#qi-hardware IRC log for Friday, 2012-01-20

kristianpaul:) http://www.flickr.com/photos/7958754@N03/6721433793/in/photostream/01:00
wolfspraulservice is back up, sorry for the inconvenience07:51
wolfsprauland the bot helps me, thanks!07:52
wpwrakgrmbl. CCFL seems to be dead for good. sigh. time to go monitor shopping ...11:01
qi-bot[commit] Werner Almesberger: m1/ledravaganza/: draft proposal and example circuit for M1r4 LED matrix (master) http://qi-hw.com/p/wernermisc/4e044fc11:14
viricwpwrak: do you know any dense dot matrix lcd (close to a monitor, it doesn't matter if it's black/white only) that does not have backlight?11:46
viricas with typical calculators, with a mirror below the lcd11:46
wpwrakhmm, isn't that qi-pixel display like that ? maybe you could scavenge an OLPC11:48
viricwhati s that?11:49
viricah looks fine11:49
wpwrakah, Pixel Qi :) nothing to do with us11:49
wolfspraulI'm thinking about putting a QR-code label on PCBs that links directly to an online failure reporting/analysis/corrective action database14:50
wolfspraulhave to find some nice free server-side software for that first14:51
mthwolfspraul: QR decoding is available as open source: http://zbar.sourceforge.net/16:46
viricI wanted to use that lib for barcode decoding, some time ago16:59
viricbut I noticed my webcam can't focus a barcode of usual size.17:00
DocScrutinizer51err, why's that?18:13
DocScrutinizer51has it auofocus that's braking on barcodes? (I've seen that more than once)18:14
DocScrutinizer51if so, it usually helps  to rotatae either narcode or cam by >45°18:15
DocScrutinizerwolfspraul: make sure you get the QR in silkscreen, not in copper. Also consider silkscreen usually tends to bleed out so your QR might get illegible when too small and np correction factor for shrinking silkscreen print19:52
wolfspraulmy idea was different19:52
wolfspraulfirst I don't like any serial numbers that are not easily removable19:52
wolfspraulsay lasered for example19:53
wolfsprauland then, the idea here is to have the link directly to the data for this particular board19:53
wolfspraula unique address for that board, like the MAC address19:53
wolfspraulbut as a gluable label, not in silk or lasered or so19:53
DocScrutinizerwhat's the use of that? nser#?19:54
wolfspraulkeep track of defects and technical history of that board19:55
DocScrutinizeryou might consider looking into CD inkjet-printers19:55
viricDocScrutinizer: no, without autofocus. fixed focus19:56
DocScrutinizerI mean such a cheesy dualuse (also) CD printer will easily take non-circle-shaped PCB after smallish mods19:57
DocScrutinizerdepending on the way you do things, your metrics of "easily removable" might either not get matched or vastly over-fulfilled19:59
DocScrutinizerI.E. "don't touch the printing, or it will smear and vanish"19:59
DocScrutinizeror "use a knofe or sandpaper to scratch it off"20:00
DocScrutinizerI for one prefer to HARDlink such properties to the tagged object20:01
DocScrutinizerso no fool sends me a PCBA with a fake/swapped/borrowed SER#20:01
DocScrutinizerviric: so why couldn't it focus then?20:02
DocScrutinizerit never can focus ;-)20:02
DocScrutinizeror do you mean "QR is too small to read it out when held far enough from fixed focus webcam to be in-focus"?20:03
viricDocScrutinizer: well, it focuses at a distance where the barcode does not have enough resolution to be detected20:04
DocScrutinizerwhich would be kinda .oO(???) as well, the usual webcam at least should have 640*400, no?20:04
viricI use "focuses" as "has a fixed focus to", instead of any active verb20:04
DocScrutinizerso you should have several pixel per datsel of QR20:05
viricwith more than 45° aperture I'd say.20:05
viricfocused at 50cm20:05
DocScrutinizereven in a good distance, unless for microscopic QR20:05
viricI'm talking about barcodes.20:06
viricthose of supermarkets20:06
DocScrutinizershort calculation: 1m dist = ~50cm width, /1024 = 0.5mm/pixel20:07
viricand you think that's good for a barcode?20:10
DocScrutinizerunless there's too much of aliasing, I.E. more than 2pixel / datsel/bar, I'd expect it to work20:10
viricit's not a Hubble lens20:10
DocScrutinizerafter all barcodes are robust to read out20:11
DocScrutinizeryou just need one semi decent scanline in one of 50 frames each second20:12
DocScrutinizerfor barcodes20:12
viricthat does not happen with this webcam :)20:13
viricif you have a webcam in your notebook.. you can try. 20:14
DocScrutinizerI don't20:14
viricok :)20:15
DocScrutinizerprobably your lib is crap, hoping for optimal input data to get away with a over simplistic approach for decoding20:16
DocScrutinizerusual barcodes always have a calibration header&trailer20:17
DocScrutinizeror fixed width binary coding20:18
DocScrutinizerwhere you see only bars of 1* 2* 4* width20:18
DocScrutinizerusually even both20:18
viricwell, seeing the images live, I can't distinguish the bars either20:19
wpwrakhmm .. choice 1) spend the weekend with only two screens. 2) go out and buy a new one, replacing the one that's broken. minor obstacle: power is down in part of the building, already for 12+ hours. this means: no elevators. and it's 14 floors at 35 C ...20:19
viricit gets like some grey20:19
viricwpwrak: 'only' two screens?20:19
DocScrutinizerhmm, now that's actually a tad too little on info then20:19
virica tad?20:20
DocScrutinizerindeed, a tad20:20
DocScrutinizeryou'd be amazed what a good filter can get out of what you think is uniform grey20:20
virichm tad == small child?20:20
DocScrutinizerad tad == a little20:21
wpwrakviric: trust me. once you've had tree, you don't want to go back to two :)20:21
DocScrutinizerI keep my 2nd screen in office off all the time20:22
DocScrutinizerwhat I can't do one one screen isn't worth getting done ;-D20:22
DocScrutinizerviric: look, do you see the single pixels?20:23
DocScrutinizerno? then why do you think you should see the 1 pixel wide bars20:24
DocScrutinizer1, or 220:24
DocScrutinizerfor a decoder that makes a hell of a difference, and a proper image with 1 and 2 pixel wide bars is perfectly decodable20:24
DocScrutinizereven when the bars bleed out to neighbour pixels, you still can decode them20:26
DocScrutinizeras gray pixels are supposed to be white when next to a black pixel20:26
DocScrutinizeryou see what I mean?20:26
DocScrutinizeryou *know* it's a barcode you look on, hell you even can use kalman filters ;-D20:27
DocScrutinizerwill pull your barcode from true noise20:27
DocScrutinizerask the GPS guys if they could see their SAT signal on a scope, by watching the receiver IF20:34
DocScrutinizerI thik the signal is some 30..60dBm *below* receiver noise there20:35
DocScrutinizerI'm absolutely sure you can get barcodes from a video signal that has pure white noise for all you could tell by visual inspection, by just applying same decoding principles20:37
DocScrutinizertakes quite a bit of computation grunt, but hey that's el'cheapo nowadays20:38
virichm maybe :)20:41
viricin any case, zbar failed20:41
DocScrutinizerbraddead lib I guess20:41
DocScrutinizerbraindead even20:41
DocScrutinizerI've seen pictures getting algorithmically deblurred20:43
viricin Tannhauser20:44
viricDocScrutinizer: ok, you want a picture ;)20:45
DocScrutinizerfor sure not20:46
DocScrutinizerjust saying there's definitely always a smarter method to decode a barcode than the one you're jst using - no matter what might be that currently used method. Further saying you might get barcodes from video material you'd not believe there's *any* info on it20:47
viricI understand20:48
DocScrutinizerI dunno if I could find the linke once more, but there's been a report from a university where they got accurate data for an individual's heart pulse rate, blood pressure and O2 saturation (or maybe it was onky two out of those 3), by analyzing simple standard video material for otherwise not noticeable color changes of the skin21:26
DocScrutinizerrather nifty stuff indeed21:28
--- Sat Jan 21 201200:00

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