix on the list, and I downloaded it from Jon Elson's site, as I use RC46 here...
Plugged it in, and had the Z problem....
Most likely, I had it all along, and just noticed it when testing the new code....
had the Z problem all along...
I'd buy that
amazing what one stumbles across.
actually the change makes a different for all moves :)
when forced to do some testing
it were missing every blend earlier :)
In the voltage mode, it had a really bad stopping clunk...which was related to it developing a .2 inch (!) following error on upward movements...
anon... so it blends better now
it didnt blend right before at all :)
although at higher speeds it stops the machine..
err it did
I found that out this morning...Then switched over to the torque mode, per Les's suggestion from last week, and the large following error went away...
all I saw was the funny bump at the sign change...could not demostrate on co linear moves
And have been working on the final tune up here with Ray and you Dave...
I might need to make a effort to tune my machine tomorrow..
tunig makes a real difference!
I need it for the USC to work nicely :)
although I am not running with feedback yet..
you'll get there
I hope so
I am kind of in a hurry since I've got parts to make
well, basically one need a machine to help support the other machines
speaking of making parts ... I need to go work on one
Bye Dave ...
hopefully this will be the last change in the process.
bye guys... later
03paul_c * 10emc2/configs/ (client.nml server.nml emc.ini emc.nml): Commit NML configs for server/client networked systems.
* paul_c nudges anonimasu
* anonimasu is getting nudged
do ya wanna set up emc2 for a remote GUI ?
hm, for testing or something else?
for JMK to play with
just need an ssh port & port 5005 forwarded on.
thoose ports are open now
but do you need 5005 forwarded from outside?
yes - That is what the GUI uses to communicate over
unless you prefer another port
it's forwarded now
and the ssh is alive
looks like there might be some compatability issues here....
sodit. Something in tcp_srv appears to be borked - It may be because of the bug JMK is tracking...
well.. I've done my bit. Time for tea.
les: are you around?
nah man, les is sqaure dude
just came in. was washing cars, trucks, and tractors.
and hitting wite balls into lakes and woods.
I am sun blind
needed to wear sunglasses.
they have this new invention now called sun glasses
les can you hear me ?
my sunglasses prescription is getting old
hang on...bad audio
I got a headset for $0.50, just testing it out
c ya Jacky^
dou you think the internal pc circuit that control cpu fan rpm or chassis fan can be used to read rpm from external device like router ?
uhm, may found: http://www.ugcs.caltech.edu/~dstaff/sysmeter.html
Oh bummer.... skype probs...
Who's got skype accounts here ?
paul_c: i've skype
* paul_c is wondering if it is worth setting up a conference call.
never tried ..
anyone wanna try ?
.... Boise, Idaho ?
yes..hello from boise
jymmm and I are on skype now
never tried confrernce
* Jacky^ don't speak english :
i can send music :)
Jacky^: I'll let you in to a little secret....
I don't speak Italian either ;o
paul_c: my username is jack_hack
add to skype, and i will send some mp3 :)
do you like KC and sunshine band ?
Lawrence has skype.... call ve7it-
LawrenceG: dsl ?
how much fast ?
dsl dropping out...storms
i'm 4 mb
60kbytes upload, 300-500kbytes download
les hang up
I love it
some day i want to try videolan :P
* LawrenceG waiting for a phone call
LawrenceG: are you waiting me ?
Chris was looking for victims
paul_c: need to test for multiconference if i think..
LawrenceG: lets try .. i add you
ok... my user name is ve7it-
i get problem tith audio device ..
calling you now
probably wrong dsp device
try a connect to echo123
it will help setting your audio levels
are you running linux or windows version
i'm looking to rexima..
Vol its ok, Pcm too
mic is 92% [R]
it should work ..
is it ringing?
i see the call
but can't ear sound
click green button
probably is the player i'm running
yep... need to shut down other sound players so skype gets access
working very well... Itally to Canada
well, so is skype is running mplayer or other players can't work ..
anyway 100 % good connection
LawrenceG: if i change the [R] source in the mixer..
should i be able to send directly music ?
darn.. thought I might catch jmk
jmk_away is now known as jmkasunich
jmkasunich: just the man I needed
I set up a function to write dac's
and it's added to a thread
a pretty fast one (tried 1000,100 usecs)
yet it seems that the dac output is written to with a frequency of 1.48 Hz ?
1.45 - 1.48 secs
I don't have a scope unfortunately
but hooked up a meter
and it's got a frequency counter
and on voltage I have an digital bargraph (I can see that bounce up an down)
the amplitude is ok (if I change the volts written to the adc pin, the amplitude increases)
what are you using for your signal?
I connected constant.0.out through a signal
in that case, the output should be constant too
but the same stuff happened with siggen.0.triangle (very low frequency)
that's why I changed
constant should be constant
and the signal seems constant
is the code on cvs, or do you have uncommitted changes
watching it with halmeter
but.. you cannot see it inside HAL
so output is changing, but hal value isn't
smthg like that
looking at the code
stg_dac_write actually writes to the board
and stg_dacs_write is the function registered to HAL
what bugs me is that it is very constant
1.453 HZ and 80.000 duty cycle
so 80% of the time it is outputting the correct value?
what about the other 20%? outputting zero?
like I said.. no scope around
and the meter is not fast enough for that
but I think so
how do you know 80% then?
* cradek loans alex_joni a scope
the meter reports that
even an analog meter is better than a digital one for things like this
[22:19:29] <alex_joni> http://www.xlsmess.de/html/body_digitek.html
how many channels are there?
lemme try with 8
same problem on all 4?
and see if the frequency changes
nope.. same stuff
you are driving one of the four channels with a signal, and the others are at zero?
should I drive all of them?
does the meter show this strange signal only on the channel you are driving? all the other outputs are zero?
night an0n - I'll be trying to work on the bug again later
lemme look at the other channels
others are on 0
all are zero?
ok, disconnect the signal from the first channel, move it to another one, and see if the output is the same on the new channel
also... do a halcmd show all and dcc it to me when you get a chanve
that's a bunch ;)
bits are cheap
samba? wtf, you running a doze box or something?
the laptop is on doze ;)
and mIRC too
and the test box isn't networked?
it is..but through this one
and.. no X running to have a decent speed up
* alex_joni likes console
and HAL for not needing X
hmmm.. just read smthg
from the stg utility:
dac test (watchdog needs to be disabled for this to work)
lemme guess, wd is 1.5 seconds or so
lemme find it
I had a printout of the stg manual laying around here, can't find it now
lol: If the jumper is out, no additional event will occur (besides zeroing of the DACs)
jmkasunich: thanks for the tip
wanna know what was wrong?
the jumper for the watchdog was set right
but it was oxidated
and didn't close the contact
I had it with one of the address jumpers
that's why I found it so fast :D
still something strange tho
if the wd timed out and zeroed the dac, why didn't the hal thread reset it a mS later
didn't look at interrupts yet
100uS later actually
it should get reset by watchdog writes
but.. maybe that should go through the other jumper possibility
like it was the jumper wasn't set at all
it's a 2 position jumper (3 pins)
you should either set it 1-2 for auto-reset
or 2-3 for sw reset
or floating for mystery-reset
I think through 2-3 the signal from the writes gets through
I set 2.0 Volts
and got: 2.0081V
I think that's fairly ok ;)
what do you get when you ask for 0.000
* alex_joni tries out the offset param ;)
the motenc board was a little worse IIRC
got it down to 0.0008V
but at least part of it's offset came from poor ground trace routing on the breakout board - the encoder supply current was flowing thru the analog ground trace
how many bits is the DAC anyway?
pretty ok for 13 bit I reckon
+/-10V / 8192 = 2.5mV per count
getting down to 0.8mV offset was just lucky
now ask for +8V and tweak the gain till that is right
then try -8V and see how well it does
going from 9.940 to -9.90 V without gain tweak
now with gain adjusted
8.0019V on 8.0
and -8.0036 on -8.0
not bad - about 1.5 LSBs
I could probably tweak gain even better
to fit both values
but I'm too tired for that..and it wouldn't help me anyways :D
any one have any servo sites i can take a look at?
cleaning up the driver and commiting soon
I wouldn't bother...
tweaking that is
heh..one more thing ;)
liable to drift more than that just from warming up and such, unless they have a very good reference
the voltages seem reversed :D
8V gets -8V
and -8V gets 8V
that's not what you said a minute ago
probably inverted a sign somewhere ;)
yup, and I see it
well... didn't look at the sign on the meter.. just value
10 + volts
ncounts = (short) ((10.0 - volts) / 20.0 * 0x1FFF);
probably should be volts - 10
on my copy it's actually 10+volts
hmm. that sounds right
I feel sorry I didn't get the software for my meter :(
I'd try writing 0x0000 and 0x1FFF to the dac and see what the actual polarity is
a meter with software?
told you it's a nice one
wow... very fanch meter
was about 90 euro's
[23:00:54] <alex_joni> http://www.teledata-update.de/sw3.htm
* Imperator_ good night guys
03alex_joni * 10emc2/src/hal/drivers/hal_stg.c: minor tweaks to get DAC's going too. ADC's and encoders still to be tested.
alex_joni: The DAC calcs causing probs ?
nah.. working now
watchdog was causing problems
because of an oxidated jumper
[23:32:46] <Yuga> http://clickautomation.com/products/index.php?func=list&cid=300
<--- is that the normal price for servo's??? cause friggin hell that is expencive
so those are cheap?
servo isn't for those with shallow pockets (which is why I don't have servo)
For "Brand new, direct from supplier.", the prices are about the norm.
gosh... was told to use servo's... but i cant afford that
for economical servo, you need to buy surplus
and where would i get that besides ebay?
and where would i get that besides ebay??
i am guessing that servo is more than big enough for a router?
Les had some SEM servos for ~$150
arg... and thats normal... he arent here when i need him :P
hi Yuga :)
wow - so that's what I'll have to spend if one of my servos breaks :(
hi there Jacky^
Jacky^... servo's suck!
what you found ?
have u seen the prices???
Price up high power stepper+amps
what kind of power do you need?
can't you get away with cheaper motors?
maybe a G340 and some simple DC's
well the cheaper ones arent much cheaper :)
alex_joni: what you mean as 'simple dc' ?
alex_joni... thought of that... but realy need recomendations
Yuga: you have to start with knowing how much torque and speed you need
that also mean, what we want cut..
* paul_c does steel @ 5-10m/min
jmkasunich.. thats the prob... i dont know what i need
largest lump on the mill to date... 'bout 6" thick
[23:44:39] <alex_joni> http://xianghe.en.alibaba.com/product/50065388/50296250/DC_Gear_Motors/DC_Gear_Motor.html
paul_c probably has a mill the size of my garage :)
Yuga: dont you need a big machine for wood ?
yep... 2mx3m cutting area
Yuga: what kind of a machine? router, mill? what are you trying to cut? steel, alum, plastic, wood?
paul_c... got a website on your cnc?
alum and wood
alum arent realy that important... would just be nice
paul: what depth/width of cut when doing 5m/min?
Yuga: and what is your budget?
jmkasunich... about 20ooo rand
On the lathe, probably no more than 0.5mm DOC
* jmkasunich doesn't know rand - what is that in euros or us dollars?
6.35 Rand / Dollar
but i can get alot of metal and that for free
is your friend)
Yuga: can you send some lot here ? :PP
so a something under $1000 per axis, for slides, screw/nut, motor, and drive
hehe... shipping will probibly cost more than shippin
Yuga: say that again?
using rack and pinion
for me the big problem is the slides, and ballscrew
hehe... shipping will probibly cost more than buying the metal there
ahh.. ok :)
yeah... 'shippin' is always more expensive than 'shipping'.
Yuga: I know..
R&P means fairly low speed (large distance per turn) I think, right?
well i think the fact that i am not using ballskrews makes it alot cheaper
so what diameter is the pinion?
wait 1 sec... got it all written down
R&P is ok for big distances
2x3m router in this case, if I understand correctly
20degree pressure angle
or rack pitch?
1 more sec
approximate is fine - just doing rough calculations here
1mm DP rack ?
gah.. DP - wrong term.
sounds reasonable, 18mm pitch diameter for 18 teeth, 20mm outside dia
jmkasunich... thats it
still... A pretty small pitch
assuming that is true, then 1 rev = 56mm
yeah - fine teeth, not terribly strong
Yuga: how wide is the gear and rack?
the gear i am looking at now is 8mm wide
not that wide
that seems rather weak for a machine that big
there is 2 rack and pinions
especially if you are thinking about cutting alum
one on each side of the machine
ok, let's carry on
SWPadnos is now known as SWP_Away
18mm pitch diameter = 56.54mm per rev = 177 RPM at 10m/min
well the width arent the issue... that is easy to change... just buy thicker
that is a very slow shaft speed - you need low speed and high torque, or you need to have belts or something so the motor can turn faster
10m/min... isnt that kinda fast?
at more moderate speeds, your motor will be going even slower
was planning on using a 3 to 1 ratio
using timing belts
ok, so that means 56.54 / 3 = 18.84mm per rev
better, but still a lot
(compared to a screw that might be 5-10mm per turn)
so how much u recon i gear it?
assume a 200 step/rev stepper with x10 microstepping for a moment
that means 2000 steps/rev