#qi-hardware IRC log for Friday, 2012-02-10

kristianpaulmorning 01:24
zrafakristianpaul: you are not in morning :)02:46
zrafakristianpaul: or yes? Maybe you just started your day? 02:47
kristianpaulzrafa: he just following the regard :)02:48
kristianpaulno no mmorning02:48
kristianpaulbtw the cheapestway to send a package to AR is 35usd02:50
kristianpaulmax 1KG02:50
kristianpaulbut no tracking no nothing02:50
kristianpauland 30day deliver02:50
kristianpaulEMS costs around 47usd02:50
kristianpauls/regard/greeting02:51
zrafakristianpaul: ah.. thanks for the information :)02:53
kristianpaulEMS take 6 days to arriv02:54
kristianpaule02:54
wpwrakfedex from digi-key: USD 4002:55
wpwraktakes about 30 hours to arrive at customs. then another 30 hours to clear them :-(02:56
kristianpauland pay $$$$$$$ ?02:56
wpwrakyeah, that too02:56
Action: kristianpaul sigh02:56
wpwrakindeed02:56
kristianpaulabout cheap, well if it is less that 100gr (considering 40gr of packaging) it will be 26usd :/03:00
kristianpauland is expensive considering i got from US 15usd air mail shipping.. for jeenodes board/kit the other time03:01
kristianpaulokay this time should work, ratio get down from 150 to 33 ;)03:13
qi-bot[commit] Xiangfu Liu: icarus-miner: record time on u.log and export to webserver ROUTER_IP/u.log (master) http://qi-hw.com/p/openwrt-packages/ed8a0b703:27
qi-bot[commit] Xiangfu Liu: cgminer: make it compile. only enable the bitforce FPGA support (master) http://qi-hw.com/p/openwrt-packages/f65d64c03:27
qi-bot[commit] Xiangfu Liu: cgminer: add libpthread libncurses jansson to DEPENDS (master) http://qi-hw.com/p/openwrt-packages/83c6a6603:27
kristianpaulah nice, 8 minutes, wonder how slow it will ;)03:30
wpwrakkristianpaul: yes, air mail is even less. but also takes longer.03:33
qi-bot[commit] Xiangfu Liu: cgminer: enable-cpumining for play (master) http://qi-hw.com/p/openwrt-packages/e114a4e03:40
Action: kristianpaul trying to setup minet.py04:07
wpwrakwolfspra1l: hmm, on https://sharism.cc/ the "Sharism Presents Milkymist" looks a bit odd, because it's placed like a title for everything, including the Ben04:07
wolfspra1lhe04:08
kristianpaul and soon at ben/usb !04:09
wolfspra1lagree, on the todo list somewhere for 'the big cleanup'04:09
wpwrakkristianpaul: yeah ! :)04:09
kristianpaula name in a webpafe tittle is not that big ;)04:09
kristianpaulwebpage*04:09
wpwrakwolfspra1l: btw, someone asked if you also got tuxbrain's UBBs ?04:10
wpwrakof course, it's rather easy to make them yourself ;)04:10
wolfspra1ldid not, forgot actually04:10
wolfspra1lone by one04:10
wolfspra1lI don't even have the atben/atusb yet04:11
kristianpauli bet tuxbrain is using UBBs with its arduinos04:11
wpwrak(one by one) sure, just wondering04:12
kristianpaulxiangfu: RPC getwork error04:14
kristianpaulsouns familiar to you?04:14
xiangfukristianpaul, yes. :)04:21
kristianpaulxiangfu: what does mean?04:22
kristianpauli already create minin worker account..04:22
xiangfukristianpaul, it maybe missing python files. or something wrong with your account.04:22
kristianpaulah FYI i'm fpgaminer core on the M1 board04:22
kristianpaulhmm04:23
xiangfuno. it not react the FPGA yet.04:24
xiangfuI mean it must be something with the python script file or network.04:24
kristianpauli need confirm fpga works ok, had you dont this before?04:25
kristianpauli dont need minining yet..04:25
xiangfufor confirm what you needs send some data to fpga. wait let me found the detail.04:25
kristianpauli'm using this https://github.com/progranism/Open-Source-FPGA-Bitcoin-Miner/blob/master/projects/Verilog_Xilinx_Port/sources/software/miner.py04:25
xiangfuyes.04:26
wpwrakrunning fpgaminer in some unused corner of the fpga may actually be a nice gag. "M1 pays for itself" ;-)04:27
kristianpaulyeah ;)04:27
kristianpaulbut fpgas dont have unused corners at least not that shape i bet..04:28
wpwrakwhat kind of bitcoin production could we expect with M1 ? anything significant ? like the equivalent of USD 1-2 per day ?04:28
kristianpaulor usable shape04:28
kristianpaulwpwrak: dont know04:28
wpwrak(unused corner) why ? couldn't you have some bitcoin-mining device in M1 ?04:29
kristianpaulyes04:30
kristianpauli already have it04:30
kristianpaulbut not sharing resources with the SoC itself yet..04:30
wpwrakah, i see. so you run it instead of the SoC04:30
kristianpaulso i would not place in the corner04:31
kristianpaulyes wpwrak 04:31
kristianpaulfor now :)04:31
kristianpauli need to know if it is alive..04:31
wpwrakyeah. so you still need to find a nice corner for it :)04:31
kristianpaulexactly ;)04:31
kristianpaulxiangfu: have a payload example?04:32
wpwrakin terms of resource usage, in the configuration you've synthesized, does it use nearly 100% of the FPGA ? or a smaller fraction ?04:32
xiangfukristianpaul, send '33c5bf5751ec7f7e056443b5aee3800331432c83f404d9de38b94ecbf907b92d' to fpga04:33
xiangfuit should return you '063c5e01' immediately04:33
kristianpaulxiangfu: had you tested it already?04:33
kristianpaulok let see04:34
xiangfukristianpaul, sorry what you mean payload?04:34
kristianpaulno problem, you already seems provide one :)04:34
wolfspraulwpwrak: at most 10-20 cents / day ;-)04:34
kristianpaulwolfspraul: you should not said it that to him  :D04:35
wolfspraulthe reason I work with ng (the designer of the open Icarus board) a little is that if we are lucky, we can share some designs and bits and pieces of tech with him04:35
kristianpaulwpwrak: free dailly candy !04:35
wolfspraulfor example his board uses a 12V power supply, and boots from SPI04:35
wpwrakaii .. so it won't pay for itself in less than ... ~9 years04:36
wolfsprauland I'm sure he would design a new one around artix-7 the moment he can get hold of one04:36
wolfspraulbitcoin mining is extremely competitive, like unleashing a storm when people sense money, I guess ;-)04:36
wolfspraulbut Icarus is a nice design and all files are open and I will see whether there are missing pieces etc. and hopefully ng keeps his stuff open in the future04:37
wolfsprauljust another one of those experimental activities like the led power supply (and others before that ;-))04:37
wpwrak(competitive) yet another gold rush :)04:37
kristianpaulxiangfu: how do you send it and wait from imediate reply?04:41
kristianpaulargh no reply..04:43
kristianpaulwpwrak: bitrush04:44
wpwrakbitcrush :)04:49
xiangfukristianpaul, try this one: d29b709fbce49b83ed9d404f38c2341330083eea5b344650e7f7ce1575fb5c33d4a8d866c832353f1ebaa88e0000000004:51
xiangfuyou have to decode them to hex04:51
xiangfuthen send to fpga.04:51
xiangfuit should reply 063c5e0104:51
kristianpaulsend, just copy and paste on the terminal?04:51
xiangfucheckout the line 68 of miner.py04:53
xiangfukristianpaul, no. that is not real hex.04:53
xiangfukristianpaul, echo d29b709fbce49b83ed9d404f38c2341330083eea5b344650e7f7ce1575fb5c33d4a8d866c832353f1ebaa88e00000000 | xxd -r -p is the real hex04:55
xiangfumaybe cat `echo d29b709fbce49b83ed9d404f38c2341330083eea5b344650e7f7ce1575fb5c33d4a8d866c832353f1ebaa88e00000000 | xxd -r -p` > /dev/ttyUSB0 04:55
xiangfuthen cat /dev/ttyUSB0 | xxd -p 04:56
wpwraki would call that "binary"04:56
xiangfukristianpaul, I maybe wrong again. needs check more about line 68 of miner.py :) but test it first. :)  back online later. 05:01
Action: kristianpaul zzzzz05:06
cladamwa(AMPEN/GPD4) Does anyone know that if set this pin high will let NN with speaker out under there's no earphone plugging ? correct or no ?10:05
cladamwaas i know even under condition in plugging earphone, then if you pulldown POP/GPB29, you won't hear sound playing on earphone. This should be like this. 10:11
whitequarkwolfspraul: hi11:40
wolfspraulhi11:40
whitequarkhow do you think, if I could get a PCB I'd like to be traced and sent it to you11:41
whitequarkcould you give it to a PCB fab so they'd trace it? like the picture on the wiki11:41
wolfspraulI suggest you get some sandpaper and do it yourself11:43
whitequarkis it so easy?11:43
wolfspraulhe11:43
wolfspraulI speak about it without ever having done it with my own hands11:43
whitequarkmhm11:44
wolfspraulbut that's just because there are so many thousands of tasks on my todo list, I haven't gotten to it yet11:44
wolfspraulbut yes, it is easy11:44
wolfsprauluse corning 120 to go through the layers, and corning 400 for polishing before scanning11:44
wolfspraultwo types of sandpaper11:44
whitequarkwell, there's one thing11:44
wolfspraulexperiment with some garbage pcbs11:44
wolfsprauland then onto the one you care about11:44
wolfspraulI think it mostly has to do with learning how much pressure to use, so you don't accidentally sand through the next layer11:45
whitequarkhm, I think you are right11:45
whitequarkthanks11:45
wolfspraulbut that's worth learning, once in life11:45
wolfspraulalso there's a practical problem of how to fix the pcb11:45
whitequarkthe PCB I may want to be traced is this: http://mobilephonerepairguides.blogspot.com/2011/03/samsung-galaxy-s-4g-pcb-board.html11:46
whitequarkI think I may find some dead ones in a local repair shop, if I'd ask them nicely11:46
wolfspraulsteve talked about this here http://en.qi-hardware.com/irclogs/qi-hardware_2011-02-21.log.html11:46
wolfspraulso he thinks about 25 minutes per layer, without scanning11:47
wolfspraulmost of the time using corning 12011:47
wolfspraulthen 30-50 seconds polishing with corning 400 before scanning11:47
whitequark(fix) this may be easy. I could use a board and some nails, then put the pcb on it so the nail heads would fit into the holes11:47
wolfsprauland he writes "not damaging the board on the borders is hard"11:48
wolfspraul"you should always just remove the copper of one layer until you see the coating of the next layer, otherwise it's hard to tell the copper of 2 layers apart and you may remove too much"11:48
wolfspraulbut I think once you went through one or two test pcbs, it should all be clear and easy11:48
wolfspraulfor pcbs, sandpaper is the way to go11:49
wolfspraulfor an IC, you need something else :-)11:49
whitequarkICs are easier, personally for me11:49
whitequarkI could easily get 100% H2SO411:50
whitequarkand I think I know where I could find a polarizing microscope11:50
whitequarkI'd like to take some photos of the baseband chip too11:50
wolfspraulsome more tips from steve (see the link)11:50
whitequarkI wonder if the GPS in it is a standalone core or actually connected to the modem part11:50
whitequarkthanks, I'll read it!11:50
wolfspraul"the best way to prevent damaging the borders is to have some other material with the exact same thickness around the pcb, so you're not damaging the edges (which will become round otherwise)"11:51
wolfsprauland "use a few rows of gaffer-tape as an underlayment, since the non-adhesive side is quite resistant to the sandpaper"11:51
wolfspraulall small mechanical tricks, like I said I am sure after 1-2 pcbs you have your own setup...11:52
wolfsprauland one day I really have to do this myself!11:52
wolfspraulnot just talk about it...11:52
whitequark(gaffer-tape) err. I see the pics in wikipedia11:52
whitequarkbut by dictionary does not know such a word11:52
whitequarkand I never seen anything like this in Russia11:52
wolfspraulI think he means the tape you use for wrapping boxes11:53
wolfspraulyou find some other way11:53
wolfspraulthe edges make sense, otherwise you apply more pressure at the edges naturally (when sanding)11:53
wolfsprauland the pcb has to somehow be fixed11:53
wolfsprauland then - sand sand sand11:53
wolfspraul25 minutes / layer :-)11:53
whitequark(wrapping boxes) that thin polyester very hard to tear by hand tape? yes, I know this11:53
wolfspraulhe just wants to avoid scratching/damaging his desk surface, I think11:54
whitequarkah. makes sense too11:54
wolfspraula cutting board from the kitchen may do as well11:54
wolfspraulthe key 'technology' is the sandpaper11:54
wolfspraulthat's what all Chinese places I've seen offering this service will do as well11:54
wolfspraulthey just put some kid on sanding through the layers :-)11:55
wolfspraulI guess you could also use more high-tech, but why...11:55
whitequarkthe chinese way11:55
wolfspraul8 layers = 4 hours, plus scanning and preparation/cleanup, let's say 6 hours11:55
wolfspraulhourly wage let's be generous 3 USD, so 6*3=18 USD11:56
wolfspraulthe boss will ask about 100 USD from me, so 92 for himself, company, overhead, customer service, etc.11:56
wolfspraul:-)11:56
wolfspraulsorry 8211:56
whitequarkmhm. yes, it's more cost effective to do it myself. also, skills.11:57
wolfspraulyou don't easily beat the economics with any laser cutter or whatever delamination device you may have11:57
whitequarkreversing is fun! now, it's only a question of finding a scrap pcb11:58
whitequarkebay has none, I've checked11:58
whitequarkso, my only hope are local repair shops11:58
whitequarkthanks again for the info! I'll write how it will go11:59
wolfspraulyes please11:59
wolfspraulI am only relaying words, unfortunately12:00
wolfsprauland I have no time to try this again right now, always an excuse12:00
whitequarkI neither. but at the Saturday I'll go and find the board12:00
whitequarkweird12:04
whitequarkBP and AP are connected via USB12:04
whitequarkand the BP is a multi-descriptor device. it has PHONET for control and media (audio) for sound routing12:05
DocScrutinizerwolfspra1l: (fixing PCB) now that's the easy part: just use one can of cyan-acrylic glue and a wooden or glass board19:43
DocScrutinizerwhitequark: nailing the PCB to the support won't fly, the last layer is way to flimsy19:48
DocScrutinizeractually the best approach would be to use two PCB and sand down each one from one side until half the layers are gone, use the other board for the rest of layers, from the other side19:50
DocScrutinizeractually I think that's what they did for the photos on wiki19:51
DocScrutinizerthe just mirrored the layers from PCB #219:51
DocScrutinizerthen*19:51
whitequarkDocScrutinizer: thanks for the suggestion20:01
whitequarktest20:31
whitequarkthere's something really FUBAR with IPv6 at this host and my logger bot :/20:32
DocScrutinizerwhitequark: when doing a PCB teardown to RE the circuit, you may want to take ULTRA-highres scans of each layer, and maybe even manually mark all vias you can find, on each layer20:59
DocScrutinizervias are most important key part for RE, and most PITA to detect and find on scans/photos21:00
DocScrutinizerI prefer the idea of doing an electrical RE. Apply voltage to one pad, see which other pads also have voltage. I got quite a bunch of weird ideas how to simplify resp automatise that21:01
DocScrutinizere.g you could do all that under water, use DC, and see where gas bubbles from electrolysis appear on the connected pads21:02
DocScrutinizeron good macro photos you probably can see the bubbles quite easily21:02
DocScrutinizeranother possibility: use phenolphtalein or another idicator and see where pH-level of water changes due to electrolysis21:03
DocScrutinizeryet another idea:21:03
DocScrutinizerdo the same with high voltage in evacuated gas containment, to build a gas discharge setup where the pads start glowing in the dark21:04
DocScrutinizeror use kirlian (high voltage high frequency) and see the mini tesla-transformer alike sparks and StElms discharge on the pads in plain air21:05
DocScrutinizeror use galvanic effect in liquid, to form a (removable) coating of a clearly colored metal on all the connected pads21:06
DocScrutinizercopper probably not that good on gold pads - silver might work, esp when you convert it to silver-sulfid which is deep black21:07
DocScrutinizeror cover the whole PCB with some low quality varnish (low adhesion, quite some pores) and then use AC to make that varnish pop off on pads where gas micro bubbles build up and vanish every 1/100 second (at 50Hz AC)21:10
DocScrutinizersure, it's unclear whether any of those weird ideas will work for those BGA footprints where yo got >2 pads / mm21:12
DocScrutinizerbut it's also unclear how wolfspra1l 's equation for expense to do PCB teardown and RE would look like, when we count in the work to actually create a connection plan from those layer photos21:14
DocScrutinizerand then a true semi decent schematics from that connection plan21:14
DocScrutinizerI guess the work for that is like 100 times the work for the sanding down and doing scans/photos21:15
DocScrutinizererr missed a 0, meant thousand times21:15
DocScrutinizermaybe a mix of automated coarse and manual fine analysis/search of pads that are connected may yield good results: use a fine long line carbon brush (kinda similar to those formerly used to clean vinyl records) to spot X and Y coords where to do detail search for a connected pad21:19
DocScrutinizerdamn hell FFS, those 0402 footprints are monsters compared to the friggin fine pitch BGA21:24
DocScrutinizermaybe this one helps a bit to get an idea21:29
DocScrutinizerhttp://maemo.cloud-7.de/Gallery-N900-exploded/platine_01.jpeg21:29
DocScrutinizerit's already not good enough to do a proper RE21:30
DocScrutinizervia dents almost invisible21:30
DocScrutinizerwhen you zoom in so a 0402 footprint is the size of a fingertip, you can guess them via dents21:31
DocScrutinizerat zoom 250% you actually can see most of them, but you also see the artifacts21:33
abushcrafterforgAnyone tried audio editing on the nano note?23:49
DocScrutinizermaybe that's already a little bit too demanding a task for that tiny machine? regarding the missing audiocard, the missing USB hostmode to plug in a audio card dongle, the limited gfx on screen23:53
DocScrutinizersox might work just fine23:53
DocScrutinizer;-)23:53
DocScrutinizerthough maybe not in realtime23:54
--- Sat Feb 11 201200:00

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