Sounds like you are both. I dabble in c, but really do math. Aero engineer.
i see good reasons to get away from trapezoidal velocity profiles, though. high end machines might want to accelerate slower at higher speeds
Cornell Math degree huh? wow.
More accurately, I can read math (mostly) and write C.
icee: trapezoidal planners have long been obsolete for modern high speed machine tools.
After all, it's been almost 40 years since that degree.
lerman: you've been playing with the interp.. right?
icee: there are branches, or there were at least. EMC_MAIN and EMC_HEAD
or something like that
Afirmative. Reading the interpreter was duck soup.
that's the same one fenn
most of my expertise is CS/EE. i'm -basically- competent in differential equations and dynamics, but haven't really applied it much
lerman: I have a question for you
you know how Tx and M6 are handled?
fenn: yah, branches are nice, but there still needs to be process to determine when branches get merged, how long they're maintained for, where what kinds of features go
you seem rather competent yes icee
heh "icee for president!"
No, I don't.
gee, I'm so nice today...
lerman: too bad :(
les: haha. I definitely have a lot to learn about mechanical systems, and motion control and machining in particular. But it's one of my primary focuses now
But what do you want to know -- maybe I can find out.
what happens if Tx and M6 are sent at the same time
from what I see 2 NML messages get sent (TOOL_PREPARE and TOOL_LOAD)
I think Tx M06 basically gets a message "ok I did that."
but the first one (TOOL_PREP needs to finish imho before TOOL_LOAD gets sent)
les_w: not anymore
Doesn't the spec tell what should happen -- there is a list of the order that things are done.
les_w: just working on that, so Mazak will have automatic toolchanging (with position and all the fancy stuff)
lerman: any idea where=
* icee drools over automatic tool changing
M06... spindle stop... orient.. move to pos'n in z
dmess: not what I am after right now
on plc side... read t code... roll chain to that pocket... pick up tool and wait for in posn from nc
Tx -prepare next tool for changing
plc can read the too any time...
preping tool is a plc out command
yes, with a feedback
that prepped ok
but.. the problem I'm having is pushing that feedback back through NML
before the M6 comes
The interp has a list of the order of G codes -- but I see nothing about the order of M codes. I believe I did see that in a doc somewhere though.
do you have an encoder on the toolchange chain/rack??? or it might just be proxy counts
yes to monitor the pots... set At 0 or 1 and count 1,2,3,4...
im simple... sorry
classic user/developer perspective clash
other toolchangers have servo t/c
plc needs input... what are your plans...
plc stuff works fine, it's figuring out what EMC is supposed to do that is the problem
fine what do you take inputs FROM???
* alex_joni takes inputs from above
alex_joni: comments in the interpreter funtion execute_block():
Actions are executed in the following order:
1. any comment.
2. a feed mode setting (g93, g94)
3. a feed rate (f) setting if in units_per_minute feed mode.
4. a spindle speed (s) setting.
5. a tool selection (t).
6. "m" commands as described in convert_m (includes tool change).
7. any g_codes (except g93, g94) as described in convert_g.
8. stopping commands (m0, m1, m2, m30, or m60).
ok.. so 6 should wait on 5 ?
alex_joni: machining without the expected tool gets funny ;)
you gotta link the to in a fer places to make a REAL machine
alex_joni: I would let the plc work and when it gets to the tool change time, check with it if the tool is right or break the program/allow the operator to manually put a tool in
alex_joni: how does that sound?
rigid tapping needs it too..
anonimasu: if you figure out how to do that with existing NML.. go right ahead
alex_joni: you dont need to ack the both at the same time, just the m(tool change)
it won't go to the M, if not ack'ed the Tx
I just figured that out
but it's a bit more complicated
yes but why not ack it right away and check if the tool's there?
in the actual changing routine..
these are synchronous packets
you get a command, you ack with status
I modified that (to push estop)
because you can't wait for the next command to check if estop has been pushed, right?
so that part probably is screwing the numbering
btw.. every message received needs acking with the same number :/
so it easily gets awky
does TxM06 send one or two NML messages?
but only if the Tx ack'ed ok
not sure if that's if or when
alex_joni: I find that a bit strange
so if i estop you beter have a GOOD recovery procedure
let me think some more
alex_joni: when you do a tX that causes the machine to be suspended until you receive the ack right?
motion should still work
or anything with higher priority then Tx
not related to IO
io should basicly be blocked
but this IS io
but if you push Estop.. it should work :D
*needs to read up on nml soon*
i need a CLUE
holloweens comming remember all... children in the dark cause accidents... and accidents in the dark cause children
I need to sleep
ditto.. but I'll stay a bit longer
My sample order of dspics won't be here until monday
I'm ordering a lot of stuff lately. UPS truck comes often. Upgrading the machine shop and electronics lab.
electronics lab will get a new building soon!
this has been bugging me for a while: download a fresh emc2 cvs, open 3dchips.ngc, and run it. you get a following error when z reverses.. if i change default_acceleration to 15, then it works
to 15 from what?
fenn: neat. sounds like you're seeing similar problems to what i have
ok i'm gonna go ahead and commit my changes then
since i think i understand it :)
fenn: you have a fix for it?
I run following error settings really tight. It is my primary crash protection really
Too bad my expensive new spindle cannot be used fully with contouring work due to the tp.
03fenn * 10emc2/configs/emc.ini: default accel cannot be the same as max accel or following errors will result, since the planner has no headroom to correct for timeslice aliasing errors (gee that sounded good huh)
I have some time though, because current production uses relatively long arcs and lines
so I can go fast.
if default accel is 20, and max accel is 20
I thought default just did nothing
I have them set the same
well apparently it does do something cause it don't have any following errors now
you found a problem?
I will have to check this out
* alex_joni takes out the RCS Handbook
question about geckodrive
has the G320 the same pcb circuit as G340 but without steps multiplier ?
les_w: it also has following errors when you are jogging an axis and click twice in a row real fast
I have not seen this.
may be a new "feature"
you have lead/lag...
like setup and hold times?
no real lead lag from a scale pos'n??
this come back to actual and command pos'n i was on about the other nite
and you can buy the stepmultiplier later if you want
dmess: i'm just running emc's default config for steppers w/no machine hooked up
so it should all be perfect ZERO
having default and max set to the same is foolish
fenn: yah, i've seen the exact same problems with acceleration levels set low
or smart.. if you know you can always do it..
fenn: what do you have your individual axis accelerations set to, and default and max acceleration?
fred had a note somewhere about that
max acceleration has to be much greater than default acceleration for things to work properly.. and even that doesn't guarantee it
icee: all at 20, change default_accel to 15 to make it work right
i figure max shiuld be tied to mass of motion...
fenn: yah.. you need headroom over the default accel in case the machine falls behind in velocity
which the stepper stuff seems to do slightly
was originally intended for machine vs joint space I think
need to have some accell availabe to catch up
uh, what's the diff between machine and joint space?
machine = commanded position?
none... rotary axis
for cartesian, not much fenn
how does that have anything to do with default and max accel?
but for hexapods or puma or something...big difference
for R's in an arm do the kinematics
all need to catch up sooner or later...
fenn, for cartesian it has little to do with anything. It's just an artifact I think
* fenn is a little slow today
rule of thumb max at 1.5 x default on 3 axis cnc boxes i got params for
les_w: how do you plot accel with gplot?
let's see .... worst case wouldn't max end effector accel be root 3 axis accel?
for a cartesian system of course
works for me
and providing axis accels are equal
I keep them the same
even though my z can accel several times faster than x
les ... your drives matched??
keeps circles circles...although not the same size...
the big machine uses 3 copley amps
the x works hardedst
the motor drivers the same
yes the amps are the same
90v, 20 amp peak
why it works for you...
if I was doing a lot of raster 3d work, I might set z axis accel high...it can do several g easily
you usually have 1 drive slowing it down
x is limiting by far
what was wrong with your sign contouring?
btw was that raster or waterline
but you could up the drive side...( in 10% shots) and find the line
waterline is a little easier on z accel
wcut loc z and ro;; on monitor 2 axes
oh the sign contouring....I think it is showing a blend algo breakdown when no cruise phase exists
this is still a problem
a big problem
basically, emc cannot do high speed contouring machining with close together points
how many pts to establish cruise phase??
close together= like 1mm
so NO GOTO....
it's a problem
a big problem
les how do you get emc to spit out accel logs?
i just get empty log files
is it CPU ... or motion controller
haha I get empty files too!!!
hell and damnation
were comin in
to hell and damnation
already there if you try to do contouring at even 200 ipm
beats up the machine too
would more drive power do it??
I have 5.4 kva peak.
cant do the #'s fast enuf??
it is a breakdown of the blending that occurs at the no cruise threshold
did you see the SHAPE manual i handed out...
well in general...and it's a separate problem I think...servo update and therefore block rate is slow by today's standards
dmess: no. shape?
we SHOULD be at buss speed for 3's right??
i wonder if i will be awake enough after dinner to write a bug report
well my machine is built to be pretty doggone fast. emc makes it go slower.
Also I ran a design for jacky today....10 meters/min, 0.5 g
emc cannot do that...
not with contouring close points
ever trid TURBOCNC or any of the others on it for SHITS AND GIGGLES ??
turbocnc doesnt do servos
have tried some others. Looking closely at galil accelera or whatever it'scalled
try 5 at 1 g
seemingly, no cnc control software for it yet.
fagor is cheap and works
emc rs274 interpreter-> galil accelera....hmmm
24 microsecond servo update
do you get your money back if you aren't satisfied? :)
I doubt it
will emc handle it?? 24ms?? i doubt it
emc maxes out at about 400
but just use the interpreter...and no problem
400 microsecond servo update
emc I can get 90IPM, TCNC 180IPM (steppers)
17 time slower.
100 microseconds or bust
it's definitely attainable
400 microsecond is fine for something like a bridgeport...no problem at all
may require a dual processor or dual core machine to do it
not good for HSM though
build and ship 6 PROTOTYPE DRIVES....
icee...Paul seemed to think it could be faster...a lot
do we have a strong enuf core to run 1 machine from 2 boxes???
until one runs into a brick wall using general purpose processors for dsp work
general purpose processors are fine at this kind of 'dsp work'
if you say so
* icee is pretty sure so.
especially if you can get 100 microseconds
General purpose operating systems may not be.
I don't see any real advantage you'd get using something like tms320c6415t -- ti's top of the line DSP
let plc and hal deal with a network to io..
over a high end pentium M
I have a motion board here that uses it
but it's not a machine control
Jymmm is now known as Jymmmm
does use a subset of emc
Jymmmm is now known as Jymm
emc would run so much better under qnx instead of linux
rtai/adeos is a joke, based on what i've seen of it.
and icee would live so much better under a bridge instead of a roof
alex: are you saying I'm trolling? ;)
trolling for better performance is ok.
a qnx context switch is about 2500-3000 cycles on x86.
* alex_joni shuts his eyes
don't want to see this :)
i have NO PROBLEM with dual cpu boxes.... what WIL it solve??
and pretty much everything is done in user space, very few lock interdependencies, etc, so SMP works well
you know, I think that new galil uses 320c64 something
dm: Well, in principle, on a dual cpu box, you could just dedicate one processor to the realtime loop, and never context switch.. just busy wait
we thru out 2 skids of HP- o2 boxes last week
ooh yeah baby...
what? they are great bricks for building shelters
artificial reefs too
no the small short ones
takes too many... i know
only good for shelter for small short people I guess.
anyways, 2500 cycles is 1.7 microseconds at 1.5GHz; so if you have a task you need to run 10,000 times a second, you still have 99.6% of the CPU left over
after context switches, to share between that task and whatever else needs to run
short ppl ... thats a song ist it
assuming single CPU.
well...I am gonna go to town and pick up some refressing beverages....
but icee...sounds good
the only downside is qnx is commercial
do so les.... so whhere's the bottle neck??
though i think for personal use you can get it free.
les: have fun ;)
need a dr. qnx or something
well, i did play a lot with neutino 6.1
hack and smack
although I know little about such things, a free, fast RTOS sounds good...
les_w: if you go down that road you'll end up on proprietary HW
which isn't bad..
just not where emc started
qnx runs fine on x86 desktop hardware
sure, you can build a BSP and go to several families of embedded hardware, too
and all the realtime stuff can run in userspace, and not be stuck in kernel space
but would it migrate that easily??
dmess: no. I'm afraid of emc, to be honest.
it would suck to port.
so how hard to migrate qnx int linux
dmess: the problem is that there'd have to be a lot of rewriting on the hardware drivers
or, at the very least, putting in compatability shims
qnx expects you to run as a user space process, and ask nicely to get privileges to touch the hardware
shims... LOL... fixup..
well, it is possible to abstract out the operating system interface
so the same driver will run on multiple operating systems, etc.
it's just difficult to go back and do that
Jacky^ is now known as Jacky^afk
so...I have this motion card in my hand. It has a TI TMS320c6711 dsp on it. It's insanely fast, and cheap.
320 ain't shabby...
seems that the dsp could run a whole roomfull of cnc machines
has 4 lmd18200 servo amps on it too. With a little fan.
Tooo coooooolll :-)
let me show you a link
[02:29:53] <les_w> http://www.dynomotion.com/
Whew *furry* and at those prices, should be :-)
seems like a lot oh hardware for 899
Yep it is
uses emc interpreter by the way
I have been talking to the guy. Trying to get him to turn it into a full blown cnc control. I think it would be a big hit.
he has agreed to make a version with bigger servo amps
and expand the g code usage
Hehe, I'd be glad to test a few out :).
well , I am. I got it free.
Some folks have all the luck :) while I build axis up from all thread and "L" brackets :)
but...I have considered marketing a line of low cost high performance cnc machines.
Very nice if you can !
So, I am trying to convince this guy that he would make a ton of money if he made the control
next closest thing to that performance is thousands with software
so...it's just a software thing
les: what kind of unit volumes do you anticipate?
the whole thing depends on a low cost control. If the mix is right, hundreds an at least
what's 'low cost'?
was hoping for an wmc control
well, for that low of unit volumes, there's a definite plus from trying to massage emc into doing the right thing
but ...let me ask this...does insane overkill processing power let one lower software dev costs by writing relatively dumb code?
definitely. and so when you're amortizing things over a small number of units
you throw a ton more hardware at it.
well he is already using the emc rs274 interpreter.
has a crude but functional TP
I would like to say thousands...but will be conservative and say hundreds
what kind of unit cost would you need a controller to be at? that kinda sets the parameters for everything else
I don't have a good feel for this market
well, quan 100 might be $2000 or so for something HSM capable right now
this thing could be $600 WITH servo amps
hmmm. tough budget to meet.
leaving me some room to do good mechanicals
well the card is there...the software is only part way.
what's bom of the card, with amps?
did you see the link?
no i didn't, one second
hm.. i wonder how he's making any money
those DSPs alone are $60-70 at 1ku
unless things have drastically changed in the past few months
he isn't. He has a day job. This is a bedroom/garage operation
03alex_joni * 10emc2/src/emc/iotask/ioControl.cc: fixed that problem, now toolchanging should be ok. (there was a bit o a hassle to figure out how to enter RCS_EXEC, when and how to exit back to RCS_DONE).
if you really want to do this cheap, les, we ought to just cure the EMC performance woes
Seems him telling me $600 quan 100 is unrealistic on his part?
dang that was a bit of a ruff job
al for fixing emc...but gosh it has been many years now
Then I could put my simple dspic servo implementation on a PCI card
or PC/104, or whatever you want to use
les_w: what has been so many years?
it will get better
the TP has been worked on since 1999
no one worked on it, to be honest..
at least not with commited code
I had the author of emc and segmentqueue here working on it!
and then you have an umbilical that you use to hook up to the appropriately sized amplifier board.
servo amps can be cheap...Mariss has proven that
les: if you're serious, I can figure out how to get you 10kHz out of EMC.
and making a PCI card at qty 100 with the architecture we're talking about is pretty easy/cheap too.
I need it. anyone doing HSM needs it. Folks running sherlines or even bridgeports do not. (95% of the user base?)
les: yah, I don't need it. I will try to get it, just out of vanity
* Jymm just finished dinner.... yum
but hey, I'd definitely be interested in working with you.
chicken n dumplings!
It'd be a real learning opportunity for me, and I think my side of it would be pretty easy for me to be honest.
can't say what paul has or hasn't done in that direction
I know abou SQ
anyways.. slowly emc2 is getting better
I really think so.. you might want to try it out some day :D
anyways.. enough for today
catch you later
* alex_joni has got to catch some beauty sleep
sounds cool. The low cost high performance cnc project has basically stalled due to emc issues.
about 3h ;) should be enough.. I hope
alex_joni PLEASE take 48 hours worth
les: do you have anything written up on what you're contemplating?
alex_joni you need ***ALL** the BEAUTY rest you can get
say...$1k for a control. Give me $4k more and I can make a pretty decent small machine
sell for under 10
the magic number.
well, let me finish the servo implementation and prove myself
* fenn cracks the whip
repackaging it, etc, would be pretty easy.
I'm not really interested in tooling up to become an electronics manufacturer, but I could certainly line up manufacturing and assembly and then revenue share.
I mostly design machines. I think I could make a blazing fast small (1x1 meter) cnc.
With a fast control.
what's the intended market? tool'ndie?
les_w and repeatability tolerance being .001 miles
compete with roland
also some proto makers
fuck roland.... mostly for model jewlery makers
out of hobby price range I think
1" Z travel
I really feel I could blow roland away...and they sell a LOT of machines
is roland partnered with sherline or something?
not that I know of
what does he sell?
just a sec
wonder how i missed that while staying at his place for a week
les_w Roland the company, or something/one else?
les_w This is $1999 ---> http://www.cadcamcadcam.com/
[03:02:58] <les_w> http://www.rolanddga.com/products/3D/modelers/default.asp
les_w Yeah, that's the roland I thought.... screw em
$24K for this?! sheeeeeeeeeeeet.... http://www.rolanddga.com/products/3D/modelers/MDX-650/
look at those prices!!!
look at the poopy engineering of the machines
optioanl tool changer
for $24k it better not be optional
but they SELL
les: so what's your plan to get exposure and to reach the market?
for them $$$'s im going for a toshiba
it looks neato, and they have a pretty website
even if you've got good hardware, and a good price, it's difficult to move product
if i were spending someone else's money, i'd buy one
if no one's buyin'
les_w This is $1999 ---> http://www.cadcamcadcam.com/
well, I am involved with a fairly well known cnc router company.
les_w : Since when did Penthouse magazine become a cnc company?
and Chinese attempting to market ok made but very badly engineered machines
still the 200 ipm cieling
there are some.. few.. very well made chinese machine..
miniMess name one
right I have been checking the little $200 machine.
so what can it do?
les_w no clue, just saw it mentioned on CCED
dah lih box style #50 macine i'll rock against the best with
[03:09:24] <miniMess> http://www.ipnews.com/archives/metalcutting/jan03/maxtec_dah_lih.htm
how about 15 meters/minute and 1g as a design goal?
ive ran 4 different models....
not with cat 50 and 48" x 30" table
les_w: how about neato plastic shrouds as a design goal? :P
El Monte... I know where that is
they were old school... well designed... machines
is dah li Korean?
i stuck a lead screw after about 96 hrs of constant up and down rapids..
ROC if iremember... tiawan mabe
I did that on a tiawan cnc once.
ROC == Republic Of China
with .150" @ 30ipm at the bottom for the part...
japanese leadscrew though
was talking to icee
blah...about the end of the day...looking at my whiteboard schedule....2 out of 5 things got done.
I put too many tasks there!
sign painted and report completed
oh and an improptu sessionat the surface grinder
I just felt like grinding something I guess
I like doing that with my angle grinder.... sparks go EVERYWHERE!
les_w try to go for distance on the sparks =)
aw was grinding cast iron...few sparks
on the boyar 612 surface grinder
but I have a milwaukee anglle grinder
it makes sparks!
catches clothes on fire.
lol... especially if open can of fuel is around =)
Interesting... a "Home defibulator" http://www.tvdepot.com/heartstart/buy_now.jsp
03jmkasunich * 10emc2/src/hal/components/blocks.c: added 'match8' HAL component, compares two sets of 8 bits and indicates when they match
Jymm is now known as Jymmmm
Jymmmm is now known as Jymmm
03paul_c * 10emc2-auto/wiki/ (11 files in 8 dirs): "Auto update wiki from a cron job. Thu Oct 27 05:30:01 BST 2005 "
icee: get a usc
icee: or write a pulse generator
anonimasu: Hey let me go to that out of my dumpster diving jems from behind the chinese fast food place!
just build a pulsegen
anonimasu you make is sound SOOOOOOOOO easy; buy yet to see massive amounts of them out there.
*goes to work*
Jymm: I was referring to how ice would get 10khz out of emc
10?! you mean 100 dont you?
or 10 per axis you mean
ah no I lost the log
14:06 < icee> les: if you're serious, I can figure out how to get you 10kHz out of EMC.
work work work..
Well, I thought you already could get 10
ok see ya in the morning
anybody know the debian source locations for emc these days?
someplace is a url that paul_c maintain(ed/s) that does all the magic to let apt-get install emc and stuff... but alas, all I'm finding is old dead links ...
hi guys... the 10khz icee was talking about was servo update rate... not related to steppers at all
heh.... funny... I found the answer in the archives, from the last time I asked it :) https://mail.rtai.org/pipermail/rtai/2005-January/009848.html
goodnite... heading for bed
ditto.. G'night LawrenceG
yah, lg is right-- servo update rate
Jacky^afk is now known as Jacky^
anyone know if this PS can be used with Gekodrive? : http://digilander.libero.it/jackydgl0/supply.jpeg
transformer is toroidal 500 VA
out is 50 + 50 + 10 V, ac
that means you got 500VA overall
its from an HIFI mosfet power amplifier
so if you use one of the 50V outputs, you'd have only 200VA
alex_joni: i think abou 6 ampere DC
at 50V that's 4A
what kind of motors do you have?
i'm looking for the motors,
i just know the nM
that won't help
X3.5 nM, Y 1, Z 0,5
you need voltage / current
they can be 20V or up to 100V
i can choose it
but the supply in the pic is usable ?
you'll need about 50V or so, to use the 70V in the supply (in the pic)
but I recomend to use only half of the transformer
as in the 0-70V supply
not the others
might be a bit much voltage
gecko's can handle up to 90 iirc
only one for all axis ?
but that's extreme
its enough ?
well.. you should have only one for all axes
but building a powersupply of your own should be fairly easy
and a lot cheaper
get a toroidal transformer
add a bridge
and condensator (for rectifying)
10.000 uF / 100V might even be a bit big
[14:37:27] <Jacky^> http://digidownload.libero.it/jackydgl0/supply0002.jpeg
I used 1000uF just fine
i asked, because already have this ready ..
and the transmoer is not so cheaper ..
if i remember well
transgformer is about 20 EUR here
i payed about E. 250 for thr transformer only
a 300 VA 24V
that's too much
way too much
it was a kit
250 W power amplifier on 4 ohm
[14:40:26] <alex_joni> http://www.toroid.com/images/rectifier_circuits2.jpg
the first one
225 7 110x55 2.20 L30.25
225 VA, 110x55 mm, 2.20 kg, 30.25 UK pounds
well, i will get more info about the motors too ..
i think ampere it have should be enough
down to 9 pounds / piece if you buy 100 pieces :D
that's 15 EUR
as voltage i'd prefer high voltage around 80 V, but
lets see the the motors type ..
80V is dangerous
it's close to the 90V where the geckos will blow
24 V seem to me few ..
so if you have big moving mass, and stop the motor it will generate power back to the power supply
i meant 80 V max
and the voltage will increase
oh .. the peak maybe
when the motors are breaking, they are functioning as current generators
btw, I was talking about DC motors
so, 24 V transformer should be a good choice ?
I know you have steppers: higher voltage
yeah, DC motors, is what i want use
then 70V is definately a LOT
I have some 42V motors (about 200W)
and the 24V transformer gets about 35V outs (after the rectified part)
generally you need to multiply by 1.5 (30V transformer gets you 45V rectified)
i know, im using 24 V transformer right now for the steppers
well.. not rectified, filtered
alex_joni: good, thanks
[14:47:00] <alex_joni> http://cgi.ebay.com/39-39V-700VA-Toroidal-CNC-Servo-Power-Amp-Transformer_W0QQitemZ7556183427QQcategoryZ4660QQssPageNameZWDVWQQrdZ1QQcmdZViewItem
50 USD brand new
that's more like it
uhm, yeah, but far away..
a lot of weight to ship
you can find some in europe
this was just a quick search.. for prices
found something ..
[14:52:04] <Jacky^> http://www.robotfactory.it/CncAccessori/Trasformatore.jpg
150VA 220V 20V E.38
maybe a bit small ..
id like to use small capacitors too instead of a big 1000 uF
1000 is not that big
you need 50V or so, if the transformer is 20V
and 50V cap. is a lot smaller than 100V cap.
I have some 10.000 uF 100V
but i read that using small capacitors values instead of a big one
fit in my hand
its better for fast time
not for the space
3 capacitors of 350 uF are more fast then only one of 1000 uF
not sure about the advantages in cnc use
but i want to try that
you don't have any advantage from speed here
20V is a small voltage to run gecko's stepper products on
you need a bigger cap. to keep the voltage constant
jepler: 20V transformer AC, gives you 30V filtered
should be OK
hi jepler , my drivers are G340
I've ran a gecko of an 12V transformer
Jacky^: you need at least 18V input
alex_joni: yeah, ive seen the user manual..
trust me .. it won't work on 18V
if you get a voltage drop the drive will reset
not very nice :D
need more info about motors.. voltage and rpm ..
* alex_joni goes home
newbie question: what the advantage of a RealTime system ?
respect to a normal one..
Jacky^: with a "normal" system, how do you ensure that something happens "at the right time"?
Jacky^: you can't, because the kernel doesn't provide any guarantees about when a particular process may run
so, is the 'priority' in execute process right ?
jepler: excellent explanation, thnaks :)
did a few tests. Looked at tp.c and tc.c some more. I can't follow it.
so many intermediate variables
just too many for me
Jacky^ is now known as Jacky^afk
you know if you google "discriminate based planner" the only hits reference tp.c and tc.c
I have never heard of a "discriminate based planner"
a discriminate is a stitistical function
should this be "disciminant" instead of "discriminate"?
i.e. quadratic formula
I bet, because it makes parabolas
and the code looks like a quadratic discriminant
I just can't follow stuuf like " give credit for"
I wonder id a storage scope on the amp current monitor would shed some light
wouldn't you see the same thing as with logging?
(what exactly are you trying to fix?)
my logging seems to have problems...either that or me. I get garbage inthe file.
What I am trying to fix:
the motion is much smoother since the tp.c change but....
as I run real files at higher speeds I still get nasty accel spikes
improper velocity adapting
so I have a $2200 new electrospindle that I cannot run at proper feed with contouring programs
with long arcs and lines no problem
well when we did the fix we knew there were still some issues lurking there
I've been pretty sure all along that the latest change was not a fix
nothing like a new spindle to highlight them huh
I couldn't understand it well enough to do better.
chris the latest change might have been a band aid
sure, I agree
that solves one problem and creates another
net effect was better....
seems perhaps something needs to be changed rather than remmed out
or #ifdeffed out
hm that requires somone to understand it..
currently we are setting prevmax and preamax to 0 always
whatever that means
I think he did that because he didn't understand what they do
(I know I couldn't)
well the comment " might be needed" etc kinda drew us to that section
so, seemingly with the #ifdef we are clampinf curent velocity to vmax and current accel to amax always
which wouls surely not be right
/* this one is either accelerating or constant-- no
room for any more blending */
preVMax = 0.0;
preAMax = 0.0;
and we do that all the time.
Jacky^afk is now known as Jacky^
Have we #ifdeffed out the nasty glitch AND blending?
no blrnding would look better that the glitch we had
les_w: ever heard about a schaublin mill?
only the name
I am wondering if they are good
don't know...let me check
[17:19:11] <anonimasu> http://www.blocket.se/view/6448380.htm?caller=nbl_s&l=0&c=1&city=0
seems very high quality swiss
[17:21:54] <les_w> http://www.smsa.ch/pl-102b-e.html
I woder if I should buy it..
les_w: what do you think about it?
anonimasu: Got Forklift?
anonimasu: Get this instead --> http://bpt.com/index.asp?pageId=657
how much do you think shipping is on 5000kg to sweden?
no more than $30USD
for the invoice alone.
I wish I could find 1/8 downcut that have a 1" cutter
Jymmm: I have several..
Jymmm: but I cant ship anything like that
anonimasu who makes them?
I think the schaublin is a better deal..
anonimasu the bits
Jymmm: why cant you just do multiple passes?
fenn 1" thick material.
Jymmm: cut out 1/2" at a time
Jymmm: seriously, cutting 1" at a time isnt something you do on hobby machines..
fenn:You dont get it.... it's not the amount of material. it's the shank hitting the wall
I dropped them a mail about the machine
Jymmm: is it because you have too much deflection on the first pass?
fenn: 1/8" cutter on a 1/4" shank - the shank will hit the material
fenn because it is not long enough
oh a 1/4" shank
well you kept talking about your 1/4 to 1/8 adapter collet
Yes guys... I want to take a butter knife and cut a chunk of RC60 steel in half in under 2.3 seconds.
that can be arranged... with a railgun maybe
making some calls...time to enlarge the shop.
that's about what I want
* Jymmm scrolls up.... I dont see any mention of an adapter.
les_w rent out the extra space
les_w Hey you can give the tractor a new home indoors
well it would be nice to have a separate electronics lab room, mechine shop room, and spray booth
les_w dont forget power coating and massage room
get a quonset hut
monolithic domes are cool too
well 24x36 with 9 foot ceilings plumbed with slab on grade floor was only $12k 5 years ago
kinda hard to beat
oops that's 1/16
[17:59:51] <fenn> http://www1.mscdirect.com/CGI/NNSRIT?PMAKA=09885963&PMCTLG=00
oh hell, nevermind.
any issues with freud bits?
what's cookin' ?
oo i found one http://www1.mscdirect.com/CGI/NNSRIT?PMPXNO=9107369&PMT4NO=0
fenn: close... flut length 1/2"
freud is ok.
had a bad dream les?
Jymmm: as long as you dont buy the cheapest bits you will be fine
Jymmm: it's not like you are running production..
what's up? got your spindle back together?
I threw a mail on a new mill
where the fsck can I find a suse ftp mirror?
* alex_joni is really pissed at novell
can't find one damn ftp with 8.2 on it
SUSE is commercialized..
[19:28:00] <Jacky^> http://digilander.libero.it/jackydgl0/snap.jpg
* Jacky^ is a bit confused ..
les_w: around ?
looking at RX130H, RX320E, RX530F,
Jacky^: you'll need gears on those
i thought the point was not to need gears
alex_joni: why ?
for the inertia ?
dunno. RPM seems a bit high to me
but then again, what do I know :)
I am around jacky
looking at some motor
[19:34:30] <Jacky^> http://digilander.libero.it/jackydgl0/snap.jpg
what about RX130H, RX320E, RX530F ?
fenn: lol :)
Coppia is Torque ..
Coppia stallo should be Stall torque
it's pretty easy to figure it out since they use SI
looking at rx 530 for x
if i remember well Z need a brake
inertia is high I think
les_w: remember he used G340's
not more than 90V
fine on terminal voltage because he needs only 1000 rpm
looking at torqur demag current is 30 amps?
seem a lot to me..
the gecko will handle 10 I think
alex_joni: not 20 ?
ok...2700 at 117v....1730 at 75 volts so ok there
you need 1000
20? might be.. my memory seems rusted right now
lowest voltage then
for a gecko..
if you drive with 90v..
anonimasu:thought it was up to 1.5 kW?
now max torque (peak using gecko 20 a) is...
let me check.
les_w: damn.. high inertia
they write low inertia motors :/
7.4 nm ....more than stall so the gecko is fine on both peak torque and terminal voltage
high moment of inertia
what did I say for x? 7x10^5 kg m^2?
i've the log saved..
it might be
please check that number
mshaver is now known as mshaver_away
be patience.. surfing a log
20:51 < les_w> ok. about 0.5 nm with 10^-5 kg -m^2 inertia
20:51 < les_w> so x...3.5. Y...1 Z....0.5 nm
looking for X ..
I found it too 59x10^-5 kg-m^2
20:27 < les_w> for y, 1 nm m torque 10^-7 kg m^2 rotor inertia
and this motor is...
accel reduced to about half for the same current but
you can get 7+ nm peak (for a short time) with this and gecko so
looks like it will work. It will use more current than planned.
inertia mismatch about 3.5
there is another serie
but i think not cheaper..
low inertia= expensive
[19:54:34] <Jacky^> http://www.eurothermdrives.it/brushless/brushless_cc.htm
the axem series, at the bottom of the page
brushless= no gecko...
amp would cost much more
ok, will search much
at this point without ask quotes for that
ok. well the 530f will work...but is not ideal
les_w: thanks !
it will provide 10 m's and 0.5 g
will want to double check before buying
one thing jacky...
I agreed ;-)
ok.. giong for dinner
Jacky^ is now known as Jacky^afk
I see one problem...we are designing a fast high performance machine for jacky....
but emc1 as it is cannot control it
oh well. we'll fix it right?
alex_joni_ is now known as alex_joni
Jacky^afk is now known as Jacky^
-> les_w> oh well. we'll fix it right?
I hope :)
btw i'm using emc2 at the moment
Jacky^: that's the way to go
Unfortunately I don't understand Italian at all! I enter your website in order to watch and download diagrams and charts- I find it very interesing and helpful.
If you can write in english, spanish, russian or polish I'd be greatful smile
lets see if he can run irc :)
* Jacky^ looking some nice video to http://www.infodelta.it/filmati.htm
i like the Vbit job :P
anyone have opinions on servo cards to use with emc ?
etla: i'm a newbie, but i think all step-dir servodrives types can work
etla: no idea then :(
what the advantage ?
more control and intelligence in emc, less in hardware
you can look at the following error all the time
not limited by max step output freq. of pc
thats cool :P
got no info on linuxcnc website ?
yes, but I'd like comments from users
it seem to be ive seen some HW compatibles page
and the servocards are quite expensive, maybe homebuilt parallell port instead...
how much expensive ?
maybe starting at usd600
$ 400/axis ?
for axis ?
no 600 would be for the servo card.
it takes encoder inputs
and generates +/- 10 V velocity or torque signals
the +/-10V goes to the servo amplifier/drive
commercial drives are around 400usd/axis
but I'm building my own
I also like homebuilt drivers
but im not able to do it
also, no schematics are online
if I get it done I will publish the plans
it's going to be for AC servos
like many, I have the surpluscenter motors on order
ive seen that motor
very nice price
yes. there is still the 300W model left
maybe a bit small
what you choosed ?
I got the 400W model
because the 1kW model was already sold out
yeah, I heard someone talking about..
400W should be enough for mani apps
yep, with gearing
the rapids wont be lightning fast but for machining should be ok
doh, after weeks of battles, my cats are sitting togheter.. lol
* alex_joni is gone
g night alex_joni
hmm.. rmmod hal_parport hangs...
no log ?
well, does't hang the computer, but uses 99% of the cpu and doesn't finish
and kill -9 doesn't even work
same thing on other hal modules as well
* fenn gives up and reboots
* A-L-P-H-A pokes K`zan
[22:51:09] <Jymmm> http://www.woodcarvingillustrated.com/images/wci32/Pete_LeClair_10.jpg
Jymmm: what is ?
Hey A-L-P-H-A , how goes.
[23:12:07] <k4ts> http://www.portanapoli.com/Eng/Culture/crib.html
[23:13:54] <Jacky^> http://www.foto.portanapoli.com/FOTO/presepe_LB2.jpg
have you seen wallace&grommit yet :)
bedtime, night all
Jacky^ is now known as Jacky^afk
what'd I do?