#qi-hardware IRC log for Sunday, 2011-08-28

wolfspraulupgrading indefero (projects server) right now03:21
wolfspraulof course it doesn't go as expected :-)03:21
rohwpwrak: about your opto-isolated inputs.. there are cheap-ish (5E) dc-dc 1000V galvanic isolated 5V-5V transformers05:44
roh4 pins.. dc in, dc out.. small sil-package.. 1W power rating05:44
rohenough to drive an optocoupler, some line driver or similar hw.05:45
qi-botThe build has FAILED, see log here: http://fidelio.qi-hardware.com/~xiangfu/compile-log/openwrt-xburst.full_system-08272011-0801/05:50
wpwrakroh: maybe you mean the analog devices parts, signal plus power. but what i want is a wide input range, plus insensitivity to reverse voltage. also, the goal is to keep this simple. since there doesn't appear to be a decent (yet simple) electronic solution, using optocouplers as fuses will have to do. at least they'll survive TTL levels, which should cover 99.9% of all cases.07:06
wolfspraullarsc: xiangfu I upgraded Indefero but unfortunately it gets stuck creating the html view of qi-kernel.git11:41
wolfspraulso for now I disabled anonymous access, I hope that's only in the web interface... but maybe also the git: access is blocked now11:41
wolfspraulas for tracking down the bug, I saw a few Apache processes going in endless circles in the javi-roman branch, seemingly11:42
wolfspraulis there anything we can do to qi-kernel.git to rid if of some unnecessary history? any unusually cruel git actions we performed on it?11:42
wolfspraulall other repositories work fine11:48
DocScrutinizerwpwrak: honestly, why do you actually need galvanic separation?12:10
DocScrutinizerwouldn't high Z do as well (and provide way better secondary properties)?12:11
Action: lekernel is happily using github. the pesky m1 problems give me enough pain and frustration without adding system administration tasks on top of them.12:11
DocScrutinizer5 pcs 220kR in series (to deal with Vmax per R), both in input and GND, plus some decent highly sensitive detector circuitry plus OVP behind that...12:13
DocScrutinizerhi lekernel 12:13
DocScrutinizerwpwrak: using this high-Z concept, sth as simple as a mosfet + zener could give you all you asked for, plus more: detection range 0.5...300V, input current 1uA/V...12:16
DocScrutinizerno reverse blocking yet, but at 1uA/V12:17
DocScrutinizerI'd wonder why you need that at all12:17
wpwrakDocScrutinizer: (galvanic separation) because i don't know the ground potential of the other thing12:17
DocScrutinizerso as long as your circuit has floating GND and you connect to DUT's GND via 1MR12:19
wpwrakmy circuit connects to USB, so one ground is already given12:19
wpwrakand there can be different DUTs12:19
DocScrutinizeryou're trying to build sth that outperforms $$$$ equipment properies-wise, and you are not willing to spend >2$ per input on the most important part of the whole thing12:29
wpwrakwhy ... there are tons of USB opto-coupler cards. it's not rocket science that i'm attempting12:30
DocScrutinizergiven your specification goal are not completely crazy but rather desirable for every lab nerd using such thing, tell me how likely is it you'll succeed in building such a thing for less than china, with way better specs than china's things can offer?12:32
wpwrakeasy. china doesn't do some features i want to have in there :)12:33
wpwraksw features12:33
DocScrutinizermeh, forget sw features. I'm just talking about arbitrary black boxes with GP-inputs12:34
DocScrutinizeror are you suggesting you get your OVP and sensitivity and galvanic separation via nifty software?12:35
wpwrakyes, but then i'd have to hack the firmware. and probably mod the case12:35
DocScrutinizerno you don't, as there ARE NO such devices from china12:35
wpwrakno, but i want the sw to be a little more intelligent than what the USB-to-relay/opto board i've seen can do12:35
wpwrakokay, maybe they make them in malaysia :)12:36
DocScrutinizeryou missed the point12:36
DocScrutinizeryou're trying to build sth that outperforms $$$$ equipment properies-wise, and you are not willing to spend >2$ per input on the most important part of the whole thing12:37
DocScrutinizernobody has achieved on that before you12:37
DocScrutinizerotherwise there was sth fainly similar from china already12:37
wpwrakit's just a few stupid opto-couplers !12:39
wpwraki'm sure those other cards have very similar properties12:39
DocScrutinizeryeah, and a stupid (sorry) design that uses them way outide their specs12:40
wpwraksee, tha's the secret when competing with china :)12:40
DocScrutinizerno, there IS NO card with input sensitivity 100uA@1.5V and OVP up to 30V AND galvanic separation12:41
DocScrutinizernot at the price tag you're aiming at12:42
wpwrakmaybe 200 uA then. well, i haven't measured yet what the coupler can do12:42
wpwrakinteresting. cvpcb cripples the net names in .net files. when pcbnew reads them, this makes it lose its net classes. lovely :)12:43
DocScrutinizeryesyesyes, and for sure you can operate that FM radio with 3 instead of 4 AA cells, and place a bridge in for battery #412:43
wpwrak(the work-around is to save the netlist again from eeschema after running cvpcb)12:44
wpwraksee :)12:44
DocScrutinizerI honestly wonder what's your initial product requirement specs and if they are reasonable12:46
DocScrutinizerif I want a general purpose logic analyzer the I won't go for optocouplers without protection that are swappable like fuses12:47
DocScrutinizerand I won't opt for operating a key component out of specs for the intended major usecase12:48
wpwrakusb-attached. 2 relays and as many opto-isolated inputs/outputs i can cram in. front panel has buttons and LEDs to control the relays. MCU runs a user-defined state machine to implement different action profiles. e.g., button = on/off. or button = on while pressed, then at least 10 s off. and so on. can cover multiple channels.12:48
wpwrakno, it's for switching power and "manual" signals (buttons and such)12:48
DocScrutinizeryou're aware that 1.5V@100uA is already in a range where you get false positives from 5m of open-ended wire in a cable that also runs a 220V~ life wire?12:51
wpwrakyou mean a cable that combines 220 V with a low-voltage signal ? sounds like generally a bad idea to me :)12:52
wpwraki would expect connections to the box to be kept reasonably separated12:53
DocScrutinizeryour design specs sound like a poor set of parameters to me12:53
wpwrakof course, anyone can take their chances ;-)12:53
wpwrakforgot one: SIMPLE ! :)12:54
wpwrakit's already taking too long to make the bloody thing :-(12:54
DocScrutinizerjust missing "shall withstand hitting it with a sledge hammer"12:54
wpwrakthat's part of the user interface :)12:55
wpwrakthe external form factor is given, by the way. a box i can easily source and that has removable faceplates.12:56
DocScrutinizermy last staement on this: 1V5 at highZ sensitivity and full galvanic separation and low BOM doesn't go together12:57
wpwraki'll let you know how my experiments go :)12:57
DocScrutinizerscratch one of those 3 parameters, or split on different inputs or switchable input properties12:58
DocScrutinizerwpwrak: actually as long as you won't run a full fledged industrial grade set of test against 1000 of those devices, for EMI and temperature range and long time stability and whatnot, I'm actually not very much interested in your results13:02
DocScrutinizeryou could tell me it works like a bulldozer and I'd stll be reluctant in getting one13:04
DocScrutinizerand then taking into account you consider to actually control safety relevant things powered by 220V with it, makes the whole concept even more scary13:05
wolfspraulall I can see is that Indefero gets into an infinite loop right at the first few git commands it sends. hmm.13:22
wolfspraulI have one random long shot idea in that the branch name with a '/' in it confuses Indefero. there is a branch "javiroman/ks7010" - can we rename that to javiroman-ks7010?13:24
wolfspraulthat's a desperate attempt without any real trace in the sources, but it's the only branch name with a '/' and if we are lucky that's just the problem :-)13:25
wolfspraulhmm. Indefero sources say '/' in branch name is supported. well then.13:32
wolfspraulworst case I downgrade to make the problem go away :-)13:33
qi-bot[commit] Werner Almesberger: labsw/: added spacers; shifted USB connector a little (master) http://qi-hw.com/p/wernermisc/f3e9d7513:46
qi-bot[commit] Werner Almesberger: labsw/: changed all ring-to-hole ratios to 2.5; placed spacers at 1 mm from edge (master) http://qi-hw.com/p/wernermisc/17d247813:46
qi-bot[commit] Werner Almesberger: labsw/: changed flyback diodes to SOT-323 (BAS16WT or similar) (master) http://qi-hw.com/p/wernermisc/302002c13:46
qi-bot[commit] Werner Almesberger: labsw/: grouped opto-couplers for socketing; improved clearance of inputs (master) http://qi-hw.com/p/wernermisc/e95d20a13:46
LunaVoraxHi everyone !15:50
LunaVoraxI'm in very deep troubles with my Ben Nanonote15:50
LunaVoraxIt won't boot any more, even after reflashing it15:51
LunaVoraxIt hangs up on setting the date and time to 01-01-197015:51
LunaVoraxNoone to the rescue ?18:31
viricnever happened to me.18:39
viricwhere have you put it?18:40
vladkorotnevhey guys, what's up? hadn't been there for a looooong time :P18:44
LunaVoraxviric, nowhere especial18:55
viricLunaVorax: then I can't help19:30
--- Mon Aug 29 201100:00

Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!