UPS is here
yea! I can make molex plugs
damm, no fricking instructions
SkinnypuppY34: still no luck! when I crack the hard injector lines I get fuel dribbles from two but not the third. no sign of it firing (but it's much colder today)
cradek: what kind of diesel?
a little 3-cylinder generator
looks like a small car engine
(I have some very limited experience with a one-cylinder Yanmar diesel
I had a bit of data - looking
hmm, been reading the tool changer posts, very interesting
had a barafaldi go completely nuts on me once, completely disapointed me and the customer at the time
[03:44:45] <cradek> http://www.martindiesel.com/Products/Generators/55-DS%20to%204405-DS%20Series.htm
it's marked 1000-DS, 10.5kW
so the part numbers have changed
thats not exactly small
but it looks just like that pic
no, I guess it's not small
nice you got it running?
cradek , no fuel from the third line?
electric start I assume?
jmkasunich: yes it has what looks like a car starter
SkinnypuppY34: and no sign of firing on the other two
can you tell if it has compression?
maybe there's still air - I'm not sure how to tell
(if it has a hand crank you can feel it, with a starter its harder)
do you know how long it had been sitting?
jmkasunich: good question.
gezar: at least two years
the yanmar's I've messed with (sailboat engines) had electric start with hand crank backup
cradek: your going to need to reprime it, more then likely fuel has gone jelly in the injectors
this one has no provision for hand crank - way too big I bet
yeah - the one I messed with was 8HP, less than 6kW equivalent
I would bleed the fuel system
Is there a clear sight bowl ? Is the fuel clear or clouded if so?
there is nice clear fuel everywhere I've looked
yes the bowl is glass just like in that photo
banjo fittings at the injectors, right?
fuel line to injector is flanged?
can you tell if your getting fuel in the cylinders? does it have that "smell" to it when your tring to start it?
the pipe is flared, a compression fitting presses it into a taper on the injector
ah, flared, not flanged
gezar: I kind of don't think so. it doesn't stink like it should
hard to tell, there's diesel everywhere
you're getting dribbles of fuel at two of three injectors when cranking, so you know you have delivery, but you don't know if you have pressure (to make the injectors do their thing)
Cradek , pull the return off the injectors and put in a small bottle with diesel in it, look for bubbles while you crank over
and one clearly has a problem with delivery
delivery to the injectors or from the injectors?
is there anything at all at the third? even air bubbles?
and the pump is working really good right?
I wouldn't be able to tell if it was pumping air. but certainly no fuel
could be a clog from sitting, something plastic may have broken down over time
gezar: I don't really know how to tell.
you can always make a mess :)
what kind of connections at the injection pump end of the pipes? can you disconnect that end and look for fuel delivery from the injection pump?
same compression fittings. I loosened the bum one at that end - nothing there either
hmm, thats not encouraging
what about the line comming out of the pump?
can you disconnect it and see if it squirts out?
the pipes are only 8" or so
I'd be tempted to disconnect all three at the pump
its like a distributer - you should get timed spurts of fuel from each of the three ports
spurts or dribbles?
you sure about that? thought it would be constant pressure at the in end of the injectors
gezar: on gas yes (electric trigger), diesel no
spurts - theres not a lot of volume per injection, but if there injector was connected it would be 1000+ psi
oh i see
if you just crack the connection, you will see a dribble as it creeps through the threads, etc
that's what I've done so far
if the tube is removed, you should see a small spurt from each port
ok that's easy to spot then
you might even try putting a finger over the pump discharge to see if you feel a pressure pulse
(after you see that it isn't generating a powerfull spurt - don't want subdermal injection of diesel fuel
yeah I'll think twice before doing that...
I have a deep respect for that pressure because I haven't worked with it before
See the delivery valve in this diagram, it's what the hardline goes into . You can take off the line and take it out to check it out.http://www.cs.rochester.edu/u/jag/vw/engine/fi/YANMARVESmall2.GIF
I'm not sure exactly what you will see in termsl of scale - it may just be one drop per injection cycle - but whatever you see should be similar at all three ports
wow that looks complex
does your pump look like that (on the outside)?
I'm what the actual make of your engine is
martin made the genset, not neccessarily the engine
I guess I can't say whether the pump has that configuration
Here's an image heavy rebuild of a rotary style pump. Yours differs instead of having a wavy washer cam it's more like a car camshaft
[04:02:31] <SkinnypuppY34> http://www.vwdieselparts.com/forum/viewtopic.php?t=6694
yes I think it's like a camshaft. I can see the driving gear under an oil filler cap
are the three outlet ports arranged in a circle or a line ;-)
[04:08:42] <cradek> http://www.martindiesel.com/Products/Generators/105%20-DS%20Tech%20Info.htm
I suspect the engine is YANMAR
I was leaning that way
stupid things like paint color - far from conclusive, but it had me wondering
[04:09:22] <cradek> http://www.bellpower.com/yanmar3tnvspecs%20oct06.htm
the 3TNV70 is a 3 cyl
oh 9L, way too big
and the HP numbers are reasonable too
oh, a typo, I see
maybe I missed a tag. I'll go look.
[04:11:58] <SWPadnos> http://www.yanmar.com/store/item.asp?ITEM_ID=97&DEPARTMENT_ID=56
the modern version
[04:12:07] <jmkasunich> http://www.bellpower.com/Engine%20Literature/TNVSeries1Lit.pdf
compare the drawing there to your engine
[04:12:36] <SWPadnos> http://www.martindiesel.com/Old/70-125.htm#MODEL%20%20%20100%20-%20DS
closer, but not quite the identical model number
I would have to check the pumps output, it probably doesnt build up any pressure unless there is something there to build it up on
model 3TN82E-HD disp 1.362 HP 13.5/16.5 (cont/peak?)
so, you know the make and model - thats a big step
incidentally, the martindiesel web site says that they have documentation for older models. you can email/call them to get a copy
Any chance it's pointing nose downhill? The third could be where the majority of the air in the pump body is
it's quite level
[04:23:25] <cradek> http://cobweb.ecn.purdue.edu/~dbuckmas/teaching/ASM345/Information%20&%20Presentations/Engines/Deere%20220%20CTM03.pdf
cradek: was the motor in known working condition, and stopped working after being in storage or something? or is it newly acquired, etc, and might be busted because somebody was messing with it?
it is totally unknown
you just got it?
it has 500ish hours on the hour meter. so it's almost new.
darn, this document doesn't have the bleed procedure
CAUTION: Escaping fluid under pressure can
penetrate the skin causing serious injury.
If ANY fluid is injected into the skin, it must be
surgically removed within a few hours by a
doctor familiar with this type injury or
gangrene may result.
that'll put the fear into you
I already had it, but yeah
[#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/"
yeesh this is what happens to your dreams when you read vinge's "synthetic serendipity" novels http://fennetic.net/dreams/the_google_project
[16:32:05] <skunkworks> http://www.youtube.com/watch?v=O3d0fueNr30
love the MOVING ESTOP button
WOW, instead of rails, he's using bars
for the Y
Hey, does everyone have photos of their controller up some where that I could look at? I'm looking for ideas.
JymmmEMC: I have nothing
it doesn't get any better than that
skunkworks: the 2nd link is real slow
yes it is. He is self hosting I think
this is stuart stevensons stuff
hello machine hax0rz, how are this fine day?
you guys should all beg twingy for mousing support, and then put the GCAM in the emc
gcam is a cam program, why would it go 'in the emc'?
you can specify filter programs to use for certain file types already
cad + CAM + controller (emc) == complete pkg
OS + cad + CAM + controller (emc) == complete pkg
maybe the request should have been, 'put it on the liveCD'
JymmmEMC: get some cardboard, make various mock-ups and try out different operations with your hands
stick the buttons on the cardboard if you like
fenn: That's what I have been doing. It's not so much the front panel as much as inside the chassis that I need ideas for.
oh. well.. anything's better than a rat's nest
its all a rats nets - it is just how well you clean it up ;)
i took apart a tv once that had these u-shaped clips that hold the wires together.. hang on lemme get a pic
fenn: I have a full tower PC case that I considered, but the layout isn't much better than these other aluminum chassis that I have. So, I just wanted to see what others have done.
cnczone search would probably bring up a lot of ideas
have you looked at jmk's setup?
It's just going to be tight as hell either way, it seems.
skunkworks: no, url?
brb, have to grab the other PS for dry fitting
[18:43:12] <skunkworks> http://jmkasunich.dyndns.org/cgi-bin/blosxom/shoptask/box-work-02-13-07.html
he has more pictures - you would have to look around
good for if you're going to be taking it apart and putting back together a lot
they also make those things that mount to a panel (snap in or threaded)
because moving the cam into the controller package is new era sexy stuff
I'm working on a debian package soon
time is limited with grad school
youre scaring me
i need to make a robot that runs the cnc machine
whenever it makes sense to put it on the live cd works for me
did your next semester start yet?
yea, almost half way through it
its quarters or semesters?
I still need to figure out what I'm taking this summer
quarter system at pomona looks neat, everying in BSEE is really specific
I'd like to go to school in australia where they don't really do tests
its all lab and documentation evaluation or something?
makes more sense
yeah, for engineering and science at least
well for anything really
more interaction = good (if you don't already know the subject)
testing is good when you don't want to spend a lot of time and already know most of the subject
yeah all my non tech classes getting all group style
teachers barely lecture now, they just kinda kick it with the class
ren, do you have working compile env now?
tech still lecture
but then you do lab and youre pretty much by yourself with a machine
fillet stuff is committed and just about done line stuff
no im reinstalling windows this weekend tho
basically organizing some media stuff and throwing on DVD and portable hdd
everyone is reinstalling windows
my gf is complaining her windows box crapped out and has to reinstall
its beenmaybe a year
i deserve it
I keep telling her to run ubuntu but she says she needs photoshop
you dont have to reinstall
its just such a bitch to get it back to like it was
reinstall is like 10x faster
well she's gotten used to using all this software she didn't pay for
and has no way to get it back if she reinstalls
so I'm like, go buy it or stop pirating
she's like I can't afford it
i keep confirmed working shit i know ill reuse
no one can
then I'm like use open source equivilents
she's like gimp is crap
$1K for a box and $3K for software, f that
I'm like, whatever
she doesnt take time to figure out the dialogs
its almost the same for 75% of stuff
she's used gimp for some time
she says stuff like her sketch pad won't work
and 25% it isnt the same functionality, you prob dont need
i cant stand adobe stuff anymore, tho
i can totally scale crop and color correct 5 or 6 images in the time it takes ps to load
and if im just scaling screen caps and exports from CAM programs, i prob dont want another monolith app running =\
CAD is like only thing isnt very good for free, cross platform
qcam is decent autocad clone, but i dont think any has it ported for windows
it doesn't really have 3d support though
it's all gnuplot based
and most of the stuff I do is assembly based
bending sheet metal and such
i dont use acad for 3d
fenn: Yeah, those I've seen, even peel-and-stick too
like, i guess it does it, but i just use it for making 2d drawings
I'd use brl-cad before qcad, but I use solid works because brl-cad doesn't even come close
jmkasunich: I was reading your cooling writeup, how did you calcualte the wattage produced in the enclosure?
jmkasunich: and the back pressure is really that significant on a 120mm "pc fan" ?
120VAC not 12V
re: wattage - PC is pretty easy, the power supply sets the absolute upper limit, most PCs use less than half the power supply rating
VFDs are usually over 90% efficient, so assume 10% of spindle power is losses
jmkasunich: The PC is going to be extrnal from the enclosure for me (no room)
gecko losses are in the datasheet
2 PS's, 3(4) drives, relays
the PS gives you an upper limit on the total power used by anything run from it (like relays)
the 24V supply is a 100W unit, probably 80% efficient
so that's an absolute maximum of 120W or so
that limit may be way high tho, since some of the loads are outside the box
(125 actually, but these are estimates :) )
for example, the supply that runs your motors delivers most of its power to the motors which aren't in the box
firstname.lastname@example.orgA = 264W
yes - the lights for example will radiate out of the box
the 24V supply isn't for the motors
but wont the drives disapate heat too?
yes, but their losses are the only thing that heats the box
the power delivered to the motors is outside the box
is that 48VDC supply a switcher too?
geckos dissipate something like 18 watts MAX, even when delivering a couple hundred watts to the motor
SO what you guys are saying is that there can't be more heat than what the two PS's can (potentially) provide whisch is a total of 364 Watts?
if you assume 80-90% efficiency for the supply, then the worst case is ~50W from the supply, plus the 18 or so per gecko = ~100W from the motor drive system
add up all the power that enters the box, and subtract all the power that leaves the box
that's an absolute max upper limit, but it's also unrealiztically high
if you are doing a stepper machine, and don't have the computer or a VFD in the box, I'd be surprised to see more than 150W, and most likely it will be under 100
The absolute values give me a baseline to work with. Now to calc the disapation from jmkasunich's writeup
hello people. I have a M100 program that gets called in my g-code. Does it matter what kind of executable it is? Is it okay if its perl?
owhite: doesn't matter
I should clarify: if you are doing a stepper machine with geckos... L/R drives can dissipate a lot of power
owhite, emc doesn't care
jmkasunich: They're actually PArker OEM750 drives
still chopper based, right?
it's Parker OEM7xx or OEM6xx drives, so should be similar to geckos, with different numbers substituted :)
thanks cradek, SWPadnos. One more question, where should the M100 file reside? how will emc find it?
uh - in the nc_files dir?
the best place for me would be config/
oh. that makes sense.
I think that's where they go. look for the sample M100 and M101 code
thanks. I think they should go in the /config directories, given that my M100 is making calls to signals which are specific to the .hal files. :-)
cradek: how goes the genemotor?
...but I aint complainiin'
owhite, that makes sense to me as well, though there are also uses for external codes that may not be as specific to the machine (maybe user interface related or something)
jmkasunich: I made sure (again) everything is bled. still no dribble from the third port
this is with the lines disconnected at the pump?
one just doesn't do anything
and you do see pulses of fluid at the other two ports?
jmkasunich: Any suggestions other than a car air filter (like you are using), I don't have THAT much room to play with =)
not pulses, but some fluid
JymmmEMC, maybe look at some flat AC filters or something
JymmmEMC: depends on the amound of dirt
there are car filters that are flat as well, for example maybe 6" wide x 8" long, and pleated for a thickness of maybe 1"
it really depends on how much dirt you expect - I tend to overdesign things
yeah, most are flat now
you're doing wood, aren't you?
you probably won't need as much cooling as jmk anyway - he's got the PC and VFD in there
1" would be cool. Like the old carbuator filters.
oh wait - I don't see the VFD
SWPadnos: mine? it ain't in there yet
still running the original single plase motors
but the cooling is supposed to account for it?
is there a way to calc the airflow thru some of these car filters?
B&S lawnmower filters are about 3" x 5" flat paper
I'd go to the local autoparts store and get the biggest thing you can find that will fit in your situation
the bigger you make it the less restriction you'll have
cradek: you get that thing running?
I wonder if a small (flat) furnace filter would do as well.
if you are doing wood (very dusty) you will want good filtering
and get those nice ones you can hose off - they're 2x the price, but you can reuse them :)
and then remember what "car" you have
if you are doing metal, there isn't really that much airborne dust
my filter is probably overkill
cradek: no doubt =)
Well, I know the filter on my shopvac is 3x bigger than the one your using jmkasunich, and it gets clogged pretty quick.
Though it's a hepa filter
well, the application is a little different for a vacuum cleaner
I hope you won't be using your cooling system to clean up the shop
.03micro or some crap like that
Heh, no. Just the only comparision I have atm
But I'd suspect the same particles will be in the air
try sticking the shop-vac nozzle in the air, and see how long it takes the filter to clog
yuo'll be there a long time
how do you like those DIN relays?
cute little buggers, huh?
SWPadnos: No, it's pretty clogged already.
gwt a new one first
SWPadnos: What amperage are the contacts on the tiny ones?
SWPadnos: Nah, just hose it off =)
don't remember - they say on the side I think
I just opened the box today
I was afraid of that. damn.
12.5A peak is what I need =(
the big ones are 20
what are you trying to switch?
router and vacuum motors
get a small contactor
I sent a couple of those as well :)
relays aren't for heavy loads
these are contactor-style, dual-contact relays
jmkasunich: they are contactors
then why'd you call them relays
the others are MONSTERS
it confuses my feeble mind
I called them relays, sorry :)
somehow I doubt they are monsters
the others are, relative to small machines
weight at leat a lb each
we have a few multi-hundred amp ones in the dumpster at work, want one?
CA3SK here: http://ecatalog.squared.com/catalog/173/html/sections/21/17321013.html
jmkasunich: only if it comes with it's own enclosure =)
shipping cost is too high ;)
JymmmEMC: if you have true contactors, not relays, you might be able to use them for your router
contactor (and relay) contact ratings are non-trivial to interpret
jmkasunich: the link SWPadnos provided
if you are comparing the continuous rating of the contactor with the peak inrush rating of your motor, you are being _very_ conservative
those are just the small ones. I'm looking for the page with the actual contactors
jmkasunich: I honestly dont know the difference in what the contact can handle.
jmkasunich: peak -vs- cont that is
what is the router? 120V AC? how many HP? what is the rated (continuous) current?
it seems the contactors are rated 7200VA for make, 720VA for break
2.25 HP, 12 Amps, 8,000-25,000 RPM
make shouldn't be a major issue, but break is marginal - you won't get great life
[20:00:45] <JymmmEMC> http://bosch.cpotools.com/routers/router_combo_packs/1617evspk.html
you want something a notch bigger for that
LC1D09 on this page: http://ecatalog.squared.com/catalog/174/html/sections/18/17418002.html#58350
(it's actually an LP1, but that's just the DC coil)
the top line in the table?
not gonna hack it - only rated at 1/2 HP at 120VAC single phase
but those numbers are very different from what's on the side of the contactor
even thought the router is in sears HP, its probably an honest HP, and a half-hp contactor isn't a good idea
SWPadnos: is that the large contactor specs?
did I send LP1D09 or LP1D18?
the D18 would probably be fine for your router
the D18 is the same size, except about 1/4 inch taller
if you need those instead, I have a couple
how many poles are the ones he already has?
it's always funny when you see breakout boards with their 10A / 20A relays
makes you wonder how long they work
they're all 3-pole +1 aux contact
hmmm - parallel the wires?
for a single phase motor, if its hard-wired so you know which conductor is hot, you can set it up to break hot only
then parallel three poles for that one conductor
not quite per code, but for a personal project it will improve reliability
biggest problem is at break, because one pole will be last, and it will get to break all the current
hmmm. on that first page (with the CA3SK), they rate inductive loads at 35% PF - isn't that a bit conservative
I've done the paralleling thing in situations where the contactor never makes or breaks under load
thats not for motors
especially for something that's supposed to be plugged into a household AC outlet
its probably about right for contactor coils
ok, they are marked separately for "heavy pilot duty"
there you go
pilot duty = controlling another contactor
also for continuous, but well, there are a lot of specs to choose from :)
that explains the 10:1 ratio between make and break ratings too
contactors have large inrush currents
higher actuation load than holding load
it is funny to be near a machine that has a big contactor
when you're not expecting the very large THUNK!
tell me about it
I always jump
no need, I'm sure
ok - the smaller ones
[20:13:59] <JymmmEMC> http://ecatalog.squared.com/catsearch.cfm?QueryText=lp1d0910&ResultCount=10&QueryTextClass=&Soptions=starts
But there is no two-letter suffix on the contactor
that's on the coul - look at the blue part under the top beige part
cradek: deisel? ck the delivery valve is not stuck open on the pump, after sitting lease get gui and get stuck open and piston will pull the fuel back down and not reload
BD means 24V DC
you sure don't need the 575V part
SWPadnos: ah on the BD
wth it says 9A on the specs I have no clue... http://ecatalog.squared.com/fulldetail.cfm?partnumber=LP1D0910BD
because 9A is the main rating
if the coil is drawing that much, no thanks =)
heh - not likely ;)
the 24V PS is only 4.2A anyway :P
jmkasunich: in the contact itself it says "Continous current: 20 A"
or 4.5 or something
JymmmEMC: I told you ratings aren't simple
there are many ratings for a contactor
picking the one that's appropriate for your load is the hard part
HP ratings, which depend on the voltage and whether the motor is single or three phase
once you know that, it's just a matter of looking at the number
Amp ratings, inductive and resistive
I do see "I <sub>th</sub>: 25A
also AC3 and AC1 ratings
Ith is a thermal rating - that menas it can carry 25A continuously when closed without overheating
don't forget IEC ratings
it does NOT mean it can make or break 25A
AC1 and AC3 are IEC ratings
jmkasunich: gotcha. Just passing on what I see is all =)
AC1 is non-inductive, AC3 is generic three phase induction motor
there are other less common ACn ratings, for different kinds of motors or applications
I hate to soudn so simple minded about this... but bugs the crap out of me when a $1.25 switch on the router itself can handle the load, yet a $50 relay/contactor can't.
JymmmEMC: this is not black and white
of course the $50 contactor can handle it - for a while
the contactor ratings are probably based on a lifetime of at least 100,000 operations under load
the router is soft-start, but not soft-stop ;)
the router has a softstart? didn't know that
I believe so, but cutoff isn't so soft
SWPadnos: It's still the same $1.25 though (in my simplictic mind)
remember - contactor ratings are NOT based on universal (brush) motors like in a router, they are based on squirrel cage induction motors
jmkasunich: Yeah, on the soft-start.
soft start helps a lot
JymmmEMC: try starting and stopping the router 100,000 times using that $1.25 switch and see what happens
see you next October ;)
I suspect that after a few thousand times, the contacts will be burned and pitted, and it won't work so well
oh, and alternate hands or you'll get Carpal Tunnel ;)
but - are you going to switch it that many times? if not, then you can certainly get away with something that doesn't strictly meet the specs
it's only breaking load that matters, due to the soft-start
I really don't know what the characteristics of a universal motor are, especially one with a soft-start
[20:28:40] <jmkasunich> http://pokoj.com.pl/files/pdf/StycznikiEng_WebOpt.pdf
it has listings of AC1 thru AC6a and b with explanations of what they are
and explanations of how contact life is affected by current, etc
that vendor designs for 1 million cycles at rated load
hmmm. being in Poland, I wonder if the prices are lower (still ~3 Zloty / $)
hmmm. I wonder if they've switched to the Euro
no clue - I was interested in the technical info, not the contactors
page 10 has the AC1-6 info
page 20 has contact life info
holy crap. the dollar is down to 2.31 PLN. that's outrageous
that's one advantage of a sliding dollar - now imports are getting expensive too :)
hmmm. woulda universal motor be somewhere between AC-2 and AC-3? (not quite slip-ring, not quite induction)?
that seems reasonable to me
Though I can't tell the contact voltage.... http://cgi.ebay.com/AEG-LS05-LS-05-MINI-POWER-RELAY-CONTACTOR-15-AMP_W0QQitemZ270174662537QQihZ017QQcategoryZ42897QQrdZ1QQssPageNameZWD1VQQ_trksidZp1638.m118.l1247QQcmdZViewItem
sorry, the coil voltage
JymmmEMC: those are no better than what you have
Do any of you guys have a laser interferometer (such as Agilent or HP)?
JymmmEMC: seriously, have you been paying any attention to what I'm telling you
lerman, hahaha - uh, no :)
cradek: ^^ ^^ sometimes you can take a small alluminum rod and with a small hammer tap on the top of the valve seat through the hole under the line coming out of the pump, you ck and see if its stuck up by measuring each hight of the valves
lerman: If you would have asked me 15 years ago, I would have said yes.
I bought some receivers and am looking at some laser heads.
jmkasunich: I'm trying to, really.
I'm curious about faking/making the cables.
just use the larger of the contactors/relays that you already have
if it fails after 10,000 operations, then replace it with something bigger
odds are it won't
odds are also that it will take you a long time to hit 10,000 operations, let alone 100,000 or a million
this is for hobby use, right?
Just curious... when they say 10,000 operations, is a single operation make AND break, or one of each is really two operations?
start and stop = one operation
yes, on the hobby use.
even production would take a while. It's only needed at startup or toolchange
man, I got a headache... and I don't even think it's from you guys either ;)
nyah nyah :)
still getting over this last bit of cold/flu crap still I think
oh - it looks like I won't be going to India ( :( ), so I'll be a bit more flexible as far as getting out there the next time
OH I got drugs right here!!!
you know you're in trouble if you're too down to see the drugs ;)
they were in the laptop bag
SWPadnos: I can't tell if that was a smilie or a frownie.... if it was me, and a trip to india got canceled, I'd be very happy
damn - now there's a precide interferometer: http://www.canon.com/technology/canon_tech/explanation/compact_interferometer.html
this one would have been pretty easy, other than the 48 hours on a plane ;)
precise, that was
the original schedule had us going through Surich, staying there for a day, then going to Mumbai, having a rest day, then two work days (not too hard - set up some cameras, shoot the actress, wrpa up), then a rest day, then the travel home day
2 days work over a week, all expenses paid, and a few $k on top of it - sounds good to me
Zurich, where the heck is Surich?
SWPadnos: That Way --->
hmmm. no, that's the back yard
which is now covered in an extra inch of ice
SWPadnos: Keep going... hoping over the fences, etc
ooooohhhhhh - now I get it
if someone here's from cradek tell to read back thanks
I never knew cradek was a place, I'll be damn
if someone here's from Canada tell to read back thanks
I know I botch the english lang a bit, and I had to read what he said a few times to understood what he meant, but damnnnnnnnnnnnn! LOL
it seems the web brings out the similarities between "their" and "they're", among others
luckily, there are a lot of people standing ready to correct those errors (usually me ;) )
whoa... format 20gb ntfs partition in 15s
write mid-drive FATs, mark whole drive as empty. done
SWPadnos: Nah, vmware =)
only when "format" means "erase FATish things, and don't touch anything else"
just wait for a sync
that should take some time
SWPadnos: no need, all virtual =)
unless you're using actual disk space for that virtual drive ...
osx -> vmware -> xp
ie, when you shut down the VM, it'll write those changes to disk
(or beore, possibly)
SWPadnos: shutdown is pretty quick actually. Though I usualyl use boot camp, yet we dont support bootcamp for users =)
unless you're using a dynamically allocated VM disk, all that data should get written to the physical HD at some point
(even with dynamic allocation, it's dependent on the XP format operation, which usually doesn't erase the whole disk AFAIK)
SWPadnos: I have no clue how vmware works, just know it does a nice job
heh - that it does
I should get it running on my big machine again, and maybe on the laptop also
though apparently even big apps like Altium will run in Wine now (with sufficient prodding)
SWPadnos: Though, since macs now have intel cpu's. if there was a firewire iface to emc, HAWT DAMN!
fsck this paraport or USB crap
I don't know if there's a similar cycle time limitation with firewire
(relative to USB)
SWPadnos: 20mbps iirc
the data rate is 400 Mbits/sec or 800 mbits/sec
ie, 50 Mbytes or 100 Mbytes
SWPadnos: might be something to look into. since it wouldn't be limited to macs and firewire cards are cheap enough.
USB is in the same realm, but the question I was voicing has to do with packet timing, not data throughput
there are servo protocols that use firewire already
SWPadnos: I totally understand
SWPadnos: first link http://www.google.com/search?q=firewire+packet+timing&start=0&ie=utf-8&oe=utf-8&client=firefox&rls=org.mozilla:en-US:unofficial
SWPadnos: I was wrong, it's 30mbps, not 20mbps
looks like an 8 KHz clock unit, which is probably fine for most motor control cases
there we go...
When should I expect to see a beta FW iface in emc?
where aer you reading numbers like 20 mbps and 30mbps? (and is that a bit or a byte?)
3 maybe 4 hours?
2112, unless you make one sooner
come on now... FW has got to be better than all this garbage we got so far
it's no better than a mesa card or equivalent
more uniniversal though.
whatever that means ;)
Well, (heaven forbid) should Mesa cease to exist overnight, then where are ya?
about the same place you'd be when the only vendor for firewire drives goes out of business overnight ...
except that you'd still have the ability to write your own FPGA configs for the Mesa
is that really needed? Isn't there enough in the fw specs so you could do it in emc?
(and there are other vendors of PCI FPGA cards, you'd just need to do a bit of work to change the PIN files for the (open-source) FPGA code
I really am just asking.
you need a piece of hardware at the other end, to talk to the motor drives or motors and all the I/O
sure, just like DV do
RT firewire has the same problem as RT ethernet (I suspect)
but couldn't it just speak DS or the other one (that eluds me atm)
you can't use normal device drivers from RT code
the Linux drivers (probably) aren't realtime, and there are too many chipsets to write our own drivers
Other than the firewire card drivers, is other drivers needed?
I thought a lot of that was in the specs
I'm talking about the card drivers
well, you need to be able to talk to the firewire controller, in addition to knowing how to talk to the end device
they are almost certainly not realtime
the FW card drivers are almost assuredly not realtime
there amy be RT FW drivers somewhere though, like COMEDI
just something I found on google http://lists.apple.com/archives/Firewire/2004/Oct/msg00009.html
imagine what would happen when you plug in a camera though ...
or is real time thread something different?
real time means different things to different people
for us, it means at least 1ms cycle times, ideally with <20 us or so of jitter
ideally, you'd be able to push that to a few kHz, maybe 10
SWPadnos: that link I gave mentions a lot of timign aspects over my head you might want to take a look though
with similar reduction in jitter - down to a few microseconds
that link doesn't seem to have a lot of useful information
most Mac users (though that looked like a Linux-related post) are interested in realtime that's related to video/audio
so if you can pump the data for a frame of video within 15 ms or so, then you're fine
it's only if you miss that deadline that there's a problem, so a few ms of jitter isn't a problem until it goes over the deadline
that's not so for us, (especially if the PC is generating PWM or step pulses - though it wouldn't be with an external device attached)
It's not that emc can't generate the needed pulses, it's the device accepting them properly, correct?
remote deivce == EMC not generating step pulses
otherwise there's not much point in anything other than a parallel card
well, when using a mesa card, what is generating the pulses?
there is no way to make a USB/firewire device generate steps under EMC control (ie, like a parport)
the mesa card would be
EMC tells it the rate needed, then sits back until the next servo cycle
you could do it with software steps, but that's kinda silly :)
ok, and the mesa outputs what? STEP/DIR to the driver?
(also, I think the Mesa digital write routine isn't fast enough at the moment)
or CW/CCW respectively
it outputs whatever you want
assuming someone writes any necessary FPGA code ;)
in a servero/stepper configuration I mean
you have the possibility of outputting whatever you want
you know what I mean
step/dir, quadrature, up/down, clock/data, PWM, PDM, direct bridge drive ...
err - well I may not know what you mean ;)
ok stepper is step/dir, what is typical of servo?
or "analog" - ie filtered PWM/dir
ok, so couldn't some device that takes FW for the input output one of those that you mentioned?
that made no sense to me
oh, not it makes sense
step back a sec
I'm sitting, so just pretend I stepped back =)
there are roughly 3 speeds at which stuff needs to be done for machine control
ok, lean back ;)
there's anothe faster rate inside the motor driver, but we'll leave that alone for now
fastest rate: motor velocity signal generation
that's usually analog, step/dir, or some form of PWM
that needs to run fast, like a resolution of a couple of microseconds is good, faster is often better
next fastest rate: the servo rate
this is the 1ms thread in most systems, and is where trajectory planning is done
ie, where are we, where should we move to, how do we get there (PID)
the next level up is the interpreter and stuff, and isn't really realtime, as long as there's enough processor power to supply the lower levels with command data
so when we're talking about realtime for EMC, we're talking more or less about two things, things that happen on a microsecond timescale and things that happen on a millisecond timescale
(for the most part - obviously some machines will have different needs)
wth me so far?
so, PCs have little trouble with millisecond scale interrupts
you could have max_latecny around 50000, and that's still only 5% of the total 1ms servo cycle
not a big deal in the vast majority of cases
to clarify: PCs have little trouble with millisecond scale interrupts IF they are running a RT kernel
even the -RT kernel patches have little trouble with milliseconds, but those are still RT, even though they aren't RTAI/RTLinux
the microsecond scale is where things get interesting, and this is why there are hardware assist devices around - the Mesa card, G-Rex, various servo cards, USC ...
so, the thing is that EMC generates motor velocity commands at a 1ms rate
(I'll say "for example" just this once - of course we know that can be changed - yadda yadda)
now, let's look at USB
USB and USB2 have a 1ms cycle time
from what I've read, a packet may only go from the PC (to many devices) or to the PC (from many devices)
this means that there is a 2ms effective cycle time for reading positions and then sending new velocity commands
new version of gcam posted
in practice, because the 1KHz USB rate and the 1KHz EMC rate won't be equal, this will degenerate into 4ms round-trip times soem of the time
so that brings our update rate down to 250 Hz, which may not be good enough
(I am aware of the cavets of USB even from a non RT environment)
looking at FireWire, I saw an 8 KHz number in there
assuming the same restrictions as USB2, that may translate to a 2-4 KHz update rate, which is enough
so firewire may be possible (I don't think anyone said it isn't), but there would be a lot of work to get it going
as for latency, the isochronous mode of FireWire may work well
but after all that, I still don't know the max jitter ofa FireWire transaction ;)
oh - the actual point I was trying to make:
the only thing you'd want to put at the end of a FW cable is something that either drives the motor directly, or does basically what a Mesa / USC does - PWM/step+dir/whatever your drive needs
SWPadnos: The latter is kinda what I was getting at via FW. Just connect your drives to the fw iface device and have everything within emc (driver wise) as needed
"everything within emc (driver wise) as needed" is a pretty tall order though
also, making the device wouldn't be a piece of cake by any means
I'm assuming a lot is in the iee 1394a specs here, and only driver really needed would be the firewire card in the PC, all other communication to the deives would be via fw spec
the trouble with these interfaces is that even an isochronous stream only has a guarantee of throughput, not one for latency. it's assumed that you have a shitload of data to transmit, and you want to dribble it out at some rate
um, there's a big gap there
SWPadnos: (think talking to a serial port in the old days, no driver is needed)
firewire is like SCSI
you have to negotiate with the controller to get a piece of the bandwidth, then with the device, etc.
note that FW is not a point to point protocol, it's meant to be deasy-chained or start networked
gah - and star
SWPadnos: I'm making a big assumtion that the driver would be developed by the mfg
so you have some of the same issues as USB - plug in a new device and renegotiate the whole shebang
well, probably so, but as a Windows DLL
with USB and firewire, there's a lot of setup to do before you can even send data to/from a device
once that's accoomplished, the actaul data is up to you, and that's where a "manufacturer driver" would fit in
it might provide functions like set_velocity, which make up a packet and send it to the right device
But if the setup is all via FW specs, once it was created, wouldn't it basically be unified and could be used anywhere?
emc, mach, tcnc, whatever
but drivers are only useful when there's some sort of standard-ish API for software to talk to, which there isn't in this case
in theory, everything works that way
So FW doesn't have a comm protocol already?
what the heck do you mean? :)
look at RS232
everyone knows how to make a +/-12V interface, and how to send bits in the right format (8,n,1 ...)
but that makes no guarantee that any old terminal program will be able to send files to any other terminal program
what if one is kermit and the other uses zmodem?
so the firewire protocol is like the RS232 spec that says "12V, 38400 baud, 8N1 ..."
but once the connection is made, both ends need to know the language being used
well, that would be related back to the device's API wouldn't it?
and firewire just being the tin cans and string
so, couldn't the device's protocol be something standardized?
but the string is a party line, and the RT messages need to be mixedin with the others
jmkasunich: I'm assuming no other FW devices are on the buss
as far as I know, there's no standard "language" to talk to servo drives
dedicated FW if you will
there are several standard protocols that provide the "tin cans and string" part (modbut, profinet, ethercat ...)
modbus, that was
but in the end, I can generically set register #101 to 300 for any modbus device (that has a register #100), but I have to know a lot about the device to know what that will do
man, I need more coffee
ok, back to the step/dir, PWM, etc... would the FW device have to "buffer" those signals, or would emc be able to shove them down the FW pipe in RT?
buffering and EMC don't play well
the 1ms task likes to know where things are, and to be able to send commands based on that knowledgs
(I know buffering isn't the correct verbage here, but it's what I use to describe the functionality of USC, mesa, etc)
also, consider what happens when you hit the feed override - how long does it take to go into effect?
actually, buffering is a fine idea, for systems that only output step/dir (no feedback), and which don't have any requirements for user interactivity (like FO...)
I guess what I'm basically asking is ther epotential for FW to break that 2mc paraport limitation.
2ms isn't a parport limitation
yes, there is potential for that
based on the little I just read
I remember back that PCI was just a clusterfsck to try to attmpt within EMC directly, that's why I was thinking FW as it's specs are more readily available
it's unlikely to be worth it though. any PC that can have a FW port added can also have a mesa card added. ethernet is a better (and less expensive) medium for communications because it's isolated
PCI is fine, though it does take a little setup to use (getting access to PCI I/O regions for example)
I don't know what the RT eth issues are that jmkasunich mentioned.
conceptually, RT ethernet is great. it's just all the work needed to do it that isn't so great
ethernet is more or less a 100mbit isolated serial port
abiet eth far cheaper than FW. FW seems to address some timing in it's specs and can be faster than eth
you just have to divvy up the time well (which is what RTNet does)
bulk data speed isn't an issue
even with 10 devices, each using a "smallest possible" packet in each direction, that's (crunch crunch) 100k packets / second
err - 10 khz update rate with 10 devices, that is
assuming each device needs a completely separate command and response packet
ok, so whats the issue(s) with RT eth device?
writing the code
and making the devices
and writing that code
and then making it all work :)
some of it works already, but since none of us has an ethernet-connected device we can program (open source) for, there isn't a lot of work being done in that area
wget ieee* > SWPadnos
wget rfc* > SWPadnos
wget aieee* > JymmmEMC
wget solderiron > SWPadnos
oh, I have a G-Rex and even the Dynamic C development package for it
but since nobody else does, why bother
I do have a Mesa 4C81 which could work, but that doesn't have enough IO to be really useful
I mean, there does seem to be a market for it, isn't there?
oh sure, there does seem to be
it does seem fairly straight foward, no?
not some conveluted propritary crap
but most of the market is interested in dumb mode controllers - no feedback
it's a real pain in the ass actually
simplisity always is
for one thing, you need a second ethernet port
on the pc? no need to be fucking surfing while cnc'ing damnit
sorry, just a pet peeve
you then need RT drivers for every ethernet chip you want to use - you can't use the normal kernel drivers for the RT port
you miss the point - you can't let the kernel touch the ethernet chip at all - it has to be controlled exclusively by the RT driver code
so you'd need to reboot and move the cable around
the reason you need a second eth port is that all traffic on the RT segment must be RT controlled traffic
lets say dedicated box, no networking
you may not be able to use any switches (hubs only)
ok, I buy that
switches may introduce unpredictable latencies
a point-to-point connection would be best, but that doesn't work so well for multiple devices
4p hub = 4 axis! woohoo problem solved!!!!
I don't think there are any major theoretical issues with RT ethernet, it's a matter of time and energy
3 axis - PC + 3 drives ;)
energy / effort, that is
there are cheap eth chips, that I know
10Mbps should be fast enough, no?
oh, certainly per axis
that's about 1 kbyte per second (or ~600 bytes after protocol overhead)
err - per millisecond
well, if using your hub idea, one eth chip controlling one asix of your choice of output (step/dir/cw/ccw/pwm, etc)
EMC wouldn't care what the output type is
Heh, imagine ARP'ing your devices =)
depends on the stack you implement on them, doesn't it? ;)
SWPadnos: No damnit! phyical layer or bust! no need for a protocol on time of that! LOL
actually.... RS-232 over ethernet???
RTNet uses raw ethernet packets, IIRC
or too shitty?
what about the stuff that HP uses between test equip?
I think the main issue with large numbers of simple devices is that the minimum ethernet frame is 64 bytes long
GPIB is what i was thinking
64bytes is a lot
so when you have 6 bytes to send to a device, you waste a lot of time transmitting useless stuff
is a frame at the eth level?
I'm not sure if the controllers enforce that though - that's to insure that a packet will be on the wire everywhere within a maximum allowed collision domain
there are frames at every level ;)
whats at teh ethernet level?
* JymmmEMC not got ccna yet
ethernet frame contains IP frame which contains TCP/UDP frame which contains data frame ...
there are other protocols other than tcp/ip tyvm
yes, in those cases ethernet frame contains IP frame which contains ICMP frame (for example) ...
we dont need a routable protocol either
[22:20:48] <SWPadnos> http://en.wikipedia.org/wiki/Ethernet#Physical_layer
the ethernet frams doesn't have to contain an IP frame
SWPadnos: Fine.... tokenring!!!
but I used that example because most people have experience with IP
[22:22:01] <SWPadnos> http://en.wikipedia.org/wiki/Image:Ethernet_Type_II_Frame_format.svg
payload 45-1500 bytes
45 being the minimum
err - 46
better than 64
as I said, I'm not sure if the controller enforces that
46 byte payload + overhead makes a 64 byte packet
14 + 46 + 4 = 64
SWPadnos: stop doing the dman math =)
that's only needed if you want to make max-sized networks: 100m cable + hub + 5m interconnect + hub + 100m cable
SWPadnos: MAYBE... use the eth type for < 46 bytes
if you have smaller collision domains, you could get away with smaller packets (if the controllers allow the small frames)
there are no eth types for <46 byte payloads
that 512-bit packet size is part of the physical protocol, to allow for large cable runs
SWPadnos: no no I mena use the eth type bytes as data for anything < 46 bytes
but it may be possible to use smaller packets if you can guarantee that the cables will be shorter
you're missing the point. the ethernet physical layer spec *REQUIRES* that you send 64 bytes per packet. end of story
what you call them is up to you ;)
with smarter controller chips, the CPU doesn't have to do as much work, but it may also not have much flexibility in doing out-of-spec communications
like 30-byte packets (18 bytes overhead + 12 bytes data, for example)
well, the big overhead is tha MAC addresses
and those are pretty much required
the last three octets of each at least, not so much the firtst 3
in a closed system
sure - there's no guarantee that you'll get different numbers in the low 3
you need all 6 to guarantee uniqueness
it would be easy enought to check to make sure initially, then go from there for better timing
even with 512-bit packets, 10mbit ethernet gives you 20k packets per second
is that a good thing?
that's not a real limitation
at a 1ms servo cycle, that lets you send/receive 20 packets per cycle
is that a good or bad thing?
20 pkts @ 16000 bytes, seems like a lot to me (good ting)
um - you couldn't get 20 150--byte packets through
I'm also assuming that the normal turnaround time specs are ignored, since there's RT control over transmissions
lerman_ is now known as lerman