i found a nice splash screen for emc today
it's very "splashy"
[00:43:15] <fenn> http://www.sandboxdp.com/images/jpgs/drill.jpg
i sent the artist an email asking if we could use it
sigh.. gotta uninstall kde 3.4 and reinstall 3.3
ValarQ: you still here?
mshaver is now known as mshaver_away
Woohoo!.. This new toolpost is the shiznit.
Hi all you late night party people :)
asdfqwega: I got the new toolpost mounted.. It is tight. :)
Does anybody no what this error means, I got this while trying to verify a program today and cant figure out what to look for. It's "all axes missing with motion code".
bpmw_: you still there?
do you have a G0, G1, G2, or G3 without a X, Y, or Z?
just my guess
I'll check. one sec.
Hi dan,yes that is the problem the last half of the program all the G1's are fine all the G0's are missing a value. Stupid Cam program. It crashed on me 5 times over the last 3 days.
The funny thing is I ran the simulation in the Cam program and it seemed fine!
anyone here actually use the webersys stuff on the BDI cd?
Hi Weyland, not me.
bpmw: hey there
anyone know anyone that *does*?
Thanks Dan, I'll have to go and fix now so can run in morning. Bye All.
did you get your mill going?
with the help of a few good people
used the sh|t oughta it today
is cutter comp working now?
it appears to be, with alex's changes
THAT was a relief
so you're using emc2?
do you like it better than emc1?
for people like me...
I don't really see a difference, other than it runs my mill a LOT better
or should I say motors?
are they running smoother than emc1?
I've changed it up to use the mini gui
HUGE difference there
I had no idea
that my mill could run so well
let me put it this way -
I bought a small lathe yesterday, for the sole purpose of converting it to EMC use
do you have home switches on your table?
think I will eventually, but not yet
I have a mill/drill too and use dial indicators at end of travel
for peace of mind
so far, it's been really good about that
send it back to point where dial zeros out
I ran some small repetitive parts today, at a pretty good speed, and it repeated all day
right now, I'm slowly getting the mill back together, it's not quite there yet
moved across country twice in 2 1/2 years
I was in that same place, but made the decision to say "fsck it" and just get to the business of business
oooohhhh that sux
used to have nice little side business two moves ago...
just getting my shit back together
well, that's how my shop happened...
work was getting the way of my side work, and starting to cost me money for showing up to work, so I quit
ha. that's a good reason to leave!
so you got a lathe to retrofit
best "fsck you" I ever gave
you're going to do the contouring on your parts with it
in the mean time, could you use your mill with some lathe tools?
bought it jsut for the brass nutz, but happened (more like stepped in shit and came out smelling like a rose) across an upgrade to a larger one, so took the opportunity
chuck up parts in collet and clamp tool holder in vise
I thought of that~!
used to do it when I was gunsmithing...
I've done it a long time ago
made many a firing pin that way
on that same mill
almost 20 years ago
those mill/drill tables look a lot like the tables on Wasino LG-8 CNC lathes
I've toyed with the idea of buying a mill/drill just to rob the table off it
was gonna get one of those little 7x10 mini lathes
mount a Dunham or similar 5C headstock perpendicular to it
but happened across a 9x20 for almost less money
that's a neat idae
throw the column and head in the pile for a future drill press
or just buy an LG-8 on Ebay ;)
yeah, there's that...
which is why we buy the little stuff and do what we can...
check this out: http://mitglied.lycos.de/dieter096/photoalbum.html
that is KEWL~!
did he make the tool changer?
someone on IRC earlier gave us the link
yea I think so
check out some of the other pages, I think he has some design work (maybe in SolidWorks)
hold on,,, trying to read it with my limited forgotten german
[04:48:55] <dan_falck> http://mitglied.lycos.de/dieter096/photoalbum.html
nice... very clean design
I really like what he did with the tool changer
makes good sense
I may just have to do something like that for my lathe
I do have that extra stepper...
a solenoid engaging a locking pin ...
mount it all on the cross slide...
oh crap.. .who am I kidding?
I have too much work already
yes, he has an incredible toy there
I think the mill/drill table loaded with tool blocks would be the easy way
we have all kinds of different cnc lathes at work
turrets, tables, multispindles
truth be told, while that way is simpler, his way really makes life easier
the little LG-8s do pretty well
I used tobe around them, but not for a while now
the little table is actually faster
you can set up the tools really close to the work and come in from either side
I mean, yeah, it's *faster*, but you really can get limited by having so many on there
that tool changed makes life *SO* much simpler
but, you could probably mount 4 or 5 tool blocks on a table
sure you could
the turrets sure are cool though
but space becomes a premium, and coding has to work within that emvelope
the turret gives so much more flexibility
oh, gods, the wheels are turning...
we've got some machines (Sl-15s) that have one turret perpendicular to another
vertical movement to align with the chuck?
so the one on the front has it's rotational axis parallel to the chuck
the other is perpendicular to that
how does it move between the two?
up/down, or forward/back?
well they actually are mounted on their own sets of XZ ways
the chuck is in the center
the two different turrets are controlled by two different controls
they hand off motion to each other with M codes above 100. ie M100 etc
man, that guy's done some real work over there...
it gets kind of hairy looking at two programs for one part
you see his headstock and collet set up?
on which page?
hold on, lemme get it
[05:03:20] <weyland> http://mitglied.lycos.de/dieter096/photoalbum0.html
made the tool holders on his lathe
* Phydbleep still wishes Babelfish would translate those pages.
looks like he's making a mill spindle with toolchanger
well, my german is pretty fsck'n rusty, but I'll try if you want
phew.. finally got kde all stitched back together
Actually I wish he'd move that off Lycos which should solve that problem
phydbleep, babelfish worked for me
er, googlefish or whatever actually
[05:07:28] <fenn> http://www.google.com/language_tools?hl=en
wow, neat stuff
so what does everyone use to design their parts?
still looking for a good (free) cad program
fenn: thanks for the google translator url
actually, I used the qcad thatcomes with the BDI today, and wasn't too disappointed with it
it aint solidworks, but its okay
qcad's alright for what it is
it crashes sometimes, so save yer work
don't really understand the synergy thing
try out the demo/tutorial movies
have you used it?
like it? hate it?
no, i've only watched the demo movies :)
it looks cool from what i've seen
yeah, that seems to be the problem, too
can't figure out how to use it though
"looks kewl" but I can't find anyone that has been actually USING it
I have no idea how to use it either
I keep trying to get SW to run in wine...
websys: are you logging?
websys = synergy's representative ?
websys is the programmer of synergy
take a clue...
it's probably a great system that none of us understand
I didn't find it nearly as intuitive as most other stuff I've tried
but then, so was smartcam...
yes, I remember smartcam
for its time
was YEARS ahead of everyone else
they're back now
Really. No shit? Wow.
the original guys got control of it again
pro'lly too late, tho
they're just down in Eugene Oregon, south of here
I mean, everyone is out there, now
I still have a copy of it
SmartCAM 10, I believe...
haven't seen it in years, but it's here, somewhere
er, nevermind .. that's the current version
we used to use the DOS version
wow, you're OLD
huh? you mean people actually used DOS?
I remember dos autocad
I don't think they ever had a C64 version ,but I might be wrong
those were the days
so what do you use, dan?
I use Vector on winders here at home
how do you like it?
we use paper and pencil at work
they have a linux?
no linux version , but if you buy win4lin, it will work under linux
the only reason I have a win machine left is to run sw
you can probably run Acad under win4lin too
I hate acad
left it in 2000
or maybe SW
would LOVE to run sw in linux
I think the guys who make win4lin are called Trelos
hmmm... lemme go google
it's been about 4 or 5 years
windows 98 ran faster and more stable under win4lin than it did on it's own
which was kind of weird...
dan_falck: Yeah, I noticed that with VMWare/98 and WineX/98. :)
$120 for the Pro
if it ran SW, I'd pay it
gonna send'm mail and ask
weyland, why exactly doesn't SW work in wine?
fenn: I really don't know, I'v been working with others that know wine, sending them the errors and logs after we try something different each time
are you a wine guy?
maybe you should send me a copy of SW and i'll see if i can get it to work :)
never used wine in my life
I have used it with Rhino in the past
but people keep raving about solidworks and i wanna see what the fuss is all about
but my computer was too slow
fenn: the only reason I have a win box is to run sw
that's something to be proud of
and that's all that box does...
weyland: It's still 100% win?
Phydbleep: as far as I know, yes. there are always rumors and such, but I've never seen anything real for th linux community
* Phydbleep has thoroughly lobotomized his XP box with XP-Lite, LiteStep and Cygwin...
weyland: No, I was talking about the pc. :)
* Phydbleep has 4(I think) of the 87 required windows components left. :)
poker. heads up. Oppponent has KK, I have A6 suited. Preflop, all in. comes up something, J6, J, 6. Full house!!! Won the game with pure luck.
IE gone, Outlook gone, Messenger, Media spayer, Windows Update.. All gone. :)
[05:38:21] <A-L-P-H-A> http://www.nliteos.com/
speaking of bloat removal, i switched from kde to icewm today
suggest you try it if you've only ever used kde
time 4 bed, thanks for the insights boyz
what's the point of a slant bed lathe?
fenn? uh... makes good cones?
chips drop onto a conveyor and don't cover ways
chip removal is a big deal with production
oh!!! production slanted ones.
i thought maybe it was stiffer or something
I know whatwe're talking about now... instead of just talking out of myass
talk to you guys later. gotta go
* anonimasu yawns
got woken up by customers
really? bastardss eh! ... bet they wanted to buy thngs too?
just support :/
they had some questions :)
does it work?
[ yes | no | sometimes ]
did once, but we smacked it on a tree?
yes it does
but I have some changes/improvements to make
I thought my solution was better
but, there are probably laws agaisnt dangling children by their ankles with a bag of tree seeds in one hand and a spade in the other.
10 of them in a row on a big frame ...
they would still suffer from the not working if you smack them on a tree stump bug though
yeah and kids wont work 24/7
they run 3 shifts with the machine today
proably need some sort of autochanger then
well they have a crane ;)
but I doubt that will do
is it out there planting now?
nice. how many units?
one right now
and a spare?
I got the other one on work waiting for the customer to puck it up
so I have access to it today and tomorrow
hehm I learnt someting on the gecko list
mount geckos vertically, with the blue connector at the bottom
common cause of failure is bits if wire and stuff ending up on the board, shorting stuff out anc causing fires
oh, isnt that common sense?
I just cant remember which way up they are in many of my machines ...
we finished the Bridgeport conversion this week too.
removed about 2m^^3 of cabinets from it ...
did you get emc to run on it?
replaced with a small box containing 3 geckos a toroid and a Baldor ESB
nah, went for a Baldor inthe end
we needed to do a lot of contouring work
was not convinced the emc TP would cope
the baldor was like, 500 quid
how many $ is that?
has all the IO and n channels of stepper or servo
too bad.. it looks like the tp will work wonders soon
things are moving on
but this has to make parts this week
what speeds do you run it at?
not fast ... 1200mm/min // but paths made from oodles of teeny segments
stopping is not an option
how fast do you run your spindle?
ahh, "subject to ongoing experiment" :)
so you havent milled any TI yet?
some, test peieces just to see what happened
I thought you needed 14000rpm and slow feed to get it to cut nicely
that would kill it
it work hardens
you need to keep it cutting, dont stop
if you let it rub, then it hardens
yeah I know.. but alpha was talking about it earlier
with CNC it is supposed to be easy ...
we know a firm here in the UK that does noting else
chatted to their machinist
easiest metal to work once you get it right ... machines beutifully, but people like to spread scare stories about it, thats OK, we wont stop them :)
or words to that effect
hope it is easy, or we hav problems :)
but I guess you still need a super rigid machine to cut it..
well, the bridgeport is up to it.
and its oh so smooth with microstepping
oh, I took some photos of the toroid winder as we wound the txfmr for it
now, how to get them of the phone ...
hm, I guess I should go to work and code some so I can take tonight off
I am waiting for the parts I have to make to get finished so I can re-build my mill
now I have managed to persuade the laser to cut 8mm steel, I want to try 10mm
how hard can it be?
I keep thinking af the stuff I can make with that ...
10mm plate means you can make *strong* welded assemblies. A cnc mill would be a possibility
I thought you could cut heavy stuff with your laser..
Woohoo!.. This new toolpost is SWEET!
* Phydbleep has been playing with test cuts with the new toolpost.
8mm is the most ive done ... it cuts perfectly
well go for it
Phydbleep: drop on tool holders?
* anonimasu got a toolpost a bit ago
one thing I do know .. you can never have too many tool holders for it
6 is good, 10 is better
4 is not enough
anything is better then a normal toolpost
I find out you wont use that many in real life..
4 was a pain, 6 I find I occasioanlyl have to swap, 10 would be ideal
got 7 at work.. it's enough
I usually swap the small boring bar to the larger one though
robin_sz: Groz type.
but thas not that often
[09:32:38] <Phydbleep> http://homepage.ntlworld.com/v.ford/gt.htm
I need to order a boring head soon
got it, looks neat
that's boring ;)
With a 3-way carbide cutter I can surface like mad.. Maybe 5 minutes with emery cloth to get good seals on steel.
robin_sz: All steel. :)
* ValarQ is searching for the reason to his 'joint error'
did you try changing max_accel, and max_vel ?
ValarQ: Wet papers?
ValarQ: just remebered
mine has a Dickson T2 toolpost
you had ferror set for inches
ValarQ: now for mm that's incredibly low
ValarQ: Never mind.. Bad joke. :)
ValarQ: try increasing ferror and min_ferror
[09:37:27] <robin_sz> http://cgi.ebay.co.uk/ws/eBayISAPI.dll?ViewItem&category=633&item=7521391984&rd=1
ValarQ: min_ferror = 0.002 (that's a bit low in mm ;)
make that 0.5 or so
did you guys hear about keyhole?
alex_joni: i'm testing 0.127 now
WTH are the triangular carbide cutter chips called?
I just call them tccc's :))
eBay is a good source
I guess they fit easily in a pocket at the end of a shift
robin_sz: Then send me some inserts and a set of holders. :)
robin_sz: you wouldn't :)
[09:44:04] <robin_sz> http://cgi.ebay.co.uk/ws/eBayISAPI.dll?ViewItem&category=64819&item=7520341893&rd=1
[09:45:52] <Phydbleep> http://www.metalwebnews.com/howto/toolholder/toolholder.html
That's almost the same as the one I have.
[09:46:33] <robin_sz> http://search-desc.ebay.co.uk/carbide-inserts_W0QQcatrefZC6QQcatrefZC6QQcoactionZcompareQQcoentrypageZsearchQQcopagenumZ1QQfromZR10QQfrtsZ50QQfsooZ1QQfsopZ1QQftsZ2QQsacatZQ2d1
loads of inserts ...
given that you can often pick them up on ebay for like 30 dollars .. its easier to buy the real thing ?
robin_sz: did you check out keyhole?
[09:50:25] <alex_joni> http://www.keyhole.com/body.php?c=popup&h=support&t=coverageListDetail&map=europe
KH-x series satellite images.
you need to install some SW
unfortunately not free
ValarQ: any luck?
but 7 day trial
alex_joni: nope, min_ferror=.1 now
try even higher ;)
.5 is up
ValarQ: also you might test ferror higher values
and there's deadband
alex_joni: i'm using ferror=.5 now
and min_ferror ?
looking at mill_mm_freq.ini (that's mm settigns for the minimill)
MAX_VELOCITY = 36
DEFAULT_ACCELERATION = 60
MAX_ACCELERATION = 60
FERROR = 26
MIN_FERROR = 0.260
i don't have any mill_mm_freq.ini
well.. it is included on the BDI4
but try the values I posted
btw. the ini's are in /usr/local/emc/*.ini
03paul_c * 10emc2/src/emc/kinematics/trajectory.h: Header to provide new trajectory planner. Not all functions implemented, others are just wrappers around new functions.
alex_joni: i'm trying to run the 3d_chips file from nc_files
i changed the MIN_FERROR to 2.260 and it seems to work a lot better
robin_sz: this is me: 45�44'24.85''N, 21�12'40.00''E
ValarQ: iirc MIN_FERROR is the ferror for speed
cradek: are you there?
03paul_c * 10emc2/src/emc/kinematics/trajectory.h: New functions declared - The old ones will go as soon as the required changes have propagated up to the top level.
tryng to test emc with chips3d i get a nice penguin but is too small, can i get it in a bigger size ? depend on tool used maybe ?
i guess the length is hardcoded in the ngc file
output_scale in the ini-file might provide a solution but it's not right (at least i don't think it's right)
i'm confused, because..
on the top of the file i read the part is cut from 100x100x50 mm
the size i get, instead, is about 15 mm x 30 mm
the diameter of tool i used is 5,6 mm, can i specify it on tbl file using mm units ?
it depend on what i choosed as units in .ini file ?
i'm not quite sure :/
ok :-) np
i have only been playing around with emc2 for a couple of days
it was a little challange to change everything to mm
* anonimasu yawns
ValarQ: I've been working until now
how are you doing?
I'll be getting engraving endmills soon
anonimasu: there was more vars to change than UNITS
Trying to find a flight to Romania.
heh.. coming over?
I think that I'll only make it to Germany.
Have you thought any more about the menuconfig system and EMC?
ValarQ: I know :)
I've got engraving endmills now
err engraving cutters..
rayh: I prepared it for now
now I just need to gear them 1:8
maybe later I'll add the menuconfig stuff
err gear the spindle
still thinking if it's not too much for emc2
so far we only have 3-4 modules
so it doesn't really pay off to select which to compile
when there'll be more than 10 drivers it'll make sense
rayh: when will you be in germany?
July 18 -> August 8 or so
steve_stallings is now known as steves_logging
I'll be there in september
Oh. Maxine has to be back for school then.
yeah.. going to essen
there's a big welding fair there
Okay. Where is it to be held?
Been by there. Not stopped.
So Kbuild or whatever is not going to have very high priority.
Saw some of your numbers on shmem -- impressive.
numbers... Where ?
seen your commits today
care to tell me what to do?
Ah, right - The gross waste of shmem and those &@!% structs
that's not correct paul
it should read those &@!%_t structs :)
alex_joni: You could start on the hal refactor that jmk has been promising....
Then there is the question about the ability to link bool pins to int signals
paul_c: I heard you say that the status stuff a waste. How would you change it?
paul_c: I'm afraid I don't know enough about HAL to do the refactor
but I've seen zwisk has done some work in that area
rayh: It's not the Status data that is the problem...
It's the mass of "debug" data and tp/tc stacks in shmem
moving the tp/tc stacks out of shmem will save some 152K straight off.
How would we make a diffferent debugable rt.
First question - Do the tc/tp stacks contain any usefull debug info ?
short answer - No.
well, int DEBUGINFO;
* alex_joni will be back later
is it a waste to try engraving anything at the wrong sped?
I'll read the logs
paul_c: if you want me to do some work, let me know
drop me an email or smthg
Isn't it possible to build emc without debug?
In it's current form, no.
What would we need to do to make that possible, or is selective removal a better approach.
paul_c: how are things looking with the tp?
03paul_c * 10emc2/src/emc/kinematics/trajectory.h: Add some doxygen & LaTeX comments.
SWP_Away is now known as SWPadnos
rayh: First, identify what is Status, what is Debug, and what should never be seen outside of the code.
'eck - They're all turning up now. !
run paul run...hide
Is it possible to have a separate NML channel for debug info?
which may not get used
Once you have separated out what is what, yes.
For example, does knowing what the Nth double in the tc stack help debugging ?
How does a person download an anon cvs from MS xp
use turtleCVS - it's very nice.
or would it be better to know how full the stack is ?
rayh: Install BDI over the top of M$
any information helps debugging - it's a question of how much :)
If you're really keen on debug info, use the printk ring buffer.
Turtle is it.
Hell... If you really wanted to, map each and every variable to a hal pin and use halscope or halmeter to monitor them.
well - that is one option, but it sure adds a lot of code to the modules ;)
Not if jmk made pin/signal registration a library function instead of inflicting it on every module
the trouble that neither of us have been able to get around is type checking in a struct that could represent a HAL pin of any valid type
type checking at compile time, that is
so treat everything as a single data type.
but then there's no type checking at compile time ;)
there is already a function to create a pin of any type, but it's not as safe to use because everything has to be cast to oid*
Two data types - Floats & ints. avoids all that lark with hal_type_'riggin_foo_t
well - how about hal_type_'riggin_foo (I know how you hate _t )
struct_t is only part of a larger issue with the code.
Got the package, thanks Matt.
I got up early for this, took care of some stuff, then fell asleep...
oh pooh ... one of the servers inthe rack seems to have gone off :(
good! I'll bring your tools to roland's place
as soon as i can get steve there
i want to be there mon morning, but steve says we leave mon note
Oh. I thought he had decided to fly on thursday for the second weekend
if so, i'll get there with you
I plan of Friday night.
i'll tak to steve & see what he wants to do
did i miss the main discussion today? kinda quiet
There was a bit of talk about tc/tp and debug structs.
Not much about my tasker.
* mshaver is getting coffee brb
hmmm .. I wonder if there is any *really* good job shop software .. you kow, order tracking, attach drawing files, schedule it for cutting, produce production plans, print delivery notes
I know a shop that installed "jobshop" and they had to add a guy to manage it.
yes, but it costs $3 million or so :)
no one pays that much anymore
Sugar CRM gets close ...
and its free
but not quite close enough
do you need full ERP?
* mshaver is back
mmmm. probably not
thats the full materials ordering, machine useag planning etc etc?
yes - "Enterprise Resource Planning"
people, machines, power requirements, conference rooms - everything
ick .. no
1 laser ..
you may just want to use something like a project manager
SWPadnos: i can probably use SugarCRM, and assign each job as a case, but its not perfect ...
ugh - that looks like it's a replacement for the venerable Telemagic...
ie, useless for getting actual work done, but great for contact management
look here: http://www.compiere.org/
yeah, its great for telesales
don't know if it applies
Compiere I already rejected
far too complex
XRMS was rejected too
that probably can do exactly what I need
but its totally undocumented
nummer that Paul and Alex aren't around now :(
where are they?
Compiere probably can do it .. but its absolutely huge, needs a J2EE compliant server etc etc .. a bit like SAP I guess
Paul just left (<1 hour ago)
Alex a bit before that
anyone have any ideas about the task / IO / motion emails to the dev list?
not at the moment
Generates about as much enthusiasm here as there, eh.
anyone have experience using dlopen() and friends?
been straining my brain trying to fix a problem with stepgen
ah - what's that?
if you send it a move that is right at it's accel/velocity limits, it generates some overshoot at the end of the move
I suspect that is behind some of the following errors people have been reporting with emc2 and steppers
sounds like a math issue
stepgen generates overshoot? that seems wrong
its a nonlinear thing....
if you send it a signal that is well withing its limits, it tracks it perfectly
actually, I think you guys have some good ideas - I just need more time to digest what you've said - but I do think that you have identified an area that needs the type of improvements suggested in your postss
if you send it a step, it generates a smooth move that respects the limits and doesn't overshoot
should that type of thing be handled in a PID or limiter module?
but if you send it something that it can almost track, it suffers from a kind of integrator windup effect and overshoots
and stepgen just makes steps at a high RT rate
well, that would be one way
PID plus freqgen
(freqgen takes a frequency command, needs an external position loop)
PID on the non RT side?
but when you do that you need to tune the PID
no, RT PID
or just accel limit -> stepgen
so I came up with stepgen, which takes a position command directly, and doesn't need a PID loop at all
(PID tuning is a continuing gripe for stepper folks - it shouldn't be needed if you are running steppers)
trouble is ... users seem to have no trouble setting up v limits and accels, but alwyas seem to come unstuck with PID and steppers
exactly... so stepgen uses the vel and accel limits only
works great for things that are well within the limits, or that exceed them by a lot (like steps)
but an input right at the limit generates an overshoot
nonlinear stuff is like that
you mean a postion command that moves right up to the limit of an axis travel?
time is discrete, position is discrete, accel and velocity limits make it non-linear....
matt: no, postion limits aren't a problem
just velocity and accel limits
speaking of which I got some following errors on the laser today
trajectory planning problem maybe?
what EMC are you runnning?
turned out some bit of it was smacking into another bit ... solved with a large hammer :)
matt: nope... (at least I don't see it that way)
the stepgen vel and accel limits are coming from the ini file, so they are the same as the traj planner's limits
that means that rapids generate moves from the TP that are right at the limit
which is exactly where the stepgen hasproblems
weyland did report that his machine was much smoother with emc2 than emc1
if the tp already applies limits .. why does stepgen apply them again?
well that'�s true
emc2 is great!
yeah. I'd like to think that emc2 has something to do with it, but in reality I think he just had poorly tuned PID
surely if the TP is doing its thing, everyting going towards stepgen should be "right" anyway??
robin: because 1) the TP may generate s limited signal, but then you add backlash comp and possibly other things to it, and 2) stepgen is a generic module, and I insist that it be able to work standalone
How much overshoot are we talking here?
varys with the config, scaling, etc
a few thou mostly
With how many steps per thou?
enough to sometimes generate a following error when FERROR=0.020 and MIN_FERROR=0.005
several thousand steps/inch
it also depends on the move
jmkasunich: sounds like your PID + freqgen notion fits that ideal better, then with a proper TP that didn;t exceed its own limits, you could unplug the limiter from stepgen if you wished
the worst case seems to be a move that is just long enough that it accels, and immediately begins decelling
longer or shorter moves have less overshoot
robin: PID for steppers is simply evil
jmkasunich: I agree, but it seems inherent i9n emc
better that I strain my brain for another week and get stepgen working right, than have every user strain their brain trying to tune PID for steppers
could it be rounding errors that accumulate as steps during the ramp?
inherent in EMC1, not in EMC2 ;-)
ray: little more complicated than that, but the fact that position feedback is in discrete steps is certainly part of it
so backlash adds non-limited moves to the traj?
the TP knows nothing about compensation
comp is simply a short move at max velocity upon direction reversal
the TP shouldn't know about lash, IMO
need to go in before the TP to my mind
that gets ugly fast with non-trivial kins
lash is in joint space, not tool space
of course, TP in general gets ugly with non-trivial kins, that's why we have things like the mixed angular/linear move weirdness
well, not many people use non-trivial kinematics anyway
part of the problem with stepgen is the different sample rates
the only reason there was pid with steppers is because i could never figure out a way to "jumper it out" in emc1 - but john is right, there's no reason for it to apply to a stepper system
always made my brian hurt
you step, therefore you move...
yet something is needed there... not only to deal with signals that might exceed accel or velocity limits but also for things like steps being deferred because of direction setup and hold times and such
infact, I'd go as far as to say it was one of the things that put me off peresevering with emc on the bridgport
I wonder if it would make more sense to do everything in "unit space", then convert to step units only at the end
EMC2 does everything in unit space...
the input to stepgen is in position units, and gets converted to steps inside stepgen
I see pos_err in steps, not units
that is a key HAL issue... HAL signals are in machine units, inches, mm, whatever
theres no reason for anything outside stepgen to know how many steps per unit I guess
pos_err the stepgen parameter?
right - it may need to be converted later in stepgen
I've maintained for years that running emc near the upper bound prevents any accumulated following error to be fed back into the pulse stream.
remember a couple weeks ago I asked you about traj limts vs. axis limits?
I'm looking at the update_freq function
my thoughts were that traj_limits should be a few % lower than the actual axis limits, so that you always have a tiny bit of headroom
several of the stepgen parameters are in steps, but they are used for testing (halscope, etc) only
Yep. I think that I suggested that in the old FAQ the Henkka maintained.
the actual pins that interface to the rest of the system are floats
yes, but the internal PID parameters look like they're being based on steps
(or PID-like parameters)
yeah.. the input is converted to steps, but it is still done in floating point, so its not quantized
probably to allow them to use no-lfoating-point math?
it uses FP
weird, so no speed advantage there then
update_freq runs once per servo period, and uses FP to calculate the proper output frequency for the next servo period
make_pulses runs in the high speed thread, it actually generates the pulses and is all integer
it lloks like that's wher ethe problem would be (plus there's a FIXME there :) )
the FIXME doesn't matter in this case... the default accel limit (which is too high) is being over-ridden by the limit from the ini file
the accel is calculated in step space, then converted back to unit space (sort of)
I suspect that the speed of execution would go up if the step conversion were held until the end of the function
perhaps... but right now I'm more worried about correctness
besides, accel isn't converted back to units
accel is in steps/sec^2
yes - correctness may be easier to check if there are no unit issues during the calculations
used to generate freq, in steps/sec
and freq is used to generate addval (an integer)
addval is added to an accumulator in the make_pulses(), every time the accum overflows or underflows we generate a step
correction... freq is used to generate newaddval
the high speed thread ramps the old value of addval to newaddval at the max accel rate
IOW, it even does ramping within a single servo period, it doesn't step from one frequency to the next every servo period
right - it's doing an integger derivative per period, and a bresenham error calc for the pulses
I prefer to think of it as direct digital synthesis rather than bresenham, but the two are somewhat related
heh - OK.
it's identical to a Bresenham line, where time is the X axis and rate is Y
true DDS would be if the high order bits of the accumulator went to a sine lookup table
instead of simply looking at the highest bit and generating steps
actually, steps/time is the Y axis
X is time, Y is position
sure, since position is directly related to steps output
the pisser (and one of the complications) is that you can't just use the accum directly
because you could have direction reversals too close together
assume a small positive velocity (only a few hz step rate)
reversals that want to violate dir hold time?
the accum slowly increases, and finally overflows just at the end of a servo period
on the next servo period, we command a small negative velocity
the accum just overflowed, so it has a small number in it, and the small negative vel can immediately cause an underflow and a step in the opposite direction
right - you get an immediate pulse in the other direction
so instead, there is code that enforces direction hold and setup times
03paul_c * 10emc2/src/emc/kinematics/trajectory.h: Some more fun with Doxygen & LaTeX.
the bad news is that because of the setup and hold times, not every overflow/underflow immediately generates a step, so the accum is no longer usable for position feedback, you need to actually count the generated steps (in rawcount)
what I would have preferred to do is make the accumulator 64 bits
and use the whole thing directly for position feedback
high 32 bits counts steps, low 32 bits counts fractional steps
avoids quantization error on the feedback, and simplifies the "PID" loop
you already have that, since the current accumulator is the fraction
sort of... accum is fraction, rawcounts is integer part, but I can't combine them into a double value safely
both because of atomicity issues, and because rawcounts isn't always inc/dec'ed on every overflow of accum.
it's depressing how much step/dir uglifies the code.... all the other stepping types are so much cleaner
jmkasunich : Did you see what Mariss is pondering?
you mean the "unstallable stepper"?
or the one wire signal scheme?
yeah... I don't like it
not good at all for software generated signals
I like it, except for the noise potential
good luck meeting the 2us/4uS criteria - you'll wind up busywaiting
ok, that too =)
it works if you are running a dedicated micro, but not on a PC
I'd rather see either quadrature or up/down signals
Well, since he actually does have a controller board now, that's probably why.
* Jymmm just got done installing 6 fans into my driver case =)
jmkasunich I'm not familure enough with methods other than STEP/DIR to even make a smartass comment.
Good grief Charlie Brown..... Matt is here !
amazing isn't it
ohhh, and not even lurking or away from the screen ;-)
* fenn lurks
just looking at this Kmotion trajectory planner
les: You have LaTeX installed ?
Ho Hum... Guess the latest round of commits won't be readable for you then.
les : http://www.deviantlatex.com/
well I can browse cvs
* paul_c takes Jymmm to one side for a quiet slapping. ;)
paul_c : Hey, it's good stuff! Make for some interesting / creative photography.
maybe a little duffing up too.
So on this card the emc interpreter and the trajectory planner send up to a 40,000 segment motion queue to the dsp via USB 2.0
paul, I am looking on cvs but don't see anythin new that you are doing
you have to use your login - the anonymous server is delayed
hmm this Kmotion planner looks like segmentqueue junior
Les and Paul (and anyone else interested in motion smoothing) - have you seen the papers at http://www.cs.wisc.edu/~deboor/usreadme.html
paul_c (btw, the soda can worked out good for shim stock - thanks!)
cool. the smooth and splerr papers seemed interesting (until I got past the word "Introduction" :) )
SWPadnos: Smoothing & blending isn't the real problem.
cubic & quintic blending is more than good enough for machine tools
I thought that smoothing the jerk was one problem, and getting a reasonable curve from discrete segments were the problems
"reasonable curve" is part of trajectory planning - i.e. look ahead.
yes - and getting the max accel and vel for following a path
much of the time, you don't want to be applying max accel
most of the time accel is very small with a smooth curve
if the programmed path has sharp corners, then I expect you will want to use max accel to go around them with minimum rounding
constantly applying max accel means unbounded jerk
V^2/r where r is the local radius of curvature
here we go again
les: exactly... it the path is truly smooth (meaning you use arcs at every intersection of non-parallel lines) then you will have moderate accel
but any intersection of non-parallel lines means zero radius and infinite accel
surely "constantly applying max accel" would be no jerk at all?
*suddenly* applying ...
inexact path mode, you have no choice but to stop and start again in the new direction, probably at max accel
robin_sz think about changing direction
in contour mode, you accept that the corners will be rounded, but you want to minimize the rounding, so again, max_accel
but you dont have to apply max accel from the start ...
I was referring to a calculation to determine the max accel needed to follow a curve at the programmed speed
so that the feedrate could be reduced if this exceeds the machine capabilities
that where you use les's V^2/r formula
well v^^2/r , what les said
* robin_sz nods
Having bounded jerk would of course require another param in the ini file.
I think jerk limiting is pretty much "expected" these days
it's a little different when the path is a cubic or quintic - you have to calculate a tangent line or equivalent to find the velocities and accels
paul: another param in the ini is the least of the issues with jerk limiting :-(
well to me bounded jerk=quintic
yep - possibly quartic, but quintic should definitely handle that
as in limited jerk, not zero jerk
(that almost sounded like I know what I'm talking about :) )
There are other algorithms that limit jerk.
you can never have zero jerk
zero jerk is impossible
not if you want to move at all
* robin_sz nods
that's fallout from the fact that this deals in discrete time
quartic and all even polynomials have asymetric blending problems
ah - OK
les: sytax error, too many big words in one sentence
that's why quintic is the lowest order with bounded jerk
discrete time also means we never have "infinite" jerk as well
and they tend to get instable (so do polynoms higher than 5th order)
robin_sz: they don't always work good.
SWPadnos: ah right ;)
jerk limiting is the key to modoern high speed machining imho
jmkasunich: I'd look at it the other way - discrete time means that we always have jerk - it's only bounded by the range of the variable used for acceleration
(data type, that is)
at bridgeport speeds, you can get away without it ...
but on somehting big and nimble ...
alex_joni: I'll dig up a card & send it to your cousin tomorrow
les: you going to fire up that controlelr card 2mrw?
I have to get some motors
small ones will do
I presume it hasn't got the capapbility to work with external drives?
I am looking at the software now
Yeah I think it can
but it is not a cnc controller out of the box
* alex_joni reads the logs
however it can read g code
but it needs some macros to make it useful?
but only a few
this was less than 1K?
sorry, distracting les ...
back to jerking on emc
yeah 699 quan 10
* robin_sz nods
[17:56:38] <les> http://www.dynomotion.com/Help/index.htm
ooh, it can drive steppers as well as servos ..
it has how many output channels?
uh let me see..
4 or 6?
so 4 DC servos or 2 steppers
yeah but has 8 3 amp h bridges
oh, 8 bridge drivers
Eight Full Bridges are controlled ...
thats 4 steppers
meant for 4 axes
servo loop for all axes is 90 us
wow ... 600mflop DSP
It ain't slow
that's $399 of the proce :)
ISTR a cray 1 was 400 mflop
that puts things into perspective
1.2 GFlop, I think (though I may be thinking of the X-MP)
Anyway the first things I am looking at is what else is needed to turn this thing into a cnc control
has plenty of Dio
'76 cray 1 160mflop
yeah, but crays make nice seats afterwards :)
that is a problem I think
back in a couple hours
jmkasunich is now known as jmk_away
why would 512K be a problem
well that means it cannot hold many full programs
I doubt it will hold any
the flash will be for the OS on the card
well the emc interpreter and the TP are on a non rt host box
I suspect that means the "programs" are on the host then
sorry, I distracted everyone from jerking on emc
noyt sure whether the 40,000 segment queue is in flash or what
but it is on the dsp
rememebr flash is slow, and dies if you use it
its dead at around 10^^6 write cycles usually
ok so the 16 meg SDRAM
nice feature that you can talk to DIO within a g-code program using special comments
The source code for the planner etc is on the site
much easier to follow than emc
it uses an emc.var and emc.tbl as well
SWPadnos is now known as SWP_Away
wow .. they kept quite a bit of it then
and keeping the interp and task open helps integration
It doesn't send position coords back to the host does it?
I wonder if that would slow things down
Good afternoon all!
Would anyone like to take a crack at a g-code guestion?
I made a test program to mill some different size holes in a bloch of pvc. I went to verify the code spit out by BobCad. And I get an error message that says " all axes missing with motion code"
I'm just learning to code and I'm not sure what to look for!
alex_joni. I gave your message to Weyland and he said to "THANK YOU, THANK YOU, THANK YOU".
No takers Eh, Ok thanks anyway. Catch you later!
Weyland, how are you at g-code no one wants to take a crack at my question.
I'm just a clunky pencil paper kind of guy
what's the quest?
[13:42] <bpmw_> I made a test program to mill some different size holes in a bloch of pvc. I went to verify the code spit out by BobCad. And I get an error message that says " all axes missing with motion code"
I bet that means there are some lines with G0, G1, or G2, that have no X Y Z A B or C on them
that's exactly what I'm thinking
ie, Move at this speed, but I don't know where to go...
how many lines of code is it
... spit out by bobcad ... - probably large :)
Yaa pretty much all of them, it uses I and J
there still needs to be an X,Y for a circle, I think
i,j show the center point
yeah, it's gotta know where it's going to and coming from
x,y is the end point to move to
Theres 349 lines.
from is easy - it's "wherever I am" - to is easy, it's specified as X,Y, but with a circle, you have to know how to get there, and that's what I,J are for
I and J only tell the distance of the origin of the arc from the place the move originates from
Obviously BobCads emc postprocessor doesn't work!
Ok I guess I'll have to do some research, i thought it might be something simple. Thanks guys I appreciate your help!!!
Bye for now!
* alex_joni is back
cradek: you here?
get me mail?
how's it going?
oh shit... http://news.yahoo.com/news?tmpl=story&cid=581&e=1&u=/nm/20050605/tc_nm/apple_intel_dc
the one from 03.06? yes
fighting python right now
weyland got prybar?
weyland: sounds fun :)
weyland ; 'got jaws-of-life?'
lol, no ... and I'm not a programmer and don't play one on tv
nor have I slept in a holiday inn express
trying to use one of Chris
weyland python == big ass snake!
trying to use one of Chris's programs
and I can't get the thingto install
python setup.py install ?
yeah... and that's where it erros out
the image 2 gcode one
python is fun to code :)
ValarQ: not really ;)
alex_joni: you can't have list comprehensions in asm :P
how bout machine code?
what does everyone else use to convert images to code?
alex_joni: only fun on older cpus
alex_joni: like Z80
I just love doing those: 04 23 A2 6B 17 12 00 01
* ValarQ actually knows some Z80 machinecode
nice... I didn't really like the Z80
at least they were better than 8080 :/
ValarQ: this is what i was thinking for controlling loads of steppers from one parallel port: http://www1.atwiki.com/gingery_machines/pages/bus
that site is geared to non-electronics people, so it's kinda over-explanatory
better than the other way around
fenn: that's a bit of overhead
but it's ok for IO stuff
why not go with an microcontroller?
and use command, data
i'm dumb, that's why
you'll do it in SW, and you'll be able to have a lot of devices
a nice micro is about 2-3$
fenn: maybe you should buy one of my cards ;)
* Jymmm invites fenn to /join #CheapBastard (you'll fit right in with the rest of us)
Jymmm: you'll make him join ;)
les: still around?
alex_joni: how do i program the micro?
fenn: load shouting helps
there are cheap programmers (about 10$ or build one of your own)
fenn basic stamp?
usually you can build one for <5$
parts that is
what's the refresh rate of a parallel port?
wait n/m we already went over this
it takes 1 usec per i/o right
fenn IF you decide to play with uC, be careful when using a laptop or a PC with APM enabled.
okay, so, if i have six steppers, i can output 1/.000001 steps/sec * .0005 inches/step = 166 inches per second?
fenn via parallel port?
via my doohicky
fenn you keep your doohicky in your pants young lady!
fenn you mean uC?
parallel port -> stepper driver bus -> motors
sorry that calc was wrong.. need to divide by three (two steppers per byte of output)
parallel port is like 112KB/s, but some uC you can toss in your own clock 24MHz =)
okay that's in the same ballpart
uC can be MUCH faster
that's 24 MB/sec?
I think Mb/s
what i'm saying is i dont really much faster than 112kB/sec
KB or Kb ?
parallel port speed is 112 kilobytes right?
let me check
swp was saying that it takes 1 microsecond for an i/o instruction to get out of the cpu
hm, you can use the multiple lines out on the parport I think..
to get more throughput
fenn well since a parallel port is on a IRQ, sure. buy a uC is like a DEDICATED cpu sorta kinda
115.2 K bits/sec
but if EMC can keep up with the limit on i/o instructions it doesn't matter
sorry that was serial. 512 kbps
fenn: right now EMC does about 50kHz stepping on the parport
fenn parallel port is a bottleneck for any controller
you'll divide that by 2 for your dohicky
SWP_Away is now known as SWPadnos
50kHz is pretty low
SWPadnos: hey stephen
alex_joni: its the ay it does 50Khz that is the problem
alex_joni: yeah, OK, 50khz is the max freq .. and the next, discrete, frequency below that is?
weyland is now known as weyland_away
well.. didn't look at the code
but I would assume smthg like 49.5
well, its 25khz
where does this number 50 khz come from?
fenn: from the time you need to do the outputs (outb)
so to do 49.9999 khz it has to do a bunc of 50k pulses, then one 25k one ..
well... it would work better if you could program the time between RT sleeps
but .. that's buggersome
4.77MHz * 10 ?!
and.. you have 6 axes
the 50 KHz number comes from a bunch of things - task switch time, I/O time, calculation, etc.
it's based on thread execution rate?
50kHz = 20 usecs in time
and as Robin said, the way it works is that a step or a pause between steps is an integer multiple of the PERIOD
50 KHz is actually pretty hard, it needs a PERIOD of 10 uS
this can be a problem at high speeds
which needs a pretty fast CPU
motors dont like doing the dirty pulse trian thing
it saps torquw
so the minimum step duration is one period high, one period low, or 50 KHz
and at high speeds, there is only just enough to go around
the next step is actually 33.333 KHz - one up, two down (or 3x PERIOD)
robin.. if you're running steppers
alex_joni: well, that what 90% of emc users do I think
if you're running geckos on DC
as the gecko does some PID
... you're still screwed
it's not optimal, but it is better
yeah, a little
and if you have one of the G340 it's even better
what SWPadnos said :)
yes, its better
as the PLL does some filtering too
actually, using the pulse multiplier in the Geckos makes it better - it kind of smooths out the pulse train
I agree that a HW pulser is even better
its not just a *bit* better ...
but.. who has HW pulsing ;)
besides the G100
its an order of magnitude better
jon elsons stuff
how about running a STG?
or jon e's stuff
the big thing with the G100 and the USC, and other hardware step generators (but *not* software microcontroller ones) is that the PERIOD is at 10 MHz or so
well, if i was worried about smoothness at high speeds i should be using servos anyway,right?
so the resolution of a pulse is about 100x better than on a PC
fenn: no, thats wrong
alex_joni: right :)
fenn: wrong ;)
fenn: what you need to understand with steppers is the speeds you have to use them at
fenn: you kow there is a point at which the torque begins to drop off .. its flat, then it slopes off
thats the "corner frequency"
since the rotor can't keep up with what the coils are doing due to inertia, right?
so ... do you use them above or below that speed ??
nope - you're using more and more power to get it to get out of the detents
and less is transferred to the output shaft
well, that's a limitation of the motors.. nothing you can do to make it go any faster
typically thats at about 600rpm
they spin fine to 3000rpm
they can still go faster, they just don't have much push (torque) at higher speeds
if you blast the hell out of it with current, will it generate more torque or not?
no, it'll burn out a winding
or catch on fire
that does work for servos, until they catch fire
above 600rpm they produce constant power
SWPadnos: only for DC's
robin_sz that's what i thought
below 600rpm, the power slopes off to zero at standstill
on AC motors that won't work either
if you want to get full power out of them you must spin at least to the corner freq
right, since 3-phase synchronous AC motors are essentially 3-coil steppers :)
and gear down
a stepper is just a 200 pole, 2 phase AC motor
right - my point exactly
so .. you have to let em spin
that's just about the opposite of every hobby cnc machine out there
so, rule 2
the world is full of idiots
that's a nice rule
and they keep getting better at it :)
well, either way works... get bigger motors or get better drivers
design the machine better
there's a limit for either strategy
fenn: talk to Mariss if you like, but this has been doen over and over so many times .. they need to spin, most hobby machines get about 10% of the availableperformance because the people building them wont listen to good advice
in know it sounds wrong
but whne you do the maths ..
in any case, the question was originally about step rates (and then smoothness of the pulse stream)...
thats the problem
emc can struggle a bit generating the rates needed
or hardware can do it with ease
that's why people who want smthg for their money go with servos
which also reduces the PERIOD requirement, since emc only processes rate updates, not pulses
I wouldn't bet my life on steppers
alex_joni: no, worng
in some circumstances steppers do a better job
right for now - wait for Mariss to finish...
well.. I won't trust a stepper to be in the precise position
low speed / high torque is better with steppers
well, what about a stepper with an encoder?
so you're saying output velocity commands to a microcontroller, and get feedback from the micro about position? (no encoder situation)
they are generally accurate to within 5-10% of a step position
but if you need high speed, good torque
you'll go AC
(full step, not microstep)
if you have $3500 per axis
weyland_away is now known as weyland
hey, does emc2 have a way of supporting steppers with encoders, other than the Mauch board, right now?
SWPadnos: might be a bit more
weyland: what board do you need for feedback?
I built a simple encoder readout board (and it has an hal driver ;)
I have a Mauch board, but don't have a ISA slot anymore
it's smthg like the DRO
well.. PCI is a pita
you really don't want to see my ISA card ;)
is there a way to do this?
to do what?
steppers & feedback?
use encoder feedback in emc2, without an ISA slot?
buy a stg board, run the feedback through it :)
but.. stg driver is not done
although the encoder part might be useable
well that was a dumb idea anyway
I wrote that, and never got to test it.. as I don't have a board yet
will get one soon
weyland: anyways.. what do you need encoder feedback for?
worried it's not precise enough?
jmk_away is now known as jmkasunich
was reading back, and I have the same question.. why do you want encoders on steppers?
to check for missing steps ;)
in critical apps
and do homing
based on zero pulse
prob is once you miss a step, what is the code gonna do? try to make up for the missing step by going faster, right? which will just make things worse
Mariss's unstallable stepper idea is something more sophisticated than that, but it requires dedicated hardware
didn't read on that
fenn: let a stepper be a stepper, don't try to make a servo out of it
Imperator_: huh? i'm not the one trying to add an encoder to a stepper
alex_joni: jusat thought it was a good idea to avoid lost steps
stepper -> cheap because it don't have feedback
ah, ok sorry
the idea is to slow down the feed rate (on all axes) to keep within the power curve of the motor
you normally can't tell if you've exceeded the curve, because there's no feedback
stepper with feedback is still less expensive than servo (arguably), and has no brushes to replace.
and it's guaranteed to be less than AC servo :)
Mariss is basically treating a stepper as a 100 pole synchronous AC motor and building a vector drive to run it
weyland : take a bigger stepper
cool idea - could be good for hobby level and light industrial machines
Imperator_:tried that the first time around, and it failed miserably
jmkasunich: how long do you expect to be around?
had some nema42's that SUCKED
on and off the rest of the day (dinner and such)
then agian, that was the first attempt at cnc and with emc1 a few years back
OK - I'm about to drive people around for a bit, but I may have HAL questions later in the day
I may try and use them again on the lathe
jmkasunich: care to tell me in a few words what's needed on the hal refactor?
more than fits in a few words
by the way, what do people think of this lathe: http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&item=7519302812
weyland: Blast.. And I was going to say that you should send me those evil Nema-23's. :)
have some, too
(sorry to interrupt - weyland reminded me of lathes)
SWL: not good for CNC'ing ;-)
jmkasunich: well.. then in as many words you can afford
I've been wathcing that lathe...
Ray said the same thing - what tells you that?
I know that model
good god lathe, my wife would say
manual lathes with half nuts and such don't lend them selves to CNC
I have a southbend 13"
ah - half nut on the X drive screw...
jmkasunich: but could be converted rather easily
OK - I'd replace it with a ballscrew anyway
motorizing an old style lathe can be done, but why?
that's what I'm gonna do with the one I just bought
* Phydbleep recognizes that lathe..
it has a half nut
I had that hernia's little brother here till my shop started sinking into the yard. :)
youre gonna scrap the original leadscrew, threading gearbox, and all the guts of the apron
I stil have it's little brother
Heve a 10Heavy
heavy 10's are nice
jmkasunich: almost, but not quite
not too big, not too small
jmkasunich: goona leave the geabox, just take out the driving gear
You'd be better off milling a new saddle and keeping the original parts just in case.
IOW, the only thing from the original lathe you are using is the bed, headstock, and tailstock
be back soon
and the spindle bearings on that lathe are plain bearings (bronze or babbit)
SWPadnos is now known as SWP_Away
Yep, VERY familiar with SB bearings
that's a decent manual lathe, I wouldn't want to try to CNC it
1.5" hole thru spindle...
weyland: Add the leadscrew handwheel mount on the tail end of the bed and drive it from there.
wonder what collets it uses?
what d'ya mean
drive the tailstock end of the leadscrew with a stepper for CNC
Ah. But why?
I'm just gonna drive it with the stepper. Why have a handwheel?
so you don't have to muck around with the QC box at the other end
weyland: Because then you can put the treading box in neutral and direct drive the leadscrew
I'm not understanding you
I think he means ti use a pulley or gear instead of the handwheel
oh. okay, thatz the plan
but I'm gonna disconnect the box anyway
jmkasunich: I'm actually going to add both to this Logan.. Handwheel for manual ops and stepper for cnc.
and less to go wrong
don't need it
gutting the apron too?
if I need to thread by hand, I have my southy
get rid of the weight
and provide a place to mount the cross-slide motor
* robin_sz agrees with jmkasunich
stripping a myford is probably a bad plan
we're not talking about myfords
thinking to put a ballscrew in place of the acme, and mount nuts on the apron
we're talking about WW2 vintage 14" southbends
well, I'm talking of the 9x20 I just bought
oh, chinese 9x20? or older american?
non american, unfortunately
couldn't pass it up though
brand new, in the crate
lees than most 7x10's
I have less qualms about converting a newer chinese lathe than I do about converting an older quality brand
Very true. which is why I didn't hesitate
a ww2 southbend, new in a crate?
weyland: 9x20 for <$400. :)
robin: sorry, thread mixing
robin_sz: No.. Not a SB.
SWP is looking at the WW2 SB 14"
Phydbleep: yeah, $375
wayland already has a 9x20 eastern import
both folks are thinking of CNC conversions
I'll have it Wednesday night
the SB is on ebay, $355 with 5 hours to go
the eastern import, yeah strip it ..
the SB would be a poor choice I feel
I'm playing with mounting it like a slant bed...
couple of big linear ways?
nah. pretty small, actually, but I'm thinking that if I build the table right...
don't tilt it so much that the V-ways aren't pointing up anymore
you can probably pick up a hradinge auto for similar money and cnc it easier
or an old mazak
once the high side of the V gets flat, the saddle falls off
oooohhhh, that completely passed my thinking - good point
hey, heres a thing ...
looks liek we picked up a nice contract for the laser
no, I've been looking for an old hardinge, prices aer ridiculous
different question: did anybody know how noisy are this DC-DC Converters ?? Want to place a 3W 5V to +-15V Converter about 10mm next to the reference voltage source for my 16bit DAC
weyland: for the HLVs yes, for the hydraulic servo copy / auto lathes? no cheap.
theres a company in the town that makes all the chip managment systems for
weyland - I've seen hardinge second op lathes go pretty cheap
mazak, mori-seiki, matsuuro etc
apparently they ship the mills from the far east
but the chip collection anc ocnveyors are made in the destination country
anyway ... theres a shit load of laser cut plate on em
and it looks like we got the work :)
justcan't fin'm down here, or even close enough to be worth it
Imperator: use different layers on the PCB
and this 9x20 came along at a good price
well, if its new far east crap, feel free to cut it up with a clear conscience :)
[20:42:22] <jmkasunich> http://www.hgrindustrialsurplus.com/Detail.asp?PartNum=03-747-006
a hardinge would be GREAT though, cuz the spindle hole is nicely large
that's why I was looking for one
that and the dovetail make them great
[20:42:58] <jmkasunich> http://www.hgrindustrialsurplus.com/Detail.asp?PartNum=20-175-039
oh .. tiny hardinge :)
no leadscrew on that one
03paul_c * 10emc2/src/emc/kinematics/trajectory.h:
Further experiments with Doxygen and embedded LaTex commands.
Extended the number of formulae that the algorithms in trajectory.c
and planner.c are based on. Will add the bib. refs. at a later stage
along with additional text (probably included from the source files).
fsck me running...
but how much to get it to fort laudedale?
I've got $375 in this oen :)
and it's local
[20:44:48] <robin_sz> http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&category=97230&item=7520958146&rd=1&ssPageName=WDVW
the rust belt is full of machines, but the people who want them aren't in the rust belt
robin: that's already 1 order of magnitude to high, and there are 8 days to go
what I was going to say ...
saw that one too
I saw a hardinge of that sort of size and shape
go for 400 GBP a while ago
[20:46:10] <jmkasunich> http://www.hgrindustrialsurplus.com/Detail.asp?PartNum=20-273-004
[20:47:24] <robin_sz> http://cgi.ebay.co.uk/ws/eBayISAPI.dll?ViewItem&category=92083&item=7520128064&rd=1
geez... monarch's are built like brick sh!thouses http://www.hgrindustrialsurplus.com/Detail.asp?PartNum=40-297-004
headstock's nearly as long as the working space
nice robin... shame it's on the wrong side of the ocean tho
holy shit, I'd take either of those
the mazak or the hardinge?
I'd pay him at least 10USD to bring it here
we see them at that sort of price often .. USA cant be that different
it's the us, after all
we're stupid in so many ways...
in CA at one point it was pretty much scrap price
that point is gone
weyland: may I quote you?
you may get beaten by the masses for doing so, tho
shrug ... you may think that point has gone, actually, you just reached a flat bit on the slope, it gets steeper from here on in
weyland - hou bout a 10EE for $999
it's a double edged sword, actually
missing the compound
jmkasunich: maybe in a few months
now that machine prices are going back up
I'm hoping it's an indicator that we're actually using them agian
two things ...
2) doubling of fuel prices in the next 12 months
and not just a relfection that they're actually worth more now at the scrap yeard
jmkasunich: if I were just opening a shop, and needed a lathe, and had 1K, I'd snap that monarch up~!
my 10heavy has served my very well for years
I've had to pass up several in that 1000-1500 price range. Too heavy
heavy lathes rock
oh, yeah... you have that basement thing going on
if I was opening a shop ... I wouldnt buy any non-cnc machines
although just if you have space..
yes they do
infact, I havent :)
heavy machine in general are great, but space and stairs make it hard to get them home
robin_sz: sometimes it's a matter of funds and availablility
robin: depends on the nature of the business you want to do, but in general you are probably right
manual machines for a shop just dont make sense these days ... you have to find skilled labor to work them .. and thats where the trouble starts
* robin_sz nods
too many variables for a sweeping stement like that
well, nowdays cad drawing stuff is so easy..
weyland: I hear what you are saying .. and I agree, if you have lots of spare cash, yeah get a manual machine
for a 1 man shop like mine where I'm really selling design, ingenuity, and convenience, not products, manual machines are OK
if you can afford the cam programs it's faster..
but for producing stuff..
jmkasunich: but you are forever stuck at the i man level
a lathe and a dro goes far..
which suits me fine at this point
robin_sz: I disagree
I'm moving forward...
I do my day job for a living, the side work pays for toys and trips to EMC-fest and such
weyland: have you ever employed people to operate machinery?
and they did it, reliably?
yes, but I'm an asshole, and anal
well, I've made so manu things on my mill lately..
fear of death goes far
many.. for work, and hand machining it would be hell
well, in my experience, only 2 things work, 1) doing it yourself and 2) having the cnc do it. as soon as you get your staff to do it, they find new and ingenious ways to fuck it up
Ah so now we know why weyland doesn't want a CNC center... too afraid the employees will stuff him in it!
sure... I'm not saying cnc doesn't open opportunities
it opens them all..
the parts you have to turn for 50 minutes, you can do in 20...
and .. they are all right
what I'm saying is that it isn't *necessarily* the only, or even best, way to *start* a shop...
wrong wrong wrong
no worrying about precision either..
well, that's the way I'd start one..
keep my job, and start doing machining on the side..
look, if you have loads of cash you can afford to subsidise a manual machine
and that's great, if one has the funds and machinery space available
but if you want profit, look at a cnc
oh, it's great if you have somone to machine for you
that's exactly how I gto started
we just spoke of this last night, ehre
well, take that hardinge CNC on ebay with a duff control
its big and heavy ...
robin_sz: does baldor do lathe's ?
and unfixable probably
weyland : I'm in an apartment. I have no space for a mill and/or lathe (even mini ones). But at least to get started, I do have room for a cnc router. There is no way in hell I could do anything reliably and within the same month manually.
but .. this is where you come in ...
I tried to mill stuff manually..
and not economical to get here, and then repair, spending time to do so, losing money while doing it
weyland: what do you have to give you an edge in the buisiness? answer .. your brain
robin_sz: yes, that's it exactly
your time is money..
you could take that lathe, drop on a replacement cnc control, and run it to its limits
make it efficient
every second you turn on the lathe is a second you could be looking for more work
you are now using your most valuabel and scarce resource ( your brain ) effectively to power your business
there is NO money to be made doing donkey work on a lathe
weyland have to ever produced a part, then get 95% complete just to find out one ting was off and you have to start all over again?
but well, I've seen shops buying cnc machines just to do it manually..
*IF* the PROJECT is within your available cash and time limits
except perhaps high quality prototyoe work
because they couldnt use them to 20% of the potential
that's what I like about cnc machines
make one, not right, adjsut drawing, hit go again
no time wasted
I'm not against CNCs by *ANY* means... don't misundersatand that
backup to cdc and thru in valut =)
or I wouldh't have done the mill, and just bought a lather
I have a manual lathe
* anonimasu dose too
what I'm saying is that I don't think it's necesary to open a succesfgul shop
its not a money maker
but machining 20 bushings, takes lots of time
and theyll all be different
it's not about the time to do it
* robin_sz falls off his chair
it's about using that time to let the machine do it and doing something else, yourself
oh yes it is
weyland I honestly feel that TODAY (10 yrs ago different) you need some kind of edge over your neighbor/china.
read the abovr
weyland: was it you asking about a hemisphere?
the edge of china is easy
weyland: I put a much-improved program on the web site http://timeguy.com
robin_sz: the larger the lathe the easier ;)
and your image2gcode program
the edge over china is easy ...
they pay the same for steel as we do
but less for labor
cradek: can't get it installed
quality is identical
our edge is simple.
weyland: can't get what installed?
the PIL stuff
find a market that needs short leadtimes
cradek: did you modify it to actually cut the pocket, or does it still do only the hemisphere surface?
weyland: but you have a working python installation?
yes, I believe so
SWP_Away: you would cut the pocket by iterating r from "small" to the finished size
ah - surface only
right - this would be good for the finish pass :)
SWP_Away: yeah, simple matter of adding another loop of r.
got my msg's ?
weyland: what problem are you having installing PIL? You should be able to just use "python setup.py install"
a roughing pass using only circles (or an outer circle and X/Y moves) in the X-Y pland, and doing a waterline approach would also work (but it's definitely more complex to code)
shold I post the error msgs?
robin_sz: China pays more for steel than we do.. They have to pay transport to get it there.
by the way, great work - that looks pretty cool!
alex_joni: yep, now I did
gcc -DNDEBUG -O2 -g -pipe -march=i386 -mcpu=i686 -D_GNU_SOURCE -fPIC -fPIC -I/usr/include/freetype2 -IlibImaging -I/usr/include -I/usr/local/include -I/usr/include/python2.2 -c _imagingtk.c -o build/temp.linux-i686-2.2/_imagingtk.o
_imagingtk.c:20:16: tk.h: No such file or directory
_imagingtk.c:23: error: syntax error before '*' token
_imagingtk.c:31: error: syntax error before "Tcl_Interp"
_imagingtk.c:31: warning: no semicolon at end of struct or union
_imagingtk.c:32: warning: data definition has no type or storage class
_imagingtk.c: In function `_tkinit':
Phydbleep: could be true
_imagingtk.c:37: error: `Tcl_Interp' undeclared (first use in this function)
weyland: install tk
you need tk
_imagingtk.c:37: error: (Each undeclared identifier is reported only once
_imagingtk.c:37: error: for each function it appears in.)
_imagingtk.c:37: error: `interp' undeclared (first use in this function)
_imagingtk.c:45: error: syntax error before ')' token
_imagingtk.c:50: error: `app' undeclared (first use in this function)
_imagingtk.c:50: error: syntax error before ')' token
_imagingtk.c: At top level:
_imagingtk.c:55: warning: parameter names (without types) in function declaration
_imagingtk.c:55: error: conflicting types for `TkImaging_Init'
_imagingtk.c:23: error: previous declaration of `TkImaging_Init'
_imagingtk.c:55: warning: data definition has no type or storage class
_imagingtk.c:57: error: syntax error before '&' token
error: command 'gcc' failed with exit status 1
"apt-get install tk" ?
stop now ...
yeah it'll probably also need tcl.
robin_sz: I found out I had to change some stuff on the seed machinery, since the positioning of the was less great in reality..
what did it do? the pasting
too much pasting
the first 3 lines would have been enough
oh, sorry, I didn't know
robin_sz: the joy of cnc is having a protytpe on your desk in 20 minutes after hearing it..
anonimasu: you mean it was being bounced around all over the place shaky shaky with the seeds?
robin_sz: oh that werent the trouble they hit it with somthing, deforming the stainless canister..
like .. a tree
the weldment broke on the cover. ..
anonimasu: Armor plate that sucker. :)
1000Nm dead stop into somthing
anonimasu: the great thing of CNC is it allows you to scale your business
err 1000nm of torque.. and 4.5t
anonimasu: Drug trest the operator?
weyland: you still here??
here we go: http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&item=7519269917
one moment ... dont run off
baq and forth between terminal
cradek: Couldn't find package python-tk
have you read the business stuff on sherline.com?
[21:14:24] <robin_sz> http://www.sherline.com/business.htm
weyland: I'm not a debianer. How do you search for a package?
weyland: try just "tk"
take a couple of hours ..
weyland: hang on a moment
err *ponders why he did write [off]*
apt-get update apt-get install python-opengl python-tk python-dev xlibmesa-dev \ freeglut-dev
python-tk should be there
weyland: you sure you're connected to the net?
on that machine?
weyland: joe martin is the bloke who founded sherline, read that article, its worth more than 20 years of hard business experience , starting fromt he msall shop scenario
apt-search <pkg name>
AHHhhhhhhhhhhh - too many threads... hold on
robin_sz: it's a good/great deal.. I can afford tools and upgrades :)
cradek: any thoughts?
cradek: tk is already the newest version.
cradek: You should have apt-get, do you have synaptic?
Phydbleep: I don't use debian
alex_joni: not really
alex_joni: I would have to get the patches you have done so far and dig into it.
cradek: What distro?
alex_joni: I don't have rt on this machine, so I can't build emc2, since there is no sim, which makes me sad.
alex_joni: also, I can't install rt, since it conflicts with the closed-source nvidia driver, which makes me even sadder.
cradek: Get the 'apt-get' rpms and synaptic, then you can grab debian pkgs.. :)
Phydbleep: why would I want debian packages?
Phydbleep: (that was a rhetorical question)
cradek: I thought you were looking for a way to deal with pkgs.. I like synaptic better than most of the other pkg managers I've seen. :)
cradek buy a real video card =)
Phydbleep: nope, I'm fine
Jymmm: I don't know of others with such good GL performance
* Phydbleep beats Jymmm senseless with a Voodoo5.
Jymmm: that cost $30 like these do
cradek add two more digits and I got soemthing in mind you can use =)
cradek: how does it conflict?
cradek hey, dual video.
alex_joni: it just doesn't work... I think the screen goes black, but I forget for sure.
cradek: Hehehe... You need the Diamond Fire GL4000 I sent to Tom. :)
hmm.. but still you can have 2 kernels
Phydbleep: what I really need is emc2/sim
and the emc2 stuff only needs to get compiled, don't need to run it :D
(but that's an aside)
cradek: it's noted
but unfortunately not very high priority :)
I know, sorry to keep complaining about it
what we ned for that is rtapi-sim
cradek: you're allowed to bitch about it
but its ugly
alex_joni: oh good
cradek: And carp, complain and kvetch. :)
as much as you want
is there any open source software to produce g-code on linux ?
cradek I got a Trident 8600 you can have =)
cradek Mmmmmmmmmmmm 8Bit ISA
Jymmm: I have *plenty* of video cards, but just this one with DVI
jacky^: I saw some somewhere I think.
cradek: what card is it?
jmkasunich: nvidia g-force xx00
I'm running a g-force here, 4400 I think
with BDI-4.20, using standard linux drivers IIRC
jmkasunich: might be a 4400
jmkasunich: is X using the vesa driver?
jmkasunich: but without accel
jmkasunich: Can you 'twin-head' with that setup?
just bought it about 3 months ago, $69 with $30 rebate
jmkasunich: I have to use the nvidia driver and reduced-blanking timings to run my 1600x1200 lcd on DVI
I'm running 1600x1200 on a sony LCD
no, on VGA, but I
i tried synergy and i losted in a labirint .. :\
(II) NVIDIA(0): NVIDIA GPU detected as: GeForce4 MX 4000
but I'm almost certain it can do it on DVI, that was one criteria I checked when buying
(right now I'm using a VGA KVM switch, but I expect to eventually direct connect the monitor)
jmkasunich: I had no luck with either the vesa or nv drivers. I could not get 1600x1200.
I wonder how I check what drivers this is using? kernel logs from startup?
jmkasunich: the X log
and less /etc/X11/XF86Config
jmkasunich: or, the X config file
Section "Device" / Driver
VendorName "Videocard vendor"
BoardName "NVIDIA GeForce 4 MX (generic)"
which X version?
XFree86 ver 4 I think?
% xdpyinfo|grep vendor
the nv driver on xfree86 4.3.0 doesn't work with this card at all (blank screen)
cradek: can you tell me the problem with the install?
jmkasunich: I bet your DVI doesn't work.
cradek why dont you use the vga instead?
weyland: huh? did you install tk?
was already the latest version
but your compile couldn't find tk.h
maybe tk devel is separate?
okay, what does that mean I need to do?
cradek: maybe it doesn't... that would be a bummer
I'm not a programmer typre
weyland: yeah try tk-dev or tk-devel or something like that
apt-search tk devel
apt-cache search tk devel
sorry.. it's late ;)
it's pretty hard to monkey up a decent python install.
"apt-get install tk-devel" did it
lemme go try the install now
huge list of errors
no, I'm not pasting them
lemme try the same with tcl
weyland: tried first apt-get update ?
yeah, did that
you should have tck and tk stuff... you were able to compile emc2, which includes emcsh and some other tck related stuff
that did it
16:08:42 < weyland> _imagingtk.c:20:16: tk.h: No such file or directory
okay, lemme go try your stuff now :)
is this the debian/python support channel?
python setup.py install-debian
apt-get install python
rpm --erase debian
rm -R /*
forgot the -f
flailing around with that stepgen thing... a break might be good
cackle.. joe martin writes: The Microsoft Word® program that I used was as much as a work of art as the marvelous cathedrals of the 17th century. We just haven’t acknowledged it yet.
hey guys, i made a test running isd.ngc, at the end the logo appear reversed like in to the mirror .. is it normal ?
lol.... the cathedrals lasted centuries, the program lasts a year if you're lucky
jacky add a - sign before UNITS
jacky^: you might be looking at it from the wrong side
er, was that input_scale?
the motors are connected right, i think..
fenn: input_scale I think
jmkasunich: about that refactor
do your jogs go in the expected direction?
as input scale i'm using 100
mm I assume?
emc1 or emc2?
i suppose emc2, i'm not sure,
unless you downloaded and compiled, that is the bdi-4 branch.. kind of a hybrid between 1 and 2
the "sherline special"
do jogs go in the right direction?
alex: about that
umpf :\ i don't know how to use manual command so..
i just move using - and + button
take a pencil or pen and draw a X-Y axis on your part
for the x and y axes it travel in the stepper direction when push - button
then use +/- to move to the point of axis interection
reverse, instead for z axes
what do you mean "in the stepper direction"?
near the motor
that doesn't help us because we don't know which end your motors are on ;-)
draw a "L" on your part
put "+X" at the end of the horizontal part
and "+Y" at the end of the vertical part
use the +/- to move to the intersection of the two lines
then do + on X axis, does it move toward +X?
do + on Y axis, does it move toward +Y?
the strange thing is
important - "IT" is the tip of the tool, relative to the workpiece, not the table movement
i tried 3d_chips and it made a penguin
i suppose normal
with isd instead, in reverse mode
chips is symmetrical, so you couldn't tell whether it was mirrored or not
ok, tnx will try tomorrow your tips
aha.. is simmetrical !
check his toes
chips has 3 toes on the right leg
also 3 toes on the left leg
but surely 3 on the right one ;)
well that's not very helpfull
* alex_joni is half asleep
1am there isn't it?
you really want to talk about hal refactor?
[01:01] <alex_joni> about ;)
aniway, today i got a positive step, finally with 100 as input scale it seem work quite well
i tried all from 1 to 1000.. LOL :\
seems like the hard way of finding the right value
at the end tuning the velocity seem work fine
I would have started using 1000, and moved X10
and measured the distance
I would have looked at the screws, belts, etc, and calculated the value
(maybe fine tune later with a measurement)
yeah.. but that wasn't available to jacky^
asked him what screws he has
not very helpfull on ipr calcs
he can't measure the pitch?
i think 1 mm
if it's standard 6 screw threads, you can just look it up
so, maybe 100 for input scale it's a right value ?
you said you had 200 steps/rot steppers?
if you tell to move 10mm, and it moves 10mm, then it is the right value
or was that smby else?
on this i'm sure
if you have 1mm pitch and 200 steps/rot then 100 might not be ok
but probably it's 0.5 mm pitch
so 200 x 0.5 = 100
0.5mm pitch is pretty fine thread
not very standard on M6
with the old value chips was about 3 cm lenght
now is 10 cm, i suppose right
jacky: are you using tkemc GUI?
some day ago..
now i'm tring to learn sherline gui
I'm not familiar with the sherline gui
is there an incremental jog mode on that GUI?
anyway also using tkemc
if so, try doing some 10mm incremental jogs and see if it actually moves 10mm
jmkasunich: will try tomorrow at morning
tkemc in inch mode can do 1.000, 0.100, 0.010, 0.001, and 0.0001 inch incremental jogs... I expect it can do something similar in mm
i will say you
now can't turn on the machine because my mammy is sleeping in the room near :\
none of these tests need the spindle turned on BTW
it make a big sound
not so good ..
maybe the steppers whine a bit ;)
the stepper i buyed from internet at low price
maybe not new stepper
too low cost..
but are working
for me the precision now is not so important, i want to learn first :D
true... but asking for 10mm and getting 5mm or 20mm is not just a minor precision problem
* jacky^ hope to learn..
jmkasunich: it's all relative
start with the basics, does each individual axis move in the right direction and the right amount
precision is based more on how your machine is built and how you use it than the accuracy of the steppers
then move on to multi-axis moves
and finally try cutting something
maybe +/- 20 mm is not a very good precision for you..
I'm not talking about +/- 20mm, I'm talking about 2:1 scaling errors
200 steps per rev on a 1mm screw means 200 steps/mm, not 100steps/mm
maybe 2:1 scaling error is not a very good precision for you ;)
if the screw is only 0.5mm pitch, then the scale will be 100step/mm
but it might be for somebody who doesn't really care what they do :)
but a 0.5mm pitch 6mm dia screw is unlikely
not sure why they still do it ;)
weyland is now known as weyland_awayplay
well, tomorrow i'll try with manual command measuring... we will know
maybe the stepper is using full steps instead of half steps or something like that
200 steps/rev is full steps, unless it's a strange stepper
beh.. a moment
i'm using 2 differente controllers
i was already 2 old controllers
3° i made from myself
but they work with half step
to use full step i should send 5V to the controller on a specified pin
to enable it
jmkasunich: here is my stepper: http://www.micromed.it/Elettronica/stepper.htm
the model is M22
passi= step ..
* alex_joni goes to bed
weyland_awayplay is now known as weyland
gotta go guys, dinner and wife calls
robin_sz: lets continue that conversation another time, okay?