#qi-hardware IRC log for Monday, 2011-11-14

wpwrakDocScrutinizer: btw, if you had some USB host port to supply and wanted to make sure inrush current of the devices doesn't kill your 5 V rail, would you consider planting a bead/inductor in the path ?04:28
DocScrutinizernope04:29
wpwrakwhy not ?04:32
DocScrutinizerbecause of the weight ;-D04:32
wpwrakhuh ?04:32
wpwrakit's the standard 500 mA USB, not the power plant edition with 500 MA ;-)04:33
DocScrutinizerhonestly you'll not find a inductor large enough, and even if you would, you won't want to place it into that power rail04:33
wpwrakit's for this kind of critters: http://downloads.qi-hardware.com/people/werner/m1/rst/m1-atusb.png04:33
wpwrakor this: http://downloads.qi-hardware.com/people/werner/m1/rst/m1-hhkb.png04:34
DocScrutinizerdo the math - you won't get happy with inductors for that04:35
DocScrutinizeruse a resistor plus a huge buffer C04:36
DocScrutinizereven better use a PTC04:36
DocScrutinizeraka recovering fuse04:37
wpwrakhmm, resistor isn't so nice. that rail's voltage is already a bit marginal04:37
DocScrutinizerthen you better use a dedicated regulator rather than some inductor that won't help04:38
DocScrutinizeror a FET04:38
wpwrakso maybe just more buffer. currently, there's 220 uF. usb standard recommends 120 uF.04:38
wpwrakthe dedicated regulator has the problem that there's nothing for it to drop :) in case you haven't guessed, it's the evil milkymist power supply again. with reliance on the external supply to provide a nice 5 V ...04:40
DocScrutinizereeew04:45
DocScrutinizerFET based current limiter04:45
DocScrutinizermaybe even with softstart04:45
wpwrakhmm, let's see if such things exist ready-made ...04:46
wpwrakah, NTC ?04:48
DocScrutinizerPTC NTC I always confuse them04:49
DocScrutinizeryou want a short protection anyway04:49
wpwrakhmm PTC would be fuse type where as NTC would be inrush limiter04:54
wpwraknot sure what happens when we short USB. the short just gets passed to the external power supply. M1 itself doesn04:55
wpwrakt try to control that rail04:55
wpwrakDocScrutinizer: i'll copy my mail to the milkymist list to you. is joerg@openmoko.org still you principal mail address ?05:01
DocScrutinizeryes05:02
DocScrutinizerbtw there's no such thing like NTC inrush limiter, that's always fuse type05:03
DocScrutinizerwell, theoretically this would work, but it's never really used05:04
wpwrakthey explicitly sell them as such :) http://www.epcos.com/inf/50/db/icl_09/ICL__B57153__S153.pdf05:07
wpwrakhttp://search.murata.co.jp/Ceramy/image/img/w_hinm/L0760E.pdf05:07
wpwrakof course, ~ 10 R isn't a lot. lemme check ...05:10
DocScrutinizerI only know those for 220V~ when efficiency and voltage drop aren't really an issue05:17
wpwrakyeah, all these seem to be designed for mains. not sure how nice they are at lower voltages05:21
wpwraksimulation says that 10 Ohm ought to help if the power supply has halfway decent load regulation05:22
wpwraksimulation also 05:22
wpwraksays that my calculation of the resulting voltage across the caps is wrong. grmbl.05:23
wpwrakweird. Cport*Vbefore^2/2 = (Cport+Cdevice)*Vafter^2/2 should hold, no ? assuming no losses to ESR, etc.05:27
wpwrakwolfspraul: as joerg has pointed out, with an inductor, he'd worry about weight ;-) (because it would have to be so big to have the desired effect)06:03
wpwrakDocScrutinizer: ah, and what do you think of chaining LDOs ? e.g., when you have one with only a limited input range. is it okay to put it behind one that brings down the input voltage a bit ? or would you try to avoid such a scenario ?06:57
qi-bot[commit] Xiangfu Liu: nanonote-files without console option make /dev/console un-readable (master) http://qi-hw.com/p/openwrt-packages/bab2d7507:16
qi-bot[commit] Xiangfu Liu: csound: make some progress (master) http://qi-hw.com/p/openwrt-packages/75ee50808:56
DocScrutinizerwpwrak: nothing wrong basically with chaning LDOs12:38
DocScrutinizerwpwrak: assuming no losses to ESR is quite simplifying thing when you have no series R like a fuse13:20
DocScrutinizerwpwrak: in that case the ESR of both buffer C (host and attached dongle) will matter in your equation13:21
DocScrutinizeryou basically get:  GND --- C1-esR1 --x-- esR2-C2 --- GND  where one C is charged and other one discharged and you're probing at (x)13:24
wpwrak(esr) yes, but what worries me there is that the simulation also uses a "perfect" cap. i'd rather have it not disagree with my perfect model that just assumes conservation of energy :)13:24
DocScrutinizera big C1 tends to have a big ESR1, and a smaller C2 vice versa13:24
wpwrakmeanwhile, this chap has been proposed as a barrier for inrush current and some more nasties: http://www.diodes.com/datasheets/AP2142A_52A.pdf13:25
wpwrakit looks quite nice to me13:25
DocScrutinizeryou get conservation of charge, not energy13:26
wpwrakapparently, reaction time is around 2 us, which ought to be sufficient to avoid the worst13:26
DocScrutinizerColoumb is the magic word13:26
DocScrutinizerno Ws13:26
wpwrakshouldn't both be conserved ? (energy = J)13:26
DocScrutinizerso your "^2" is wrong I think13:26
DocScrutinizerno, the energy gets converted to heat in ESR etc13:27
wpwrakhmm, right, Q doesn't have a ^213:28
wpwrakwait, in a perfect cap, no energy goes to ESR13:28
wpwrakso where does it go ?13:28
DocScrutinizernowhere :-)13:29
DocScrutinizerthe model is wrong13:29
DocScrutinizerin a near to ideal real setup the energy gets emitted by a monster EMP13:30
wpwrakwait .. i'm not letting go of conservation of energy just so quickly ;-) i have a closed system, two caps and a switch. the switch has Ron = 013:30
wpwrakthe switch is open. i charge C1 to V1 and discharge C2.13:31
DocScrutinizerin the end you're moving electrons, do that fast enough and watch gamma ray escape13:31
wpwraknow the energy in my system is C1*V1^2/213:32
DocScrutinizerright13:32
wpwraknext, i close the switch. now both caps become one large cap of C = C1+C213:32
wpwrak(they're parallel)13:32
DocScrutinizer:nod:13:32
DocScrutinizerelectrons move from C1 to C2 in *zero* time, right?13:33
wpwrakdoens't matter. i look at the system after they've arrived13:33
DocScrutinizermoving electrons create radiation13:33
wpwrakbut yes, there is no impedance13:33
DocScrutinizeryou can watch that at CERN13:34
wpwrakhmm. i'm not sure if that enters this ideal model13:34
DocScrutinizerbtw CERN  is for sure highly interested in your setup13:34
DocScrutinizer;-D13:34
wpwrak;-)13:34
DocScrutinizerwpwrak: either your electrons move in *zero* time and emit x-ray/gamma quite a bit, or you explain why they don't move in zero time and you are about to approach a valid model then13:35
wpwrakcan i use a friend of maxwell's demon ? :)13:36
DocScrutinizerhehe13:36
wpwrakbtw, interesting site: http://hyperphysics.phy-astr.gsu.edu/hbase/electric/capcon.html13:37
wpwrakalas, doesn't have the answer to my puzzle13:37
wpwraklemme calculate this with Q instead of E13:38
wpwrakhah, pretty exactly what the simulation says, too13:38
DocScrutinizerwpwrak: if this charge exchange between Cs would work, we could build PSUs without inductors and with 100% efficiency13:38
wpwrakoh, again, i was assuming an ideal model. of course you have parasitic effects in real life13:39
DocScrutinizerright, like electromagnetism created by electron flow13:40
DocScrutinizeraka parasitary inductance13:40
DocScrutinizerESR13:40
DocScrutinizerthere is no complete correct model that could eliminate these13:42
DocScrutinizerthe simpler your model, the more even properties of plain wires matter13:43
wpwrakbut if, for modeling purposes, we assume perfect caps and superconducting wires. then the energy lost would go into what ... entropy ?13:44
DocScrutinizerbuild a model of an ideal battery shorted by an ideal wire, what do you get?13:44
DocScrutinizerright: nonsense :-D13:44
wpwrakokay, but you can do a limes :)13:44
DocScrutinizerinto an EM field, as you build an oscillator this way13:45
wpwrakalso my simulation has an ESR of 100 mOhm, or qucs would puke :) but it doesn't really affect the result13:45
wpwrakand where does the EM field go after things have stabilized ?13:46
C-Keenspeaking of which: is it a known problem that the nanonote discharges itself rather quickly? I never noticed but I forgot mine at a friend's place and got it back 5 days later. the battery was dead.13:46
DocScrutinizerit dissipates in the environment13:46
DocScrutinizeryou build a LC resonator with your ideal wire and Cs, and it will oscillate and thus emit RF13:47
DocScrutinizeras even an ideal wire is an inductor13:48
wolfspra1lC-Keen: the normal discharge rate for a li-ion cell is about 1% / day I think13:49
wpwrakC-Keen: in theory, it shouldn't. not sure about practice, though :) (i keep all my bens tethered)13:49
wolfspra1lthe cell may be worn out substantially after a while, reducing what is 100%13:49
C-Keenyep I usually do too, that's why I did not noticed13:49
wolfspra1lalso there may be a software issue, that is the NanoNote is actually not fully off and drawing from the battery13:49
DocScrutinizerI bet all my bucks on the latter13:50
C-KeenI am not running the latest image, I should upgrade anyway13:50
wolfspra1lif you want to investigate this more, you need experiments ;-)13:50
DocScrutinizeror LEDs13:50
DocScrutinizer;-D13:50
wolfspra1lsecond battery, or even two new ones, charge both fully, leave one in the device, one outside, etc.13:50
C-Keenthe led besides the usb plug cannot be controlled by the kernel, can it?13:51
DocScrutinizerI always found it rather reassuring to have an LED on each and every power rail :-)13:51
wolfspra1lunfortunately the Ben battery charging status cannot be measured or read well, complicating any fact finding13:51
DocScrutinizerand they are cheap, giving good data at as little as 0.5mA13:52
DocScrutinizerwell, alas I have no Ben to retrofit those LED13:53
DocScrutinizerand nobody asked me to design them into the original board13:54
DocScrutinizerthough it was absolutely cheap, just needing 20 min of thinking and layout work - you can make them NC any time you want, later on in MP of the million boards13:55
DocScrutinizerto save that 30ct/PCBA13:55
DocScrutinizerthe layout is same price no matter if footprints for LED are there or not13:56
DocScrutinizers/layout/PCB/13:56
wpwrakstill seems overkill to me :)13:57
DocScrutinizerhah13:58
wpwrakbut i guess *i* would have routed those USB host signals somewhere, even if only test points :)13:58
DocScrutinizeryou're only interested in creating a safe job for you, as investigating those things *without* LEDs for sure costs more in labor than a 1000 boards' LEDs would cost13:59
DocScrutinizer;-P13:59
DocScrutinizerlabour*14:00
DocScrutinizerlabour at labor14:00
DocScrutinizernow you know why my OM business card said "HW & synergy"14:01
DocScrutinizerwith LEDs: get bug reports like "device doesn't shut down, LEDs shine" within 24h after rolling out new firmware. Without: investigate for weeks and months in SW department14:03
DocScrutinizerwhile CR gets flooded with "fsckng Ben cuts thru battery in no time"14:04
DocScrutinizer"batery is crap"14:04
DocScrutinizer"device locks up after few hours of shutdown"14:04
DocScrutinizer"device doesn't boot"14:05
DocScrutinizer"device dead, reflashing only helps for 1 day"14:05
DocScrutinizeretc14:05
DocScrutinizerand honestly, how much overkill are a few FOOTPRINTS14:08
wpwraki think you misunderstand the economics of work at qi-hw. labour has zero cost. ergo the LEDs are infinitely more expensive than even the most excruciating analysis :)14:08
DocScrutinizerno the footprints are exactly as free as are months of headbanging due to obscure power drain14:09
wpwrakat OM, shouldn't the biz card have said "HW & sarcasm" ? :)14:09
DocScrutinizerabsolutely for free14:09
DocScrutinizerthe sarcasm comes for free :-D14:10
C-Keenuhm, what are you arguing about :) and why? :)14:10
wpwrakC-Keen: just banter ;-)14:10
DocScrutinizerC-Keen: indicator LEDs on Ben's power rails14:10
DocScrutinizerI'd have insisted on getting at least footprints on PCB for those14:11
C-KeenI understood that14:11
DocScrutinizerbtw I hope tomorrow you'll "lose me"14:14
DocScrutinizerto ST-Ericsson14:14
DocScrutinizerR&D labs LTE etc14:15
DocScrutinizerconveniently located here in my town14:15
DocScrutinizerwpwrak: you know what'S been overkill? those myriad of 0R on GTA02 which easily could have been surface traces to cut, with two pads to re-solder the cut trace14:19
wpwrakDocScrutinizer: (STE) interview or already siging ceremony ?14:20
wpwraksigNing14:20
DocScrutinizerinterview14:20
wpwrak(0R in gta02) ah well ...14:20
wpwrakbest of luck then ! though we'll hate to "lose" you14:21
DocScrutinizerwell, I may still be available at what's night hours in my TZ14:21
DocScrutinizeror even beter get a job involving doing lots of IRCing maybe ;-D14:22
DocScrutinizermaybe I could push for more community oriented work?14:23
DocScrutinizer(doubt that, it's ST-E after all)14:23
DocScrutinizerguess they don't care much about "community"14:23
DocScrutinizeraah, and thanks wpwrak 14:24
wolfspra1lDocScrutinizer: wow, great news - good luck!14:26
DocScrutinizerthanks :-D14:26
wolfspra1lI'm not worried we loose you, you are indestructible14:26
qi-botThe build was successfull, see images here: http://fidelio.qi-hardware.com/~xiangfu/compile-log/openwrt-xburst.full_system-11132011-1302/18:54
--- Tue Nov 15 201100:00

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