#emc-devel | Logs for 2006-04-12

[00:00:10] <jmkasunich> re: that IR 600V "40A" part
[00:00:25] <jmkasunich> yeah, I suppose I put the FET limit a little low
[00:00:32] <skunkworks> ok
[00:01:01] <skunkworks> I was looking at igbt's and - yeh high voltage - don't need that
[00:01:02] <jmkasunich> however, that part would dissipate 176W at 40A, not counting switching loss
[00:01:16] <jmkasunich> so realistically it would be used at much lower current
[00:01:29] <jmkasunich> I have used a 1500V "4A" mosfet for a power supply
[00:01:42] <jmkasunich> actually ran it at about an amp, with a moderate heatsink
[00:01:45] <skunkworks> says it handles 500+ watts ;)
[00:02:07] <skunkworks> max
[00:02:08] <jmkasunich> at case temp of 25C
[00:02:12] <skunkworks> right
[00:02:30] <jmkasunich> IOW, the proverbial (and hard to find) infinite heatsink
[00:02:43] <skunkworks> just need a small heat exchanger ;)
[00:03:03] <jmkasunich> we typically use FETs in power supplies
[00:03:08] <jmkasunich> have to assume 50C cooling ari
[00:03:10] <jmkasunich> air
[00:03:20] <jmkasunich> and heatsinks perhaps a 2 inch cube at best
[00:03:28] <jmkasunich> so 10-20 watts is about the limit
[00:04:01] <jmkasunich> anyway, let me restate:
[00:04:21] <jmkasunich> IGBTs are the only game in town for 1000V and up, at currents much over a couple amps
[00:04:33] <jmkasunich> FETs really come into their own at low voltages
[00:05:23] <skunkworks> :) thanks - took me a second to figure out the a hexfet is jut a trademark or ir (and it is just a mosfet)
[00:05:27] <jmkasunich> if you have less than 500V, FETs are the way to go
[00:06:04] <skunkworks> of
[00:06:15] <jmkasunich> IIRC, the on resistance of fets can be very low at low voltage, but then goes up as the square of the rated volts
[00:07:06] <jmkasunich> meanwhile IGBTs have a almost fixed saturation voltage drop, varies only a little with the rated voltage
[00:07:56] <jmkasunich> FETs tend to be more rugged too - IGBTs get very unhappy with even very short overloads over 2x rated
[00:08:16] <jmkasunich> many FETs can handle 5-10x rated as long as you remove the overload before it overheats
[00:08:23] <skunkworks> from what I understand - the igbts act more like a transister when on. as far as the v from source to drain
[00:08:31] <jmkasunich> yes, exactly
[00:08:45] <jmkasunich> IGBT = insulated gate bipolar transistor
[00:09:03] <skunkworks> ah - I am pretty smooth.
[00:09:04] <jmkasunich> think transistor output characteristics with MOSFET input characteristics
[00:09:36] <skunkworks> that is kinda what I got out of it. all I would have had to do what really read what igbt stood for.
[00:11:20] <jmkasunich> one thing to watch out for with the IR2100 type drivers
[00:12:32] <jmkasunich> they get upset of the negative supply of the high side goes below the negative supply of the low side
[00:13:56] <skunkworks> what would cause that?
[00:14:06] <jmkasunich> stray inductance
[00:14:12] <skunkworks> AH
[00:14:18] <skunkworks> ah
[00:14:51] <jmkasunich> imagine the top FET is on and supplying 20A from the supply out to the motor
[00:14:56] <jmkasunich> then you turn off the top FET
[00:15:09] <jmkasunich> the motor is an inductor, so that 20A ain't gonna stop
[00:15:24] <skunkworks> I am putting free wheeling diodes in - should that not take car of it
[00:15:31] <jmkasunich> instead the voltage at the output flys down, until the bottom FET freewheel diode turns on
[00:15:51] <jmkasunich> when the diode turns on , in theory the output goes one diode drop below ground
[00:16:16] <jmkasunich> but that assumes 1) no turn on delay in the diode and 2) no inductance in the diode and low side fet
[00:17:33] <jmkasunich> the driver chip should be as close to the fets as possible, and the connection from the top of the bottom fet to the bottom of the top fet should be as short as possible
[00:18:23] <jmkasunich> it can get intereting if you have multiple totem poles
[00:18:53] <jmkasunich> all their grounds (low side sources) need to be close to gether, and the system ground needs to be connected at that common point
[00:19:34] <skunkworks> that was the next plan - pcb. (while I am waiting for parts)
[00:20:14] <jmkasunich> the bootstrap diode needs to be fast too - again, the higher the voltage the more important that is
[00:20:15] <skunkworks> lots of big fat copper traces
[00:20:45] <skunkworks> I was having trouble finding one but I did find a 100v piv schotty(sp)
[00:21:21] <skunkworks> to play with
[00:21:29] <jmkasunich> thats a pretty decent app note
[00:21:47] <jmkasunich> chapter 5 talks about negative transients on the Vs pin (what I mentioned before)
[00:21:58] <skunkworks> I saw that ;)
[00:22:26] <jmkasunich> these parts seem much more robust than the ones I recall from some years ago
[00:22:37] <jmkasunich> they would poof if you got more than a volt or two negative
[00:23:01] <skunkworks> well I need all the help I can get. alex hates them - he said he destroyed a bunch and never got them to work. ;)
[00:23:24] <jmkasunich> not knowing anything else I'd blame it on those negative transients
[00:23:53] <jmkasunich> we considered them for a 240V fractional HP drive some years back (don't recall how long, before 2001 for sure)
[00:24:03] <skunkworks> I am shooting for a tight pcb with fat traces :)
[00:24:09] <jmkasunich> they worked ok, except during a short circuit shutdown
[00:24:26] <jmkasunich> high current during short ckt leads to high dI/dT and high negative undershoot
[00:24:31] <jmkasunich> and they were toast
[00:24:35] <skunkworks> ouch
[00:25:12] <jmkasunich> for 480V class drives we use 1200V IGBTs so drivers like this are completely out of the question
[00:25:20] <jmkasunich> we use true isolated supplies and optocouplers
[00:25:54] <skunkworks> do you use other packages or build them with descrete componants?
[00:26:05] <jmkasunich> packages for what?
[00:26:11] <jmkasunich> the IGBTs
[00:26:13] <jmkasunich> ?
[00:26:23] <skunkworks> other higer voltage ir2111 like packages
[00:26:33] <jmkasunich> there aren't any
[00:26:52] <jmkasunich> 480V drives use optocouplers, not 2111 type drivers
[00:27:13] <jmkasunich> there are some optocouplers that have IGBT/MOSFET driver output stages
[00:27:31] <jmkasunich> but they need isolated power supplies, you can't really bootstrap them
[00:27:42] <skunkworks> bear with me - so your using the optocouplers to directly drive the -- yah what you said
[00:28:08] <skunkworks> neat
[00:30:07] <jmkasunich> http://www.avagotech.com/products/product-detail.jsp?navId=H0,C1,C5231,C4945,C4988,P88831
[00:30:34] <jmkasunich> fsckers... HP became Agilent became Avago.... need a scorecard to keep track these days
[00:31:01] <jmkasunich> the HCPL-3120 is a higher powered one, 2A peak output IIRC
[00:31:23] <jmkasunich> the 3150 is good for about 50A IGBTs, the 3120 up to a couple hundred
[00:31:38] <jmkasunich> we use special drivers for our stuff (1800A worth of IGBTs in parallel)
[00:32:28] <skunkworks> nice - I would be scared working around that equipment.
[00:32:58] <jmkasunich> most of the time its inside a steel cabinet
[00:33:06] <jmkasunich> I don't probe live on anything like that
[00:33:15] <jmkasunich> power off, clip scope probes, power on
[00:33:41] <jmkasunich> I blew up a scope probe in my hand 10 years ago, that was enough
[00:34:08] <jmkasunich> fortunately no damage to my hand, just scared the snot out of me
[00:35:47] <skunkworks> thats good. I think the closest thing I did was destroy a rectifier in school when I plugged the scope into an unisolated plug.
[00:35:59] <skunkworks> ground was ground ;)\
[00:36:21] <jmkasunich> you were floating the scope?
[00:36:46] <skunkworks> yes
[00:37:01] <jmkasunich> tsk, tsk, tsk... thats not a good thing
[00:37:25] <skunkworks> don't know why we were not using differential
[00:37:33] <jmkasunich> about six months ago a professor at CSU (Cleveland State Univ) killed himself with a floated scope
[00:38:30] <skunkworks> wow. like I say - I don't know why we where not using differential - maybe they where only one channel scopes. been a while.
[00:39:01] <jmkasunich> really good diffy boxes really are a recent development (last 10 years or less)
[00:40:06] <jmkasunich> jepler: did we chase you away?
[00:40:26] <jmkasunich> I'm getting really disgusted with building on the older farm slots
[00:40:31] <jmkasunich> most recent mess:
[00:40:47] <jmkasunich> emc/usr_intf/keystick.cc: In function `int main(int, char **)':
[00:40:47] <jmkasunich> emc/usr_intf/keystick.cc:2886: Internal compiler error.
[00:40:47] <jmkasunich> emc/usr_intf/keystick.cc:2886: Please submit a full bug report to `egcs-bugs@egcs.cygnus.com'.
[00:40:47] <jmkasunich> emc/usr_intf/keystick.cc:2886: See <URL:http://egcs.cygnus.com/faq.html#bugreport> for details.
[00:40:47] <jmkasunich> make: *** [objects/emc/usr_intf/keystick.o] Error 1
[00:41:30] <cradek> does it happen every time?
[00:41:48] <jmkasunich> dunno, I can kick it off again
[00:42:41] <jmkasunich> yep
[00:43:06] <jmkasunich> theres a warning before that, dunno if they might be related
[00:43:23] <jmkasunich> the error line is the closing } of main() (and of the source file)
[00:44:03] <jmkasunich> the warning is 1400 lines earlier, somehow I don't think they're related
[02:16:43] <jepler> I'm back now
[02:17:35] <jmkasunich> I guess I was just expressing my frustration with the old builds
[02:17:42] <jmkasunich> I certainly don't expect you to fix it
[02:17:53] <jepler> oh
[02:18:09] <jepler> this is on the bdi2 slot?
[02:18:15] <jmkasunich> yeah
[02:18:27] <jepler> you should do like it says and submit the bug report :-P
[02:18:49] <jmkasunich> egcs-2.91.66
[02:19:05] <jepler> it was a joke
[02:19:08] <jmkasunich> I know what they'd tell me to do with my bug report
[02:19:16] <jmkasunich> I know ;-)
[02:19:24] <jepler> I get it on my bdi2 chroot too
[02:19:56] <jmkasunich> at one time (ages ago) keystick _did_ compile on that bdi
[02:20:06] <jepler> with emc1
[02:20:15] <jmkasunich> heck, keystick was probably last used on that bdi ;-)
[02:20:20] <jmkasunich> yes
[02:20:47] <jepler> by compiling it with -O0 it succeeds
[02:20:55] <jepler> it fails later on when 'msgfmt' is called
[02:21:06] <jmkasunich> fsckit
[02:21:16] <jepler> does your system have 'msgfmt'?
[02:21:20] <jmkasunich> I'm just not in a fixing mood tonight
[02:21:23] <jepler> cradek: msgfmt is another build-dep to add
[02:25:37] <jepler> nevermind
[02:27:01] <jmkasunich> what part of "I certainly don't expect you to fix it" did you miss?
[02:28:07] <jmkasunich> ;-)
[02:28:21] <jepler> hah
[02:28:24] <jepler> I just got you to the next message
[02:28:25] <jepler> er, error
[02:28:51] <jmkasunich> I must say, I like the new farm script, starts much quicker after a commit
[02:29:05] <jmkasunich> now if I could get it to report on cia....
[02:32:18] <jepler> you can install the other needed program with this command: rpm -ivh ftp://archive.download.redhat.com/pub/redhat/linux/6.2/en/os/i386/RedHat/RPMS/gettext-0.10.35-17.i386.rpm
[02:32:42] <jmkasunich> stand by
[02:33:58] <jmkasunich2> could you repeat that please?
[02:34:07] <jepler> rpm -ivh ftp://archive.download.redhat.com/pub/redhat/linux/6.2/en/os/i386/RedHat/RPMS/gettext-0.10.35-17.i386.rpm
[02:34:53] <jmkasunich2> shame I can't copy/paste from a tty
[02:35:11] <cradek> I think you can on modern systems
[02:35:22] <jmkasunich2> this isn't a modern system
[02:35:40] <jmkasunich2> I can transcribe it into anothertty
[02:36:18] <jepler> anyway, with the Makefile change and the gettext package installed, my bdi2 chroot builds to completion
[02:39:10] <jepler> I don't suppose we'll know the outcome on your system for another 10 minutes or so
[02:43:03] <jmkasunich2> we'll know it after I get done farting around with the ftp
[02:43:13] <jmkasunich2> first I had to poke a hole in the firewall
[02:43:19] <jmkasunich2> then it failed anyway
[02:44:04] <jepler> oh
[02:44:06] <jepler> well heck
[02:44:29] <jepler> I guess I'll have to wait until morning to see the results
[02:44:43] <jmkasunich2> strange
[02:45:04] <jmkasunich2> I was able to login (anon) to the ftp site, and cd all the way down to the RPMS dir
[02:45:11] <jmkasunich2> can't ls tho (permission denied)
[02:45:22] <jmkasunich2> and when I try to get the file, that fails too
[02:45:27] <jmkasunich2> (permission denied)
[02:46:19] <jepler> I don't know if it'll work, but try giving the URL to rpm instead: http://emergent.unpy.net/sandbox/gettext-0.10.35-17.i386.rpm
[02:47:01] <jmkasunich> I have another plan
[02:47:11] <jmkasunich> just downloaded it to this box (which worked fine)
[02:47:15] <jmkasunich> now I'll scp it over
[02:47:41] <jepler> xcok
[02:47:42] <jepler> er
[02:47:43] <jepler> ok
[02:50:07] <jmkasunich2> copied, installed, build started
[03:00:40] <jepler> yay
[03:01:38] <jmkasunich> thanks
[03:02:04] <jmkasunich> and the old systems survive another brush with obsolesence.....
[12:53:25] <SkunkWorks> logger_devel: bookmark
[12:53:25] <SkunkWorks> See http://solaris.cs.utt.ro/irc/irc.freenode.net:6667/emcdevel/2006-04-12#T12-53-25
[14:48:08] <Roguish> good morning, anyone here?
[15:18:47] <jepler> Roguish: I am but I'm not paying much attention
[15:21:11] <Roguish> thanks, i have just done a completely new install of ubunto and emc2. i have also been going through the instructions in the wiki. the cvs reports 'no Testing'
[15:22:17] <Roguish> also, it is not obvious where i should put my customized .ini file.
[15:22:43] <Roguish> for it to show up in the config screen.
[15:26:39] <cradek> hi Roguish, do you intend to develop emc2 or do you just want to run it?
[15:27:14] <Roguish> just run right now. i will customize my mi520.ini file
[15:27:23] <cradek> the easiest way to get the last TESTING release is to use the packages, no cvs is required
[15:27:42] <Roguish> sorry, good morning cradek, and thanks for the incredible work!!!
[15:27:56] <cradek> hi, you're welcome
[15:28:16] <cradek> for customizing a configuration, just follow these instructions on the wiki: http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl/emcinfo.pl?CustomizingConfigsOnUbuntu
[15:29:03] <cradek> the executive summary is you put your configuration in ~/emc2/configs/configname
[15:33:49] <Roguish> thanks, got it.
[15:47:24] <Roguish> cradek: ok, when i run emc and select a 'sample' config, all is good. when i select my config it errors "can't open 'emc.nml'. error = 2 -- no such file or directory"
[15:48:47] <cradek> Roguish: what version are you running?
[15:49:16] <jepler> Roguish: is there a file 'emc.nml' in the new config you created (in the same directory/folder)?
[15:49:49] <Roguish> from 'help, about' 'emc2 testing-2006-04-09'
[15:50:02] <jepler> it should be there but if not you can copy it from one of the sample configs
[15:50:24] <cradek> which sample config did you copy?
[15:50:54] <Roguish> jepler, no there is not. i copied the m5i20 directory, so there was none in the original directory.
[15:51:48] <cradek> chris@buster2:/usr/src$ dpkg -c emc2_TESTING-2006-04-09_i386.deb |grep m5i20/emc.nml
[15:51:52] <cradek> -rw-r--r-- root/root 5210 2006-04-10 22:05:31 ./etc/emc2/sample-configs/m5i20/emc.nml
[15:51:52] <Roguish> wait, there is one in the original directory.
[15:51:56] <cradek> there is one in the package
[15:52:04] <Roguish> must have missed it.
[15:54:19] <cradek> does that fix it?
[15:55:28] <Roguish> yes, we are happy once again. thanks.
[15:55:33] <cradek> yay, welcome
[16:01:45] <jepler> cradek: with the old scripts it would say "X files in Y directories" or something like that. Was that "smart" not on the cia server side?
[16:02:41] <cradek> are you sure cia did that? I know the emails did that
[16:03:06] <cradek> since the script gets called for each dir, it has to write state to /tmp and sleep or something, it's horrible
[16:05:56] <jepler> yeah I am vaguely aware of those details .. I figured it was 'cia' the server that batched things up.
[16:07:25] <cradek> I don't even know how to put multiple directories in the email
[17:10:47] <Roguish_> Roguish_ is now known as Roguish
[19:17:09] <alex_joni> hello
[19:17:23] <cradek> hi alex
[19:18:24] <alex_joni> how goes it?
[19:18:33] <alex_joni> * alex_joni just got out of his massage session :)
[19:18:40] <alex_joni> nice and relaxed...
[19:18:40] <cradek> fine, but just about to go to the dentist
[19:19:30] <SWPadnos> argh
[19:19:53] <alex_joni> cradek: again?
[19:20:13] <cradek> yes, but nothing serious
[19:20:29] <alex_joni> ok then :)
[19:20:34] <alex_joni> what did I miss?
[19:20:38] <alex_joni> seen some commits
[19:20:53] <cradek> I updated the release branch to make sure it matches what I'm putting in the packages
[19:21:06] <alex_joni> ok, good
[19:21:12] <alex_joni> * alex_joni is thinking about docs
[19:21:32] <alex_joni> but I would want ray's oppinion on it, before I start to fsck *.lyx files :)
[19:21:46] <alex_joni> do we still need docs for emc1?
[19:22:07] <cradek> not sure what you mean
[19:22:18] <cradek> some emc1 docs are still relevant and better than anything newer that we have
[19:23:25] <jepler> We shouldn't try to document emc and emc2 in the same file
[19:23:36] <jepler> maybe we need a branch or a separate module
[19:24:04] <alex_joni> jepler got what I meant
[19:24:17] <alex_joni> maybe a folder for emc2 stuff
[19:24:29] <alex_joni> documents/emc & documents/emc2
[19:24:35] <cradek> ok I see what you mean
[19:24:50] <alex_joni> err.. documents/lyx/emc & documents/lyx/emc2
[19:25:02] <alex_joni> because only those are different..
[19:25:13] <alex_joni> and maybe keep generic stuff in documents/lyx
[19:25:18] <alex_joni> and include from both locations
[19:25:53] <cradek> maybe they should be separate modules
[19:26:07] <cradek> we could rename the current documents and make a new one
[19:26:29] <cradek> they are big, so it seems bad to make someone download both when they don't care about half of the files
[19:27:06] <alex_joni> afaik, no-one should want to download them
[19:27:17] <alex_joni> we should get them up-to-date then publish
[19:27:21] <alex_joni> pdf's and html's
[19:27:34] <alex_joni> people should only worry about that..
[19:27:58] <alex_joni> and the writers (how many are there...) can wait a few minutes the first time..
[19:28:01] <cradek> I don't think they should be pdfs, but I'll lose that fight
[19:28:15] <cradek> I better go, a long way to go by 3
[19:28:18] <cradek> be back later
[19:30:05] <alex_joni> ok, laters..
[20:36:21] <alex_joni> hrmm.. anyone around?
[20:36:45] <alex_joni> * alex_joni is having problems with CVS...
[20:43:43] <jepler> alex_joni: I'm here
[20:43:45] <jepler> alex_joni: what's up?
[20:43:52] <alex_joni> cvs commit -m "emc2 compile instructions" cvs commit: Examining .
[20:43:52] <alex_joni> ? Hal_Introduction.pdf
[20:43:52] <alex_joni> cvs [commit aborted]: received abort signal
[20:43:52] <alex_joni> Assertion failed: (key != NULL), function findnode, file /usr/src/gnu/usr.bin/cvs/cvs/../../../../contrib/cvs/src/hash.c, line 312.
[20:44:26] <jepler> you're lucky the assertion is that easy to read
[20:44:43] <alex_joni> that happens when I try to commit something to documents/lyx
[20:44:59] <jepler> every time?
[20:45:01] <alex_joni> yes
[20:45:04] <jepler> does 'cvs update' succeed?
[20:45:21] <alex_joni> yes
[20:45:39] <jepler> (here's the assertion I once saw cvs give: http://emergent.unpy.net/software/01132614188-best-assertion-ever)
[20:45:40] <alex_joni> same error afterwards
[20:46:00] <alex_joni> ouch.. I know that one.. seen it on your page
[20:47:05] <jepler> I cam guess from the filename that it's happening on the server side; I'm not sure what to do about it
[20:47:42] <alex_joni> here's what happened:
[20:47:58] <alex_joni> I added a new file (Int_EMC2_Compile.lyx)
[20:48:04] <alex_joni> using cvs add
[20:48:17] <alex_joni> then I wanted to commit (cvs commit -m "message")
[20:48:45] <alex_joni> but I changed my mind imediately and stopped it (wasn't sure if I hadn't any other changed files in the folder)
[20:48:59] <alex_joni> so I tried cvs commit -m "message" file
[20:49:07] <alex_joni> but since then I am getting that error
[20:49:35] <jepler> http://ximbiot.com/cvs/wiki/index.php?title=CVS_FAQ#Error_in_Win32_version.3B_hash.c.312:_findnode:
[20:49:49] <jepler> it sounds like this error could be the consequence of the interrupted "add"
[20:50:41] <alex_joni> hrmm.. wonder if you have access to the repository?
[20:50:48] <jepler> I don't
[20:50:52] <alex_joni> you could delete the entry, then I'll add again..
[20:51:00] <alex_joni> ok, hope cradek will fix it then..
[20:54:57] <alex_joni> jepler: I managed to remove the newly added file from CVS, and I added it using another name
[20:55:20] <jepler> ok
[20:55:23] <alex_joni> heck.. ascii art on the loose again
[21:03:19] <alex_joni> ok, I'm off to bed
[21:03:24] <alex_joni> g'night all..