#emc | Logs for 2010-04-17

[00:26:36] <MarkusBec> MarkusBec is now known as MarkusBec_away
[03:18:48] <calcite> quiet
[03:24:57] <steverob> anybody home?
[03:25:55] <steverob> Guess not... going to bed ;-)
[07:47:45] <MarkusBec_away> MarkusBec_away is now known as MarkusBec
[07:50:54] <MarkusBec> MarkusBec is now known as MarkusBec_away
[07:51:09] <MarkusBec_away> MarkusBec_away is now known as MarkusBec
[09:43:20] <MarkusBec> MarkusBec is now known as MarkusBec_away
[09:43:27] <MarkusBec_away> MarkusBec_away is now known as MarkusBec
[10:10:35] <MarkusBec> MarkusBec is now known as MarkusBec_away
[10:22:30] <MarkusBec_away> MarkusBec_away is now known as MarkusBec
[10:49:03] <PCW_> PCW_ is now known as PCW
[10:55:28] <MarkusBec> MarkusBec is now known as MarkusBec_away
[12:26:50] <JT-Hardinge> spring has sprung around here
[12:59:19] <elmo40> long time ago.
[13:02:56] <MarkusBec_away> MarkusBec_away is now known as MarkusBec
[13:46:06] <JT-Hardinge> I think I can clean out the Wonderwood now
[14:13:26] <awallin> ayone worked with QT here? comments vs. gtk or wxWidgets?
[14:14:52] <The_Ball> Is there any value to having seperate limit switches for each axis? (X-min, X-max, ...)
[14:21:54] <alex_joni> The_Ball: if you can spare the I/O sure
[14:25:04] <JT-Hardinge> The_Ball: yes there is value
[14:37:06] <The_Ball> Ok, I'll make room
[14:38:19] <JT-Hardinge> one advantage is EMC won't let you jog off of a limit switch the wrong way with separate limit switches for min and max
[14:39:20] <The_Ball> good point
[14:40:14] <The_Ball> what about homing, I'm using index pulses for homing, but need to hit a limit first, is that a problem, first the limit switch is activated but EMC needs to continue past that to hit the index pulse?
[14:42:27] <JT-Hardinge> as I recall you have to set the limit as shared with home in the ini file
[14:43:47] <The_Ball> Ok, so someone has done the thinking :)
[14:44:37] <awallin> do large pro machines have multiple home-switches? i.e. they can G0 at 30m/min somewhere close, and then slow down a bit before contact with the real home-switch?
[14:47:23] <alex_joni> awallin: they (and emc2 too) use a multistep homing procedure
[14:47:34] <alex_joni> there's a velocity to find the switch (home_search_vel)
[14:47:42] <alex_joni> which is usually pretty fast
[14:47:49] <alex_joni> then it backs out from the switch
[14:47:54] <alex_joni> and looks for it slower
[14:48:09] <alex_joni> then it can even use an index pulse together with the switch
[14:48:54] <alex_joni> it's usually common to have the home switch active for half the working area of the joint
[14:49:10] <alex_joni> then at power up you know on which side you are, and do the homing a bit faster
[14:49:21] <awallin> home in the middle?
[14:49:26] <alex_joni> awallin: yeah
[14:50:32] <The_Ball> that's clever, hmm so many switches... good thing the mesa card has 72 I/Os
[14:54:07] <alex_joni> and you can always add a second one
[14:54:51] <The_Ball> hehe, just in case. actually there is a spi add-on as well which adds a "ridiculous" amount of I/Os
[14:54:51] <JT-Hardinge> awallin: my Hardinge lathe has a coarse home and a fine home switches
[14:55:17] <alex_joni> The_Ball: yeah, but that's not done in the emc2 driver yet
[14:55:20] <alex_joni> afaik
[14:55:54] <The_Ball> that reminds me, I've been meaning to write the dev about the new firmwares
[14:55:55] <awallin> would spi I/O be non-realtime? I imagine there is some latency?
[14:56:11] <Valen> SPI could be done with RT limits
[14:56:13] <alex_joni> awallin: it would still be RT
[14:56:20] <Valen> its done on a set clock
[14:56:22] <alex_joni> mesa updates are at about 1khz
[14:56:34] <alex_joni> surely enough for spi's to happen inbetween
[14:56:46] <awallin> but not pwm generation or encoder counting or such?
[14:57:27] <alex_joni> that is done in the FPGA
[14:57:43] <alex_joni> so once every ms you update the pwm rate
[14:57:51] <alex_joni> or you read the encoder counters
[14:58:25] <alex_joni> for special apps you can pro bably run it a bit faste
[14:58:26] <The_Ball> I'm trying to find the spec, I read exactly how long a read takes just a few days ago
[14:58:26] <alex_joni> r
[15:00:16] <pcw_home> Most of our SPI stuff runs at 8-24 mbits/sec
[15:00:18] <pcw_home> output stuff is easy but you need to either wait for return (input) data (maybe 4 usec max)
[15:00:19] <pcw_home> get the input data 1 sample late, or have a SPI state machine that repeats reads
[15:00:21] <pcw_home> to the return data is alway up-to-date
[15:00:39] <pcw_home> (so the return data)
[15:01:48] <Valen> 24mbit is pretty fast for SPI ;->
[15:03:53] <The_Ball> mesa sells the 7I64 board "The SPI interface supports data rates to 8 Mbps so a full read/write of all 48 I/O bits takes approximately 4 uSec"
[15:04:16] <Valen> I believe it, just sayin, its pretty fast for SPI is all
[15:04:18] <pcw_home> Fast is easy in output only devices, more trouble in devices that return data (do to isolation delays and cable time of flight issues)
[15:04:42] <Valen> I figured if it was I/O then it'd need to be 2 way
[15:05:26] <pcw_home> output only at 24 Mb/S is easy read/write is limited to about 8
[15:05:41] <Valen> I cant see pushing even 8mbit more than a cm or 3 without needing balanced lines and other assorted tricks
[15:06:03] <alex_joni> pcw_home: how about doing spi in the fpga?
[15:06:11] <pcw_home> (this will get better when I add read skew compensation to the SPI interface)
[15:06:15] <alex_joni> and only have some registers for the actual i/o bits?
[15:06:28] <alex_joni> is that feasible?
[15:06:33] <The_Ball> So I guess that's realtime with the breakout boards my 5i23 could have 3x6x48=864 I/Os!
[15:06:52] <alex_joni> The_Ball: surely more than enough ;)
[15:07:38] <The_Ball> ok then, switches all around
[15:09:12] <pcw_home> alex_joni: Ive though about that as a way of hiding the specific daughtercard SPI details
[15:09:15] <pcw_home> This is how th resolver interface works, it has its own dedicated processor and a dual ported RAM for host interface
[15:09:16] <pcw_home> so it looks like any other HostMot2 device, with registers for position and velocity
[15:09:37] <alex_joni> pcw_home: right, that's what I was thinking
[15:09:52] <alex_joni> especially since the 5i22, 23 still have some fpga space left
[15:11:42] <pcw_home> Yes (the resolver interface for example fill not fit in the 5I20)
[15:12:30] <The_Ball> your working on a resolver interface?
[15:16:04] <The_Ball> any chance it will work on a whedco resolver?
[15:17:32] <The_Ball> my whedco servo drives have firmware 1.5 which doesn't support outputting motor feedback as quadrature pulses, so I have no real feedback. Firmware 2.1 added that feature, but the hardware is different so I can't upgrade it :(. If I could sniff the resolver that would be fantastic
[15:18:50] <alex_joni> The_Ball: there are also resolver/quadrature converters
[15:18:56] <alex_joni> I know a couple in here use those
[15:20:24] <The_Ball> alex_joni, yes, I've seen them for sale around $150, not sure if they use the right protocol though, I have no idea what my protocol is... Anyways I bought a new set of controllers just a few days ago, exactly the same but firmware V2.2 for $150 each
[15:21:53] <alex_joni> iirc Jon Elson had some of these
[15:22:00] <alex_joni> ask cradek I think he uses them
[15:22:36] <The_Ball> is Jon Elson jepler?
[15:23:35] <SWPadnos> no
[15:23:49] <SWPadnos> he's jmelsin, when he's here
[15:23:52] <SWPadnos> err
[15:23:55] <SWPadnos> jmelson
[15:24:35] <The_Ball> Might try my luck on emc-users then
[15:31:06] <pcw_home> The_Ball: I dont think our resolver interface would work with your drives because we expect to be in control of the reference waveform
[15:31:07] <pcw_home> I think you would want a "vampire tap" resolver interface the read the sine and cos signals that go to the drive
[15:33:35] <pcw_home> (might be doable with different ucontroller firmware ill have to think about it)
[15:52:05] <piasdom> /msg NickServ identify domsub
[15:57:34] <skunkworks> oops
[15:58:29] <The_Ball> pcw_home, ah, I'm learning about resolvers... back in school I learned that a resolver was a "binary device" so to get 1024 lines per rev you would need 10 I/O lines. Things are not like that any more it seems
[15:58:39] <The_Ball> skunkworks, long time no see, how have you been?
[16:16:51] <Dave911> awallin: >> anyone worked with QT here?? I haven't, but QT4 designer sure looks nice.. have you tried it?? Glade is ok, but it is QT4 designer appears to be a lot more flexible.. plus it is being used for a lot of serious commercial apps..
[16:18:06] <awallin> Dave911: I have been running my cam-experiments from python, but now I am thinking of a more seriosu GUI. There appears to be a VTK widget for QT which is good. I did not try the designer yet
[16:19:33] <Dave911> If you have some time .... go to Youtube and look for "QT4 designer" I think that will let you find the videos... I was impressed vs Glade
[16:20:29] <Dave911> Once I grind through my existing project I may go back and move it over to QT4 and see how that works ..... just takes time..
[16:22:04] <awallin> 'qtdemo' runs fine on my desktop which has an Nvidia card and drivers, but it's just a mess on the laptop with integrated intel graphics. Both are Ubuntu 9.04 installs
[16:22:15] <awallin> Dave911: are you working on something emc2 or cam-related?
[16:24:00] <Dave911> yes ... a custom filter program for EMC2
[16:25:34] <awallin> I'll try the cutting simulation which is slow on python and see if it is reasonalby fast in C++/QT
[16:26:26] <Dave911> Let me know what you think ...
[16:27:17] <awallin> gtkmm is supposed to be "purer" c++, but I'm not sure about the VTK widget, and I need that
[16:46:37] <MarkusBec> MarkusBec is now known as MarkusBec_away
[17:01:31] <MarkusBec_away> MarkusBec_away is now known as MarkusBec
[17:07:12] <MarkusBec> MarkusBec is now known as MarkusBec_away
[17:21:42] <MarkusBec_away> MarkusBec_away is now known as MarkusBec
[17:27:50] <aystarik> awallin: http://www.libqglviewer.com/ -- might be closer to what is needed
[17:29:47] <awallin> raw opengl might be faster than vtk, but for now I am sticking to vtk since I roughly got the hang of it...
[17:38:47] <awallin> is there an automatic way of generating a header-file for a shared librar (e.g. ocl.so) ?
[17:39:00] <awallin> how there are a dozen or so .h files which are needed to build ocl.so
[17:39:23] <awallin> but I wouldn't like to distribute them all to people who want to link their c++ to it
[20:43:57] <JT-Hardinge> I'm convinced that Hardinge technicians were contortionist and had tiny upside down hands
[20:44:10] <JT-Hardinge> I did get one way cover changed though
[20:46:12] <JT-Hardinge> oh crap I think this side was the easy side :/
[20:52:01] <toast_> toast_ is now known as toastydeath
[20:53:37] <frallzor> * frallzor is tires
[20:53:39] <frallzor> *d
[21:06:47] <alex_joni> JT-Hardinge: heh
[21:06:55] <alex_joni> don't forget to take pictures
[21:18:17] <JT-Hardinge> I had to degrease myself just to type on the computer LOL
[21:28:48] <alex_joni> JT-Hardinge: I know, but remember you only do this once
[21:34:55] <JT-Hardinge> yes, because I know better now
[21:44:44] <geo01005> anybody have an idea of how much you would increase the inductance of a wire wound resistor by putting steel allthread down the middle of it?
[21:46:41] <steverob> Anybody home?
[21:47:31] <geo01005> a few..
[21:47:39] <JT-Hardinge> yep
[21:48:57] <steverob> just wondering if anyone has any experience with a rockwell milling machine?
[21:50:23] <steverob> Found one in pretty good condition for ~$850 and wanted to know if that was a good price?
[22:04:36] <JT-Hardinge> manual mill or cnc mill?
[22:05:13] <JT-Hardinge> hmmm steverob appears to be away
[22:16:30] <steverob> Sorry had to step out for a few minutes.
[22:16:37] <steverob> The mill is a manual mill.
[22:18:05] <steverob> I had been looking for a mill that I could afford and figured this was a good candidate.
[22:19:50] <JT-Hardinge> can you inspect it?
[22:20:01] <JT-Hardinge> I'm back and forth too
[22:21:02] <steverob> Yeah... It's 200 miles from here so, I've gotta be pretty committed before going to see it.
[22:22:25] <steverob> This is a similar machine. The one I'm looking at isn't this clean but "supposedly" everything is still mechanically sound.
[22:22:29] <steverob> http://www.bid-on-equipment.com/uploaded/37000/37019.jpg
[22:23:36] <steverob> I could use a few hints on how to inspect the thing.
[22:25:40] <JT-Hardinge> check the endplay on the screws and look in the middle of the ways for wear if a heavily used machine ways will be worn in the center.
[22:26:15] <JT-Hardinge> if worn can still be good for small parts but will not face off a long part to the same thickness
[22:27:46] <frallzor> http://cgi.ebay.com/DSP-CNC-Motion-control-system-Controller-Handheld-box_W0QQitemZ180476518937QQcmdZViewItemQQptZBI_Control_Systems_PLCs?hash=item2a053d2219 anyone tried using these things with their setup?
[22:29:31] <steverob> I've never owned a mill so, figured this might be OK to get my feet wet.
[22:29:50] <alex_joni> iPad printing: http://regmedia.co.uk/2010/04/16/ipad_printing.jpg
[22:34:02] <JT-Hardinge> steverob: my basic rule of thumb is if it looks like crap it might be if it looks well taken care of with normal wear and tear then well it's normal... if it has been dolled up look real close at it
[22:37:33] <steverob> From what I understand, it has not been dolled up. Supposedly, a little elbow grease and it'll be ready to go to work.
[22:38:36] <steverob> I'm not looking to do production work so, a machine that's "pretty good" should more than satisfy my needs.
[22:39:02] <JT-Hardinge> make sure everything moves like the top turret etc. my BP that I bought a few years back was so rusted up that it took us 2 days to free it up
[22:40:44] <steverob> The machine does have a broken handwheel on one end but, I'm told the table still works smoothly (using the handwheel on the other end). Obviously, I'll want to take a close to make sure the screws weren't damaged when the wheel was broken.
[22:43:51] <JT-Hardinge> yep
[22:45:02] <JT-Hardinge> just look it over real good and move everything that is suppose to move
[22:45:30] <steverob> Yeppers.
[22:45:48] <JT-Hardinge> well time for me to clean up and get the Kalbi ribs on the grill
[22:45:58] <steverob> See ya!
[22:46:42] <alex_joni> check for backlash too
[22:47:14] <steverob> Hmm... Just wondering how much would be acceptable?
[22:47:25] <alex_joni> for cnc none ;)
[22:47:33] <alex_joni> for manual.. depends on the work you're doing
[22:47:55] <alex_joni> if you do climb milling it will pull the part
[22:47:57] <JT-Hardinge> if you fit a dro to it and there is a cost effective kit for one you can deal with the backlass
[22:48:16] <toastydeath> check the tightness of the handle through the machine's x-y travel
[22:48:38] <toastydeath> center both axes, roughly, and put enough pressure on the locks so that there's a bit of drag on the handle
[22:48:47] <toastydeath> then crank it to one side, and back to the other
[22:48:58] <toastydeath> the change in drag on the handle tells you, roughly, if there's a lot of wear somewhere on the table
[22:49:01] <toastydeath> er, ways
[22:49:46] <steverob> The machine does have some years on it so, I'd be shocked if there wasn't some backlash.
[22:49:53] <toastydeath> the least amount of wear usually happens out by the edges of the table, so that's why you start in the center
[22:49:59] <toastydeath> backlash is adjustable
[22:50:09] <steverob> Ah...
[22:50:10] <toastydeath> the biggest concern is differences in backlash
[22:50:31] <steverob> From one end to the other / middle?
[22:50:33] <toastydeath> if there's more backlash in some areas than others, that's wear on the screw
[22:50:50] <steverob> OK.
[22:51:07] <toastydeath> the backlash test is simple
[22:51:18] <steverob> Obviously, uneven wear would be difficult to compensate for?
[22:51:21] <toastydeath> loosen the gib locks on the table, and turn the handle until the table mobes
[22:51:24] <toastydeath> *moves
[22:51:45] <toastydeath> zero the handle, and then turn it back until you meet some resistance (i.e. it catches the other side of the screw)
[22:52:17] <toastydeath> very difficult, unless you're using cnc'ing it
[22:52:25] <toastydeath> then there's the ability to create an error map
[22:52:35] <alex_joni> and even then,it's easier to replace the screws
[22:54:18] <toastydeath> on a knee type mill, also, expect backlash between 5-10 thousandths
[22:54:50] <steverob> Gotta feeling parts aren't going to be easy to find. That means serious bucks to fix anything that is too badly worn.
[22:54:55] <toastydeath> and if you buy it, don't adjust it below 5 or so
[22:55:08] <toastydeath> also don't freak out about worn stuff too much?
[22:55:30] <toastydeath> think about the parts you are making and how much accuracy you need in that envelope
[22:55:51] <toastydeath> and then consider it's a knee mill and isn't going to be all that critical to begin with
[22:56:13] <steverob> I'm guessing even a sloppy machine will be plenty good enough for the work I'd ask it to do.
[22:56:42] <toastydeath> yeah, this is why i recommend just making sure the slide wear isn't too bad
[22:56:50] <toastydeath> because very bad wear on the ways will cause chatter and break tools
[22:57:14] <steverob> OK
[22:57:15] <toastydeath> but if there's only a bit of difference in the drag on the handle, that's fine
[22:58:21] <steverob> My other concern is moving the machine. I've never moved one before and am wondering how to secure it for the move.
[22:58:40] <toastydeath> check on practicalmachinst.com
[22:58:54] <toastydeath> use the search, there's a lot of advice in the archives on how to properly move a knee mill without a huge amount of fuss
[22:59:05] <steverob> I'm hoping to put it in my PU truck and strap it down.
[22:59:12] <steverob> OK. I'll check that out.
[22:59:42] <frallzor> anyone around into DSP controls for cnc? :)
[22:59:54] <steverob> I guess my concern it that it looks awfully top-heavy. I'd hate to have it move around too much.
[23:00:14] <toastydeath> steverob, they are a bit top heavy, yes
[23:00:22] <steverob> Screw up the mill and my truck!
[23:00:26] <toastydeath> that's why there's a lot of discussion on how to set them up to move
[23:00:41] <steverob> Supposedly about 800# total.
[23:00:52] <toastydeath> sounds light for a knee mill
[23:05:04] <steverob> I had been looking at some of the desktop and smaller mills and just can't see spending >$2000 on a machine with so much less capability.
[23:06:01] <toastydeath> if you have the space, a knee mill is the way to go
[23:41:59] <MarkusBec> MarkusBec is now known as MarkusBec_away