well, I have a plan (for my worn motor shaft)
just made a sleeve, 0.2497 ID, 0.312 OD for most of its length, 0.322 (rough) for the last 0.3" (where the ball bearing is gonna go)
the sleeve is long enough that 1/4" of it will bear on undamaged 0.250 shaft (0.0003" press fit)
gonna use retaining compound to fill the gap, and the press fit to keep it centered
then after it hardens, turn the bearing seat to 0.315 (bearing ID is 8mm = 0.3149)
turning the bearing seat between the original centers on the original shaft will ensure concentricity
eh why not just weld some material back onto the shaft?
its a precision balanced 10,000 RPM motor, 1/4" dia shaft
I'm still stuck on the 0.2497 ID part
heating it would A) ruin the motor windings B) warp the shaft
doubt I could reliably turn an OD to that size
its a press fit
I can't either
I have a 0.2497 reamer
and a 0.250 reamer... wonderful for dowel pins, use the small reamer thru both parts, then the big one on the removable part
"fits like a glove" ;-)
find anything neat at hgr?
yeah, the thing I'm fixing
dumore 1/4 HP toolpost grindr
oh - *scrolls up*
bit of a fixer-upper, but...
cradek: I could probably make an OD pretty close to that size
for OD, turn till close, then polish to size
I can take off a couple tenths polishing
[01:13:12] <cradek> http://timeguy.com/cradek-files/emc/x-motormount.png
aw, I thought you were gonna show us a finished part ;-)
still working on the gcode now that I have a design laid out
gotta admit, I'd not like to make that part without CNC
no, it would have more square corners without cnc
and still take two setups
one with an axis parallel to the slots, and one with an axis parallel to the top edge
well that's not counting the holes
not sure how you'd get them in the right place
X-Y coords, from the CAD
I guess you could center drill in the mill
I've never been able to step drill and keep on center
I always mill holes that have to be right
you mean the big hole?
how big is it?
yeah that would be the problem
I have a 5/8 drill bit that centers pretty well in a smaller hold
that last 50 thou would be a challange tho
I'd probably just spot drill the hold, then put it on the faceplate and bore to size
you'd have to bore it on the lathe
(if ID and or location is critical)
cnc is sure good for some things.
is the top edge of the motor bracket flush with the top of the table?
you might want to think about clearance cuts so you can access the tee-slots then
I thought about that, but the other open end is only a few inches away
someday you're gonna have something already on the table and want to put something else on the far side of it
that's probably true
and I'll sure curse about it
I'm starting to wonder about this grinder
whether it's worth it?
I think the previous owner was a moron
lots of previous owners are morons
the motor shaft, and both ends of the spindle shaft, are threaded 1/4-32
three of the four nuts in the box were 1/4-28
and the threads are somewhat buggered
that's a special kind of moron
something tells me I'm gonna learn alot about high speed precision spindle design
(which isn't a bad thing, but I'd rather not have spend $150 for the lesson)
however I bet the 1/4-32 nuts are right next to the 5-40 screws in the typical hardware store
1/4-32 is an oddball
but mcmaster has a tap, $5 or so
yeah that's what I meant (oddball just like 5-40)
5-40 is _not_ an oddball
its a standard national coarse
its the standard thread between 6-32 and 4-40
I know, but still it's relatively hard to find
its an oddball wannabe
I think I want to take this spindle apart
step 1: build a pulley puller ;-/
I think you bought more project today than you intended...
pulley puller built already (just some creatively arranged mill table clamps, and some steel bars)
however, applied a fair amount of force to the pulley, no come off
threaded on? doesn't look like it, but maybe...
some good news
a munged up thing that I thought was part of the spindle shaft turned out to be a little sleeve.. easily replaceable
still can't figure out the pulley tho
its a two step flat belt pulley
1-1/8 and approx 3/4
the larger step is slightly crowned, as you'd expect
the smaller one isn't
i'm going crazy here trying to get gcc to see the include files it ships with - it should search through /usr/include and /usr/local/include and emc2/include right?
it looks badly worn
it searches thru whatever directories are specified by -I options on the command line, I think
right now the compile dies with stdarg.h: no such file or directory, from rtapi.h
even though stdarg.h is in like 5 different places
sounds like you've got a very porked gcc install
can you just start over with whatever rpm?
he had 3.4, build the kernel with it, then replaced it with 4.x, then reinstalled 3.4
so yeah, its borked
i should be able to build an rt kernel with 4.0.2 right?
last I tried, I got a warning when I used gcc4 to build rtai
it wont actually be running anything
we need non-rt sim don't we
argh I think I've screwed up radius comp for mills
[01:45:29] <cradek> http://timeguy.com/cradek-files/emc/wrong.png
blue is the toolpath?
where did you start?
[01:46:50] <cradek> http://timeguy.com/cradek-files/emc/right.png
this is what I get (the right thing) if I just change the g41 to g40
it's going clockwise
trying to remember the diff between 40 and 41
41 is left comp, 40 is comp off
where does the path start?
the white line is the feed down
its on the part edge
then it jumps up and goes back down on the blue line
oh, thats up/down, not horizontal?
don't you need a lead-in move from outside of the outline?
the blue goes up to the previous Z then back down by the end
fenn: yes, it's above
in x/y plane
it's the dashed one to the right before the white downfeed
thats where I was going fenn
if you're gonna do comp, you gotta start at least one tool radius outside the cut line
it does, above in Z
you can turn comp on, then cut down to go around the part
well, except I broke it
if you are doing comp in XY (which I think you are), then Z should be irrelevant
you start outside the part in X and/or Y
no, listen to me
you can cut up or down in Z with XY comp active
it's perfectly fine
so you can start "outside" the part (but also above)
then cut down
but you're not doing that, you're starting _on_ the part (but above)
i don't see the lead-in move on the backplot, but even so it shouldn't mangle the toolpath like that without an error
[01:52:58] <cradek> http://timeguy.com/cradek-files/emc/leadin.png
blue is the leadin move?
this is left comp, so it makes a convex corner, that's the only requirement
so you're leading in before you plunge, then plunging, then beginning to cut the profile
why do we need a lead-in move again?
I guess I agree its borked...
fenn: to tell the tool how to transition from on-the-line to next-to-the-line
but I still don't understand why you don't plunge in the waste, get that out of the way (tool deflection, etc), then leadin and cut the profile
I could, I'd have to leadin with an arc
but it would hide this bug :-)
couldn't you lead in just like you are now?
just plunge before you do the blue move
no, I'd cut off that nice radiused corner
the one at the end of the path
ok, I'm clueless
can you post a top view, the perspective has me all screwed up
I thought the blue line started out in the waste, and ended up with the (compensated) tool tangent to the arc
[01:57:17] <cradek> http://timeguy.com/cradek-files/emc/top.png
from above it looks right, you can't see the Z problem
ok, what am I seeing here... blue is the left hand edge of the part?
the blue is generated from the highlighted gcode line (a g1)
it makes an arc (to go around the convex corner) then a line parallel to the original part edge
and I can't see the actual part, so I can't tell what the comp is doing
the problem is the endpoint between the generated arc and line is at the wrong Z
the white is the end of the path, also compensated? so the actual part edge is somewhere to the right of the path?
so in order to do this kind of leadin at this spot, I have to be above so as to not cut off that corner
regardless of the Z problem, seems to me the leadin move should start _outside_ the compensated path
yeah, generally, but if you're above it doesn't matter
if you make a convex corner like this, you don't have to mess with getting an arc right
but I am probably just confusing things, I've never done comp
from outside a part without any sharp corners (like this one) you have to leadin with an arc otherwise you make a concave corner
duh, I think I get it now
it's NOT obvious
when its cutting the leadin move, it thinks the part has a sharp corner there
right, if you leadin with a line from outside the part
so the arc it draws is just the arc it would do to cut a sharp corner
ok, now you got me straightned out
go fix the Z bug ;-)
the problem is I've broken and "fixed" this a couple times now
was something recently hardcoded to look in /usr/local/bin? i've never seen this error before: Can't execute server program /usr/local/bin/emcsvr
well, fix it right, dammit... what are we paying you for?
(i'm running from /pub/emc/emc2 btw)
grep /usr/local/bin/ ./ -R
Binary file ./objects/hal/utils/halcmd.o matches
Binary file ./objects/hal/utils/scope_horiz.o matches
configured for run in place?
oh whoops that might be it
i still think run in place should be the default
only developers run in place
let's say we get a sim mode working
* cradek gleefully attaches the debugger to the userland interpreter
and you want to compile and run axis on a machine you dont have root priveleges on
fenn: not gonna happen
sim mode will still load the code into kernel space
you still need the "sudo make setuid" step
but non-rt sim will not use kernel modules right?
sure it will
the code will still be in kernel space, it just won't be running hard RT
why kernel space?
I think the whole point is to get it away from the kernel
(the alternative is to build all the code for userspace, IMO that would be a gawd-awful mess in both the makefiles and the individual source files
the point (I thought) was to get away from requiring a patched kernel)
we obviously have different goals for it
my approach would be pseudo realtime
if it's not for realtime I see no reason to involve kernel modules, and doing that messes up any benefit for developers.
yes, and bugs don't oops your kernel
and you can run without root
you can run without root now (can't build without it...)
and you can run on any unix OS
"or setuid binaries"
yeah, you know what I mean
hal as it exists will never work in userspace
it's irrelevant how you get your superpowers - they are still required
because one thread of execution (rt or not) needs to invoke code that lives in many modules
that works in kernel space, but not in user space
sure, that is possible
fwiw I have nfc about that
if you want to have two completely different approaches, and piles of ifdef's everywhere
if we aren't going to get away from kernelspace, I think we're wasting our time doing anything.
thats a developer who likes gdb speaking
the users that I've heard gripe about it are upset about kernel patches, don't give a shit whether its kernel or user as long as they can run it on a vanilla kernel
like you said before, different goals
I _don't_ think their goal is "wasting our time"
as long as they don't think they can run a machine with it, that would be OK
I do think two completely different models for the two modes adds complexity for minimal benefit
the main goal I've heard is to be able to look at a simulated machine, and be able to look at part previews in axis
if getting it into user space is that critical, then we should do it in userspace all the time
for that, we can fix axis to not use emc
I'm talking about a simulator, not a gcode previewer
I think they may also want to "run" a part
IMO, the simulator should be as close as possible to the real thing
running in userspace isn't
IMO, the simulator should not require superpowers, should not be able to crash the machine, and should NOT be able to run hardware
no drivers at all?
no output, input is fine
whats the difference?
as soon as there's output, someone will try to run a machine with it because patching is "hard"
ugh especially with encoder counting
let me rephrase
and their machine will be all jerky, and we'll tell them "you need realtime you idiot"
output requires root access, so is incompatible with "no superpowers" anyway
I'm thinking of parport here, there may be other things I haven't thought of
what do you expect to be doing in sim mode?
simulating an emc run
its becoming very clear that we have completely different goals in mind
you want to let people control machinery without realtime?
to the point that we need two "simulator modes"
does 'simulating emc" include running classicladder and hal components?
what I have in mind and what cradek has in mind are incompatable - differnent goals, different approaches
cradek: yes, I do want to let people control machinery without realtime
machinery that can safely be operated with the kind of latencies that you get with vanilla linux that is
that may not include _any_ CNC machine, but I'm not limiting it to CNC machines
classicladder running at the jiffies rate can do plenty of usefull things without a RT kernel
if you believe people will go through extra hassle for a realtime kernel if they can use their favorite distro instead... well..
a lot of people won't
and that would be unfortunate
I believe people aren't stupid (which is probably my mistake)
if somebody wants to run christmas lights with classicladder and the parport, I intend to let them do it
(using a regular kernel)
well i guess this boils down to the same debate about estop
* fenn motions for the board to come up with a policy on whether or not to trust the user with safety decisions
and the answer is yes
the alternative is to assume all risk ourselves, and I'll be damned if I'm going there
we do _not_ tell anyone "if you do this you'll be safe"
we tell them "it is your responsibility to ensure your machine is safe, and btw, don't trust software for safety"
right now there seems to be precious little documentation on safety, is that intentional?
it does need to be better
from hal.h (and many other files):
THE AUTHORS OF THIS LIBRARY ACCEPT ABSOLUTELY NO LIABILITY FOR
ANY HARM OR LOSS RESULTING FROM ITS USE. IT IS _EXTREMELY_ UNWISE
TO RELY ON SOFTWARE ALONE FOR SAFETY. Any machinery capable of
harming persons must have provisions for completely removing power
from all motors, etc, before persons enter any danger area. All
machinery must be designed to comply with local and national safety
codes, and the authors of this software can not, and do not, take
any responsibility for such compliance.
those words need to appear somewhere in the integrators manual, not just in the code
are "local and national safety codes" the sort of thing that would be on the internet that we could link to from a wiki page?
maybe, but again, its not our job to say "if you follow these cods you'll be alright"
the codes depend on where they are
in europe, rules are different than in the US
yeah but pointing people in the right direction is better than nothing, isnt it?
rules are differnet for personal machines vs. machines that employees are gonna use
you're thinking of hobby type people
if somebody is designing a machine that is capable of killing people, he damn well better know what he's doing
if he doesn't, I don't want even the tiniest little hint of a possiblity of a chance that his widow's lawyer can come to me and say "you said if he did X, Y, and Z he'd be safe"
one thing we could do if someone were to donate money would be to hire a lawyer to write up a disclaimer :)
there is a long history of declaimers for software
and the one in our code goes farther than that
but giving out safety advice tends to counteract the disclaimer
that just seems wrong somehow
welcome to the 21st century
BE SAFE (but who knows what that means)
BE SAFE: but only you know exactly what that means, because only you fully understand _your_ situation
believe me, I don't want uninformed people getting hurt
but my approach to that is to tell people who have no business designing dangerous stuff "DON'T DESIGN DANGEROUS STUFF"
not "go read this and then you'll be qualified"
because some of the people I've seen, they could read a bazillion pages and still not be qualified
going back to: are "local and national safety codes" the sort of thing that would be on the internet that we could link to from a wiki page?
most of the codes are written by engineer types, for engineer types
most of them are _not_ on the net, the standards bodies that make them up sell copies
right, i figured, thats why i asked
and they are thick
why do you say that?
i've been drowning in STEP buzzwords all week
i dont imagine any other "standards" documents are any better
step isn't a safety code
the national electric code is a very long, very complex document
i know, but its made by the same type of bureaucracy
but every rule in there is the result of a fire, electroction, or some other bad thing
there's a new one every 3 years too...
the complexity results from the need to explicitly define every possible situation (which is of course an impossible task)
right, there are rules, then subsections, exceptions, variances, and fine-print-notes
I've used the NEC in my work, its just one of probably a half dozen standards that our products need to comply to before we can sell them
99% of that is irrelevant to the basement guy
the pro's already know what standards they need to follow
anyway, one place where we do agree: we need more safety info in the manual
we _might_ want to cover the basics, to keep the idiots from frying themselves
but we need to be very very clear that we are only covering the basics, and the responsibility for a safe machine lies with the machine builder
time for bed
hm. program g81 z-1 r1 f0
it went to z1 at a speed much higher than 0
the retract move is supposed to be at traverse, I think
but for me it sometimes sticks (never moves, never accepts the next g-code) and sometimes it goes directly to z1 (retract height), not to z-1 first
jepler: rain detector?
cradek: no, moving the computer back to the regular room
turns out sometime in the last 6 years, linux got broken so that if you don't have a ps/2 mouse attached when you boot, you can't attach one later
so I had to reboot bald
after going through all the trouble of moving UPS, CVS server, and gateway as a single unit to avoid powering anything off
you got the ceiling fixed already?
heh it's always something
no, I'll have to move it again
or maybe we'll just leave it
"larry"'s phone has been disconnected
if it's not too in the way, you can cover it
wonder if he's moved away
who is larry? ;-)
jmkasunich: the guy cradek told me does good drywall work
I had to do our kitchen ceiling after the upstairs bathroom drain leaked
yeah, this is a ceiling
you are smart to get someone else to do it