#qi-hardware IRC log for Saturday, 2011-10-15

qi-bot[commit] Xiangfu Liu: new package: listener automatically records when sound is detected (master) http://qi-hw.com/p/openwrt-packages/f3d79bd02:36
qi-botThe build was successfull, see images here: http://fidelio.qi-hardware.com/~xiangfu/compile-log/openwrt-xburst.full_system-10142011-0131/06:05
DocScrutinizerzrafa: more like 3 years older than 30: blit 1981, X11 198413:41
whitequarkDocScrutinizer: thanks a lot for your suggestion about solder mask/silkscreen removal21:53
whitequarkI've improved it a bit: if you use a sharp needle, it's easier to not accidentally split some important trace in two21:54
DocScrutinizerwhitequark: sorry I don't actually recall what I suggested and for what21:54
Action: whitequark looks up the logs21:55
wpwrakDocScrutinizer: if he used a knife, it was probably me. if he used the dark side of the force, all the credit it yours ;-)21:56
wpwrakDocScrutinizer: that suggestion was about removing stuff that covers traces21:56
whitequarksorry, that really was Werner21:56
whitequarkfor some reason I remembered you as the author of the advice...21:57
DocScrutinizerwell, maybe because it is quite probable that you'd get such kind of advice from me21:57
whitequarkwpwrak: yep, I've tried with a knife. using needle is more precise, i.e. I do need the solder mask for 8WSON a lot, but please, no pad-covering silkscreen21:58
DocScrutinizerI'd nevertheless use a small "round" knife and carefully scrape (NOT cut!) soldermask away21:58
wpwrakif you solder manually, it actually doesn't mater all that much. it's the flux that does most of the magic. but yes, silk screen on pads = evil ;-)21:59
DocScrutinizersoldermask on pads == BUG22:00
whitequarkwpwrak: yes, flux is very important too, but if I have exposed traces, solder tends to leak to the uncovered places and makes weird surface tension patterns22:00
whitequarkDocScrutinizer: that was a $25 extra-cheap prototyping (for 10 10x10cm boards)22:00
whitequarkeverything except the damned solder mask is great22:00
wpwrakwhitequark: when that happens, try thin solder wick. (thin: <= 1 mm wide)22:00
DocScrutinizerfor my education: I thought solder mask is the even green coating, while silkscreen is the printing (usually white)?22:01
whitequarkand the latter is off by ~0.8mm22:01
whitequarkargh, that was a typo22:01
whitequarkof course the silkscreen is off22:01
whitequarksolder mask, holes, vias and every other thing is perfectly done22:01
wpwrakoff by 0.8 mm ? nice ;-)22:01
whitequarkmaybe even 1mm, it varies22:01
wpwrakwhy not just do their own creative painting instead ? ;-)  well, thinking of it, they kinda did :)22:02
whitequarkthey've also added some internal number on a random place on the board22:02
whitequarkI've rechecked: they actually do that for every PCB22:02
whitequarkas you can see, the printings inside the top connector aren't on gerbers22:03
DocScrutinizerwpwrak: seems they really do "creative painting" then ;-P22:03
whitequarkbut well then, it's $25 and they did it in single day22:04
whitequarkI'll rather scratch off some silkscreen from two boards I do rather, for example, pay twice of that, or wait for a week.22:05
wpwrakand then let's count the hours you spent figuring out what exactly they had done ;)22:05
wpwrakif i need quick but unreliable, i do it in my kitchen :)22:06
whitequarkwpwrak: I think that was a meaningful experience. sometimes things go wrong anyway, and it's better to know what the hell may happen there instead of trying to figure it up in a hurry on some much more complex and expensive board22:06
wpwrakyeah. you already know where not to go for complex boards :)22:07
DocScrutinizerwpwrak: you do ds+via in your kitchen? respect!22:07
whitequarkDocScrutinizer: ds?22:07
wpwrakDocScrutinizer: double-sided yes, vias with a bit of wire.22:07
whitequarkthat's very impressive, yes22:08
DocScrutinizerwired vias suck :-D22:08
whitequarkI am barely able to do a 1-sided board with 12mil traces22:08
wpwrakDocScrutinizer: i have a drill and wire combination where the wire stays in place by friction. takes a lot of pain out of the soldering :)22:08
whitequarktried to make them for ~half of a year intermittently, but finally gave up22:08
whitequarkit has consumed a whole lot of time with almost no outcome22:09
wpwrakwhitequark: maybe you should reconsider your approach to vodka :)22:09
DocScrutinizerwpwrak: I used to use small copper 'nails' which I riveted with a 20g hammer and a bolt then22:10
wpwrakpheew ...22:11
DocScrutinizerfinal soldering is utterly simple, and almost not needed at all22:11
whitequarkwpwrak: anyway, PCBs without any solder mask conflict with my aesthetic feelings22:12
wpwraksame with my wire. i just add solder paste on both sides of the vias, put the wire into the holes, heat, done22:12
DocScrutinizerwhitequark: I'd redesign my silkscreen anyway22:12
DocScrutinizerit's too tight22:12
wpwraki can also do it without solder paste, but it helps in difficult situations22:12
DocScrutinizersilkscreen usually isn't that precise 22:13
whitequarkDocScrutinizer: you've meant my silkscreen, I guess. or not?22:13
whitequarkthat's Eagle's default silkscreen for most of the components. I think I only designed it for IC222:13
whitequark(sorry, I'm too stupid to use KiCad or some other FOSS tool)22:14
DocScrutinizerJP1 JP3 and IC1 silkscreen looks errr odd22:14
whitequarkahh yes, also I've made JP2 from scratch, yes22:15
whitequarkthat's for a very perverted Hirose connector22:15
whitequarkthe really funny thing is, when I've tried to plug in the LCD, the connectors did not match anyway22:15
DocScrutinizererr is that JP2? I just see JP1 is at the bottom, not next to R322:16
DocScrutinizerumm JP422:16
DocScrutinizerJP2 looks not all that bad22:16
whitequarkpitch is the same, but my one has "blades" (i.e. flat and wide pins), and the one at LCD _almost_ likes my mating part, but it expects small round pins instead, and a bit different keys22:16
wpwrakyeah, a bit tight. but what really screws it up is the HUGE offset. i don't think the rest would be a problem.22:17
whitequarkDocScrutinizer: JP3 and JP4 is from Eagle's pinhead library. I think I'll use SparkFun one in the future, it's quite better IIRC22:17
DocScrutinizeranyway, why print any silkscreen so close to the soldering area at all? allow silkscreen to move a little, like +-1mm22:18
wpwrakexcept, of course, the absence of thermal relief. you want to fix that before you make the next 10M units ;-)22:18
whitequarkI understand it _now_. But that's my first PCB with silkscreen/solder mask, as the title states22:18
whitequarkit's not surprising that I screw it up :D22:18
DocScrutinizere.g. I'd never ponder to print silkscreen in between adjacent pins of a post-connector22:19
DocScrutinizerand what eagle does there feels kinda odd22:19
whitequarkDocScrutinizer: do you add all silkscreen on your boards manually? or only the outlines?22:19
whitequarkor do you just design all the components from scratch? :)22:19
DocScrutinizererr, I usually only add component label silkscreens, or if I choose to add frames etc for e.g. smt resistors etc then I make sure they have at least 0.5mm clearance to the solder area22:21
DocScrutinizerit's not exactly unusual for silkscreen printing to be less carefully done than all the other steps of PCB manuf22:21
DocScrutinizerafter all silkscreen is not supposed to be any "functional" part of the PCB - it's for mere convenience22:22
DocScrutinizersolder stop mask otoh is usually very accurately done22:25
DocScrutinizereven the paint used has different properties it seems22:25
whitequarkI think it's because of different processes, i.e. solder mask is photoresistive too--isn't it?--and silkscreen is done, well, with silkscreen process22:27
whitequarkthe latter seems much less precise for me22:27
DocScrutinizeryes, exactly22:30
whitequarkdarn. still does not work :/22:30
whitequarkI wonder why22:30
DocScrutinizerthat's why for silkscreen there's just one mandatory rule: keep clear of soldering pads ;-D22:30
whitequark10 minutes earlier, EN was effectively isolated by silkscreen. it isn't now...22:30
whitequarkahh of course. it won't work with FB either.22:34
whitequarkits thickness is enough to prevent contact if there's even a small island, less than 0.1x0.1mm22:35
whitequarkhow do you think, is it enough to place a small (8WSON, it is like a small LGA-8) chip on the pre-tinned (i.e. with a small amount of solder on it) pads and then heat it with hot air and let surface tension do its work22:42
whitequarkor should I press it gently on the top to push it to the board?22:42
wpwrakhow would you prevent it from getting blown away without pushing a little ?22:44
wpwrakand i hit air fails you, you can always fall back to flux + solder + iron22:45
whitequarkit doesn't get blown away even on the top air speed of my hot air gun22:45
whitequarkI use an RMA flux-gel, that may be the cause22:45
wpwrak(not blown away) lucky you :)22:45
wpwrakokay, RMA is very sticky :) but still ...22:45
whitequarkit doesn't get at all, really22:46
whitequarknever had that problem22:46
whitequarkeven with 0603 resistors or 3x3mm cases22:46
whitequarkit's our special russian hot air :D22:46
wpwrakhmm. maybe your is just better than mine :)22:47
whitequarkmmmkay, still does not work.22:48
whitequarkI hate my life.22:48
whitequarkI guess when it _looks_ like it's a bit off, it is actually not soldered at all.22:48
wpwrakif you can see a problem, there are probably two others already that you don't ;-)22:49
whitequarkok, I've pushed it to the board. looks like _now_ it _is_ soldered22:51
whitequarkso I should cool it down and check...22:51
whitequark(the reason behind me not wanting to push the component is that yesterday I've moved it quite far away from the pads. It's better today by some reason.)22:52
wpwrakthree spools each of 0.5 kg of solder later ... ;-)22:52
whitequarkI don't add more solder, it isn't needed22:53
wpwrakfor pushing, what works best for me are angled tweezers22:53
whitequarkand I've used almost no flux, maybe several mm's from the syringe22:53
whitequark(tweezers) yep, I'm using them. they're most convenient for this IMO, too22:53
whitequark(on the other hand, these extra-super-small components are fast to heat too, so they _could_ survive more solder-desolder cycles :)22:54
whitequarkfuck yeah. IT WORKS!22:55
whitequarknow, the other board.22:55
wpwrakcongratulations ! :)22:56
whitequarkat least, now I understand the implications of using a random TI-component-which-has-right-characteristics and not-looking-up-the-case-in-advance.22:57
whitequarkit looked _much_ bigger on my notebook's screen.22:57
DocScrutinizerbuilding dummies helps a lot22:58
whitequarkDocScrutinizer: do you mean printing the PCB on paper?22:58
DocScrutinizereven if it's only cardboard or paper22:59
wpwrak(it looked bigger) ;-))))22:59
DocScrutinizernah, dummy of e.g. a 0402 component22:59
DocScrutinizerI constantly fail to figure how small those critters really are23:00
wpwrakbuilding dummies seems hard. printing is easy. also gives you a feeling. although, on paper it always seems more compact than in real life.23:00
whitequark0402 is slightly less than the distance between tips of my tweezers, so I don't use them23:01
whitequarkalso, (seriously), it's nowhere to be found in Russia23:01
whitequarki.e. there is one very expensive (and by "very" I mean 10x-50x more expensive than in the right place and/or in something like DigiKey) shop which has them23:01
DocScrutinizerhelps for other components as well, esp chips23:01
wpwrakwhitequark: you need better tweezers, too :)23:02
whitequarkin the places where you can get components by sane price (and I mean radio markets), there are no 0402 ones. 1206, 0805, yes. 0603 -- harder, but still possible. 0402 and less is missing at all. looks like it's not only me having problems with them23:02
DocScrutinizerwhitequark: well at real values of a dozen per penny they will need massive increase in price to make any money on selling them23:03
wpwrak0402 isn't too hard. but you can't see very well at that size. so you just have to have faith :)23:03
DocScrutinizerwhen sold in single qty23:03
Action: kristianpaul feels bad about his soldering equipement23:04
kristianpaulI guess 30W is not enought?23:04
kristianpaulIs for this board http://gnss-sdr.ru/media/1/20101001-front_image.jpg23:05
whitequarkDocScrutinizer: at market I can buy some regular components (like, for example, 0603 22p ceramics) for ~$0.016 or even less, down to $0.005 if >50 pts23:05
kristianpaulalso i dont know how to handle the GPS RF IC, i dont want to damage if i touch it?23:05
whitequark(there are some things like really fat and small ceramics in 0603/uF range, which is still expensive and/or missing, but that's different things)23:06
whitequarkand that shop (called Chip & Dip; don't go there if you'll see it) sells them for no less than $0.2 each, regardless of qty23:06
whitequarkMAX232 is $10 instead of $1 as it should be, etc.23:07
kristianpaulwpwrak: do you have inventory of your soldering equiment somwheere? ;^)23:07
whitequarkalso, they frequently fail at selling the right component, as their consultants are quite stupid.23:07
wpwrakkristianpaul: hmm, i'm afraid not. it's a collection from all over the place23:07
wpwrakwhitequark: seems that digi-key will get your next order soon :)23:08
kristianpaulwhat i should get? a 90W iron? twexer of course, flux pen ..23:08
whitequarkwpwrak: radio market for passive components and terraelectronica for ICs are a really good combination, but they don't have everything. when I'm doing something for myself, I try to pick from the components which are in stock in both, when someone asks me to make a board, I just say that they will finally pay for DigiKey order and shipment :)23:09
wpwrakkristianpaul: a temperature-controlled iron mainly23:10
whitequark(terraelectronica is a company name, of course. they have some _really_ good prices, like selling from qty 1 for the price of 100s)23:10
kristianpaulcan i bought that too from digi-key? cause i'm really thinging in make my first big order :)23:11
kristianpaulnv, i can buy that locally i guess :)23:11
kristianpaulwhat about handling RF ICs, any advice?23:11
wolfspraulgood morning23:12
whitequarkI'd definitely begin with an antistatic bracelet23:12
wpwrakwhitequark: (price from d-k, parts from terra) good idea ;-)23:12
wolfspraulwhat do you guys think about making an open/copyleft LED driver board?23:12
wpwrakkristianpaul: (soldering iron) you should be able to get this locally. they're not too exotic.23:12
wolfspraula Chinese friend of mine is working on one and I'm wondering whether I can help him in any way23:13
wolfspraulsame as Jon's speakers there't not that much software on it though :-)23:13
wpwrakwolfspraul: what kind of LEDs ?23:13
kristianpauli need LEDs to iluminate the backyard :)23:13
whitequarkwpwrak: hah, I only buy components in d-k when I can't find them in terra. d-k is quite painful to pay and also shipping is _not_ straightforward23:13
kristianpaulor at least i want give a try toe LED tech :)23:13
wpwrakkristianpaul: phew :)23:13
wpwrakkristianpaul: i'd stick with halogen for a few more decades :)23:14
kristianpaulbut they got damaged and... oh well23:14
wolfspraulI don't know that much about it, I think 'normal' LEDs?23:14
wpwrakwhitequark: well, now you've done d-k once. probably is smoother now.23:14
wolfspraulthe board attaches to a package with 70 leds I think23:14
wolfspraulbasically it's an intelligent power supply23:14
wpwrakso what do the leds do ?23:15
wolfspraulwith an Altera epm570T100 cpld controlling the power supply23:15
kristianpauldisplay cool 128 x 128 bit images?23:15
wolfspraula lamp23:15
wpwrakkristianpaul: halogen is cheap to replace - no matter what the damage is :)23:15
kristianpaulwpwrak: sure ;)23:15
wolfspraulfirst I will try to get some more data :-)23:16
wpwrakcpld and this kind of circuit sounds a little unusual. i could imagine an mcu with pcm, though23:16
wolfspraulmaybe Adam can help him with a small run and we can learn something? don't know yet23:16
wolfspraulno only that cpld, plus LM139 comparator to monitor the output, plus some mosfets23:16
kristianpaulsounds fancy for just a lamp, but i'm curious23:17
wolfspraulhe made a prototype board and will show it at a LED fair in guangzhou in a few days23:17
whitequarkwpwrak: why not use CPLD if you need a lot of pwm's, but nothing more?23:19
whitequarksounds like a sane choice for me23:19
whitequarkI guess it is not :)23:21
whitequarkwpwrak: now that the second board works too, I think I've found a flaw in my soldering habit which has caused all this crap23:22
wpwrakwhitequark: well yes. if you need lots of fast PWMs :) but an MCU can provide quite a lot of not so fast PWMs via software. LEDs don't need a PWM running at MHz ;-) but perhaps there are requirements we don't know. also, maybe he just likes cplds :)23:23
wpwrak(soldering) heh ! :) experience makes the master23:23
wolfspraulhere's a bit of high-level info I could get so far http://pastebin.com/FbRHWMgU23:23
whitequarkwpwrak: I'm holding the hot air gun at a small angle, and the component, no matter it's just 1mm high, blocks the air flow to the other side of it, causing it to not solder to the board 23:23
wpwrakhmm, i just go around the component with the hot air23:24
kristianpaulwpwrak: do i need a hot air gun or the iron should be okay?23:24
whitequarkwpwrak: when I beginned to do the same, the boards suddenly began to work23:25
wpwrakkristianpaul: i've only truly needed hot air once so far, and that was for the atben//atusb crystal23:25
wpwrakwhitequark: ;-)23:25
wpwrakkristianpaul: because for all the rest, an iron works, too :)23:25
whitequarkwpwrak: how much years/centuries of experience do you have?23:25
wpwrakkristianpaul: the crystal has large pads at the bottom23:25
kristianpaulso heat went away easilly?23:26
kristianpaulwolfspraul: 70w wow !23:26
whitequarkwpwrak: how can you solder something with pads underneath the chip like LGA or WSON with an iron if you have a board with a solder mask where you can't heat the traces directly?23:26
wpwrakwhitequark: heh, not all that long ;-) must have been around 2005 when i got interested again (i played a bit with electronics as a kid but didn't really get much done)23:27
wpwrakwhitequark: WSON has a bit of metal on the side23:27
Action: kristianpaul strugles in getting a bom23:27
wpwrakwhitequark: and for DIY, i have no solder mask anyway :)23:27
whitequarkwpwrak: indeed it does, but I was unable to solder it with just the metal. I've tried getting a solder blob on the tip of my iron and touching the WSON sides with that blob23:28
wpwrakDocScrutinizer: does "H-bridge" in wolfspraul's spec make any sense to you ?23:29
wpwrakDocScrutinizer: except as a self-destruct mechanism for the LEDs perhaps :)23:29
kristianpaulargh, that .net file is too uglgy to print in raw --23:29
DocScrutinizerwhat's "wolfg specs"?23:30
wpwrakwolfspraul: even 250 kHz is well within MCU territory :)23:30
wpwrakDocScrutinizer: http://pastebin.com/FbRHWMgU23:30
kristianpaulwolfspraul, fyi i got two boards from gnss-sdr.ru also some components, now i if get this build i can have a second source of data and compare results easilly ;)23:31
kristianpaulto give you a quick update, i'm currently working on the tracking algorythms for namuru23:32
kristianpaulnot big deal? but i need re-enforce my C knowledge a bit :)23:32
wpwrakwhitequark: (soldering WSON) hmm. looks similar to QFN to me. they're quite doable. the ones without metal on the side are evil, though23:33
DocScrutinizerwpwrak: hardly23:33
whitequarkwpwrak: yeah, it's quite like qfn23:33
wolfspraul[likes cpld] yes he's an IC designer23:33
wpwrakkristianpaul: any response from the DIY GPS guy in the UK ?23:33
wpwrakwolfspraul: that would explain it :)23:33
DocScrutinizerunless they plan to use a cascade multiplier to drive a lot of LEDs in series23:33
wolfspraulI'm sure pretty much everything in electronics can be built in 10 different ways, so?23:33
wolfspraulyes, I've heard 'series led' somewhere23:34
wpwrakDocScrutinizer: tesla would be so proud ;-)23:34
kristianpaulwpwrak: yeap, http://paste.debian.net/hidden/87191b1a/23:34
kristianpaulwpwrak: links most of milkymist, qi-hardware related23:35
wolfspraulthe cpld currently uses 242 logic cells, with some more functionality should still stay < 1k, or could be reduced to ca. 100 for a low-cost derivative23:35
wpwrakkristianpaul: oh wow, from tomorrow. he's really ahead of his time :)23:35
kristianpaulDate: Sat, 15 Oct 2011 09:50:20 +010023:36
wpwrakkristianpaul: anonymous posted at 2011-10-16 01:34:46 23:36
kristianpaulsorry i missed on the pastebin23:36
kristianpaulX-Mailer: Microsoft Outlook Express 6.00.2900.5923:36
kristianpaulof that make you happy :)23:37
Action: whitequark has just unsoldered that hirose connector back, just for the sake of doing it23:37
whitequarkI haven't melted it, yay!23:38
wolfspraulok I will gather some more info about the led driver board...23:39
wolfspraulif we can open it up, maybe it's worthwhile to help with a little small run23:39
kristianpaullamps are very wellcome over the world i guess, so worth the try i think23:42
kristianpauland the battery is a good plus23:42
wpwraklamps are also still difficult economically. leds are still too expensive.23:42
DocScrutinizerI can't see use for a cpld there, for the one-line dimming etc it's too low level, for controlling the H-bridge it's overkill. Any stupid 4bit or 8bit uC will do just fine23:45
kristianpaul10 + 1 ways ;)23:46
wpwrakyeah :) i'd pick an MCU, too. but then, if you're a hammer, everything looks like a nail, right ? :)23:46
DocScrutinizeryou don't need "lots of independent PWM" for controlling an illumination LED array23:46
wpwrakdepends on what else the LED array is supposed to do. i wouldn't be surprised if it, say, grew USB at some moment in time, too :)23:47
DocScrutinizerUSB and cpld? hardly23:48
wpwrakthe usual answer to this is ftdi :)23:48
wpwrakand the rest is a software problem :)23:48
DocScrutinizerthe whole thing is mainly just a PSU23:49
DocScrutinizerthere's just so much you could possibly innovate in PSU design23:50
kristianpaulenergy efficient : 95% ? :)23:51
wpwraka random number23:51
DocScrutinizerprolly, 23:52
DocScrutinizer>>a novel design realized with no radiator<<23:52
wpwrakat least he's engineer enough that he knows it can't be 100% ;-) but the rest sounds more like marketing speek23:52
DocScrutinizersure, with only as little as 3.5W to dissipate ;-D23:52
wpwrakyeah, particularly that :)23:52
wpwraknaw, "delivered power" 70W. are those LEDs _really_ so efficient ?23:53
kristianpaulis it china, no?23:53
kristianpauli mean, i wonder what world of kind of LEDs there are23:53
wpwrakagreed. he didn't specify for how long this is suppose to work23:53
kristianpaulhe :)23:54
DocScrutinizerwpwrak: I guess that's power TO leds, not OUTPUT power from leds23:54
wpwrakmaybe it's an array with 10 cm of metal structure between LEDs. that would work nicely :)23:54
kristianpaulat least during the fair ;-)23:54
wpwrakDocScrutinizer: doesn't really matter - you'll still burn a lot in LED losses :)23:54
wolfspraulok we know little :-)23:55
DocScrutinizeras with any other light emitting technology23:55
wpwrakkristianpaul: is was more thinking in terms of fractions of a second :)23:55
wolfspraulI think conceptually yes, it's just a smart power supply23:55
DocScrutinizerwolfspraul: no matter how smart it gets, finally you'll find it kinda looks like http://www.analog.com/static/imported-files/images/functional_block_diagrams/ADP1653_fbl.gif23:59
DocScrutinizers/one or two LEDs/a random number of LEDs/23:59
wpwrakthat's already quite fancy :)23:59
--- Sun Oct 16 201100:00

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