#qi-hardware IRC log for Wednesday, 2016-03-23

wolfspraulwpwrak: hmm. is it possible that that is just this greylisting stuff?00:23
wolfspraulwhere does this error msg come from - did you get a final "cannot be delivered" back into your mail client?00:24
wpwrakso far it's just trying to connect. is it MTA running ? on turandot, telnet localhost 25 -> telnet: Unable to connect to remote host: Connection refused00:25
wpwrakand the mail hasn't timed out yet on my side. don't remember what timeout i set. probably something relatively large00:26
wpwrakyeah, 5 days :)00:26
wpwrakbut it'll start complaining after 4 hours00:26
DocScrutinizer05seems there's really stormy weather in the internets, eh?00:52
wpwraklooks more like some system upgrade at qi-hw, and the usual fallout such things produce00:53
wpwrakhmm, i wonder if AAA + diode + polyfuse would work for battery reversal protection01:15
wpwrak(as alternative to an "ideal diode")01:16
wpwrakthe sub-question there is what the damage mechanism is for battery reversal. i.e., what sort of current has to flow through the reversed AAA battery to actually do damage to the circuit01:21
wpwrakbut i suspect that relatively little may be sufficient, i.e., a few hundred mA01:22
wpwrakwolfspraul: thanks ! :)01:52
whitequarkwpwrak: there are ideal diodes.02:52
whitequarkwell, "diodes" with zero drop. they're available. it's actually a fet in a diode package02:52
wpwrakgot an example ?03:08
wpwraklooking for "ideal diodes" i found some parts, but usually stuff that's far out of the range we'd need here03:09
whitequarkwpwrak: hm03:33
whitequarkso there's this 55mV Schottky https://www.digikey.com/product-detail/en/fairchild-semiconductor/SS14FP/SS14FPCT-ND/589213003:33
whitequarkI can't find the exact device I was looking for yet, but does it even matter for your application?03:34
wpwrakyou mean the 0.55 V = 55 mV type ? :)03:39
wpwrakthe golden rule still applies: if it's too goo to be true then it probably isn't ;-)03:39
wpwrakor, in the case of digi-key: then it's probably a typo :)03:39
whitequarkah right, it's a typo03:40
whitequarkanyway, this is what I was talking about: https://www.maximintegrated.com/en/app-notes/index.mvp/id/63603:40
wpwrakhere is a "smart diode". alas, expensive (and designed for higher-current applications): https://www.digikey.com/product-detail/en/texas-instruments/SM74611KTTR/296-35688-1-ND/391115503:48
wpwrakti's data sheet is from 2012. so maybe around 2030, when the patents expire, we'll see more such smart diodes ...03:49
wpwrakthe maxim circuit with boost converter is interesting. haven't thought of that.03:50
wpwrakthe FET circuits all have the problem of low voltage. it's hard to find FETs that are happy to admit ~ 1 A with a low drop and Vgs <= 1 V03:51
wpwraka radically different approach would be to block this mechanically03:52
wpwraknot sure how well that would work, though. the "nipple" of an AAA cell is supposed to stick out by >= 0.8 mm. so an obstacle of about 0.5 mm around the contact should work. however, i don't remember ever seeing such an arrangement in real life. there may be a reason for it.03:54
whitequarkyeah, the voltage is troubling04:18
whitequarkoh, I remember now the proper name for the device I was looking for04:22
whitequark"MOS Rectifier"04:22
whitequarkthey do have lower Vf than regular Schottky diodes but I think not lower enough04:23
whitequark0.28V@5A apparently04:24
whitequarkbut those are unavailable... and I've definitely seen those in production04:26
whitequarkAHA! found it.04:30
whitequarkwpwrak: http://www.ti.com/lit/ds/symlink/sm74611.pdf04:30
whitequark26 mV at 8 A04:30
whitequarkand that's not a typo.04:30
wpwraksee above :)04:30
whitequarkread the datasheet.04:30
whitequarkoh, you found that device04:31
whitequarkI was under impression that there were more of them04:31
whitequarkbut this is exactly the thing I had in mind.04:31
whitequarkthe maxim circuit is basically the same thing as that diode, except discrete04:34
whitequarkwpwrak: how about this? https://www.fairchildsemi.com/datasheets/FD/FDN327N.pdf04:55
whitequarkRds(on)=120m, meaning Vf=120mV04:56
whitequarkat your target 1A04:56
wpwrakneeds boosting. Vgs(th) is too high05:29
wpwrakafter all, that little battery can only cough out 1.5 V when fresh, and with use, it gets worse05:30
wpwrakthese are the results of the little search i did: https://defuse.ca/b/pYH84uqSKqz33KLH2YnlPT05:31
wpwraki found no p-FET that looked truly convincing05:31
wpwraki found two n-FETs that may be okay05:32
whitequarkoh right, you have an AAA battery, not li-ion05:34
wpwrakbut i wonder if there may be an easier solution. that variant with providing the gate voltage from the boost converter looks interesting. not quite sure if it would actually work, though. the path of components that are borderline operational gets fairly long.05:34
wpwrakyup. if i had 3.7+ V to burn, there would be no worry at all ;-)05:34
wpwrakthe big concern is to avoid choking the battery too early. e.g., when current demand is high, battery voltage drops in the battery and then again at the FET. the resulting lower boost Vin makes it draw even more current. worse, since that reduced voltage is also Vgs, the FET becomes more resistive, further increasing drop and current demand05:41
eintopfmhh, I like smartphones which getting hot when I turn on GPS. Then I have something to heat me up at the bus station when its winter06:53
kyakit get hot because of GR effects. You phone travels in time and space, and it is quite hot there07:00
kyakwhile waiting for a bus, don't forget what's your phone is going through07:01
xiangfuwpwrak: Hi07:21
xiangfuwpwrak: we got 15 N900. one of them is very bad. the others looks like ok. but all used devices. vc.cheng@can.... should be send one email to you. with pictures.07:23
xiangfuwpwrak: let me know what should I do next? 07:23
xiangfubut it is not easy to get 200 of them. I buy them slowly...07:23
wpwrakare all 15 from the same source ? and have they been cleaned properly, i.e., are they of similar quality as the refurbished we looked at some months ago ? or are they worse ?08:47
wpwrak(not easy) the messier the sourcing, the greater the glory ;-)08:48
whitequarkwpwrak: the boost circuit really shouldn't draw any current10:06
whitequarkmaybe microamps. it has to run the charge pump once, and afterwards its job is done10:06
eintopfit's not a good idea to glue your usb hubs on your table with a velcro fastener powerband...12:04
wpwrakeintopf: why not ?14:32
wpwrakwhitequark: yes, it all should work. but i imagine that there could be little quirks like the boost regulator trying to pull significant current in startup or such, and cycling back if it can't (because the FET isn't fully open yet)14:41
wpwrakso such a circuit would need careful testing, to check that there aren't such quirks while in borderline conditions14:42
wpwrakbut i still very much like the idea to increase the gate voltage this way. that solves a LOT of problems ;-)14:43
eintopfwpwrak: it's too strong16:06
eintopfI tried to remove it again and the whole case creaks in two pieces16:06
eintopfbut I have it now16:07
eintopfI need some knife to remove it :-)16:07
wpwrakah, the idea is to make this permanent :)16:07
DocScrutinizer05hmmm  >> At t2: When the voltage on the capacitor C1 reaches its predetermined lower level, the FET driver shuts off the FET. The bypass current will then begin to flow through the body diode of the FET, causing the FET body diode voltage drop of approximately 0.6V to appear across ANODE and CATHODE. The charge pump circuitry is re-activated and begins charging the capacitor C1. This cycle repeats until the shade on the panel is 16:20
DocScrutinizer05removed and the string current begins to flow through the PV cells instead of the body diode of the FET.<<16:20
DocScrutinizer05wpwrak: why the heck you need 1A?16:21
wpwraki need a bit less, something like 800 mA16:21
DocScrutinizer05ugh, that's much16:22
wpwrakIbat(max) = 120 mA / Vbat * 3.3 V / efficiency / duty_cycle16:22
DocScrutinizer05which duty cycle?16:22
wpwrakthe "charging" part of the inductor16:23
DocScrutinizer05there's capacitors for that16:23
wpwrakah yes, that was the formula for inductor / switch current16:24
wpwrakso for the battery it's a bit less ...16:24
DocScrutinizer05I hope so, you won't find a AAA that behav4es under such conditions16:24
wpwrak~550 mA (assuming 80% efficiency)16:24
wpwrakcheck the AAA short-circuit curve ;-)16:25
DocScrutinizer05you need to check _all_ those curves, unless you specify an approved battery brand and type for your device16:26
DocScrutinizer05iirc a 500mA are about max to draw from a AAA16:26
wpwraki mean this one: http://www.varta-microbattery.com/applications/mb_data/DOCUMENTS/GRAPHS/04103/G_04103_O_S01_en.pdf16:28
wpwrak(see my post to he qi-hw list)16:28
DocScrutinizer05that diagram only mentions Ampere but no voltage16:33
DocScrutinizer05and it's Varta, a known semi-devcent cell manuf16:34
DocScrutinizer05actually you can read the diagram as "initial internal Z of cell: 100mOhm, after ~4 minutes 1 Ohm"16:37
DocScrutinizer05cells of slightly lower quality might already start with 1 Ohm out of the blister16:39
DocScrutinizer05particularly zink carbon cells16:40
DocScrutinizer05not every battery is alkaline even today16:40
DocScrutinizer05sorry I'm not on qi-hw ML16:41
DocScrutinizer05I still wonder what draws such huge amount of power in your device16:44
DocScrutinizer05you must have a power hog you should disable by default and only power up on user request16:45
DocScrutinizer05if it's backlight, it's prolly a good measure to only enable it when ambient light is low, and definitely keep it disabled during device ramp-up and when battery voltage is low16:46
DocScrutinizer05ever pondered http://www.batteryjunction.com/10440.html ?16:49
wpwrakthe power hog is the memory card. it can draw up to 75 mA for up to 250 ms (per operation)16:55
wpwrakotherwise, anelok uses a lot less, of course16:55
wpwrakand there's no stinkin' backlight. in OLED we trust :)16:56
wpwrak(besides, OLED looks much better)16:56
wpwrak(zinc carbon) naw, i expect people to use alkaline of halfway decent quality. it's not as if they were difficult to find.16:58
DocScrutinizer05good, then don't power up the memory card until the device is fully up to provide proper bias for Vgs16:58
DocScrutinizer05besides that, a fuse plus reversal diode is the means of choice to protect from battery reversal17:00
DocScrutinizer05use a schottky17:00
wpwrak(10440-category.html) -> "UltraFire" how appropriate ;-))  http://avherald.com/h?article=495aeb3117:00
DocScrutinizer05yeah funny17:00
DocScrutinizer05the right one there is nice17:01
DocScrutinizer05only 350mA but seemingly safe chemistry17:01
wpwrak(fet charged) yes, anelok can ramp up "nicely". but "setting up" the FET is actually something that happens even earlier, long before the MCU even begins to reset17:01
DocScrutinizer05toldya, use a pushbuttom to power up, it can even generate 2*Vbatt on gate17:03
wpwrakit's kinda funny that all three refer to fire, either directly in the name or in the logo ;-)17:03
DocScrutinizer05...for long enough to boot the MCU17:04
wpwrakpushbutton wouldn't really help. this is between battery, fet, and boost converter. there is no change in the configuration of the boost converter.17:05
wpwrakthe boost converter starts enabled (unless there is usb power)17:05
wpwrakand only the presence of USB power can turn it off17:06
DocScrutinizer05I don't get it17:06
wpwrakso such a hypothetical button wouldn't affect what happens at this early system bringup17:07
DocScrutinizer05what for is tha FET then?17:07
wpwrakto protect against battery polarity reversal17:07
wpwraki was thinking of also using it to disconnect the battery, but it seems that this would get too complicated. in any case, even if i do this, it would be controlled by the presence of a good USB voltage.17:08
DocScrutinizer05you know each decent device doesn't allow to reverse AAA batteries since the button on plus is the only thing that can touch the battery contact?17:08
wpwraki use a reset generator for detecting USB voltage. that way, i won't get funny things happening on connect / disconnect, when USB ramps through Vgs(th). had issues there in the CR2032 design, and it would only get worse in the AAA design17:09
DocScrutinizer05[2016-03-23 Wed 18:00:12] <DocScrutinizer05> besides that, a fuse plus reversal diode is the means of choice to protect from battery reversal17:10
wpwraki'm not so sure about diode + fuse. it would seem that a partially spent battery may end up just producing enough reverse current to do damage, yet wouldn't trip the fuse17:12
DocScrutinizer05I think my last comment on all this was to recommend battery change only when attached to USB, to keep RTC running. When a user doesn't follow that recommendation, they will have to adjust date and time anew after battery inserted17:12
wpwrak(mechanical prevention) i wonder about that. i don't recall ever seeing that done in practice, though the mechanical design of the AAA (et al.) cells clearly suggests such a design17:13
wpwrakno, i don't want to require USB power17:13
DocScrutinizer05a schottky is supposed to be <0V5 which is well within allowable voltage range in every chip I know17:13
wpwrakyou should be able to change batteries in the field, without preparation17:14
DocScrutinizer05well, you can do this17:14
DocScrutinizer05I honestly don't see how you want to keep RTC running without a decent power source17:15
wpwrak(schottky) yes, but i also want to make sure the fuse trips. wouldn't be so nice if the battery merely fast-discharges through the schottky.17:15
DocScrutinizer05maybe get a Dallas chip ;-)17:16
wpwraki have caps to hold power for a litte while17:16
DocScrutinizer05I don't want to have to cope with such worries you seem to have17:16
wpwrakand a FET before them, to isolate the MCU from the rest of the system17:16
wpwrakwe'll get to this when debugging neo900 power ;-)17:17
DocScrutinizer05IF a battery has a button on minus (instead of flat plate) so it could make contact from minus pole to plus battery contact mechanically, AND the battery is already empty enough so it doesn't trip fuse, then you don't want that battery to get discharged17:18
DocScrutinizer05no, we won't get to this with Neo900 power17:18
DocScrutinizer05been there, done that17:18
DocScrutinizer05I'd be more worried about somebody inserting a 9V AAA than about discharging a weird mechanical nonstandard half-empty secondary 1.5V AAA17:21
DocScrutinizer05odds are the user completely depleted the battery anyway before even noticing she needs a new battery, so RTC is void in either case17:24
wpwrak(trip) let's see ... battery: Vinternal = 1.5 V, Rinternal = 1 Ohm; fuse: Ihold > 600 mA; diode: Vf@Ireversed = 0.5 V. so the short-circuit current through the diode would be (1.5 - 0.5 V) / Rinternal = 1 A.17:25
wpwrakso the difference between Itrip and Ihold would be only 400 mA. difficult to find a fuse that's that selective.17:26
DocScrutinizer05and when you want user to take care about not losing the date and time while battery swapping, provide a "swap battery" function user needs to activate before removing battery, which will store the current date and time to non volatile storage and recovers it from there during power-up after new battery inserted. will be off a minute maybe17:26
DocScrutinizer05the fuse is only to avoid damage to the device, not to protect a half-dead battery17:27
DocScrutinizer05keep in mind that with proper mechanical design it's virtually impossible to reverse battery17:28
DocScrutinizer05only needs a 10mm^3 of plastic to form a stopper that keeps minus plate of battery from touching the plus battery contact of device17:29
DocScrutinizer05unless you do extremely weird things, you _cannot_ reversa a AAA cell in device17:30
wpwrakone minute would be very long. e.g., TOTP has a default time increment of 30 s.17:30
DocScrutinizer05ohmy, then add a adjustable offset, let user guess17:31
DocScrutinizer05measure discharge of a capacitor and calculate the battery off time, simple as that17:31
wpwrak(mechanical prevention of battery polarity reversal) have you seen anything like this in real life ? even high-end equipment seems to have very reversal-friendly contacts. i wonder why17:32
DocScrutinizer05because they are fools? I've seen it in ~80% of my devices17:33
DocScrutinizer05you usually don't even notice it17:33
wpwrak(cap) hmm ! now that's an interesting idea ! i like that :)17:33
DocScrutinizer05yeah, a funny simple concept I just invented in a fraction of a second17:36
DocScrutinizer05actually best part of (18:31:57 - 18:31:31 ) taken by typing17:37
DocScrutinizer05you can start discharge at very moment you take the timestamp, before you write said timestamp to storage17:40
wpwraksomething like 47 uF vs. 1 MOhm should do the trick for times in the order of 1 minute17:40
DocScrutinizer05I guess you could even use 10M17:40
DocScrutinizer05AD-inputs are high impedance, no?17:41
wpwraki should probably only activate ADC briefly. lemme check ...17:42
wpwrakwithout ADC, input leakage is 1 uA or maybe 25 nA (25 C).17:43
DocScrutinizer05charge, measure voltage and stop charging, wait 2s and measure voltage again [1], then charge again, take timestamp from RTC [2] and stop charging, write [1] and [2] to storage17:43
DocScrutinizer05hmm, no, [1] needs to be a derived value from both voltage measurements17:45
DocScrutinizer05representing the percentage of discharge during 2s17:45
DocScrutinizer05and actually you need the absolute voltage too, in storage, to know a reference later on17:46
wpwrakspec also applies to ADC, so real-life leakage should not be much higher than 0.1 uA. so yes, maybe 2 or 3 MOhm should work fine.17:46
DocScrutinizer05so store both voltage at t0 and voltage at t2s as well as timstamp to storage17:47
wpwrakyes, the critter needs to recalibrate every once in a while. not a big deal, though.17:47
wpwrakthe trickiest bit is avoiding messing up that cap when the system ramps up/down17:48
DocScrutinizer05hmm yes17:49
wpwrakbut even if we don't take precautions, having a calibration happen while the rail changes dramatically should have a pretty low probability. prolly better than 1e-517:49
wpwrak(all that assuming nothing nasty happens for too long when the MCU is below safe operating voltage)17:50
DocScrutinizer05you should spend a 2 GPIO, one for "power" and one for "gate", where the switch (FET) charges on power != gate17:50
wpwrakyes, that could be an option17:52
wpwrakin any case, a sufficiently large cap would also be immune to very brief GPIO upsets, given that Ron of the GPIOs is around 100 Ohm.17:53
DocScrutinizer05well, or you simply connect the AD/GPIO to capacitor via 3M and have a small capaicor (for glitches) and a 47kR from AD/GPIO to GND17:54
wpwrakbut having an extra fet would also help to keep the "high" cap voltage from a powered-down MCU. it has no upper clamp diodes, but you never know what else may happen17:54
DocScrutinizer05err uh, that idea of mine would spoil the AD probing :-/17:55
wpwrakand it would make charging a lot of work :)17:55
DocScrutinizer05so use a GPIO with all that birdseed for charging/discharging and a separate AD input to capacitor via 10M for measuring17:56
DocScrutinizer05aah yes, chatging takes a while then, but then, who cares17:56
DocScrutinizer05keep capacitor charged all the time17:57
wpwrakseems easier and safer with one measure/charge pin and one to switch a FET17:57
wpwrakalso, i'm pretty limited when it comes to IO pins17:57
DocScrutinizer05maybe not17:57
DocScrutinizer05you need two pins for both concepts17:58
wpwrakif i have a direct charging pin, then i could use that as well for measuring17:58
DocScrutinizer05that's the initial idea17:58
wpwrakso if separating the two, that's ADC + charge + switch(FET). not so nice17:58
DocScrutinizer05you don't need a FET when charging takes "ages"17:59
DocScrutinizer05short glitches will be irrelevant then17:59
wpwrakah, i see what you mean17:59
DocScrutinizer053 resistors (the 47k from charging GPIO to GND for proper discharging during MCU down) and 2 pins is all you need18:00
DocScrutinizer05add a small capacitor in parallel to the 47k for even better glitch immunity18:01
DocScrutinizer05hmm maybe that's a bad idea, it will buffer any glitch and extend its duration18:02
DocScrutinizer05so no small capacitor in parallel to 47k18:03
DocScrutinizer05if you could make sure the GPIO is tied to GND via a strong pulldown in all MCU glitch resp reset situations, you can even leave out that 47k18:04
DocScrutinizer05and use one pin only18:05
wpwrakafter reset, gpios are high-Z. so, no18:07
wpwrakbut yes, the rest should work18:07
wpwrakon power loss, the MCU needs to save the whole data set to a pre-erased flash page. takes some 70 us per longword on average. the usual optimizations apply.18:09
wpwrakso that doesn't look too bad18:10
wpwraknice idea, thanks !18:11
DocScrutinizer05CC by attribution18:11
wpwrakin the post where i explain the part ? alas, there's no document explaining the whole system. not enough hours in a night :(18:19
DocScrutinizer05wherever you feel it's appropriate18:26
DocScrutinizer05it's just a ;-) thing18:27
DocScrutinizer05do you have a power-good signal which tells you when battery removed, by getting drawn to GND? If yes, you might want to connect the 47k pulldown to that signal instead of GND, so you don't get quiescent currents during normal operation (battery inserted)18:35
wpwrakyes, i thought of that, too. the boost converter i'm considering has a "battery good" signal. i think it may be used for something like this, but need to check some more.18:47
DocScrutinizer05wpwrak: (mech reversal block) http://maemo.cloud-7.de/share-service/DSCF1916.JPG http://maemo.cloud-7.de/share-service/DSCF1917.JPG AA but the principle is the same19:05
DocScrutinizer05http://wstaw.org/m/2016/03/23/DSCF1914.JPG  halfassed implementation, actually a fail19:12
wpwrakkewl. so it does exist in real life. no FET or polyfuse needed then :)19:23
wpwrakthe less gunk in the power path, the better19:24
wpwrakhmm. too good to be true ? https://www.kickstarter.com/projects/olo3d/olo-the-first-ever-smartphone-3d-printer20:36
mththe campaign period does include April 1st20:52
mthif it's a light-sensitive resin, shouldn't it be kept in opaque bottles?20:52
wpwrakmth: yes, at heise.de they have a pretty long list of doubts: http://www.heise.de/newsticker/meldung/Olo-3D-Druck-mit-dem-Licht-aus-dem-Smartphone-Display-3149561.html20:56
mthhere is a company that claim to use daylight resins as well: http://www.photocentric3d.com/21:14
mthall their bottles are black though21:14
mthmaybe the tech is real but the video is faked since the kickstarter project has no working product yet?21:15
mth(which would violate kickstarter rules for hardware projects)21:20
mthhmm, in their demo video they seem to fill the tank from a black bottle21:29
mthso the bottles shown in the main video are probably not actual resin21:30
mthalso, if it takes 4 hours to print a small object, is it actually a good idea to use a phone as the screen?21:31
mthwell, if you have an old smartphone around that you no longer use as a phone, I guess21:31
wpwrakif the thing works, it would still be cheap if you add the cost of a new smartphone :)22:34
--- Thu Mar 24 201600:00

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