cradek: yes. ! I'm so pleased.
that's a really big accomplishment, congrats
Sorry, I don't know what happened to the rest of the HAL
[00:05:04] <andypugh> http://pastebin.ca/1583469
Say, I discovered just how easily we can put handles on these things. Ought to put a handle override button that inhibits servoamps for this purpose.
I tried to get a higher-resolution plot
andypugh: I'm puzzled by setp encoder.0.position-scale 1.000000
I thought you said you had 50? lines
Was it encoder.position-interpolated you wanted?
how about both of them, interp and not
setp encoder.0.position-scale 1.000000 should be 50
It has, that setting it over-written in custom.hal
ok I see that now
Thing is, you can't really usefully plot encoder.position as it ramps up continuously
you can right after index
since it resets
I hadn't bothered homing etc and running the g-code
I have the raw data if that is any use?
[00:12:18] <andypugh> http://pastebin.ca/1583481
time for me to drive home
I thought you were in the UK?
Shoiuldn't you have been home about 6 hours ago?
The data in a slightly more friendly format is at http://pastebin.ca/1583502
The interesting column is the last one, which shows the difference between each successive encoder position
And that does look worryingly irregular at each encoder rising edge
Sorry all, but I need to log off now. If cradek or anyone else comes back with any ideas, can you send them to firstname.lastname@example.org otherwise I will miss them.
Thanks for your help so far.
cradek: When you get back, I've found an interesting thing, would like your advice as to what we are doing wrong(?) Real simple test program, middle of table, x only. Moving from X15 to X16, then back to X15.
If I put the feedrate to max (80 IPM), then X doesn't make it all the way to X16, maybe X15.750 or so, can't really measure while it's changing, but obviously not going to the end.
If I reduce the feed quite a bit, then it will make it to the end where X = 16. Any suggestions?
40 IPM will make it, maybe 60 IPM also, I didn't do any calculations.
Just watching axis display readout.
does this problem manifest itself on a program where it matters?
did you try to turn off blending?
Are we having a conflict with our max accel limits? Do max accel limits take priority over final destination? With no warning? If so, can we get a warning?
hi, I would think turning around halfway would increase your acceleration
eric_unterhausen: We are just at the point of doing final adjustments. I wanted to make a little test program for tuning to use with halscope.
I don't know what we have blending set to, but I'll look
Valen1 is now known as Valen
so i guess "save halscope plot" idea never went anywhere
sounds like a performance hog?
Bridgeport_II: are you saying the DRO doesn't go to the expected number, or the table travel doesn't?
ok after a few more re-reads I think I understand what you are seeing. the program reverses and you don't see it reach the endpoint before reversing. if it is more than an ferror's difference, you are in fact seeing motion blending - the "corner" gets rounded off because of limited acceleration. You can get less of a rounding several ways including increasing acceleration (I bet you have it VERY low if you're seeing this much deviation)
hm, wonder if that was cut off
You can get less of a rounding several ways including increasing acceleration (I bet you have it VERY low if you're seeing this much deviation)
eric_unterhausen (& cradek): It looks like we had blending set to G64. This happened by accident because we loaded the example "spiral.ngc" which has a G64, which we were pretyy much clueless about, and so were astonished when strange things happened.
running in G61 mode will make it reach the exact endpoint
running with G64 P[tolerance] will make it reach within tolerance of the endpoint
you said maxvel is 80 = 1.33; what is the acceleration?
Yes, we are playing around with spiral and G61, G61.1, G64, and G64 Pwhatever. BTW, the docs are not at all clear on the difference between G61 exact path mode and G61.1 exact stop mode
well, I think there is actually no difference and those docs are "historical"
to get very close path following without extraneous stops, program G64 P[small]
Max accel on X is 2.0. No difference between G61 and G61.1?
because G61 will come to a full stop even if the adjacent moves are along the same line
right, I think there is no difference
What was the historically intended difference?
hmm, that's most of a second to get up to speed. does it stall if you try to accelerate faster?
I think the idea was that "exact path" would not stop if the lines were colinear. But no version of EMC ever actually had that functionality to my knowledge.
I would hope for an acceleration of more like 15 from a machine that goes 1.33 (0.1 seconds to full speed)
We were having some stalling issues because John wanted to get as fast a rapid as possible, so we were pushing it pretty far. But since we went back down to 80IPM (what the original 1973 BP2 docs say) no problems. John is not amused. We may be able to get it to 90IPM before we're done.
you will get much better contouring/path performance if you can get that accel up
So we can now come back and increase the accel
good, I bet you'll be happier with that side of the compromise
have you tried various microstep settings? maybe including half steps?
ipm doesn't matter much if it takes 3 seconds to get to that speed.
yeah you'll often not even get up to it
I'm thinking about things like peck drilling - you need the accel to get the heck out of the hole
cradek: first test of max_accel from 2.0 to 10.0 brought back stalling. And X is the lightest (least massive) axis. Although it is currently carrying the 4th axis loose for testing. Dave says 8.0 seems OK. He'll try 9.0 next.
yay, that'll be much better than 2
do you think the limitation here is just the motors? if I remember right he is using the original motors?
John has got X at 9.5 now, no stalling. We'll have to run a loop move to look for missing steps. Yes, original motors, 42 size, 900 in-oz.
who is Bridgeport_II? John is the one with the 2 bridgeports and jonE pico systems boards.. right? was at witchita..
skunkworks_: Kim K and John who were at wichita
not Jon E
ah - kim K - Hi!
He's starting on Y. He's going back and put X at 9.0. skunkworks_ yes, that's the right John. I'm KimK, i came up here to help out.
John has been working on this retrofit for years :-)
where are you guys?
I thought john was Bridgeport_II but Bridgeport_II was talking about john... Sorry - trying to get rid of some champagne for recycling. :)
Speaking of which, what can you tell me about tuning (PID) for steppers? Same as for analog servos?
We are in Minneapolis. I drove up Friday night.
I have not done it, but I understand the pico board takes velocity command. I think this means you can tune with mostly P and FF1.
if they have a gain adjustment you might set FF1 to 1 and then turn the gain until the steady state following error (during cruise) is nulled
er, 'they' = pico board
(I'm not sure how it tunes)
OK, tuning that way sounds good. BTW, we ran across a parameter "PID_MAX_VEL=30", not in docs, what is that?
no special meaning to emc - so check for it being used to set something in your hal file
setp pid.0.maxoutput [AXIS_0]PID_MAX_VEL
John says he can only get Z up to accel=2, but that's a big chunk of iron, w/ air counterbalance. In fact it started stalling only on the way down after the air compressor came on, so we are checking the air assist setting, maybe too high?
looks like the default univstep config uses the same SCALE for encoder in and velocity out. so maybe you want to just tune FF1 to be whatever nulls the cruise error
are you getting lube everywhere? those darn little metering valves are always plugged
OK. What about Pgain? I've had that all over from 100 to 5000 (found it at 1500, thought that was high), with no observable change.
in my servo system I'm running P at 2.4 or so atm
I bet it won't matter much because all the heavy work is done by the FF1
dunno if thats Pgain
the values are highly dependent on the configuration of the scaling - it's apples and oranges unless you carefully match your ... apples
* Valen DCC's a beer to cradek, he's funny when he's wasted ;->
there's some evidence that the "velocity" command for ppmc is in Hz, not velocity at all
oh wait it was skunkworks that was toasted :-<
Valen: haha thanks, I've got some
yeah I'm always this way
I really need to play with the tuning on ours again
doesn't get much worse when I'm liquored up :-)
now that we have "enhanced" the driver boards it looks like we can wind the gains up somewhat
[03:03:13] <Valen> http://www.youtube.com/watch?v=4pAxqw5yeZM
Pgain, P, not sure, I forgot. John says there's lube on the ways everywhere, but we do have some seeping hoses.
Valen: looks nice!
need to square up the head, its all sqewif atm
doing some nice cuts though from time to time
.oO(does machining make sense when sober?)
Z is being finicky. It's the biggest one with the same motor size for all. And the only one not horizontal. We're working on it.
I bet the air setting is critical considering how little headroom you have with those motors
toastydeath: air counterbalance. About a 6" cylinder. About 2000 lbs force, give or take a few hundred lbs. cradek: Yes, and I wish we had real encoders so I could look at the servo lag and know what to set the air to, but we can fake it by going up and down and varying the air and looking for stalls.
can you pull the belt off the motor and balance it by eye? is it really free?
maybe you could see what psi moves it up, what moves it down, then set halfway between
hmm it'll be hugely different with and without the rotary table won't it
if it's a 12" rotab absolutely, if it's like a 6, not realy
John says yes it is free, but no, it's a job to remove the belt. We're sort of doing that now by trial and error ^N^N^N statistical scientific analysis
Yes, the book says we are allowed a 1000 lb workpiece, and the 4th axis weighs maybe 250-350 lbs(?) So, yes, I agree, a lot of tinkering with the air setting, probably. But that's OK.
So when people talk about stepper tuning, is that what they are referring to, max velocity, max accel, stuff like that? Not the usual P, I, and D stuff? If so, makes sense to me, but I got the idea that EMC2 really does some kind of stepgen pulse stream adjustment and so really used some kind of PID loop. True? False? Both?
Or maybe it's not so much EMC2's internal stepgen as external pulse gen add-ons by vendors?
the internal stepgen and the on the mesa don't use pid as emc knows it. you are right that people using these and talking about stepper tuning are really talking about vel/accel and maybe the step timing parameters for their drive
I think they work more in the position realm [I am not an expert], unlike the pico board which takes velocity command and looks completely like a servo system, complete with pid, to emc
the pico products were developed in the emc1 days and everything was a servo to emc1. you got pid whether you wanted it or not - there was no position output.
But you say the pico PID (now) only uses FF1?
no - pico takes velocity command - FF1 is the pid parameter that is proportional to the commanded velocity - so it is the one that dominates
the pico board itself doesn't know or care how you come up with a velocity command for it
I better get going - goodnight
John is pretty happy with the performance of the machine as we have it now, he says it's as good as it used to be for sure, maybe better. But maybe we should run it by Jon E before pronouncing it done, just to be sure we have not overlooked any performance-enhancing adjustments? (We still have to set FF1, but that won't do too much I expect.)
OK, goodnight, thanks for your help, and thanks for everyone's help.
We'll be here awhile yet.
for PID to work there has to be an error, if your using steppers, there cant be an error
PID works on the difference between commanded and actual positions
if you have that with steppers your kinda sunk
at least as far as i can work out, because there's no feedback for it to work off
Yes, I agree with you. But I am told that the pico board takes the generated pulses in as encoder counts and runs a PID loop. True, still no outside feedback, but has some PID tuning, I am told? I am not very familiar with the pico boards, so I can't elaborate.
Valen1 is now known as Valen
Valen1 is now known as Valen
MarkusBec_away is now known as MarkusBec
Valen: you have errors even for steppers
because you have the motion controller and the step generator
ideally they do the same thing, but the step generator will always be a step behind
emc2's stepgen uses an internal pretuned PID to do it's thing, so there's nothing for the user to do
true but a PID loop seems like a bad solution to that problem
but if you're using a pico board where you need to supply a velocity command, then surely there's an PID
there there would be
would that not be in the controller itself rather than EMC?
the controller is separate
it might have it's own thing
emc2 sees this: there's a device I feed velocities into, and it outputs position
like a true servo
some you feed torque into
that seems "truer" as far as servo goes
otherwise you need 2 PID loops
which seems messy
it seems, but it doesn't need to be ;)
traditional (old) systems most use velocity
+/-10V for vel
but I guess it depends on the amps
Valen1 is now known as Valen
micges_work1 is now known as micges
Valen1 is now known as Valen
JbLb_ is now known as JbLb
Valen1 is now known as Valen
micges_work1 is now known as micges
[11:04:48] <jthornton> http://www.linuxcnc.org/component/option,com_kunena/Itemid,20/func,view/catid,10/id,880/lang,en/
he is not stating if HIGH acceleration is commanded or what his amps are set too
I don't know enough about servos to open my mouth :)
I know enough about inertia and loops to know there is a miss match if set slow
halscope to find out
probably backlash would be my guess
with high accel the motor powers through it quickly
with low accel it'll sit there flapping for a while
yeah that would do it too
He did say linear encoders though, and didn't mention backlash
seems to me and I don't know that what Valen said high acc the motor follows the commanded position better so less ferror
ferror shouldn't be affected by accel
your right because at low accel the commanded position is different...
accel should actually make things worse because it'll have less tome to correct things
with feedback is ferror just a straight forward math commanded position - actual position = ferror?
I believe so
jthornton: from source: joint->ferror = joint->pos_cmd - joint->pos_fb;
Valen1 is now known as Valen
JbLb_ is now known as JbLb
Valen1 is now known as Valen
wires falling off--
duct tape to the rescue
archivist: Why weren't they soldered up originally? huh? huh? huh?
So, like I said.. duct tape to the rescue!!! LOL
Valen1 is now known as Valen
archivist: is this on a connector?
back of a D type, was just boxing the lathe up to stop pull in the wires
and of course that was the last straw
Heh, Lot less headaches if you spend the money and buy a crimpter.
we crimp and solder, but then we have a dodgy crimper
A PRE-SCHOOL TEST FOR YOU
Which way is the bus below traveling?
To the left or to the right?
[15:08:02] <jymm> http://imagebin.ca/view/UPtoRfv.html
Pre-schoolers all over the United States
were shown this picture asked the same question.
should have had a 50% direction response
I bet they say right, drivers say left (in US)
Well, what is your answer?
90% of the pre-schooler's gave this answer.
'The bus is traveling to the left.'
When asked, 'Why do you think the bus is traveling to the left?'
'Because you can't see the door to get on the bus.'
its not a bus?
You are dumber than a pre-schooler!
jymm: where is this study?
US of A
no I mean where can I read it?
In my email
ok I see
But I do think it's interesting that they used logical reasoning to deduce the direction due t the abscenes of a door being shown.
[15:17:11] <cradek> http://languagelog.ldc.upenn.edu/nll/?p=1767
I'd wager that if such a study ever existed one kid might have given that answer
cradek: it should be FTW :D
Allie said, September 25, 2009 @ 10:04 am Obviously they should have used "Federation of Tourism in Wisconsin".
[15:22:04] <archivist> http://www.warwickfoos.co.uk/images/h20090326/59.jpg
Well, I can't see anyone that knows/uses "wtf" would ever visit Wisconsin
heh - that is funny :)
Is it possible to put Gcode command inside of an Mcode ??
I thought I saw an example of that and printed it out but now I can't find the printout or the reference!
OK..... so is that a no ? ;-)
OK, I will take that as a no..
One more question. I saw someplace that M100 and M101 was at on time used to test an input?? Is that still present? Or was that removed?
argh, andypugh never made the plot I asked (several times) for
some peoples kids!
looking at the code again, it does look like it's probably very sensitive to bad feedback
whats the relation of max_acceleration and stepgen_maxaccel
it expects that (revs - previous-revs)/cycle-time (consecutive servo cycles) is a good estimate of current velocity
which it IS if your feedback is good
mark space/quality of encoder
but these systems where you get no change for a dozen servo cycles and then a big change of 0.05 revs will really screw with it
yes maybe even that too
his printed on paper could be horrible in that respect
I wrote it all testing on systems with good spindle feedback (500-2000 counts/rev)
er maybe the one is 4096/rev - something like that
ah that explains a thing or two
yes his may be so bad that even the interpolated output is wobbly. that is why I want to see a decent plot of it!
*for each servo cycle*
yours is low resolution but high quality, so the interpolated counter can fix it up pretty well
one has to have hardware counting for that encoder qty not crappy parports
actually the 1024 line encoder went to a crappy parport - I used one channel only, and counter mode, and could get up around 1000 rpm iirc
hmm I do have a 600 lying around
er one channel plus index of course
but was going to use it on the mill for hobbing
spose someone should poke him to get on irc at the machine and follow halscope instructions
yeah my calcs are right - assuming 20usec base period, 1000 rpm, counter mode, you could use an encoder that's up to 1500 lines, around 1000 is better for safety
I think folks who say you need very few lines haven't actually done the math
hmm, I guess with full quadrature (required for tapping) you can only have half that
cradek: With the limitation being?
but I bet still 400-500 lines is where you want to be
jymm: the speed at which the parallel port signals can be read
cradek: And mesa card (or equiv) being the alternative?
yeah with real encoder counters there is no practical limit
cradek: Think a pico board would do it?
which is why I was so oblivious to this problem :-)
sure the pico stuff works approximately the same way as the mesa
heh, well I wouldn't exactly say that, but I understand.
"<archivist> whats the relation of max_acceleration and stepgen_maxaccel"
I was wondering that also apropos gtoms following error problem
I would expect a low stepgen_maxaccell would cause following errors
well it's true - they both count the encoder signals in hardware, and then every servo cycle they tell emc the latest count
I asked the question because stepconf only asks for one number
the exact means of communication is different but they communicate approximately the same data and same amount of data
There is a program called flowsnake.ngc and flowsnake.py in the example programs.
If I follow the Integrators manual properly - the Python file is a filter for the NGC file. Are there any docs on how the filter file interacts with the Gcode file? I found the [FILTER] section on page 30 in the Integrators Manual. Or should I just look at the "Wizard" code and ask questions. I asked a question before about putting Gcode into Mcodes - I think this is what I was...
...really looking for. Thanks!
Dave911: filters are passed a program name on the command line and they send the (possibly) filtered contents to stdout, which is read by AXIS
OK, so that is something different than the flowsnake.ngc and flowsnake.py programs are doing? I thought they were interacting but perhaps not.
sorry I don't understand what you are asking
the program "python" is sometimes used as an AXIS filter. It takes the name of a file on the command line (somefile.py) and generates gcode on stdout.
this is how you can run flowsnake.py, for instance
other filters might take an image file and generate gcode
or, take as input a foreign dialect of gcode and generate emc gcode (I use one like this sometimes)
So flowsnake.pv is generating the flowsnake.ngc code ?
archivist: stepconf automagically takes the max_accelleration + percent = stepgen_maxaccel. (or the other way around)
it is likely the flowsnake.ngc code you have was generated by just redirecting the output of flowsnake.py to a file
I was hand editing the result and needing the magic, I guessed
I usually add a good 20 to 30%
OK. I think I need to try some things.. ;-) Thanks!
cradek: no, I think you're mistaken about that. flowsnake.ngc demonstrates recursion directly in emc gcode using O-call.
oh sorry, I didn't even look at it
but now I remember ken wrote it, so that makes sense
MarkusBec is now known as nenunad
>>flowsnake.ngc demonstrates recursion directly in emc gcode using O-call.
So recursion - is basically code calling itself... So is the python code creating the code to do so or is it actually interacting with the G code?
I haven't looked the code again yet
OK, forget the last question.. I looked at it again. The python program is clearly generating the Gcode file but at a different level of Gcode than I am used to seeing! Pretty interesting. Thanks.
[#emc] "This is the #emc channel - talk related to the Enhanced Machine Controller and general machining. Website: http://www.linuxcnc.org/,
wiki at http://wiki.linuxcnc.org/"
Hello, I decided to write my own GUI for EMC2 .. where should i start.. i mean to say any documentation regarding CORE LIBRARY
hello alex_joni,cradek u there??
to understand how the GUIs work, you should study the existing ones
I recommend starting with xemc or halui because they are the simplest
yes thats what i am doing and even i Learned Python for understanding AXIS..
one quick question.. i am new to linux . but experienced windows programmer.. which path should i take, GTK+ QT or anything else
i am good at C, C++ , JAVA and C#
whatever you're more comfortable with ;)
there is no answer to that question. there are many toolkits because people have different preferences.
start with a windows GUI then ;)
we have plenty of linux GUIs
all that are new for me thats why i need guidence.. which is more powerful and which have more options like IDE and all
if you use an IDE like Eclipse, I think there are modules for most any language
but that has nothing to do with what gui toolkit to use
or even what language
you will need to do some studying to decide those things
One goal you might keep in mind when choosing a GUI toolkit is to add as few new requirements as possible. Using Qt or C# would add many new requirements, since emc doesn't use them now. Using Python, Tcl or C/C++ with Tk or Gtk would add few or none.
that is very good advice
if it is something you eventually hope to contribute to the emc2 project, Gtk is the toolkit to use
yaa thats true.. C# was just example.. i am more comfortable with C++
if it is something you will only use for yourself, your options are open
If the lack of documentation bothers you, then you can help others by contributing documentation as you learn from reading the source code.
i will definately contribute..
@jepler i am good at coding not documentation
you wont belive but my Master Thesis documentation was done by my friends
that is a very common affliction
... that is not
anyways thanks guys.. so GTK will be more apropriate right??
and yes one thing i forget to mention i am planning to port all this things on Puppy Linux..
like that coolcnc guys
Himanshu12: sounds like a good idea to go with gtk then
puppy is all about small systems
I don't think I've seen a puppy with Qt on it
umm unless I'm mistaken, the coolcnc puppy-based distribution is very old and not being updated
yes thats why i choose it.. coz the machine running EMC dont need hi fi system like UBUNTU
jepler: they tried and failed
yaa i know but i recompiled puppy kernel and ported 2.3.3 on it
so if Himanshu12 wants to do a new one, then it's great
@alex_joni do you know the reason for failure.. so i can take care..
On IRC, write "<name>: ..." to address someone. "@name" is not the style.
ok jepler .. i am new to IRC too.. i use yahoo for PMs
Himanshu12: they patched a kernel which would crap out on some machines
then there was a compiler problem
causing random segfaults
then there were some more issues, and after 2-3 weeks the guy gave up (can't say I blame him)
oh i see.. i patched 184.108.40.206 .. now still i get few error like emc.nml not found
but i think it will be solved its due to the wrong library i used few library from Debian Lenny that too pre compiled
[20:20:58] <ZeroFlex1> http://www.youtube.com/watch?v=AV2MehkFcTw&feature=related
Yippie! I have permanent electricity and internet in the shop...
welcome to the 1900's
200 amps of power :)
hmm here the single phase goes only up to 16amps
now I can roll up the extension cord to the house
and 3 phase up to 30amps
what, 30 amps?
for the WHOLE building?
thats for my workshop
just 16 for whole house
that is insane, our house is a 200 amp main
or rather my parent's house, seeing as how I'm in an apt
16 is enough for some lights and a small toaster at least
230volts and max 16amps stots
2 groups of 16 amps for the oven and cooking plate
oh, that's just each breaker, whew
i thought you were talking about the maximum draw for the house was 16 amps
less then 16 amps
zeroflex1: you just said there were at least two 16 amp breakers, so that means you cannot have 16 amps for the whole house draw.
every house here in the netherlands
I dont think 16 covers it
has i think max 30 amps breaker in the street connection
25amps in the breaker closset
far too low
and per connection 16 amps max
shower a bit lower 10 amp or so just for lights
? why not
does it already over far before i was born :)
electric cookers , immersion hearter,
i can do up to 3600 watts on a 16 amp
20 a on its own, pluas any other load
they are reducing britich houses from 80A to 60A
no way 25 could handle kettle shower immersion heater and lights
dang rolling up 300' of 6-4 SO cord is hard work
but over here we have a lightnet of 230 volts
240v, 200 amp
* alex_joni did some welds today with 120kAmps
nice... was it loud
BJT_Shop: nope, the cleanest process ever ;)
no sparks, no sounds (from the weld at least)..
nope not 240 200 amp
bit of a hummmmmmmmmmmmmm from the welder
happen real fast?
only a couple volts though
zeroflex1: i am talking about MY place
nope here industry power for normal gear like a mill or a lathe is 3 phase 230v 30amps
one of our lathes can draw 30 amps easy
a normal car mechanic shop does not have more
yha my welding inverter tig is 200amp ac/dc
a 200 amp tig inverter is only 10-15kW
works on a 230v 16amp line
nope works fine even with a wire on a roll (dont know how to call it in englisch) with 30meters wire
zeroflex1: one of our machines will draw 125 amps
what herz ?
the breakers to banks of machines are all 300 amp breakers
anyone got time to help me lift up that extension cord... it weighs a ton
2-3 machines per breaker
big machines are all 400V over here..
63A breakers are most common
then there's 100, 150, etc
big machines here run on 4160V
BJT_Shop: not that big :P
what my cord?
I mean bigger than a desktop mill
you mean 460V?
ac 230v 50hz whole west europe ic on wiki
mozmck_work1: no, 4160 is for industrial applications
no 4160, 480 is very common
like rolling mills and arc furnaces
ZeroFlex1: except three phase
which is 400V
3 phase is just 3 times 230 line over here
that's what I'm talking about
never heard of 4160V other than power lines. wow!
r s t
mozmck_work1: power lines are much higher than 4160
ZeroFlex1: but it's measured phase-to-phase , and it's 400V
r 230 s 230 t 230
power lines are a lot higher than that
the 4k range is referred to as industrial service
and underground to pad mounted transformers
but ten i have a big breaker of 30amps between it (30amps times 3)
Seriously big motors tend to be very high voltage as you can have much thinner windings for the same current.
some older underground is 3k
rolling mills tend to run on those high voltages, it's hard to get a pair of 3000 hp electric motors running on 400 volts
at 4160, that's a ~600 amp load - not unmanagable
the servos we have on the latest robots are all 600V AC
or a 5000 amp load at 400 volts, and now your stuff resembles substation switchgear
I have made the connectors for a 3000 amp switch, I don't want to know how big a 5000 amp switch is
heh.. reminds me of soem weldign machine we installed (also resistence spot welding)
it was a 500kVA or so machine
had wires about 150 mm^2 per phase
that's about 15 times 0000 AWG
to convert it to funny US metrics
well that's about the same size as the barrels for the 3000 amp switchgear
might have been more than that, though
the tubes are 5" in diameter, which is about 150 mm
toastydeath: this was 150 mm*mm
not 150 diameter
probably 1.5" or so
small stuff :D
well, you don't really see much 1.5" wire lying around
we used 3 cables / phase
12,000 hp rolling mill engine, and it reverses in less that one rev: http://www.youtube.com/watch?v=89Sto6U2-ME
I worked at a smelter when i was in high school. There we switched 35 000 Amps manually. Crazy shit.
that is an insane reversal
And I _think_ my great grandad used to run it.
what is that whistling
Wow awsome engine.
is that the steam in the pipes?
Or leaking past the pistons.
love the last comment
i work with catalysator systems behind big engines, 4 up to 8 megawatt
those engines are a lot smaller than that one in the video
probably a bit newer too ;)
12k hp = ~9 mw
Anyway, now I have bribed you with steam engine videos, how about you all fix my threading problems?
yha a lot newer
a factory in the neighbour hood here, build even bigger engines (til 30 years ago i believe) for big ships
stork is the company name
As big as this? http://people.bath.ac.uk/ccsshb/12cyl/
andypugh, you need to do a halscope at the zoom that cradek asked for
How big must the milling machine be to make that crankcase?
I pastebinned the raw data, I think that is probably more useful?
not that big i think, you just need to mill a shovel to dig the hole for the mold
andypugh: did you try a finer thread?
Yes, it is a little better, but I rather need the M39x4mm one :-)
And I have scrapped a lump of 4" stainless :-(
andypugh, he needs to see the it for a servo cycle
OK, so we want a halscope image at what zoom and what channels> I can do it now
archivist: I saw it the other night, and I'm a bit lost what else to investigate
there was an post on the forum though which suggested trying imperial instead of metric for the units
although the spindle units aren't really in mm vs. inches
That was from me. I created a whole imperial setup for the machine, and it was marginally better, I think. But only marginally.
mine it set to imperial
I was wondering if the controller that handles G76 was picking a gain based on the machine accell, and not converting. It was worth a shot.
no, it doesn't know much about machine limits
Is it a software PID in there?
nope, no PID there
just tracking the position from the encoder
but if the axis can't keep up, it might feel link a pid
It oscillates like a badly tuned PID.
andypugh: maybe try to replicate your config on a sim machine, and increase z-accel and vel
sim/lathe.ini should be ok for that
see if more accel/vel helps
Worth a try. Is there a way to introduce encoder dither?
how I had halscope set http://pastebin.ca/1585007
and we zoomed to 10ms
you can run off the same encoder
have the rest simulated, only the spindle live
Ah, yes, good plan. But my quiche is burning.
BJT_Shop: err.. ?
1am coming up here ;)
the cord that ran between my house and the shop 300' of 6-4 SO
andypugh, cradek and I were discussing earlier best change is to make a much higher resolution encoder disk
it's too heavy for me to pick up :(
But then I think I would run out of p-port speed?
* alex_joni suggests growing a muscle then :P
I'll have to tie it to the back hoe I think
andypugh: what's the current base thread?
Base thread is 50uS
pick one end up, and respool it ;)
andypugh: oh, why so slow?
I think I can run 30
Not sure, it is half what stepconf chose....
stepconf choses doublestepping by default
and it choses the biggest number that still satisfies your need to output pulses
andypugh, quoting yeah my calcs are right - assuming 20usec base period, 1000 rpm, counter mode, you could use an encoder that's up to 1500 lines, around 1000 is better for safety
I need 1.5uS per step, and have 2.5uS latency
I think it doesn't think about counting encoders
lets see.. 20 usec means 50000 changes / second
I am concerned that the dither between 2 and 3 samples per encoder line will cause huge problems?
you can count about 20k transitions (to be safe) / second
at 1000 rpm = 16.67 rps -> 1200 counts
1200 counts means a 600 slot encoder
I think I have been fooled by halscope running in the servo thread and making it look like I drop counts at 20lines and 1000rpm
andypugh: can you try to adapt one of the stepper configs, and try that?
as he is running a lot slower base thread a 240 slot giving 480 changes
e.g. a config without doublestep
archivist: no reason against running a faster base thread
BJT_Shop: aren't there 2 cables in that pic?
yea you caught me one is 25' long
you can still pick up one end, and respool it somewhere else :D
but backhoe is easier
I tied it up and rolled it like a flat tire
I see a drawback with the 1000 line encoder idea, I don't have a 0.3mm milling cutter...
make two disks then overlap them :D
I would not go all the way too 100, have some headroom for higher spindle speeds
andypugh: surely there is something else going on too
Slitting saw is a plan.
there are people using 1 pulse/rev and it's working ok
Maybe I should try that too?
I bet they are not making coarse threads
your 4mm pitch needs some control
...Try #bdsm, I'm sure you'll be able to find a dominatrix for some control
I think that the problem is lack of accelleration, but it isn't that slow.
What max accell do tohers use?
(Not that I know what the units are)
I have 300 mm/s2 which is 12"/s2 Enough to get from 0 to max speed (1500mm/min) in 1mm
max acc is 12" and max veloc is .9 stepgen_max acc is 18
0.9 and 18?
.9 is the max velocity
Ah, yes, sorry.
What _is_ stepgen max accell?
300 should be enough
I tried it at 1000 in G61.1 mode and it almost worked
someone else said 512 should be enough ;)
or was it 640
IfI need more, where can I find it?
* alex_joni was joking..
Yes, but if that is the problem...
However, I don't think it is. There must be _some_ sort of controller in there that converts position error into steps. And I think the gain doesn't suit my system.
more voltage to your steppers
UG UG UG MORE POWER!!!! UG UG UG
I could rewire them parallel and get a bigger controller.
If you rewire them, you'll need a flux capcitor too
SWPLinux: Maybe double that.
inflation strikes again
SWPLinux: So, what planet are you on now?
one that's "mostly harmless"
Seriously, they are rated 2.5A series and 4.5A parallel, so I am running series to suit the 2.5A controllers I have.
waiting for my flight home
SWPLinux: Where are you?
waiting for my flight home
have a cold one on me SWPLinux
andypugh: I think you asked for the code at some point.. right?
andypugh: what voltage rating and PSU?
SWPLinux: Dang dude, you must have like a bazillion miles saved up
Motion Control Products
jymm: he spends them on flight attendents
alex_joni: Why, they get their own damn miles!
I may go to Omaha next week too
errr.. business services
nah, just "Service"
[22:18:02] <andypugh> http://www.motioncontrolproducts.com/drives/msd325-microstepping-drive.php?cat=2
you guys have all the fun
alex_joni: You do know that many of the "flight attendants" are 1) old, 2) male
now why would the word "fun" come up so soon after the word "Omaha"?
andypugh: 32VDC max
that's a bit low
jymm: not over here
Looking at the motor spec, the difference in torque-at-speed is not huge for a doubling of current
they are mostly 1) young 2) female 3)good looking
alex_joni: 4) vampires
andypugh: nope, but doubling the voltage should definately make a big difference
Hmm, I am running at 24V anyeay
except when you fly air france
SWPLinux: You'll be home for a while, or just a layover?
at 30v and parallel they will be faster
err, good question
I am not sure I need more speed though, as much as more accelleration.
And I am not totally convinced I need that.
you get better acceleration as well with a higher voltage
its rate of change of current
andypugh: the good thing is that you can easily test that
hm.. can you try a simple thing?
andypugh: are you limited to 24V power on your drive?
or is the machine down?
Easily for the cost of a bigger power supply and drive, you mean?
no.. I mean using sim
Machine is off at the moment, but can go on
your not going to get much acceleration with 24v on a stepper
I get 12" !
try a faster servo-thread and see how it behaves
now it's at 1msec, try .5 msec
also.. you can run the machine with the motor drives switched off
only spindle running
It was at 0.5mS when I came in for my dinner.
then you'll get a 'simulated' machine
FASTER BABY FASTER!!!!
then you can play with accel and speeds and see what happens
(see = halscope for the same threading move)
Yes, I will be trying that tomorrow.
andypugh: any difference at .5 than 1msec?
alex_joni: You know, that wouldn't be a bad thing to add-on... a PID tuner
jymm: there's a component called at_pid
alex_joni: Does it loop thru settings to find the optimum configuration?
(I will have to turn off the amps in the ini file, the same switch controls the computer monitor, the pwm for the spindle motor, etc
jymm: it uses math to calculate usable settings
based on feedback from a specific type of output it does
alex_joni: You seemed about to point me at the source code? I have it downloaded, just need to know which file to look in.
SWPLinux: Ah, ok.
andypugh: I was about to past a link to the line :D
SWPLinux: Oh, what's math?
let me look it up again
[22:30:22] <alex_joni> http://git.linuxcnc.org/gitweb?p=emc2.git;a=blob;f=src/emc/kinematics/tp.c;h=033198b6b0401d58edfff961aa01cb15ca97c03d;hb=HEAD#l896
Experience at work is that autotuning is rarely optimal, and we run some very exotic controllers (mapped gains, for example).
SWPLinux: I got 1.5
1+1 = 10?
(A diesel engine ECU is a hundred or so controllers all arguing with each other)
[22:34:06] <jymm> http://notalwaysright.com/a-match-made-in-size-7/1490
nenunad is now known as MarkusBec
good night all
Thanks for the souce link, it is going to take me a while to figure it out
But it does look simplistic, if I may say so.
SWPLinux: how is it going?
uh - ok
here's where I was on Friday: http://www.worldamazingrecords.com/2009/09/most-cameras-used-in-creed-live-concert.html
MarkusBec is now known as MarkusBec_away
andypugh: did you read the logs and see my comments about your problem earlier today?
I suspect the cause is your poor quality encoder and you should try 1ppr if you can't try a real encoder
also I still want to see the plot I've asked for several times
SWPLinux: As a concert goer or setting up AV gear?
making the cameras work so they could break the world record
s/they/we/, but you know how that goes
Yeah, I was JSUT thinking that actually. Too bad it's not a film, as you would be in the credits.
I've worked on films, but never been credited
SWPLinux: You wouldn't? (I was being serious)
well shit, get up in imdb and add yourself to the uncredited list!
and they didn't invite me to the press conference for this record, or have me be involved when accepting the award
heh, yeah - I ought to
I've only worked on two mainstream films though, and helped out a friend on another
SWPLinux: You need to start building in back doors
like the one where nobody but me can operate the new digital system?
SWPLinux: Or where you're the only one that get the system back online ater you've hit the button on the remtoe control
oh, I don't need that level of complexity
I just based everything on Linux, and the bunch of mac-heads can't even set it up :)
SWPLinux: I think you need to start getting credited - put it in your contract or something.
eh - it's no biggie
what I've done is "work for hire", so I don't own it
it would be nice to get some credit, but that's often not how it works
(it's always the manager / executive that takes credit for what the engineers did)
SWPLinux: Actually it is. It builds up a reputation for one, free advertsing for another.
SWPLinux: No, no, I mean in the tiny print. Not above the line or anything, but at least something.
ok - flight is still on time
SWPLinux: how long?
I'll have to go board in about 20 minutes
SWPLinux: How long was the layover? 4 hours?
about 3 hours
not planned, my original flight this morning was cancelled
they just cancel flights like that?
so they re-routed me through Chicago, with a 3 hour layover
not too often
Did they say why it was canceled?
the flight from Washington DC to San Antonio was canceled this morning, so there was no plane to fly back there
no, I don't think so
bastards.... probably you the only one on the flight!
heh. no, I think there were at least 8 seats sold :)
You coulda have 4 sexy flight attendants all to yourself
3 of them female! LOL
I get home only 4 hours later, and they put me in first class on the way here (still waiting to see if that happens going to Burlington)
SWPLinux: Do they have a private lounge for you in ORA?
I'm a Red Carpet Club member, so I've been in the business center all this time
free snacks, coffee, and internet (and power)
well, free once you pay a few hundred dollars a year
What, no reclining chairs or beds?
sure, but I chose to use the business center instead
SWPLinux: Hi, this is KimK. I'm chatting from John's in Minneapolis. If you get to Omaha and I'm back let me know, maybe I can buy you lunch or something.
(no beds anywaqy)
I may be there for not very long - we might shoot the Creed concert in (that Iowa town next to Omaha) so we have more footage for the DVD
(fly in one day, work a 20-hour day the next day, fly out the next)
yes, that's the one
You mean BFE?
Sure! Like Mexican food? I'll introduce you to Alvarado's.
email me your contact info, I don't see your phone number in my phone
SWPLinux: BFE == Bum Fsck Egypt = Middle of nowhere
no, it's right next to Omaha ;)
Nothing special, but it's open 24 hours
SWPLinux: Like I said....
and Omaha is only an hour from Lincoln
Well, there you go!
SWPLinux: Anytime you fly ove a place and it looks like a checkerbaord = BFE
that's most of the country
I stand by what I said.