#qi-hardware IRC log for Thursday, 2014-06-12

whitequarkooooh, transistor failure modes, now it gets interesting00:44
wpwrakjust when you thought they were invulnerable ? ;-)01:01
whitequarkwpwrak: not quite, I was fully expecting something weird01:02
whitequarkbut I really want to understand how exactly did it fail01:02
whitequarkit's no fun otherwise01:02
wpwrakah, so you blew it up. you did add the diodes ? :)01:21
whitequarkof course not, told you so!01:22
whitequarkI'm adding them now though, since I understand how and why it failed01:23
wpwrak;-)) i did warn you when you mentioned the schematics, though :)01:24
whitequarkI remember01:24
whitequarkI thought about it even when designing it, but it was probably too late and the rational voice was ignored01:24
whitequarkanyway, it's even good that it failed, since I now understand it better01:25
whitequarkanother thing: 1) my homebrewn boards stand up to rework pretty well (no delaminated traces) 2) investing in solder mask was wise01:25
wpwrakat least your experience puts the lingering question to rest whether this protection is really needed or whether we all just add it because "that's what one has to do", but it would actually be fine without it :)01:25
whitequarkexactly01:25
wpwrak(rework) well, they better do :) even on my little anelok board i replaced the OLED with its wide FPC twice already, without a problem. (FPCs are fragile critters and don't like a lot of mechanical abuse, so while they were just flapping freely, their life was intense and short)01:27
whitequarkhuh, yeah, both transistors are broken down entirely01:33
whitequarkconductive between any pair of pins01:33
wpwrakwhitequark the transistor slayer ;-)01:39
whitequarkummmm01:41
whitequarkI took a brand-new bc547 and C-E are shorted01:41
whitequarkwat?01:41
whitequarkI'm fairly sure this is not how transistors are supposed to work01:41
whitequarkoh, nevermind, it was my error01:45
whitequarkand such a stupid one I'm not even going to say which01:45
wpwrak(wat?) you don't know the tv series "buffy the vampire slayer" ?01:46
wpwrakah, new transistor. i see01:46
whitequark"wat?" was about the shorted C-E pins01:46
Action: whitequark looks away01:46
whitequarkI put it into the breadboard, completely forgetting that *solderless* breadboards have rows of shorted pins01:47
whitequarkbeen years since I touched one01:47
wpwrak:)01:47
whitequarknope, after adding diode it still does not work, with almost exact same symptoms02:03
whitequarkand the transistor seems fine, or at least the Vf of both junctions is the same as for new02:03
whitequarkand it's not broken down02:03
whitequarkso basically... initially, Vbe=1.3, Vce=8.7. after feeding a pulse to base, Vbe=0.6, Vce=8.002:05
whitequarkand the relay is on02:05
wpwrakyour transistors are NPN or PNP ?02:16
wpwrakin any care, your Vbe looks quite wrong :)02:19
wpwrak#s/care/case/02:19
whitequarknpn02:24
Action: whitequark scratches his head02:29
wpwrakwhat are you trying to send down Vbe ?02:29
whitequarkum?02:30
wpwraki mean what levels are expected on the base ? on = xxx V, off = xxx V ?02:30
whitequarkoh02:31
wpwrak(before the resistor)02:31
whitequark0/5V02:31
wpwrakso the initial 1.3 V should be 0 V02:31
whitequarkyes02:31
wpwrakand when you turn it "on", the base voltage drops. remarkable ;-)02:31
whitequarkI mean, I measure 1.3V with base floating02:31
wpwraka floating base means nothing :)02:32
whitequarkso I've tried to look up a commercially sold kit with a similar circuit, they have a resistor divider at base02:33
whitequarkwhich makes 0.7V out of 5V, and also pulls it down normally02:33
wpwraka BJT is a current-defined device. as long as you have 0 current, your base could be anything (well, it will still settle to some value due to secondary effects)02:33
wpwrakpull-down is a good idea02:33
whitequarkwpwrak: (current-defined) sure02:34
whitequarkwhat I mean is, I don't understand why it functions as a *latch*02:34
whitequarkthat just seems absurd02:34
whitequarkit's not a FET, when the base is floating, it must do the same thing regardless of what I did to it before02:34
whitequarkassuming it's not fried02:34
wpwrakwell, leave the base open (float) for now. then the relay is off, right ?02:35
whitequarkyes02:35
wpwraknow connect 5 V to the base (through resistor). then it should turn on. does it ?02:35
whitequarkyes02:35
wpwrakwhen you remove the 5 V and float again, does it turn off ?02:35
whitequarknope.02:35
wpwrakfascinating :)02:36
whitequark*exactly*02:36
wpwrakin russia, transistors have memory02:36
wpwrakif you force the base to 0 V, what happens ?02:36
whitequarknothing02:37
whitequarkrelay stays on02:37
wpwrakare you sure it's not a bistable relay ?02:37
wpwrakone thing to check: are your transistor pins really where you think they are ?02:38
whitequarkhttp://www.songchuan.eu/en/attachments/030_834.pdf02:38
whitequarkhttp://us.100y.com.tw/pdf_file/TR99.pdf02:38
whitequarkneither of these mention that it's "bistable"02:38
whitequarkI am sure, I've checked it with a DMM and it has two junctions between the expected pins02:39
whitequarkand it matches the DS02:39
whitequarkalso, I've tried desoldering the transistor and triggering the relay manually02:39
whitequarkso it is NOT bistable02:39
wpwrakok02:41
wpwrakforcing base to GND and still having the relay on is ... interesting02:41
wpwrakdid you connect to before or after the base resistor when forcing base ?02:42
whitequarkconnect to what?02:42
wpwrakGND to base02:42
whitequarkoh, tried either02:43
whitequarksame result02:43
wpwraksequence: 1) all open. 2) apply power to system (relay off). 3) drive base through R to 5 V -> relay on. 4) float base (relay still on). 5) drive base to GND (relay still on). correct ?02:44
whitequarkyea02:44
wpwrakcan you monitor the overall system current throughout all these steps ?02:45
whitequarkwith mA precision, yes02:45
wpwrakfor step 3, you can just use your system voltage (~ 9 V ?) so you don't need an external 5 V source02:45
whitequark24V02:46
wpwrak"system" the transistor plus relay board. let's eliminate any driving circuit02:46
wpwrakah, quite high voltage02:46
whitequarkummm02:47
whitequarkthe current is... below what my adjustable PSU can measure02:47
wpwrakuse a multimeter :)02:47
whitequarkbut the relay needs 18 or so mA to stay on02:47
whitequarkwtf02:47
wpwrakmore like 2 mA, i think (drop-out voltage)02:49
whitequarkvery interesting02:49
whitequarkif I keep the base connected via a 2k resistor (in sum) to 24V, I see 2.5mA of current02:49
whitequarkbut the relay stays on either way02:50
wpwrakwhat's the other way ?02:50
whitequarkewhether the base is connected or floating02:51
wpwrakat 2.5 mA the transistor should let something in the order of 100 mA through. more than enough.02:51
whitequarkok, the PSU is lying02:51
whitequarkit's 7mA just after turning it on02:51
wpwrakmaybe it's not floating when it's "floating" ;-)02:51
whitequark... then 25mA while the base is high02:52
whitequarkthen back to 7mA02:52
wpwrakwhat's your transistor's current gain ?02:53
wpwrak7 mA should be plenty to keep the relay on02:53
whitequarkit's bc547c, so 420..80002:53
wpwrakbut just not quite enough to _turn_ it on02:53
whitequarksure02:54
wpwrakso you'd need about 10 uA on the base to turn the relay on. at 24 V that's about 2 MOhm. time to check for sneak current paths.02:55
wpwrakhow careful were you with cleaning the board from flux residues ?02:55
whitequarkI didn't02:57
whitequarkmy DMM shows about 19M© to +2402:57
wpwrakwell, more like 40 uA, when we include the mystery 7 mA you start with. but that's still low enough for flux "connections"02:57
whitequarkthat still doesn't answer the question of why doesn't the relay turn off with base forced to 002:58
wpwrakone thing after the other ...02:58
wpwrakdisconnect power, connect base to GND, reapply power. what is the system current ?02:59
wpwrakoh, and your diode goes across the relay ? or across relay and transistor, or something like this ?03:00
whitequarkonly across the relay coil03:00
wpwrakgood03:02
whitequarkcleaned the board with dimethoxymethane, same result03:03
wpwrakis system current is still high when starting with base to ground, then the next step would be to remove the transistor and see what happens03:03
wpwraksince it behaves as if there was a path in parallel to the transistor03:04
whitequarkthe resistance went under what my DMM can detect03:04
whitequarkwell, I have second relay without the transistor03:04
whitequarkon the same board, which behaved exactly same03:05
whitequarkcan check it03:05
wpwrakyou flux is friendly, being in the 10+ MOhm range. you ought to try one of the nastier ones some day, which go below 1 MOhm. they do great things to reset lines.03:06
whitequarkouch. good thing my PSU is current-limited03:06
Action: whitequark accidentally shorted it with DMM lead03:07
whitequarknice flash03:07
wpwrakyes, that's why we use lab power supplies and always set the maximum current to something not too far above what we expect to need ;-)03:08
whitequark*facedesk*03:08
whitequarkfuck.03:08
wpwrakwhat was it ? :)03:09
whitequarkyou see, the board is nicely symmetric...03:09
wpwrak+/- reversed ?03:09
whitequarkoh no03:10
whitequarkX1/X4 swapped03:10
whitequarkit behaved like a latch because lol, I assembled a latch, essentially03:10
wpwrakhmm, i don't have the schematics in sight03:10
whitequarkhttp://i.imgur.com/c9LXLcS.png03:11
whitequarkoh. hm. nothing changed.03:12
wpwraknot sure how swapping X1/X4 would turn it into a latch ... well, if K1 is open, maybe fun things could happen with ground03:14
whitequarkthought about that, yes03:14
whitequarkoh. open03:14
whitequarkwell, nevermind, doesn't matter03:14
whitequarkit's STILL a latch03:14
wpwrakreliable russian technology. no matter how much vodka you had and where you connected things, it always works the same way03:15
whitequarkbut why is it a latch?!!03:17
wpwrakdid you do the base-to-ground-when-applying-power test ?03:18
wpwrakoh, and you have two K1 in your circuit03:19
wpwrakand three K203:19
whitequarkwait03:20
wpwrakah, that's all the relay03:20
wpwrakweird symbol without the connection03:20
wpwrakwhich one are you testing ? K1 or K2 ?03:21
whitequarkK103:21
whitequarkso I shorted base to emitter03:25
whitequarkit still eats 7mA03:25
whitequarkdesoldered the transistor. 0mA03:30
whitequarksoldered the transistor and literally left the base floating in air. 12mA03:32
whitequarkinsanity. there's thousands of variants of this circuit on the web03:43
wpwraki strongly suspect your transistor is somehow connected the wrong way03:53
wpwrakbehaviour would still seem odd, though03:53
wpwrakwhich transistor is it again you said ?03:54
whitequarkbc547c. marked as "C547C"03:54
whitequarkbtw, you can actually connect the npn transistor in reverse (i.e. swap C/E) and it would still kinda work, although with lower hFE03:54
whitequarklet me try to reproduce this on the breadboard...03:55
wpwrak(reverse) yes03:55
whitequark...04:22
whitequarkwell04:22
whitequarkif I indeed connect it in reverse, then it has an idle current of 12mA04:23
whitequarkwhich matches what I observe on that board..04:23
whitequarknow, how did this even happen? the transistor has legs bent that they have a "triangle" footprint. the board, same.04:24
whitequarkwhat the FUCK04:26
whitequarkEagle has the footprint in revers04:26
whitequarksure enough, now it works04:36
wpwraknever trust a footprint you didn't design yourself ;-)04:40
whitequarkand especially not the ones I designed myself04:41
whitequarksince I have attention span of a squirrel on meth04:41
wpwraki trust the ones i designed because i know that they may be wrong and i have to test them. so they're all properly tested :)04:42
wpwrakbesides, with fped it's pretty easy to review the critters04:42
whitequarkit's not exactly complex with eagle, it just never occured to me that BC547 can come in two mirrored footprints, both asymmetrical in a way that would prevent you from misplacing them04:43
whitequarkthis is just so fucked up04:43
whitequarkoh also04:43
whitequarkflyback diode not required :P04:43
wpwrake.g., numbering and precise geometry: http://downloads.qi-hardware.com/people/werner/tmp/kicad-libs-modules.pdf04:43
whitequarkat least the transistor survived more than 50 cycles04:43
wpwrak(that's also roughly what you see on the screen with fped.)04:43
wpwraki had my transistor experience with gEDA. in kicad i actually found that the stock libs matched each other and reality fairly well04:44
wpwrakbut you sometimes got odd shapes04:45
wpwraknot necessarily incorrect but still odd. e.g., standard package X, but from a part with weird thermal requirements, so the footprint for all X is like that04:45
whitequarkcrap. I completely forgot07:37
whitequarkthe control board has the *signal* pin hardwired to +5V, and it pulls the *ground* pin down07:38
whitequarkwhich is kind of really weird but oh well07:38
whitequarkso I basically have a logic OR between "spindle CCW" and "coolant pump on"07:38
whitequark...07:39
whitequarkseems like a rev.B is in order07:41
wpwrakwhitequark: btw, when you draw footprints with eagle, can you output your work in a way that allows for easy review ? e.g., with measurements, like on pages 4 etc. of http://downloads.qi-hardware.com/people/werner/tmp/kicad-libs-modules.pdf ?14:36
whitequarknot as far as I know14:37
whitequarkbuuut it would be trivial to write an ULP that outputs that. maybe someone even did it14:38
whitequarkwpwrak: why the question? neo900?14:48
rohwell.. you really now if all footprints are ok if you soldered the first prototype and it does work without reworks ;)14:53
rohin the end.. one would even need to check reality to documentation since the latter may be wrong from time to time14:53
wpwrakwhitequark: yup. collecting arguments against the lame bird, just in case there's an opportunity :)15:17
wpwrakroh: and yes, that's the ultimate test :) actually, not even then you're sure since you may have footprints that happen to work but aren't really good and will give you yield headaches. or the fab silently fixed some problem for you but the next fab won't be so nice ...15:18
wpwrakroh: but with all the measurements you can at least make sure you draw what's in the data sheet. big step forward :)15:18
wpwrakalso, you can see all the measurements at once. not having to hunt them down interactively with some measurement tool15:29
wpwrakttataa .... big thunderstorm coming. if that sort of weather moves north, they'll need snorkels at the world cup17:07
eintopf:/17:33
ysionneaularsc: about the RF chip I am working on, my company just published some stuff: http://www.m2mnow.biz/2014/06/12/21403-sequans-introduces-colibri-lte-platform-internet-things/17:56
ysionneauthe rf is 324117:56
ysionneauthere are some approx spec17:56
larscah yes, the latest buzz, the internet of things18:01
ysionneauyes that's just buzzword you don't care about :p18:04
ysionneauthat's a chip, you just do whatever you want with it :p18:04
ysionneauI suspect those words are only there for search engines18:04
larsceverybody goes crazy about the internet of things recently18:05
larscI guess some market research institute said that this is going to be the next big thing18:06
ysionneauit could be honestly, if you have a good idea for a cool thing to connect to internet18:06
ysionneauso far I've seen none18:06
ysionneauso far most "small devices" use bluetooth or BLE and don't access internet directly18:07
ysionneauand have small battery life18:07
wpwrakin an occult ritual around 1952, five satanists brought the Thing into the pentagram. the Thing snarls and hisses. the arch-satanist, with a loud voice, proclaims, "Things, I name thee NSA !". thus, what traitor Emmanuel Gol... errr, Snowden, revealed, is merely The Thing claiming was what always has been his by rights.18:07
Action: ysionneau gives a shot of tranquilizer to wpwrak 18:08
ysionneauthere there :p18:08
wpwrak;-)))18:08
wpwraki think the customary procedure would be to sing "warm kitty" :)18:09
ysionneau;)18:24
DocScrutinizer05sigh, another chip with embedded APE19:09
DocScrutinizer05it sounded so good19:09
DocScrutinizer05except for the OTA-management stuff, which I rather wanted to have disabled19:09
DocScrutinizer05ok, it got VoLTE, but how's about 3g/2G voice? Which bands will be supported?19:11
ysionneautalking about the Sequans bb+rf chip?19:13
DocScrutinizer05yup :-)19:13
ysionneauno 3g/2g at all19:13
ysionneauonly 4g19:13
DocScrutinizer05oops19:13
ysionneauand only LTE actually19:13
ysionneauthat's the downside :)19:13
DocScrutinizer05well, there's a reason why it's labeled "for IoT"19:14
ysionneauit's small, cheap, low power chip (don't ask me numbers, I don't have them, but it's our roadmap)19:14
DocScrutinizer05actually 3G/UMTS wouldn't even be compatible with the low-power claim19:15
DocScrutinizer05and iirc 3G royalties are insane19:15
ysionneauWhich bands will be supported? < I don't know the entire list out of my head, I heavily tested band 13 so far19:17
ysionneauI know there has been some tests on band 419:17
ysionneaudon't know about others19:17
ysionneaumainly it will be band A and C/D19:17
ysionneauhttp://niviuk.free.fr/lte_band.php19:18
ysionneauLTE bands, what a mess19:18
astrRecommendations for a tiny/mini/small speaker + amp that is as loud as the one in the nano note or louder? bonus points for even better sound quality :)21:27
astrthis is for a diy handheld computer with a 4.3inch lcd to give you a idea of the size of what the speaker is going into21:29
DocScrutinizer05that's a tricky question21:58
DocScrutinizer05generally you won't get away without some sort of resonance chamber or transmission line21:59
DocScrutinizer05unless you use speaker of at least ~5cm diameter21:59
DocScrutinizer05and even that one will sound mickey mouse when operated without any proper chanber/damping etc22:00
DocScrutinizer05general problem: the basic speaker has a front and a backside ;-) sounds trivial bit that's actually the problem you need to fix, since for acoustic waves of wavelength ~radius of speaker and longer, you get a short between front and back22:02
DocScrutinizer05so no matter how much energy you blast into your timy speaker and how far the tiny membrane goes forth and back - at 50Hz or 100Hz it just will mix the air and not produce any sound wave 22:04
DocScrutinizer05the air goes straight around the speaker from front to back22:04
DocScrutinizer05you could power a 1cm speaker with 100W and it would only heat up the air 22:05
DocScrutinizer05at 50Hz22:05
DocScrutinizer05you probably wouldn't hear anything in a meter distance22:05
DocScrutinizer05so you need a chamber at one of both sides, to contain the air in it and not allow it to just move around the membrane from front to back and vice versa22:06
DocScrutinizer05either you use a closed chamber, or an open one which basically is a transmission line then22:07
DocScrutinizer05and that chamber must get tuned to the speaker's properties, like inertia and damping of the membrane22:08
DocScrutinizer05otherwise you build a organ pipe22:09
DocScrutinizer05that only works for one frequecy, but then works extremely good on that single one22:09
DocScrutinizer05astr: sometimes you can find so called "closed speakers", they already have a chamber integrated, and on a good seller you can find datasheets that explain how to mount the speaker so it will sound good22:12
DocScrutinizer05but often they only mention frequency range, Wattage, resoncance frequency and Q of the speaker transducer, and leave it up to you to calculate a matching chamber for that22:15
astrDocScrutinizer05: thanks, very helpfull22:15
DocScrutinizer05sarcasm?22:15
DocScrutinizer05sorry! it's as nasty as that22:15
astrbouns points awarded22:15
astrno sorry chinaise way of writing has worn on me. I meant it in a good way22:16
DocScrutinizer05:-)22:16
astrtoo much reading of aliexpress english22:16
DocScrutinizer05make sure your chamber is not as perforated and soft material as the white silly thing in GTA01/02 ;-)22:18
astrcan you get effenant little speakers that use 1W or less? but still as loud?22:21
astrs/effenant/efficient/g22:22
DocScrutinizer05umm, yup, I think 1W is feasible22:22
astroh goody good, where? I guess not aliexpress hehe22:22
DocScrutinizer05I don't know where to buy small speakers22:23
DocScrutinizer05I either used spare parts for an existing design or talked directly to the manufacturer22:23
DocScrutinizer05for stuff like GTA03 speakers22:24
astrI haven't found any useful results for closed speakers22:24
astrI'm using ixquick.com22:24
astrok22:24
DocScrutinizer05yeah, they are not exactly common in sub-inch size22:24
DocScrutinizer05I mean, such chamber *always* is bulky, that's by physical principles22:25
astrI think I can work with a 5cm one, though I don't know how big the chamber needs to be. I expect my diy handheld to be bulky22:26
DocScrutinizer05ooh, 5cm is already a great size22:27
astrbukly any way ... so what harm is a fre more cm for a speaker :P22:27
astrs/fre/few/g22:27
astrthough I would prefer to keep the overall size small so I can still call it a handheld and hold it with one hand22:28
astrit's only for fun alarms, notifications, text2speech22:29
astrthe speaker that is22:29
astrthen again,I  plan to have a pezio sounder = nice and loud for alarms to wake me up hehe mwhahah :D. not sure how to control one on the olimex LIME board22:33
DocScrutinizer05typical closed chamber driver/transducer:  https://en.wikipedia.org/wiki/Compression_Driver22:36
DocScrutinizer05the horn is an acoustical transformer that creates a lot of air moving fast against low force, from a smal amount of air at driver moving slow at very high pressure/force22:38
DocScrutinizer05since only a small amount of air needs to get moved by speaker, you can use a small (but rigid strong) membrane22:39
DocScrutinizer05in a small case like phone, you don't use horns but chambers or transnissin lines 8usually a combination of both)22:40
DocScrutinizer05transmission*22:40
DocScrutinizer05http://members.shaw.ca/t-linespeakers/diy-11-apr-99/   http://t-linespeakers.org/projects/chris/index.html22:41
DocScrutinizer05http://p10hifi.net/TLS/projects/chris/images/innerView_tn.jpg is pretty good picture - of course yours will have to be smaller, but the principle stays same22:42
astrtiny tiny amp http://www.aliexpress.com/item/Free-Shipping-5pcs-lot-DC-5V-2-Channels-3W-Digital-Power-D-Audio-Amplifier-Board-Amplifier/1758901278.html22:44
DocScrutinizer05http://www.quarter-wave.com/22:47
DocScrutinizer05class-D amps there's quite some choice of chips to pick one22:48
DocScrutinizer05nice thing about class-D: extremely efficient22:49
DocScrutinizer05thus you need no huge heatsinks for lots of Watts to the speaker22:49
astryep22:51
astrI'm falling asleep, maybe in 12hours I'll resume this chat :)22:52
astrthank you ever so much22:52
DocScrutinizer05http://www.quarter-wave.com/TLs/TL_Anatomy.pdf22:54
astrDocScrutinizer05: I'll the my irc running. so feel free to dump and extra info,thoughts questions,links and I'll read them in the morning :)22:54
astrs/and/any/g22:55
DocScrutinizer05nah, i'm also leaving now, cheers! :-922:55
DocScrutinizer05:-)22:55
astrnitnti22:55
DocScrutinizer05in http://www.quarter-wave.com you really have it all22:56
DocScrutinizer05nice site22:56
DocScrutinizer05http://www.quarter-wave.com/Horns/Advanced_Back_Horn.pdf p.3 nice "horn" with coupling chamber design. Transmission line also uses coupling chamber22:57
DocScrutinizer05http://www.quarter-wave.com/Gallery/Poor_Plumbers_Pluto.jpg  :-D23:20
DocScrutinizer05nice TL design23:21
nicksydneywpwrak: https://www.indiegogo.com/projects/3dpad-touchless-gesture-controller-arduino-shield-by-ootsidebox23:57
--- Fri Jun 13 201400:00

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