See my post on timing belt reduction robin?
you were saying someting about belt reduction?
yeah check the cad cam list
okay, thats in emc1
ill have to view via cvs
ok, seen that
so there is an optimum
I dont have the rest of his original desing criteria ...
speed, accel, gantry weight
60 kg moving mass
6m / min ?
well 4 ips
100mm / second = 6000mm min = 6m min
so thats for a big 'heavy plate' sort of machine I guess
yeah... I used a 1.5mX 2.5 cm jackshaft in the analysis
suprised he need 0.25g
and .25g is plenty right?
so about 5:1 seems most economical
and you are using?
theres a thing
twice the size of pinion
I figured 2.5 cm pinion
im using a 60mm pinion
and a 20:1 box
I find it MUCH easier to think in terms of 'movement per rev'
so that would be....about 8:1 in terms of my analysis
also...your moving mass is less I think
and you are not requiring .25g
it is .. I could use a lower ratio box fo rsure
.25g gives about 2.5 mm corner radius at cutting speed
yeah...if the ratio is too high the motor just fights it's own rotor inertia
* robin_sz nods
reflected inertia= motor inertia*reduction ratio^2
catches up with you
so ... I find this a LOT easier to think of movement per rev
I just pumped ratios, moments, masses, etc into motioneering
you are suggesting about 15mm per rev I think .. right?
hmm let's see
I think im on about 5mm per ... so basically I have way too much force, not enough zip
btw I used .1 oz in sec^2 for the reduction inertia
reasonable I think for a shaft and timing belt
I'll see if they will ship some lower ratio boxes on the next controllers then
well also...using steppers changes the economy equation a bit
do a run in that software...you have it right?
I do ... somewhere
and that ohter neat one ..
beamboy is it?
handy little thing
Had an uncle visit the shop...he is 19 and just took statics in school
he has to do it the old fashioned way
uncle at 19?
the uncle bit is another story...
he came by with his dad
his grandfather was my great grandfather
* robin_sz presses the 'too complicated' button
I'm not the only one known for robin' the cradle haha
interesting calcs though ...
When I was married to that moonshiner's daughter here for a little while one time some witnesses came calling
she answered and they asked if her dad was home
yeah the calcs are telling
but makes sense
the math is right and the error generated is correct -.00413715
I just took 75% of max motor revs and geared to do that at G0 speed (15m/min)
at some point direct drive becomes too $$$ from underutilising power
and that square term gets you at some point with gearing
so something somewhere in the middle has to be optimum
and about 15mm/rev seems to be it
for that size machine yes
for a 60kg gantry and 0.25g
* robin_sz nods
and if I didn't screw up anything
gezr: sorry...which error is that?
* robin_sz adds "try some lower ratios" to his to-do list
jepler are you still around?
les the motion error
it is in fact a negative value
the output from the error when plugged into the code results negative
but It didnt for the first how ever many loops existed
so a debugging string would be needed for me this time around
where at each pass at the motion, the hold up
sqDumpQueue(sq); is logged
would probably also need to see 2 more vars
make that 3
diagnostics("plInitInc = %e plFinalInc = %e\n", s->plInitInc,
as you saw I found only one line in sqplansegment() that could go negative
yeah it can go negative, why it doesnt every time on that program boggles my mind
if the exact same g code goes in then plInitInc, plFinalInc and tl should always be the same
yeah...but it's repeatable and there has to be a pattern
need to find multiple motions that cause it and see wht's in common
the patter would be found by dumping the same info that the error generates without an error, as in a debug line
the log would be or should be, everything the same, and then the error
every time the program loops m2, all the numbers should match
if they do not match, then something within that code could be to blame, if they do match, then something external is causing a data failure
you know, the bad original planner is the only thing that keeps emc from being a state of the art system
it's kind of like a ferrari witha rubber band motor...
it could be as much as something, I dont know how to add debugging code, to print to a file, or what not, nor do I have an active rt system available, but im betting its either in the kernel, or an external math section, something is causing a data malfunction in the stream to length <0
if I was a betting man ide say a float
but I'm no programmer heheh...just a poor aero engineer
all the numbers are in xxxxxxx format, then all of a sudden xxxxxxx^-e3 pop up and err
its using doubles for the numbers too
tl is a simple enough number, its all the formating to make it all look xxxxxx vomit
why use a decimal in one instance, then not in another
I dont know if it matters, but were talking about 2000 or more #s a sec, anything could go wrong, a single bite and it could go negative
I have a hunch it is just bad math somewhere
but who knows
yeah but if that was the case, then it would happen the moment the first run takes place
whoever finds it will be the big hero in my eyes...
in jepler's example, he was able to loop through it for quite some time
this problem actually costs me money
then it would eject the exact same error
after several loops?
he was running it over and over again
then all of a sudden it would stop, and spit that error out
* robin_sz hands out some global variables
ha we have plenty thanks
perhaps something needs to be clamped somewhere
isnt that a bit like ermm
the error, say, that one was to change that to length == 0 instead of is less then, just to see, and the stream output was printed like Im thinking, then one would expect to see the error in different times(during the m2 loop) while the remaining output looks the same
there are a lot of length==0 tests in there I saw
sounds liek a fundamental desing problem
if your testing for the same ting more than once ...
why would a negative length matter anyway?
a good algoritm is continuous
look at these numbers
the code seems to just be written for positive length values
when you have to start putting tests in for edge cases, its probably better to think about your methods
the math 2*plInitInc first
I guess +ve lenghts and angles makes sense in polar
in caretesian, -ve legths makes sense
yeah, just coding style. Negative lengths are fine otherwise
but the value as printed from the output is 4.136575e-03 for plInitInc, I realize the computer doesnt care how a number is formated, so humm,
He is doing roots for magnitudes and it has to be + probably
"formatted" ... can be an indication of casting between longs, floats, etc
isnt 2*4.1xxxxxxe-03 gonna be 8.xxxxxe-03 anway?
now negative MASS is a problem haha
are these in printf's
gezr: yeah I think
bedtime here ...
see you all later
les : I honestly think what is happening is a number from hell is popping up causing the negative condition to exist
and the 2* isnt helping the issue
heh...it is a mystery
when NIST can't find it
I dont think its there code
but they cannot put much time in
you might be right
its the only thing that can happen, you said you had the error occour on large programs
how can negative length be generated by a forward or positive motion, unless a mistake in the data stream occours
there are so many global functions and variables that it is very hard to follow everything
if you are right an if and a clamp to positive might be in order
test for garbage data
you know what
and default to some band-aid thing
what if this
not very eloquent
but might work
say something has the same varable, and happens to land in memory at the exact same time, and at that instance that file, reads that value, does the math and the negative number is spit out and lenth is less then zero
that would cause the error
but it should protect is memory for the data
seems unlikely but possible I guess
maybe jepler can add in a debugging dump for us
accually you know what
lets say that the numbers never changed
well I am just thrilled that people will look into this
btw...you name and location?
what was output in the error was the same as it was always, and the error finally hit, because the math should always enter a negative state
Phillip Smith in central Arkansas
im serious, the math should always be negative
oh ok ...Les in NE georgia
I sent you a copy of bdi live once :)
ive been taking a break since then
I need to find out what or where plInitInc is
I was pondering that
cause anytime that number has more digits in it then tl aka total length, its going to error
im going to have to check out emc1
cause I need the power of grep
Well, as a background I had the authors here for a while looking at it
They realised that the emc tp was badly flawed
but lost funding before it was fixed I guess
yeah I remember that, that was right uup before my mom passes
passed that is
so Fred came down from Maryland and Rogier Blom came over from Holland
Rogier wrote segmentqueue
Fred wrote emc
and you guys could not find what was causing the random stopage
We found other problems and fixed them...but only had a couple days
can only work on one bug at a time
this one is a biggen
before it would go into a loop and thrash the disk drive
well it's the difference between emc being state of the art or crippled
well, smoth motion means a lot to me
its just sexy to see something gliding
me too...I make my living from emc run machines
I know, so we get it fixed :)
lots of talk from time to time about jerking
in the email
well better go prepare a bit late dinner
maybe Ill find something
hey hold up
when you were running your program with segment were you also looping a program?
no...I was not
pi is wrong
well, close enough I guess
no its on the ball :)
im almost out of steam
things are starting to run together tonight now
gezr: do you have a bigger file that bombs?
I didn't see all the previous conversation
in jepler's program example total moves is 3, total distance for each is
I want something long where it bombs in the middle of the file
na, cant do that, im gack
do you want his exact gcode?
jepler's? I have it
okay, were you able to recreate the error?
oh you were?
but I'm not convinced that's the real problem
did you read my ramblings right before les went to dinner?
not really - I want to work on my own just in case I'm on a different path
about how I think that something is causing a data blurb?
what info did you need to hear then?
it might be bad for us to all work together
yeah, good point, problem I have is I cant code
I need a file where it errors in the middle of a long run
not at the end of a short run, like jepler's
you know, you could use one of the example ngc files in the examples dir i bet
ok, I will try them
I dont know if they are long enough
wish les was back, I'm sure he would have one
yeah I bet he does for sure
I havent looked at emc in over a year
so im starting off the hip here
I put things in motion some times, I dont think this is a wich hunt either, its a viable necessity to have working
well, I have a small machine and the old planner works superbly for me
but I understand that's not typical
I dont understand enough about code to know how things jump around, when looking at the source its easy for me to get lost
it takes lots of practice to be able to read code
reading is possibly harder than writing
I dont understand a lot of things about it, I can follow some but when an external from an include is called, and I dont know what is returned I get squirly
you may be able to use a hpgl converter to convert a jpg or something to gcode
ok, thanks, I'm going to go work some more
I didnt realize you were off working on this
cradek: that short run came from elm320.ngc at line 900 .. I'm not sure how long the "run" up to that point was, though. And it was repeatable, it gave me the short run just by removing most of the lines of the file.
so it always bombed at that line, and that line wasn't the end of the file?
it bombed two times at line 901, which was line 4 in the small file
is your md5sum fb95...?
(not sure if I have the right file)
or, to be exact, fb9549aa1bd376d33a0f7836d2873960 elm320.ngc
Dome_test is nearly 9k lines long
8k that is
I just walked in from dinner...the ten commandments production code I have is huge and exhibited the error
in the file jepler posted, the linear distances between the points is in order: .26978290, .059264, .0041366
well if we all have the same error, I fixed or worked around it
I get past line 900 in jeff's file now
the problem is when the segment queue is clearing out
in jeff's file it was the last segment before a pause
it's just a boundary condition problem
a pause as in dwell?
les: can I give you a change and you can test with high acceleration?
I have had it do it with no dwells
you think It was trying to read the g4 into the queue?
but that's a hint at least
gezr: no, it's just that the queue empties when there's a dwell
cradek : ah okay
Ican can sentd you the no dwell file that did it with Fred, Rogier, and I
hang on let me find it
(I may have fixed only one of many problems)
my case is "move with a chance to accelerate, then a short move, then a dwell" -> negative length
I just like the fact the ball is rolling :)
is that the one you attacked, Chris?
jepler: yes, it fixes that
jepler: I think a short move (relative to the last move), then dwell, will do it every time
possibly -- it at least doesn't require that the second move actually is in a different direction
I hope that didnt go into the channel
it can all be g0x moves
like this one: g0x0y0 / g0x-.1 / g0x-.101 / g4p0.01 / m2
chris timed out on irc file transfer to you
oops I didn't see the notification
try again please
this is a 3-d one that had the problem
No solution found in sqPlanSegment
Error 1 in sqRunCycle
different than the one I mentioned
jepler: different bug!
but fred used it and had the problem mid way
hm, I got this message only with a 'g1 f1000', not with 'g1 f10'.
prob not an error really
just not enough accel for a solution to exist
so that error may need to reed, Possiable excessive feed rate, No solution found in sqPlanSegment
les: could you send me a sim.ini that represents your machine?
surely there's a solution: go slower than the given f-rate.
ideally segmentqueue lowers feedrate until a solution exists
but apparently not always
see how little I know, :)
it may also error if plane mode is switched during maching, xy then xz and so on
les: I'm at line 2000...
les: any idea where it errored before?
I am wondering if redundant zero length moves can cause it
sounds about right
zero length moves should be dwells
well cam programs often do dumb things like repeated blocks with zero movement
* cradek points at les
I just see them as a waste of space, then again im using machines with limited space
I think I agree with chris and les -- programs might generate zero-length moves. emc should keep going at best speed
les: line 3700
les: it will take forever at this speed
can you send me a sim.ini file that represents your machine?
now you know why I charge so much for those signs
well the .ini is on my web site
www.lmwatts.com...go to links then engineering then cnc
cradek : its still running?
it's not the sim.ini but I am running at 15 accel and 3 ips max vel
oh I had the pid zeroed out there for saftey if folks wanted to use ir
but the sim parameters should work
grr, can't get your ini to work
remember...pid is zeroed out
I put it back
it was commented
minimilltaskintf.cc 1017: !ERROR! Error on axis 0.
minimilltaskintf.cc 1017: !ERROR! Error on axis 1.
emcTaskInit: adding user-defined function programs/M101
minimilltaskintf.cc 1017: !ERROR! Error on axis 2.
Minimum Hardware Limit on axis 0 exceeded.
Maximum Hardware Limit on axis 0 exceeded.
oh...change to bridgeportio
my old sim.ini used simio
[EMC] IO_BASE_ADDRESS not found in sim.ini; using default 278
bridgeportiomain.cc 190: !ERROR! can't initialize IO-- privilege problems?
well the errors crop up on all sorts of different configurations
ok, let's forget this
I'll go back to my old file
yeah...I think the problem is independent of that anyway
I hope that was all it needed :)
I think im gonna start leveling down and getting ready for bed
anyway my parallel port address was in a funny place
I hope it doesnt error on that file for you les, that would be sweet
gezer: have a beer and hit the sack...thanks for your help
I had my beer for the night earlier with the pizza :)
does the old planner have this cubic acceleration profile?
...my dinner was grilled on cherry scraps created by emc
that has to be full of flavor
cradek: old emc has cubic SUB interpolation at the servo rate which is not all that useful IMHO
you need a emc controled, smoker house, so your chips go onto a conveyor, and emc trigers the conveyor to feed the box, and rotate the meat
have emc creating several TONS of cherry scraps
all emc is making for me the past what is it now, 6 years? is lots of grey hair
yeah, at least that long
Im just a good thinker I guess
ha not grey hair for me....less hair
less hair, grey beard here
well the ones that havent fallen, jumped out turned grey
heck im still a tiny 31
doubt I'll have grey on top - it won't last that long
gezr: it'll hit you any day, trust me
les: sure am getting tired of watching your sign in axis here
slow ain't it?
in the spirals there's Z motion that kills the speed
those things often take half an hour or more
what's it doing there?
les do you use a single flute endmill or router bit?
which is why they are $50/ft^2
does it go down in the center of the spiral?
That is artcam making a 3-d carved celtic spiral border
wait a moment, do any of you own a surface grinder?
using a v-cutter
I have a surface grinder
a boyer schultz 612 deluxe
recently rescraped to 25 millionths
ill remember that, I may ask you to quote me a good price on grinding some acme like threading tools, ill be making some gear hobbs, and then gashing the hobbs, all materials ill provide. easy simple stuff
I dont have but some simple things here, and hand grinding acme bits drives me crazy, and using a dremel(dont laugh) to gash the teeth
well guys, I think it's possible I fixed this
cool! tell me more
well, it's still running your sign
it's about 4/5 done with the border now
tell you what chris....if you get segmentqueue fixed...you get a free solid walnut house sign!
can I have some servos and amps instead?
3 good servos went on ebay last night for under 100
haha I sold all the servos...100 of them
les: by "fixed" do you mean "working" or "working right"?
or how about "working better"?
right now atatched to my home built driver board are 3 cdrom head positioning steppers
kept 3...I am doing the auction thing for a second vertical mill
need em for that
les were you ever able to have a complete run of that file under the new planer?
The router is actually metal capable...but now it got it's first decent production job pumping out $4 every 2 minutes
so another mill for metal
gezr: no never got that file to run all the way
I have a little atlas horizontal
an atlas shaper and an atlas lathe
im the atlas junkie some how
I have had many atlas tools
my shaper is the jem
good to have
it's on the text finally
it looks like hell, but its the strok-o-matic 2000
dangerous oh my lord
sounds like a machine a woman could love
please tell me "stroke-o-matic 2000" isn't its real name
hahahah, no its not, I dont use 2000
I call it my shaper
I am in this growth pang stage where I am having to up the hp on everything
deepest cut or pass ive taken was like .050 in stainless
gezr: mill or lathe?
your cutter may help with hp, if you use a different geometry
and...every minute emc runs...it creates a minute of employee labor sanding what it made
the deepest cut ive ever taken on a machine is about a half inch depth on a lathe
heavest part is a ton
I am using custom designed $70 carbide cnc router bits
longest part is 15'9" or there abouts
largest dia ive turned at is well I had a 25inch piece in last week, but just had to lay it out
working area on my tools is about the size of a brick
well it FINALLY finished
this stuff is where I work
cradek: it ran?
runed, were in the south les
cradek : most execlent :)
let me try it on my real machine
anxious to see what you did tothe code
it's a one line change :-P
oh scale it...that is about a 12x18 in sign
I sometimes do 10 foot ones
tommrow I talk threading :)
have to shift it and do it in halfs
no Ill save that for later
I do think slaved axises would be way cool
for say, ridgid tapping and the likes
I gave away two cnc hardinges with fried bandit controls when I moved down from Chicago
too heavy though
im really liking the fagor controls im using where im at now
Got fed up with two nasty ryder truck rides over the mountains
I can really help some one work out conversational programing methods for emc
Ive never had something generate gcode for me
and conversational is neat
gezr: if segmentqueue is fixed emc will outperform fador and fanuc
I know it will
motion like fluid
I have a few dozen thousand worth of cam software here...but this is a business
have to live off of grits to buy it though
ill probably try tubocam out
millwrite2000 is very good...$800
Ive been a turbocad user for over 10 years, I can only make straight lines and the likes, I just seem to lack motovation
thats not a bad price
signs like the one cris just run start as a 2d autocad 2000 dxf
then converted to 3-d chip carving by artcam
it pulls the v-tool up to make sharp inside corners
how many rpm are you pulling?
but now the machine is on a 2 month job that has 6 tool changes
can you bring that up closer to 30k and not need nearly the hp?
but parts are 36 up so not too bad
I dont know jack about cutting wood
but im not affraid of 200ipm in steel
right now I need just plain old hp
how many hp do you have now?
looking at a 7 hp colombo motorized 24,000 rpm spindle
ill look tommrow in the shop, we may have some drives
now I have only 3 hp
they are pretty long, look powerfull, they came off of crains, so probaly low rpm but high hp
I need just the opposite
timming belt and away ya go
oh you probably want a cartdrige unit?
the colombos are 2 pole 3 phase induction motors run by a 400 hz vfd
these things are 8inch or so in dia
do they take step type input or +/- stuff or direct speed control?
very expensive though
I wasnt going to ask, they say ifyou have to ask you probably cant :)
they just run from the vfd...which can be controlled by emc
let me give you a link
I have a very wicked tool changer idea in my head, its manual, but high tech, and part of the spindle
I have a fully automated hydraulic idea as well, and oh neat a non hydraulic idea just popped into my brain
[04:23:53] <les> http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&category=57122&item=3809305426&rd=1&ssPageName=WD1V
I would like automatic tool change, but it cost as muchas a new car
oh that is wicked
ill have to draw it up, cant find good links
well I got through les's and jeff's files fine - but going to bed now. More testing and reading tomorrow.
[04:29:16] <gezr> http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&category=45019&item=3863250403&rd=1
that is a tool holder that with the turn of one screw would hold tightly into the spindle
cradek : goodnight
they have a complete system, and the holder unit will be the spindle flooting in a bearing pack
really excited about the segmentqueue stuff
me too, but my chickens aren't fully hatched yet.
im gone as well, ya'll have fun
on closing...I have been trying to incorporate emc in a large corporation manufacturing line. I hoped I might get some funding for you guys...but segmentqueue has to work to do it.
The money will go somewhere. It would be a damned shame to give it to parker.
anyway, good night all
i was wondering if emc can be used to do blank milling so i can cleanly prototype pcb layouts without corroding ..thus using chemicals
hmm, well, ill see :) lets insert this BDI iso into qemu and take a look :)
hmm, ok, stuck at the "enhanced machine controller" popimage
Good afternoon paul_c.
* paul_c hides..
cradek: did you commit your segmot fix?
no, I want to make sure it's the right fix and not a band-aid first, and that requires more study
you'll end up the segmot expert if you're not careful
well there's a well-written paper describing the algorithm... I could have it worse
03paul_c * 10emc/ (lukas.run generic.run emc.run): Modify the generic.run script to handle the new realtime & nonrealtime plats if they are not symlinks. Replace emc.run & lukas.run with wrappers calling generic.run - One run script to maintain rather than a multitude.
I just released axis 1.0b2. see the axis website, axis.unpy.net, to download it.
I'll have a look
whoops think your site is down right now
well back to the shop
Anyone want a used Silicon Graphics 320?
I wouldnt mind :)
but I am in sweden
is here somebody how knows how precise spindels are working ???
looks like that it is a two phase motor
what kind of spindle?
it is maybe a SC62
les: hm, can you browse axis.unpy.net yet?
les: this irc session starts at the same host as axis.unpy.net so I think it's "up"
jepler_: fine from IN
jepler_: the front page still says "coming up"
cradek: OK, front-page article added (and the download should have been on the sidebar, too)
is 256k right?
seems to be complete
wow, that's small
strange can't get the site...got it a few days ago
les: what error do you get? connection refused, timeout, 404?
just doze saying it can't find the site
look at the very bottom of the doze error
does axis.unpythonic.net work?
the last line sometimes has the real error
there is no www.axis.unpy.net. http://axis.unpy.net/
browser puts that in
jepler_: axis.unpy.net and axis.unpythonic.net are working here
oh got it...browser did'nt put in the complete add.
had some crashes and it forgot preferences I think
the new improved xp sp2 crashes very excellently
les: do you have a precise spindel ???
it's a feature
I would like to think the spindles on my machines are precise yes
you mentioned the one you have is 2 phase?
i can mesure it
seems very unusual
sure it's not star 3 phase?
but i have four wires and i can mesure 2Ohm between two pairs and they have no connection to each other
cradek: 256k is just the source, compressed .. that seems pretty big
could not find manufacturer's data I guess?
the problem is i have a frequency converter but i don't know how to connet it
test to see if there is a low resistance between all possible wire pair combinations
there is nothing else
* Imperator_ hangs on the phone
if there are wire pairs that have high resistance then it must be 2 phase or a starting winding
I googled 2 phase motors and there is just about nothing other than steppers
1,3, and 5 phase but not 2
* Imperator_ is back
les: two phase motors are also possible
i don't know the english name, but the asynchron motors with the condensator is such a one
yes I saw some reference in a book about them using a phase shift capacitor
They are apparently very uncommon though
I wonder what advantage they have
Hm, for two phase motors is a complicated electronic needed.
don't know why they are doing this
I found a link
Is it marked 2-phase?
they are run from single phase
no but i have mesured it
or two legs of 3 phase with reduced stability
Not just a 1-phase that can run series or parallel on 220/440?
i'll get the link
[21:14:41] <les> http://www.ee-one.com/new-5796140-4316.html
It is a Precise Spindel which runs on a frequency converter
that guy is not correct, a one phase motor is also possible but you have no starting torque
I have an old, old book around here somewhere...
this might also help:
[21:18:24] <les> http://www.micromech.co.uk/dir_products/prod_range_slosyn_sync_motors.asp
No, he is quite correct...
les : were you able to test the changes made last night?
chris is still working on them
cool, I was just wondering
the spindel is a asynchrone motor
I am running production right now and can't play with the software much, but I will try iy as soon as he is ready
did anybody find a test case that wasn't "short move followed by dwell"?
imperator: yes I understand...funny that the 2 phase slo-syns are synchronous
that's the only case chris fixed, as I understand it
yes I have absolutely had it stop with no dwell programmed
but there may have been repeated points
I cannot see the code where it stops on the display
jepler_: I don't think a dwell is required to tickle the bug I fixed
Hm..must have left the book in my other library :P
well I can't see the next line really
I have a ton of stuff to do for now, ill be in and out
ill see if I cant get a box up and running this evening
Ill probably just dump whats on it to disk, and then boot i up, clean debian
the file I sent had no dwells programmed...for sure
but it stopped at first right?
I think the only time It wouldnt run was when he was trying to use your ini file to match your machine
that or I dont pay enough attention
hm, I wonder if a patch would be accepted so that ./sim.run xxx would pass "-ini sim.ini xxx" to the display program. axis would load the g-code file xxx in that case, which would be quite useful for us unix shell lovers...
asdfqwega: a one phase motor is possible, we had such a one at our self made grinding machine when the condensator has died. It was working well but you had to start it by hand
jepler_: if you have commit access, every patch is accepted
I have more files known to cause a stop with segmentqueue if you need them
they are huge though
im going to try and get a machine read as x and y with a spring loaded stylus so I can run tests as well, if thats cool cradek?
les: is the error always the same (negative length)?
* paul_c tries to build the latest Axis
paul_c: you'll have to make your setup.py changes again, those didn't get incorporated.
the only other thing I have seen is assigning a too slow velocity to long straight moves sometimes
I'm still studying the reason for that (even though I know how to get around it).
works, just goes slower than programmed
interesting - I thought I saw that, but wasn't sure
paul_c: I have to confess, I haven't tried to install bdi-4.xx yet, I only burned the image
it's possible that the long segment was "joined" with a very short adjacent one
yes you should have seen that in the border cut
I'm not sure the criteria for linking segments are right. I see some suspicious behavior.
right...it is doing some tiny 3-d stuff in a sharp inside corner then a long straight move
The math looks right....not sure about the code
* paul_c waves to rayh
cradek : quick question for you, if max seg lenth is 500 segments,does it complete 500 then refill or does it refresh once one is consumed?
you mean the max limit of linked segments? if so, the queue rarely gets that long
yeah, oh okay
funny thing is one really needed only 4 point look ahead to do this stuff
not true - sometimes dozens of segments are linked and planned all together
I was just wondering
for instance, a circle around a pcb pad is planned all together as a circle
(mine have 14 segments)
segmentqueue apparently looks ahead the whole program (if it is not too long)
currently 13 are linked and the 14th does something strange (related to this error)
yes, I understand it's possible the whole program can be linked in some unusual cases (maybe a spiral?)
I know it is linking many many segments...I'm just saying the cubic spline can be done with only four
but if the four are short enough that your machine can't get to full speed, you lose
I thought that was the whole idea here
* rayh waves back
cradek so basically, then the planer needs a parser to see if something even qualifys for that planer
think of a circle with huge radius (needing very little acceleration) but made of very small segments
The planner solves the boundary conditions that include machine constraints
* paul_c has a question or two for rayh
for a circle it would update linked cubics...actually quadratics (cubic with a zero first coef,)
les: ok, I obviously don't know enough to debate about that
I would suggest that if there's a simple planner that's better than tp/tc and simpler than segmentqueue, you should code it up
Here's a program that gets a "Panic: length negative in sqPlanSegment", no dwell. g0x0y0 / g1x.1 f8 / g1x.2 / g1y.002 / g1x0y.001 / g1x0y0 / m2
haha Fred and I had a talk about that....we both used the same controls textbook in school and it has the whole thing solved closed form
this is without chris's patch
on the other hand, the next line is "m2" which is maybe like a dwell...
yeah the error is at line 6 which is not linked to line 5
the only link it makes is 4 to 5
what tells the planner what is motion and what is machine?
Chris... I think Rogier is a super smart guy and he made the task such that it was a challenge...whether it was needed or not
i'm not complaining
les: write yours! there's always room for 3 planners!
I just can't follow all the code in emc
if someone would set up some structures or something to get and throw data I could do it
If you are smart enough to understand the math, I can't believe that
start with the old planner, it's very clean
I work day-to-day on a huge confusing application that I am far from understanding in full after 6 years. I just plow ahead, and still get stuff done...
there's a queue, one file adds to the queue in a canonical way
the other file monkeys with it. I understood it enough to fix helixes in an hour or two.
well if i can get data from the cannonicals and know where the queue is and how to get stuff on it I could do it
what would happen is this:
4 coefficients would be calculated for the cubic
each cannonical motion would cause those numbers to change
but they change in a way that beginning and end velocities and accel would match
it's an 8x8 matrix I think
linear...so solved closed form
So the PID always can just plug in the current coef's to find out where it is supposed to be
done for all axes of course
then inverse kin can find the eq in joint space
Now this is already done at the servo rate...but that's the wrong place
Has to be done at the planning stage
well the servo rate sub interpolation doesn't hurt...but it's too tight to do much in most machine aps
Rogier's code does a similar thing...just in an unusual way. It preprocesses many more points
After all it is planning. Actually tp could be done a few microseconds before or two weeks before...it's causal (unlike pid)
what is the location of the newest bdi release?
I wont be able to reach a private address
I'm waiting for the beta testers to break it before going public
I guess im just a casual user
You could still do some testing...
I would not have asked for newest had I been planing on using it for production, I can either clean install woody, or toss bdi onto a hard drive, It makes no difference to me how I set a system up
it was just a good place to try some install to drive options
The latest BDI is using a 2.6.9 kernel...
gezr: I'm running it here and it's very nice. Synaptic doesn't seem to work right but the rest is very good.
Got some autoconf stuff in rcslib that needs testing
so you would rather I went fresh debian install, cvs gets, match up the kernel, and have it ?
have at it that is
If you wanted to do a manual install of Sarge (or Sid), I can bung all the debs I've created into a repository here
Then you could add it to the sources.list and apt-get everything.
im flexable, however you want me to test or not test,
as long as you move the files to your mirror using apt-move it should be dandy for apt-get
* robin_sz wodners if debian has actually released a new stabel version yet
no still on woody
3.0r3 i believe
Sarge will be going stable real soon.
by the time sarge goes stable it will be old ...
it always is with debian
yeah, they said that in june :)
You could always release your own Debian based distro.
that would be madness
the trouble with distros ( and that includes BDI ) is its all too easy to go down that route, and then you soon end up spending more time on the disro than on the actual packages ...
In the case of EMC if there was no BDI, you'd spend all your time trying to help folk compile it.
Trust me I've been there.
true enough ...
I suspect the real answer is packages like rpms or .debs
then you end up having to maintain packages for kernels & RT extensions...
The balance between progress and stability and workload...
And learning curve...
* asdfqwega just compiled a 2.6.9 kernel for Debian
It's like learning to walk again, with different legs
oh yeah a real irc client :)
are steppers really supposed to be loud?
They always make noise, but if they growl or grind, something's wrong
oh, it sounds like "FRRRRRRRRRP"
at really high speed it goes away
is there anything wrong with it?
With mine, there's always the high pitch whine from the chopper drive
Which kinda blends into harmonics when it moves under load
does it aeffect accuracy?
Not in my application
in any application?
unless we are talking extremities..
Only if you're expecting too much from them :P
bleh, I'll order a new servo for that axis soon.
You know how servos 'seek' back and forth?
got servos on the other axis:es
Well, a stepper is a magnetic coupling, and it's kinda bouncy
You move a step, and the motor tries to pull the rotor into position
But it rarely stops at exact position...it oscilates(sp?)
A really good example would be a variable reluctance stepper motor, like a few I have here
24 degrees per step, very low friction and damping
I cant remember what step angle mine's at
but I gear it 6:1
so it's sufficient..
With no load, each step makes a "Pwiiiiiinng" noise
Most steppers are 1.8 or 2 degrees...usually 1.8 for new ones
I kind of fell in love with servos..
totally silent except for the slight whine
24 degrees per step? .. that cant be right
the stepper I have are a japan servo one
let me check
* rayh is away: getting a bite to eat.
What kind of drive?
You're using microstepping?
1.68 newton-meter = 237 oz-in ...that's about the size I'm using
it's lots of torque :)
I'm using Xylotex drives
I hope the axis will perform as well as I want it to�
going to try mill some molds :) for some fibreglass
That's not a lot...I've got some 670 oz-in Nema 42's set aside (that's 670 in unipolar)
Okay, what RPM are you trying to run it at?
1,68 * 6
10nm at the shaft to the ballscrew..
asdfqwega: I dont know yet, I dont have any data on that screw..
it's the only thing I havent replaced at the machine yet
I've found that I can only expect a maximum of 300rpm from steppers
After which, the torque curve is too low to be useful
But that's in my situation...I'm not using ball slides or ballscrews
I've got friction and stiction :)
I dont have slides, but ballscrews..
* an0n longs until he gets time to build a lathe
I need a very small cnc lathe
like 5dm bed..
and a very heavy chuck
and linear slides..
and a toolchanger
I've got a good idea how it should look
only problem is time and to find somone that can cast the base for me :)
anyone who casts, if you supply the pattern
hm, dont know any in sweden
cast iron perferably..
or is epoxy concrete better?
there are some spainish foundries im sure in spain
or poland, with toolmex/bison
paul_c: can it be machined nicely?
a lot of tool builders use a epoxy-concrete base, with linear slides on it
'pends what type of aggregates are used.
cincinati, and hardinge I know use something like that
that would probably be perfect..
an0n: i prefer epoxy-concrete. I want to cast a frame for a milling machine
Imperator_: is that hard to work with?
Imperator_: and can you get it for home use?
* asdfqwega can cast aluminum...haven't tried iron (yet)
i think/hope it is not hard to work with
i have made a test part last week, i was surprised how good it is
Hm...I'll have to look into that
Imperator_: very nice
i have bought a special sand mixture and mixed it with 10% epoxi
how is the weight?
rough concrete casting first, then the epoxy layer is put on top, with mounting holes left in teh concret base, so that the epoxy is machined to accept say a bolt for mounting?
oh its all mixed together neat
the wight is about 3,5g/cm�
hm, what's regular cement/cast iron?
or well, how's the rigidity..
that's more important..
what is "agreagate" ???
look at www.epucret.de
my spelling is off on that
seems very nice
is that expensive?
very expensive !
I don't know how it is overseas, but if I wanted a base for a machine, I'd find a used granite surface plate
sand mixture about 4 EUR/25kg
thats nice stuff
exopy 15 EUR/kg
well, for 7dm2..
the thing with the granit plate is also good but only if you don't have to machine it
if you can use it like it is, then it is perfect
Imperator_: can you guess the price? of somthing like that
well if it's < 400$ it's a good deal.. :
ah, i see, you need to use say a steel inlay in the form, to have something to machine,
wow...that must be some amazing stuff, to make sturdy machines
the prise is like i said for the material, the question is how many tryes you need for a good frame
and you have to make the mould
can you cast it in sand?
you dont really machine the base, you put the nuts into the casting its self, so that, after it hardens, you just have to scrape down the tops or machine , set your slide in place, and bolt it down, since the bolt holes are pre machined
ah well, 2-3 tries..
You can cast concrete in sand, but there's a little trick to it
the trick is that you have to shake the concrete after you have filled it in the mould
so you need a stable mould
vibrate.. yeah :)
I can't remember it you have to spray the sand mould with something...
dont you usually do that when working with cement?
you need to do that anyway to remove the air pockets
if you want cement to flow you use a vibrator
well the mold would be in 6 pices.. if I milled it on my machine
and 2 passes per piece..
the vibrations make it flow like water, and lets the air bubbles rise
not like water, but it was ok
well, I could cast the base as a solid square.. :p
but that isnt as neat.. as to have a structure..
you could probably get by with a woden mold, slime it a vaseline, or similiar grease, have all your ingredients,atatchment points, everything ready to go, mix and pour, put the atatchement sections into place, vibrate, add more to top it off, and your done, I would use the top of the mold as either the bottom or a non used side
that was amazing when i put the 10% epoxi into the sand, the sand was only looking a bit like it is wet. but it doesent flow. then i vibrated it and it began to flow and it filled the mould completly
the concrete has on the bottom side the best quality
so you use expxy and a cement? I keep wanting to say lime, but I cant think of the proper term sand, gravel, here I would use a bag of portland cement
* an0n has a idea
no cement !
so the epoxy is the binder?
I dont really need to make the mold that high..
90% sand 10% epoxy?
I need a ballscrew to pass under and a servo on the other side of the machine..
max 10% less is better i think in the industry they use 6-10%
* asdfqwega wants to see the look on the store clerk's face when I go to buy a 10lb tub of vaseline for concrete molds...
how much shrinkage did you get?
i haven't measured it, but at the next thest i will do that
asdfqwega : you could just load one cart up with the little tubs :)
Imperator_: thanks for the idea!
I never considered that at all, probably no stress relieving needed, nothing like that
i will report if i make the next test
or better ask me
test to see how much it grows after its been put in warm water, for thermal expansion, and check shrinkage stuff like that
i hope it is not so far away from steel
if your going to do another pour, if you can put 2 steel rods in the mold, so you can measure expansions between to cast in points
you just saved me a ton of money, I was going to go with castings
because i want to put some steel profiles in. on this steel profiles i place the slides
can you add a cement die to colorize the base?
or would that cause problems with the epoxy?
you can colorize the epoxy
the newest technik is to cast the frame with a tollerance of less then a mm. Then put a special surfacer on that aerea where the slides have to be and press a reference element on it. The surfacer doesent shrink while it gets hard
then you don't have to machine anything on the frame
that company(Epucret) sells a nice little book about all that
but it is in german :-)
that's a large problem
there are nice pictures in it :-)
but that dosent help or does it?
* an0n is very interested in casting
I'd love to cast aluminium :)
can you do that ?
I dont know how to do :
i can help here ..
1) get aluminium
i think casting concret is not that complicated. I will make my frame out of four parts. then i machine the steel profiles and then i stick them together
2) melt it
3) pour it into something
heat n serve
then throw it away
isnt there more to it then that?
* an0n has a side project for the car
not really, im sure a learning curve is needed for near net shape parts
molds are fun
and some dangers with all of it
well, I am not molding anything final
I'd be machining it to tolerance anyway..
they need to be VERY dry, or the moisture boils out and ruins the cast
ideally they need to be porous
it has a reason why a casting company normaly makes only one type of casting !!!!
roller throttle body..
robin_sz : vent holes right?
like on little airplane motors?
hm, dont know how they work..
but on a normal throttle body you have a shaft in the middle..
gezr: well, I was thiking more of some of the neat vacuum tricks you can do
I have a buddy that has a nice forge, I guess he works with aluminums, and some other things,
robin_sz : oh yeah, almost making an investment casting then hu?
an0n: on butterfly throttles, yes. high performance carbs are almost always flat-slides
yeah, but the flat ones are hard to design..
gezr: polystyrene investment, in a sand tub
robin_sz: and roller ones also :)
thats where part of the mold its self desintegrates right?
well all of it i guess
gezr: blow into sand, push polystyrene in ... revers motor to vacuum, sand locks down tight
pour alloy onto spur and job is done.
lost foam casting :)
yeah hahaha, too cool
an0n: not seen roller ones used except on really tiny carbs for model airplane motors, they are bulky in bigger sizes and its easier just to desing a 3% bigger butterfly carb to take into account the support rod
robin_sz: they cost like $700 to buy a pair of them ;)
an0n: flat slides were introduced for special class restriction racing .. where they spec 'no carb more than 32mm bore' .. then its worth the hassle of a flatslide to get the max performance from a specific bore
[23:59:06] <an0n> http://www.lumenition.com/roller.pdf
ok, see you guys