MarkusBec is now known as MarkusBec_away
If I took one of these 3 section shelf rails, and shortened the inner slide to 1/5 length, and the middle slide to 3/5, then I could use it for the X motion
ah, I'll find out "the truth" about shelf rails when the table is done
if the table works well tho, these things are cheap
I've come to the conclusion that you always need at least 3 per axis, and they come in pairs
they are plenty stiff in one axis (when new) but springy in the other
using them as linear rails?
and springy in torsion, so 2 in T shape isn't good enough even
of the build
In fact I will tomorrow. not much is built but the end drilling rig (lever press bolted to workmate) is quite cool
my drawings are way out of date now. it was a plywood box but now the horizontal rails are on beams
all the work so far is turning bits of found wood into faux extrusion, which will just bolt together
then if it's worth it later, gradually relace wood with aluminium
or steel. I can weld flanges and tap bolt holes in that now
Almost bought an Epilog Zing today
I was at http://cms.thewoodworkingshows.com/cms/Home/Dallas10231025/tabid/160/Default.aspx
and I got to see one in action
Almost cried, its incredible
why didn't you? forget to bring the pink slips for the wife & kids?
I need to hunt for the best deal
Maybe a used one
I don't know yet
a laser table is just an xy table...with a laser
make a Torchmate and decide whether you need a $10k laser later
steering the laser beam is very 21st century tho :)
$150 a month terms
what's that compared to a house in your neighbourhood?
I'm making a Joes CNC
mine is most like a solnya
I want to add a plasma torch, laser, and a few other attachments to mine
but I didn't exactly build in space for a cooling tub
mine is made of wood, so no water
air cooling, from the blower, is all it'll get
you can get tubs of all sizes. stand your machine in one
Well, my cabinet base is MDF, everything else is unistrut
reservoir, sump & footbath all in one :)
[03:13:45] <El_Matarife> http://lascolinascnc.com/
is what my machine looks like
[03:16:07] <MattyMatt> http://www.solsylva.com/cnc/13x13x5.shtml
2nd one down is closest to mine now
oh I've seen one of those
Bob Campbell had one at the North Texas CNC Users Group
mine will have planks at the sides, to carry the vertical shelf slides, although I'm thinking of ditching that idea and having them on the inside of the beams, attached to a box with the nut in it
these shelf sliders have very poor access to the mounting holes once it's all together, that will make alignment hard unless each slider is on its own beam or plank
sorry :) thinking out loud
I read that the ideal place to provide vertical support is at 2/9 and 7/9 of the width
odd, .2222222 and .777777, some resonant thing?
dunno. I got it from an old workshop practice book, for straight edges (those narrow surface tables for measuring)
I assume it's a least-total-deflection thing
[03:46:05] <cradek> http://en.wikipedia.org/wiki/Airy_points
ah cool. so it doesn't only apply to cast iron :)
Ruben is now known as Guest96409
[04:33:42] <MattyMatt> http://en.wikiversity.org/wiki/Introduction_to_Elasticity
I just hurt my brain on the bit about tensors
ctrd & 5/9 between, yep thats at 2/9 & 7/9
I haven't found any explanation why yet :)
bookmarked 'where to put boards under long straight things'
it's important to note the airy points are not really useful in most situations
especially for things like straightedges
but it would be better to have the rails under a table there?
the airy points preserve the parallelism of the end faces of length standard
i.e. on a one meter length standard, if you put the supports at the airy points, you will measure the same length value anywhere on the faces
as you move the points around, the faces will begin to droop or raise, and now you have no idea where on either face to take your measurement
so what's best for preserving the stiffness/flatness of the top surface?
it should be machined or ground flat with the full torque spec of any restraining bolts.
trans. sanded after glue+screws :)
the issue is with any "point" or kinematic based mounting, that once you get to a loaded situation the whole thing goes out the window
so the idea is to machine it in place, with all applied loads
when would the end face parallism be important on a long skinny thing?
tom3p: length standards
(or why isnt it a big fat thing )
ok, I've had to stack rods to measure huge dia's
yeah, this isn't that kind of standard
these are meant to be use horizontally
so , the beginning of this was the idea that, say 2 bearing for a slide should be separated at the airy points... not true?
indeed, not true
it does not preserve the parallelism in that direction.
I'm using 4 rails, and each one is effectively 2 points
it preserves the end parallelism of a length standard
why did my old EWP book say that straight edges use it then?
whats a good place? ( how wide is a saddle and how much hangs over the bearing )
straight edges use two points to rest on
on the saddleback
because cast iron settles over time
and as long as you return the straightedge to the same two force points, it will tend to remain flatter longer
the straight edges my boss uses are supported 1/4th from the end
tom3p: two points is not a great idea - you want it supported all the way if possible
that's why even the surfaces ways or linear guides will mount on are ground or scraped
or sanded >:)
oh, i think i misread your question, sorry tom3p
I would try to support it at the extremes
well, two roller bearing packs...(std machine building today is 2 roller rails seperated by some distance, yes on a ground surface, one rail up against a ground 'L', under a slide that is wider than that distance )
and if the "extremes" get too far apart, the machines we have add another rail
that's my worry. a saggy middle
dont touch the 3 rail!
then I'd move them 1/4th of the way in or thereabout
if it's really a problem
is this a mill or a lathe or what
1/4 2/7 it's all the same to a woodworker
also you need to get that out of your head
because adding 1/4th of the distance halves the rigidity of the structure
i.e. a 1" bar is half as rigid as a 1.25" bar
the horizontal supports would go at the ends
ah, I'll just build it, and worry about how it could've been better next time. wood burns
also for a concrete answer
in machinery's handbook they have the equations for deflection
with some rearranging, you can find the position from the ends where the sag in the middle will equal the sag on the outliers
That'll happen some time after I add FEA to blender
i did it once and forget the result, but it was a little bit further out than 1/4
not 2/9ths far, though
good idea, thats what that little drawer is for! ( between .22222 and .25 seems pretty picky )
the ways on really, really big mills
I see again as being at the very edges
and the tables/pendants have a large arch under them to make up the rigidity
like, on the planer table that's just hanging out, the ways are at the extreme outsides
and the distance to the working surface is pretty thin
and then all the crap in the middle is thick and hangs down below the ways
yeah keeping the ways above CG gives free rigidity
perhaps that's something you could do with your table, put the linear guides up on like, 5x5 boxes
and then add a bunch of trussing to the underside of the table itself
also i guess the main frame, too
if the ways are above the table, it'd be easier to move the gantry
wait - is this machine purely a gantry?
or does the table have some movement to it
the table does Y, the gantry does X&Z
why are you doing Y
on the table
that's quite a distance to bridge, usually the gantry does Y and Z, and the table X
if the table's going to move
same thing probably
originally, the shelf sliders would retract with the table, for easy domestic storage, but now I've got a machine 2x the size of the table
so now I'd prefer a moving gantry but I wouldn't trust these shelf sliders with the weight
the sliding table approach is more accurate
and if the gantry is detachable, I could do 3d scanning too
or put on a really high gantry and have a rostrum camera
i liked some machines i saw with 'concrete' side walls, the tops were the y rails, the machine tool was all overhead. fast 5 axis machines, knuckle heads /nutating.
the really, really big mills?
well 4m x 10m x 2.5m strokes
it was moved into TIMTOS for the machine show, they may have larger\
that's pretty large
seen bigger taurus
even for texas
the ones I am talking about are like, warehouse type?
haha timtos taiwan
where they are like a gantry crane, except a milling machine
yeah, that lets you drive into the working area
but the machine is just dropped on top of these walls , seemed smart
that's pretty cool
i've seen CMMs like that, but never a machine
that's exactly what I want to do with my friend's garage
use the garage door rails ? ;)
it would do to support a camera tho, which is one use for a room sized machine
okay so there's a cool way to create a straightedge
for machine calibration
if you put a long shaft, fairly straight, on bearings on both ends
and spin it
put an electronic indicator on it, and then scoot the machine down the axis
you can calculate the center of rotation of the shaft
which is a line between the two bearings
no matter how bent the shaft is
or how saggy
this depends on straight ways for the scooting?
this is so you can calibrate a machine
so if you want to see how the X axis affects Y position, you set this "straightedge" up along X
and as you jog down X, you can see the Y position fluctuate against a "true" reference
so if somebody was able to come up with a hobby verson of the software required, using hal or some such
anybody would be able to construct their own master calibration straightedge to set the error correction tables
there were apparently a few nonsensically accurate CMMs created using these as references
integrated into the machines, and constantly checked
could you sketch that up? i'm not seeing it.
will MS paint work
I am. if the rotating shaft is parallel to the ways
even roughly parallel, just enough to keep it in the range of the gauge
i get the idea of the indicator knowing where a tru ctrline by touching hi pts, anywhere on its wobbly length...
tom3p: that's all it is
we are talking about kinematics, rite?
it's two indicators, one on either side
mireland: the opposite of kinematics
i thynk i get it
the software compares the high/high and low/low for each indicator, 180 apart
i am terrible at drawing but i am trying.
not the roughly parallel bit... i dont get that
that was me
right, my bad
i can't draw this, sorry
np, keep it in mind, i'll try to draw it up, then pdf or jpg it
I'm about 3 machines away from needing it
mk1 will make me happy if it moves without seizing and carves to the nearest mm
best i can do
it should be aligned to an axis master rail, and would show the deviation first ( how straight the master rail is to motion of same axis )
the machines they used this on, the axes were on calipers
so that the ways weren't actually parallel to each other
and that perfect line, could be later be used to check y against x... im beginning to see this
and as an axis would drift, the calipers would move and keep the whole thing parallel to these rotating standards
the dwg is fine
it would be trivial to align the two ends of the bar
to a machine
do cheap digital 'dial' gauges have outputs?
yes, a slow serial format 'spi' i think
dunno how "cheap" i'd call an spi caliper
,uh , the cheapest have no outs,
so I won't be trying it this side of christmas :)
'sears better' has 'em
and it's more like "press a button to record value"
rather than a continuous stream
my mics have SPC but i've never tried connecting it to anything
little audio cable jack
back to the spinning bar, good idea, a laser might be nice too, but not as easy to tell it's 'edges' ( not as easy as the spinny stick )
yeh serial is often 3 wire, sometimes 2
you can do it with a laser if you're willing to build your own interferometer =)
can you do that with a piece of CD? :) (joke)
any fairly monochromatic laser will work.
iirc, you can get all the parts to build an interferometer on ebay fairly inexpensively
programming the software and getting the right mirror setup, not so much
or easily, rather
if it requires the kind of bed stiffness that holography does, then I'm out
no, it doesn't
doesn't require any stiffness at all, really
all the interferometer does is measure relative tilt of a mirror
not many machines use them now because they're so complicated to set up to measure anything
why wouldn't an interferometer setup need high rigidity?
...why would it need high rigidity?
tilting mirrors... i suppose flex of one sort or another is the thing to be measured
I had the idea that distance could be measured very accurately by interferometry
for that, rigidity would seem to be a plus
sure, it is
but it's not a huge requirement
toastydeath: do you have a pic or a link which describes this kind of interferometer? I built one interferometer for measuring the shape of a telescope mirror which I hand-grinded/polished
awallin: not off the top of my head, all this is from personal conversations and books
the beam setup is split and bounced off two parallel mirrors (or two spots on the same mirror)
iirc they had a specific curvature to them so only the distance changed, never the angle of reflection
i think there were a few other setups that did the same thing
this is for screw-comp? not measuring titled-kinematics?
mireland: earlier high-accuracy machine tools used interferometers for all machine movements
awallin: no, it's for measuring tilt/parallelism errors in a live machine tool
well, it's not like I have extra time for this anyway :) ...
it's often useless, the machines they used that method were square to .0001" over the entire travel
not sure how well an interferometer of any sort would work on a machine less accurate than that
plus dust or anything else
the bar method is really resilient to that kind of thing
ibm apt360 manual example 1, thru aptsketch, to emc .ngc file http://imagebin.ca/view/pZlsUTh.html
hello :] who uses Gammu ? :]]]]
tilting mirror is a collimator.. interferometry is the interference of wave with change of distance
I wonder how fast one would have to have an interferomiter run to keep up with the vibration in a live mill
if you inclined the inteference plate at the end you can get direction
hard part is making sure the moving mirror doesn't change its angle and send the laser off to whoopwhoop somewhere
that and measuring the frequency of the laser, they tend to wander some
by a few %
well that would also measure the error in height of the plane
nah, a straight error in height won't move the reflected laser spot
you need a sensor on the carrage (or a laser there) to measure that
hmmm, unless you intentionally angle it
that might work
see how a collimater is used to do just that,
combining the collimater and an interferometer
still wont help with a pureley vertical offset
you would need to angle the mirror on the slide
the angle of mirror on the moving carriage varies as it rides up and down a hill
ahh true, I hadn't thought of that, I was thinking of the mirror remaining vertical but moving up and down
its a basic long lathe bed test
I think i'd rather shoot a beam directly from the saddle at a CCD
do that at both ends and you get angles and heights in all dimensions if you combine it with a distance measurement
I have a small collimator and a precision prism, have tested dividing heads, and they are often crap
I was wondering how you can measure 360 degrees of rotation optically
hmm, projecting a line across a CCD would probably work
I was testing over smaller angles as the error is per tooth of the worm wheel
all those half covered pixels would give you lots of precision
yeah, I was trying to see how you could make an encoder that had "no errors"
anyway movie time bbl
its morning !
MarkusBec_away is now known as MarkusBec
tomp is now known as tom3p
[14:46:28] <skunkworks> http://www.cnczone.com/forums/showthread.php?p=681922#post681922
sounds like papabear had some issues with emc (maybe I am reading too much into it)
he only said he recommends Mach3 is somebody else is paying
a lot of people are happy to pay for the right to shout at someone later when it goes wrong
heh - good point.
I try to yell at the emc guys - but they just laugh at me. ;)
I'm doing machines for money and I don't recommend any mach3 use ;) I don't like to be shouted all the time ;)
I was appointed Evil Producer for a 'mock commercial' game programming project, but it was an impossible job without either physical access to their tender parts, or the power to starve their families
commercial pressure can help, but you can't fake it
if you're running Mach3 and your OS ends up being buggy, who do you yell at?
the Mach3 people of course. That's the deal when you sup with the devil and write WinAPI
Mach3 can yell at the OS writers, if they want
what do they do for real time support?
employ more staff, with the profits from the software
naturally this means you are less likely to speak with someone who actually wrote the code, but c'est la vie
that happens even worse with open source sometimes, when the author goes AWOL
Radisys and a few others had real time kernels for windows or hypervisors
ah sorry, I thought you meant hotline helpdesks :)
did BeOS ever get a foothold in the CNC market?
[15:09:36] <L84Supper> http://www.tenasys.com/products/demos/intime-jitter/
the last time I had adequate real-time performance from a PC was running DOS
otherwise, if I wanted to control something with µs accuracy, I'd rather use a games console
or any embedded system with a known spec
[15:16:21] <skunkworks> http://www.electronicsam.com/images/KandT/testing/goal3.jpg
[15:16:39] <skunkworks> http://www.electronicsam.com/images/KandT/testing/goal3screenshot.png
the very idea of latency testing with "try to create the worst case scenario" advice is what I mean. computers are supposed to be deterministic
but are stachostic
skunkworks: sounds like papabear had some issues with emc (maybe I am reading too much into it)
No, you are reading too much into that. I know Poppabear. He just thinks that Mach3 is best with a Galil card..
I've used Mach3 on commercial machines - but not in a typical CNC app - more of a motion control application. Right now Mach3 is going through
a rough spot in the road. Many bugs, but not many fixes.. Time will tell...
* archivist_emc has heard stories offline about lack of bug fixes in mach3
what kind of bugs does it have?
screw cutting is a major bug
bug fixing is no fun
hello to everyone !!
i have black sky problem!!
cradek: for now I like bugfixing, but who knows how long ;)
and sometimes the thing that looks like just a bug is the result of a fundamental limitation because of the whole way the software works - for mach I suspect bad screw cutting is one of those
fundamental limitation = bug
no, not at all
i-pink: here also sky is black
well maybe to a user, but to a developer they are very different
well, it depends
no use arguing semantics, sorry
no no.. i mean i have 2 LPT port, one is on board, and one via USB
here, the sky is sort of lit today
i-pink: with emc, you can't use a usb parallel port. get a pci card instead.
hehe PCI to the laptop...
then you need a different computer too, sorry
one sec i start to soler it..
PCMCIA, EXPRESS CARD
you can usually tack on new geometry representations in 3d programs. cartesian is a universal protocol for exchange
so e.g. you could always add a screw function to mach3d, whatever the internal limitations
The Mach3 bug list is very long. There are some fundamental issues in the way the core software was put together that they have been struggling with for a long time.
They are trying to fix that right now, but it is a huge job and they don't have a bunch of guys doing software dev for it. Threading is a minor issue compared with the core problems.
i see the address of the on board is 0x378, but how i can get the address of the USB?
i-pink: again, you cannot use a usb parallel port for emc
not even if the driver fakes a port?
Dave911: how many people do they have working on bugs?
cradek, ok but how i can see the adders??
i-pink: it doesn't have an address
cradek: There is really only one main developer, and he is totally consumed with redoing core Mach issues. Hence bug fixes have been put on the back burner. All bets are being put on the next big release.
Yes, not a good position to put one's self in. If the next big release doesn't work, where are you at?
are you talking Art as he has sold it so not working so hard on it
Heh, got out from under it in the nick of time (so to speak). lol
Brian is the guy now. Art is not out of the picture, but he is on the sidelines. Brian is "the guy".
I got the impression art is still working on stuff he wants to work on in retirement.. While brian is doing everything else. :)
it's not retirement, it's equity release
Brian is a cnc manager seen him, not sure how good at software he is
Brian is a sharp guy, but he is one guy with a lot of irons in the fire. He is doing a lot more than sw dev.
Compare this with the "group" who is contributing to EMC for a moment. The peer review and idea kicking around that goes on within the EMC2 dev group is a very,very powerful thing.
I think you can maybe have that in a closed-source project too, but you need more than one or two people developing
I have been using emc2 for quite a while. Still very very impressed with the advances over the last few years.. Wonderful work
here people sometimes help fix the thing that's bothering them - all users have the same empowerment
cradek: I totally agree.
I notice it's after noon and I'm not making any progress getting my coolant tank cleaned out...
(can I write software to do that?)
cradek: No, but you can pay the neighbor kids to do it for you
yes to control a robot with a water hose
jymm, and how much is this INtime addon? :)
I still think all the RT stuff should be offloaded to an embedded cpu running a real RTOS with no gui except ssh
it should be smart enough to cut a smooth curve unattended by the client
* MattyMatt is newb telling pros whad they shoulda
it would be the best solution, yes
you need to understand whats under the hood
MattyMatt: that can be done with emc2
have the GUI on another PC
yeah I don't know how high level emc goes, or whether it'll ever split cleanly between raw motor control and high level
oh, it does already
ah nice, so it just needs compiling for arm linux :)
if you have a smart controller via USB or whatever, you need loads of smarts embedded
including handling a lot of not so trivial stuff
like feedoverride handling
adaptive feedoverride (changing in realtime while moving)
synchronized motion like threading, rigid tapping, etc
or MIPS even. cheap routers use MIPS
all those don't fit a 'smartish' controller where you offload some commands into
yeah synchro motion is the limit of what I'm thinking. the embedded cpu runs all the motors
i'd love to see some kind of a simple open source program for an AVR (or similar) to be connected to a pc's serial port for doing the raw motor control... to only do the very RT thing, just following line segments or so
celeron55_: check the list above for reasons why that won't properly work
you still dond see that there more than an avr's worth of realtime
i think it'd work for a hobby mill controlled by a laptop that has a bit of too much latency
celeron55_: I'm sure it is
it's just not emc2 ;)
so anybody interested in that can take whatever parts of emc2 they see fit, and complete the project
(obeying the GPL of course ;)
and there are at least 2 projects doing that
i might try doing something like what i described sometime... but probably not 8)
I will certainly try it. I expect the overhead of talking over eth is less than running an X gui
mine is getting a 500Mhz cpu
it's an intel nic it'll have, so low latency fo sho
the slowest I've run emc2 on was a 200MHz SBC
so a 500Mhz arm should do nicely too, when it's ready to go in rom
or a 200Mhz mips from a reflashed router, although they have no obvious parallel ports
the problem with an ARM is finding a RTOS for it
it's not that common and streamline as for x86
routers already run VxWare usually
RTAI has an ARM port, but for some older ARM chips, and it works with ancient kernels
arm may be risc but it has some horrors
ARM (StrongARM; ARM7: clps711x-family, Cirrus Logic EP7xxx, CS89712, PXA25x)
there's no particular reason to choose arm. no cnc machine needs to worry about power use
so a nano itx is probably the way to go
although mips & ppc are obviously cheap enough to use in consumer gear
I'd chose one of the Atom boards
I know lots of people used them
successfully with emc2
Intel D945GCLF2 mini-ITX
yeah they were offering samples of atom + bridge for $45. I'd have liked a cnc mill then
take the reference design, add a few ls373 for extra parallel ports. job done
would a 66Mhz NDS do the trick? no RTOS just hard coded interrupt handling
ah I can find out :) you can get linux on the NDS now
right I gotta find a working nic for the emc machine. bbl
I have used the d945glf2 - works great - eric on the list built a 2 core kernel for it that has a lot better latency
With the stock ubuntu install It isn't steller (a bit under 20us)
I think eric was getting sub 10us
unde 3us iirc
maybe someone can work on an optimizer for the kernel scheduler that could be tweaked for given applications
well, RTAI runs outside the linux context
so the kernel scheduler is irrelevant
right was thinking of RTLinux
well RTAI also has a scheduler so I don't understand your point alex
so it's bus mastering that breaks RTAI? hmm, maybe intel nic not ideal after all
killer NIC would be the thing. it handles all the network traffic
bus mastering is one lock you can't override with any priority
although you probably could with a modern chipset with multiple ram paths
idle speculation again, sorry :)
idle speculation is good reading when you have nothing else to do :P
no it isn't. if I'm wrong I pollute your brain and this channel
IRC = clear question + informed answer. hrrmph
you just have to remember it's just speculation :P
MattyMatt is now known as IdleSpeculation
IdleSpeculation: at least it's supposed to ;)
IdleSpeculation is now known as MattyMatt
how i make print in ruby?
tomp is now known as tom3p
MarkusBec is now known as MarkusBec_away
does anyone know who used to run nerdulator.net, or where this page has been moved to? http://18.104.22.168/search?q=cache:3sXxM9k3UwIJ:ns2.huboftheearth.com/9x20/cnc.html+ngineering9x20_CNC.html&cd=1&hl=en&ct=clnk&gl=us&client=iceweasel-a