#emc-devel | Logs for 2009-08-07

[02:38:52] <jepler> seb: ???
[03:44:46] <cradek> jepler: I slowed X,Y both down to 6ips and they home and follow relatively well, but neither sounds right
[03:45:09] <cradek> I tuned Z very easily and it sounds perfect
[03:45:36] <cradek> P,FF1 only - super easy
[04:28:13] <cradek> darnit, I forgot to go to bed
[10:57:43] <gezar> alex_joni : do you by chance have the link to the old nist kinematics transformations like the one you linked yesterday?
[10:58:29] <gezar> wait that link may just be enough to get me going
[11:00:34] <gezar> I finally understand that stuff, or at least most of it at least, huge steps forward for me im excited.
[11:02:11] <gezar> and yeah, it was gauss-jacobi using 4x4 blocks and I was not able to finish my code on that specific one, the ps3 is a very interesting platform, the ability to do work on 4 floats at ones in interesting. the speed ups for calculating 1000x1000 matrices was insane.
[11:18:26] <alex_joni> heh ;)
[11:18:43] <alex_joni> gezar: I can push some code later, maybe you'll be interested in looking at it
[11:18:50] <gezar> I would love to
[11:18:54] <alex_joni> right now I'm fighting with a stupid ttyS0
[11:19:13] <gezar> Ive been trying to help with all of this for well, a very long time,
[11:19:43] <gezar> granted my c is not that good yet, but it keeps getting better and better, same with my math
[11:20:33] <gezar> however my linear class was more proof based then actual application
[11:21:47] <gezar> what is ttyS0 doing?
[11:21:53] <alex_joni> nothing :/
[11:22:04] <alex_joni> it's /dev/ttyS0 - first serial port
[11:22:15] <alex_joni> nothing coming out of it
[11:22:19] <gezar> is it on in the bios?
[11:22:22] <alex_joni> brb.. trying a reboot
[11:22:25] <alex_joni> (it used to work :/)
[11:27:11] <gezar> working now?
[11:30:16] <alex_joni> yup
[11:30:19] <alex_joni> no idea why it hung
[11:30:34] <gezar> strange
[11:52:35] <alex_joni> strange indeed
[12:02:24] <jepler> cradek: but before swapping the amps, you had X tuned well too didn't you?
[12:16:25] <SWPadnos> argh. it's very hard to design PC boards when your mouse doesn't work
[12:21:21] <alex_joni> I'm sure there are text-mode pcb design softwares somewhere ;)
[12:21:51] <gezar> snake, and then use a screen capture ? :)
[12:22:10] <SWPadnos> if I knew the scripting language well enough, I could do it in Altium in text mode
[12:22:36] <alex_joni> see.. there's always a solution
[12:22:42] <alex_joni> (or get another mouse ;)
[12:22:49] <alex_joni> logger_dev: bookmark
[12:22:49] <alex_joni> Just this once .. here's the log: http://www.linuxcnc.org/irc/irc.freenode.net:6667/emcdevel/2009-08-07.txt
[12:23:20] <SWPadnos> the problem could be more accurately stated by saying that the USB subsystem is fubared
[12:24:46] <alex_joni> PS2 ?
[12:25:14] <SWPadnos> I could try that, after a reboot or three
[12:26:13] <SWPadnos> this is a Windows 2000 machine, so I can't even install the driver for the brand spanking new Logitech MX-518 mice I have
[12:26:28] <SWPadnos> and the MX510 driver doesn't work (eliminating about 5 other mice I have here)
[12:26:54] <SWPadnos> which is very frustrating, considering that they're basically the same as the MX500 I have used for the last few years
[12:28:16] <alex_joni> mine's an MX400
[12:28:19] <alex_joni> works great
[12:32:28] <SWPadnos> lucky you
[12:32:48] <gezar> how has emc2 been performing lately?
[12:32:54] <alex_joni> gezar: it really was gauss-jordan resolution
[12:33:21] <gezar> ?
[12:33:30] <gezar> yeah, gauss-jacobi
[12:34:08] <gezar> we also had a gauss-sidel that worked with the most current value to solve a 1000x1000 for Ax=b
[12:35:20] <gezar> the course was a tad over my head, I lacked a lot of the number theory stuff, and all of the linear algebra necessary, it was a very interesting study on SIMD computations
[12:35:50] <gezar> single instruction multiple data
[12:36:00] <alex_joni> gauss-seidel
[12:36:15] <gezar> how does emc do the math?
[12:39:59] <skunkworks985> logger_dev: bookmark
[12:39:59] <skunkworks985> Just this once .. here's the log: http://www.linuxcnc.org/irc/irc.freenode.net:6667/emcdevel/2009-08-07.txt
[12:42:31] <skunkworks985> skunkworks985 is now known as skunkworks
[13:10:10] <cradek> jepler: yeah, I guess so, but it was never as smooth as Z is
[13:10:28] <cradek> maybe it's just because Z is much heavier
[13:18:47] <skunkworks> cradek: what does it sound like?
[14:04:15] <cradek> rough/gravely/squawky
[14:04:58] <cradek> one amp doesn't have a stable velocity loop
[14:05:09] <cradek> the other thinks its fuse is blown but it isn't
[14:06:18] <cradek> I think the one that thinks its fuse is blown is fairly stable
[14:10:35] <cradek> also, the spindle drive has faulted with "overvoltage" and also tripped its breaker for some reason
[14:10:43] <cradek> progress has stopped being easy :-/
[14:12:11] <skunkworks> yecky
[14:12:34] <skunkworks> so many little problems - sounds like it got struck by lightning
[14:13:12] <skunkworks> do you have any specs on the servo?
[14:13:24] <skunkworks> volts/amps?
[14:16:40] <skunkworks> are you running this on single phase or your phase converter?
[14:20:26] <cradek> phase converter
[14:21:26] <cradek> I think the servo power supply is 300V 40A
[14:21:42] <cradek> so the motors are maybe 15A?
[14:21:49] <cradek> * cradek pulls a number out of his ...
[14:22:17] <jepler> 300V? I thought I saw a rating like 1xxV on the motor nameplate
[14:22:27] <cradek> hmmm
[14:22:40] <jepler> but I don't remember
[14:22:58] <cradek> I'm sure the power input says something like 270-310
[14:23:36] <cradek> I don't know what the motors say
[14:26:21] <cradek> skunkworks: how much for three skunkworks special amps?
[14:28:00] <skunkworks> heh - I was thinking I could just send you the one I have made up but it is pushing at 200v. Then it would be acutally tested ;) (the bootstrap diodes are only rated at 200v) cap at 350iirc
[14:28:33] <skunkworks> but - all we would have to do is get jepler to finish up his machine and he could mill out some boards...
[14:28:47] <jepler> too bad the voltage is so high. the big mesa amp is 22.5A per channel but only 165V (180 absolute maximum)
[14:28:56] <jepler> skunkworks: yeah yeah
[14:29:25] <skunkworks> :)
[14:32:08] <SWPadnos> well, the good news is that I figured out which switches you need to repair flaky mouse buttons
[14:32:15] <SWPadnos> (for a lot of mice anyway)
[14:33:37] <SWPadnos> DK SW501-ND or SW502-ND, depending on how stiff you like your buttons to feel
[14:34:12] <SWPadnos> cradek, I have a few 80V/30A analog amps
[14:35:37] <cradek> http://www.yaskawa.com/site/dmcontrol.nsf/(DocID)/TKUR-5EJQ5Z/$File/TSE-C717-13E.pdf
[14:36:05] <cradek> I'm not sure which one (current) I have, but they are 270v (270-310) input
[14:36:34] <SWPadnos> +-200V output
[14:37:25] <SWPadnos> the Yaskawas I have don't change voltage, so you need a 200V input drive to run a 200V motor, and you can't run a 100V motor from a 200V drive
[14:38:03] <skunkworks> ah - so you have a dc bus.
[14:38:21] <SWPadnos> plus 220V control power input
[14:38:23] <cradek> on page 5 you can see the power unit
[14:38:23] <SWPadnos> AC
[14:38:40] <cradek> it's not isolated, it just rectifies the 3ph 230
[14:38:52] <skunkworks> scary
[14:38:55] <cradek> yeah
[14:39:33] <skunkworks> the trace isolation of my amps might also be a little on the narrow side... ;)
[14:40:30] <SWPadnos> geez. make sure you correctly identify the grounded phase
[14:41:32] <cradek> I don't have one...
[14:41:41] <SWPadnos> you do now
[14:42:05] <cradek> ?
[14:42:22] <SWPadnos> they identify the lowest leg of the 3-ph input as the grounded leg
[14:42:40] <SWPadnos> I don't know if they internally connect that to ground
[14:42:45] <skunkworks> lowest leg to the floor? ;)
[14:42:51] <cradek> apparently not
[14:43:09] <SWPadnos> lucky you
[14:51:19] <skunkworks> http://www.cnczone.com/forums/showthread.php?t=87332
[14:53:00] <cradek> ?
[14:53:58] <skunkworks> that output stage is similar to mine.. I actually isolate the motor ground from the logic (with the ir driver ic) and I have blanking.
[14:54:22] <skunkworks> that is 250v working
[14:57:30] <skunkworks> I was actually thinking of testing to 30a by paralleling the current sense resistor. (they are only rated at 22a max)
[15:01:21] <skunkworks> isolate is a strong word - separate would be better.
[15:02:11] <skunkworks> The ir chip has a motor ground and a logic ground. This allows for a 5v difference between the two - great for low side sensing.
[15:06:21] <skunkworks> the uhu has the motor ground and logic ground tied togather. I asked why - and he said - because he never had any problems with it. :). also - without blanking he has to really filter the piss out of the current sense signal to get rid of the turn on spike of the mosfets across the sense resistor.
[15:07:18] <skunkworks> and that is about all I have learned so far. ;)
[15:07:57] <cradek> I'd definitely buy your product first
[15:08:48] <skunkworks> heh - I wouldn't
[15:08:59] <skunkworks> (buy either)
[15:18:04] <SWPadnos> hmmm. I wonder if it's intentional that a lot of design documents of the LHC are indexed by google
[15:42:23] <micges> cradek: on your toolchanger tool number is saved between runs?
[15:45:09] <cradek> yes it will be - I have some changes on the random_toolchange branch that I will use
[16:03:05] <alex_joni> this is interesting
[16:03:13] <alex_joni> just got something nice from Fred P
[16:03:24] <alex_joni> it's called go motion, similar to emc, but mainly for robots
[16:03:29] <alex_joni> does jerk limited TP
[16:03:36] <alex_joni> and proper joints/axes separation
[16:04:07] <micges> alex_joni: go on...
[16:09:39] <alex_joni> micges: gotta run ;)
[16:09:56] <alex_joni> it's (based on Fred P) not ok for CNC operations, but targeted towards robots
[16:10:06] <alex_joni> and I wonder if that wouldn't be the better approach for emc2 too
[16:10:15] <alex_joni> no g-code needed for robots, etc
[16:10:53] <micges> cool
[16:27:01] <mozmck> alex_joni: have you looked at orocos? if so what do you think about it?
[18:25:19] <alex_joni> mozmck: nope, I haven't
[18:32:04] <mozmck> alex_joni: http://www.orocos.org/ It's specifically for robot control. Looks pretty interesting to me.
[19:32:42] <alex_joni> this is also interesting, one of the byprojects: http://sourceforge.net/projects/usarsim/
[19:32:53] <alex_joni> using Unreal Tournament for simulating robots
[21:36:12] <mozmck> alex_joni: looks like fred proctor is one of the developers on that
[21:37:28] <SWPadnos> Fred is one of the original EMC developers at NIST
[21:38:10] <mozmck> yeah, I know. does he still work on it now?
[21:41:58] <SWPadnos> I haven't seen much from him for the last few years
[21:42:15] <SWPadnos> but it does look like he's working on some similar stuff at NIST still :)