#qi-hardware IRC log for Saturday, 2010-08-28

DocScrutinizereither you 'tune' the circuit, means in-Z == trace-Z == out-Z. Or you feed 'asym' with very low source-Z and proper termination, so the missmatch has no adverse side effects00:01
DocScrutinizerwpwrak: sorry, I'm not cmpetent to answer specific detail questions wrt that00:02
wpwrakhmm, on rx, the differential impedance of the transceiver is 100 Ohm.00:04
wpwraknothing specific to find about TX.00:04
DocScrutinizerthere's programs to calculate such things, and in appnotes the chip manufs clearly signal which parts need to be tuned. This is based on their knowledge about chip fan-out/fan-in or Z properties, and on simply analyzing lab proto boards00:04
wpwrakthe docs are a bit fuzzy. and contradict each other :)00:05
DocScrutinizeras a general rule of thumb, you don't care about tuned lines, when the loss caused by mismatch is irrelevant. Then you just make sure you don't run into other truble by line ringing (reflections or standing waves)00:07
wpwrake.g., ti don't specify anything. they give a balun using transmission lines for inductors in their reference design. if you dig further, they do mention an integrated balun in one web forum. and one rf component maker even has a balun supposedly designed for that chip.00:07
wpwrak(information is similarly confusing with atmel, although with less secrecy)00:08
wpwrakokay. what would be typical mistakes to avoid when ringing may be an issue ?00:09
DocScrutinizerRF design is black magic - you're supposed to have learnt that lesson :-P00:09
wpwrakyeah yeah :)00:09
wpwraki don't aim at cooking up a perfect fcc- and etsi-pleasing design that's ready for MP :)00:09
DocScrutinizerwpwrak: typical mistakes: too long traces, too high terimation Z, to high source Z00:10
wpwraki'll be happy already if i emit more power around the carrier than at the harmonics and get a bit of range :)00:10
DocScrutinizermissing clamp diodes (if digital signal)00:10
wpwrak(z) one end is given by the antenna, 50 R. the other is defined by the balun. feed line is ~10 mm in total.00:11
wpwrakthis is the pcb: http://www.almesberger.net/misc/ben/wpan/first-pcb.jpg00:11
DocScrutinizerhmm, so good PCB layout progs have a function to make a trace match 50R Z00:12
wpwrakand here with components: http://www.almesberger.net/misc/ben/wpan/board-100813.jpg00:12
wpwrakkicad doesn't have impedance matched trace. i know they should be about 1.6 mm wide, though.00:12
wpwrakhere's some of my research for now: http://projects.qi-hardware.com/index.php/p/ben-wpan/source/tree/master/atrf/ecn/ecn0004.txt00:13
DocScrutinizerso design it that way. If you want to do a fancy, design 5..10 PCBA with a range of trace geometry, and check which performs best00:13
wpwrakyeha, "try a few things and see what works best" does indeed look like the way to go00:14
DocScrutinizer(in the end of the day that's what industry at large does as well)00:15
wpwrakluckily, i can replace some of the small chips with an integrated balun, so there will be fewer knobs to turn. should also generally perform better.00:15
DocScrutinizerif peak performance is at either end of your range, you should go for another set00:15
DocScrutinizeralso, as a catch-oops, usually you see unpopulated Pi-filters at start and end of tuned traces00:17
wpwrakyup. right now, i have suckish reach. barely makes it room to room. interestingly, one direction works much better than the other. (the two boards have some small component variations, that's probably why)00:17
wpwrak(pi) that may be for fcc compliance00:18
DocScrutinizerif after start of first batch of 10k PCB it turns out the mismatch is too severe, you always can throw a bead and 2 C at it00:18
wpwraki saw this approach in many app notes. first they design and simulate the perfect circuit. then they build it and throw in filters until the unwanted harmonics and stuff are small enough to pass certification.00:19
DocScrutinizeryep. usual SOP00:19
wpwrakwell, my board is so small that the unpopulated pi filter would already cover most of the feed line :)00:20
DocScrutinizerthe Pi are for SWR / Z matching though00:20
wpwrakthere is actually one at the end00:20
DocScrutinizeryou see these (usually) unpopulated Pi at every RF design antenna foot point00:21
DocScrutinizerin case antenna impedance changes later on, to what it was during proto tests00:22
DocScrutinizerreally *every*00:22
DocScrutinizereach BT or WLAN module, each GSM design00:23
wpwrakso they place a 0R ? or use a transmission line as inductor ?00:23
DocScrutinizerand the 2 C are NC00:25
wpwraki've seen one antenna design recommend a LC termination ("half-T")00:25
DocScrutinizerhmm, you do this if you know in advance your antenna is detuned00:25
wpwrakthey do show remarkably better performance with the LC in place ;)00:26
DocScrutinizersometimes you simply cant get a lambda/4 or whatever00:26
wpwrakah, and how close can/should the antenna be to the transceiver ? is there a minimum clearance one would normally try to observe ?00:27
DocScrutinizernot to my knowledge00:27
DocScrutinizerusually the closer the better00:28
DocScrutinizerfootpoint that is00:28
DocScrutinizerantenna tip should be max free air around in all directions00:28
wpwrakshould short feed lines be bracketed by GND ? or is the ground plane underneath sufficient ?00:29
DocScrutinizershould suffice00:29
wpwrakhmm, the definition of "antenna tip" may be a bit tricky in my case :)00:29
DocScrutinizeryeah, sontimes it is :-D00:30
wpwrakti's design [1] feeds the antenna from the long side and has free space on the left. [1] http://focus.ti.com/lit/an/swra117d/swra117d.pdf00:31
wpwraki have the left side blocked with the circuit and the feed line goes around. not sure how horrible this is in the general scale of things.00:32
wpwrakof course, nobody has a useful antenna design that's fed from the short side :-(00:32
wpwrakpinged the wrong bot ? :)00:35
DocScrutinizerwhat's wrong with mgedmin's logbot? the log is laggy like hell00:36
DocScrutinizer(wrong bot) nah, 24h forced disconn00:37
DocScrutinizerqi-bot: ?00:38
DocScrutinizerqi-bot: info00:39
wpwraki think qi-bot is only logging, nothing else00:39
DocScrutinizerdon't think so, or maybe... >Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!<00:40
DocScrutinizereggdrop doesn't look like povbot00:42
DocScrutinizerwell, maybe I'm wrong00:42
DocScrutinizerprobably irclog(c)eggdrop, http://en.qi-hardware.com/irclogs(c)mgedmin00:43
wolfspraulDocScrutinizer: qi-bot is only logging, the entire setup, every last configuration option, is documented here http://en.qi-hardware.com/wiki/Server_setup#eggdrop00:46
DocScrutinizercompare http://mg.pov.lt/maemo-irclog, it's povpot there that does the logging00:46
DocScrutinizerwolfspraul: :-D, yep thought as much00:46
wolfspraulthen irclog2html picks it up, it's equally documented http://en.qi-hardware.com/wiki/Server_setup#irclog2html00:47
DocScrutinizernow for the 3 lines of wpwrak I missed during downtime00:47
wolfspraulcron job every 30 minutes, finally the result comes out at http://en.qi-hardware.com/irclogs00:47
wolfspraulit's a bit delayed, first eggdrop only flushes the log file every few minutes I think, and then mostly the cron job only runs every 30 minutes00:47
DocScrutinizeraaah, so the cronjob is introducing the enormous lag00:47
wpwrakoh, i could have repeated them :)00:47
DocScrutinizerI didn't know there were any, wpwrak :-)00:48
wolfspraulI can change the cron to run faster00:48
wolfspraulthe next thing I plan to add to qi-bot is an opt-out list of nicknames, so the bot will automatically discard all msgs from people who don't want to be logged00:48
wolfspraulbut it's not the #1 on my priority list right now, so I'd say it will be a few months until I get to it, unless someone else steps up00:49
wolfspraulwell but for discussions like the one between you and wpwrak, I think it's really great to keep them archived :-)00:49
DocScrutinizerwolfspraul: put logbots /ignore list to purpose ;-D00:49
wolfspraulso I hope you won't make use of the opt-out feature once we have it00:50
wolfspraulDocScrutinizer: does freenode already have something that prevents your msgs from going to a particular (logging) nick?00:50
DocScrutinizernow for ti.com00:50
DocScrutinizerwolfspraul: I'm agnostic of the particular ways /ignore is implemented, but wouldn't be surprised it's server side00:51
DocScrutinizer/mode #qi-hardware +i wolfspraul00:51
wolfspraulk I'll look into it one day00:51
DocScrutinizercan't remember right now00:52
wpwrakthere's no feature under the sun someone hasn't implemented in irc yet ;-)00:52
wpwrakhmm, interesting. ti have an 8051 with usb and transceiver. that's the one the used for the antenna circuit. rather nice.00:54
DocScrutinizerhmm, http://toxin.jottit.com/irc_commands suggests /ignore is a UA command00:55
wpwrakah, but it's one if quirky non-ieee 802.15.4 designs. darn.00:56
DocScrutinizerdarn, closed irclog, wpwrak can you repost the ti url please?00:56
wpwrak| ti's design [1] feeds the antenna from the long side and has free space on the left. [1] http://focus.ti.com/lit/an/swra117d/swra117d.pdf00:56
wpwrak| i have the left side blocked with the circuit and the feed line goes around. not sure how horrible this is in the general scale of things.00:56
wpwrak| of course, nobody has a useful antenna design that's fed from the short side :-(00:57
wpwrakthese were the last three lines00:57
DocScrutinizerwhat's short side and left side in your world?00:57
wpwrakif you look at my pcb pictures, left is left, etc.00:59
DocScrutinizeron a general note - PCB antenna designs are very sensitive to *any* change of geometry00:59
wpwrakleft and right are short, top and bottom are long.00:59
wpwraki didn't touch the geometry. just the surroundings :)00:59
DocScrutinizeryou need to keep the whole design incl feedpoint and adjacent gndplanes01:00
DocScrutinizere.g you can't add gndplanes to top, right or bottom, and you need to keep the full gndplane (edge, at least) on left side01:02
DocScrutinizersorry didn't look at your design01:02
DocScrutinizernow I did. Nope that won't pan out01:03
DocScrutinizerthe GND plane parallel to the short fat antenna branch is way too close01:04
wpwrakhmm. feed and all that are where they should be. how close can i go with gnd on the left ?01:04
wpwrakand would it help if i put the gnd zone at, say, a 45 deg angle ?01:04
DocScrutinizermax ~1.5 of the general structure detal size, I'd guess from my guts. So about  size of one of the "squares" of meander * 1.501:06
wpwrakalso, does the gnd that's too close affect rx and tx in about the same way or does one suffer a lot more ? (e.g., i would suspect that it hurts tx a lot more)01:06
wpwrakokay, that should be feasible. parallel to the short side or angled ?01:07
DocScrutinizersorry, please rephrase01:08
DocScrutinizer(affect) both virtually same01:08
wpwrak1) in the current design, with the gnd too close to the antenna, would that gnd distort/weaken emissions and receptions roughly in the same way ?01:08
DocScrutinizer(affect) both virtually same01:09
wpwrak2) we assume the lower left corner of the antenna is at coordinates (0, 0), should the edge if ground plane run parallel to the short side, (-d, 0) to (-d, h) ?01:10
wpwrakd = distance gnd plane to antenna. h = "height" of antenna01:10
wpwrakor would it be better if i make it (-d+h) to (-d, h) ? i.e., at a 45 deg angle, but the closest point of the gnd plane would be closer to the antenna this way01:11
wpwrakkewl. that's easier :)01:12
wpwrakbtw, i don't think the crystal in the board on the picture worked :) had to reflow the critter for better alignment. first time i actually had some successful use for my hot air pen :)01:14
DocScrutinizerSmall changes of the antenna dimensions may have large impact on the performance. Therefore it is strongly recommended to make an exact copy of the reference design to achieve optimum performance01:14
wpwrakyup, that's what i did01:15
DocScrutinizerwatch D1, D4 and GND Layer 201:15
wpwrak(hmm, need to add a filter for backup files to fpd2pdf)01:18
wpwrakd4 ?01:18
DocScrutinizeralso amazing: even while there's (of course) no gndplane under the antenna structure, still they recommend >> It is also recommended to use the same thickness and type of PCB material as used in the reference design. Information about the PCB can be found in a separate readme file included in the reference design. To compensate for a thicker/thinner PCB the antenna could be made slightly shorter/longer.<<01:19
DocScrutinizeryour design is missing upper gnd plane01:19
wpwrakah ... D1 as a dimension01:21
DocScrutinizerpage 4, figure 301:22
wpwraki keep that clearance. but it's a lot less than what you're suggesting01:22
wpwrakthe upper gnd is a bit mysterious01:22
wolfspraulwhat if the board Werner is designing is integrated into another bigger pcb (the one we currently have behind the lcm)?01:23
wolfspraulif thickness matters - I think the PCB that currently is behind the LCM is rather thin01:23
wpwraki read figure 3 as not having gnd there. but then, on the picture of their real board, there is a gnd plane01:23
DocScrutinizer>> The size of the ground plane affects the performance of the PCB antenna. Connecting the USB dongle to a computer increases the size of the ground plane and thus the performance is affected. Figure 5 shows how the performance is affected when the USB dongle is...<<01:24
wpwrakwolfspraul: it;s very close to the 0.8 mm i use01:24
DocScrutinizer(lot less than what you're suggesting) yup. Even I can fail01:24
wpwrakDocScrutinizer: (size of gnd plane) yeah. big computer = better. that's why i want to be close to the lcm. it's not much but better than nothing.01:24
wpwrakDocScrutinizer: so you think keeping it that close is acceptable ? ti don't say specifically that it's okay to put something nearby. they just strongly suggest that nothing should be in that area.01:26
wpwrakor, rather, nothing but the pcb dielectric01:27
DocScrutinizerI'd keep it away, though in theory it might be irrelevant, given D1 keeps minimum of 0.50mm01:27
DocScrutinizerin the end that fat trace has a GND via, which I wasn't aware of first instance01:28
wpwrakokay. a few more mm can't hurt anyway.01:28
DocScrutinizerso parasitary capacitance to nearby gndplane might be less of an issue01:29
wpwrak(gnd via) rf is spooky to measure. all things are "shorted" to ground ;-)01:29
wpwrak(to measure with a digital multimeter)01:29
DocScrutinizeryeah X-P01:29
DocScrutinizerput your DMM in your attik as long as dealing with RF01:30
DocScrutinizeruseless equipment01:30
wpwrakokay, so we have the transmission line and the upper ground plane. i also need to check their pcb thickness. i have it something in the back of my head that 0.8 mm are okay, but i'm now no longer sure if i took that from ti or just from compatibility with the lcm's pcb.01:30
wpwrakit's still good for testing traces ;-)01:31
DocScrutinizerwpwrak: you can have a short (about 2..3 times trace width) parallel to gndplane 'serife' at long end of antenna (like a L ) and then dremel away successively until antenna is tuned01:33
DocScrutinizer(next to "W2 <->")01:34
wpwrakyou mean extending from the end of the antenna towards the right ?01:34
DocScrutinizernext to arrow tips of W2 and L1, to the right, length suggested 2 * W201:35
DocScrutinizerparallel to "<->" ow D301:36
wpwrakand now the big question: how do i tell when my antenna is tuned ? :) that is, with my modest measurement apparatus ?01:37
DocScrutinizerthat's the most challenging part, and the reason we got antenna houses doing that for us ;-D01:37
wpwrakgrmbl :)01:38
emebwpwrak: what test equipment do you have?01:38
DocScrutinizerprobably measuring exact dBm at a given never changing distance with a good receiver will do01:38
wpwraki have an usrp that can operate at 2.4 GHz and 5 Ghz. so i should be able to do *something* at carrier and 1st harmonic01:39
wpwrakno calibrated antenna, though01:39
DocScrutinizer1st harmonic - nevermind01:39
emebthat's actually not bad...01:39
DocScrutinizercalib rcv antenna - nvm01:39
wpwrak(1st harm.) gcc and etsi love that one ;)01:39
DocScrutinizerjust make sure you got a clearly defined never changing testbed01:40
wpwrak(1st harm) also, the more power i see there, the less i'll have where i want it to be :)01:40
DocScrutinizerwpwrak: (1st harmonic) there's just so much you can do about it anyway01:40
wpwrakthat's where your pi filter comes into play :)01:40
DocScrutinizernot exactly01:41
DocScrutinizerPi filters are Z matchers01:41
DocScrutinizernot filters01:41
DocScrutinizerfirst approach at least - of course they are also lowpass01:42
DocScrutinizerwpwrak: also note there's antenna detuning even by plastic enclosures, so final antenna tuning has to be done in final product anyway01:43
wpwrakthe one i have is used as such. to make up for all the mistakes atmel made in their balun :)01:43
DocScrutinizerhumm, I see01:44
wpwrakDocScrutinizer: exactly why i don't see wolfgang avoid rf issues by using modules :)01:44
DocScrutinizerthe issue doesn't get much better by using a module. It will see same detuning :-P01:45
wpwrakyup :)01:46
wpwrakand then you're back to tweaking little caps and inductors while staring at your measurement equipment :)01:46
emebeveryone who works RF has to do that...01:47
wpwrakyup. there's no hiding place.01:48
emebjust need to avoid the iPhone 'death grip' :)01:49
DocScrutinizerdamn, a nice antenna design: 350MHz bandwidth01:49
emebhow much BW is needed for the wpan? - only a few 100kHz, no?01:50
wpwrakthe band is ~100 MHz01:51
wpwraknot quite sure about the chanbnel width01:51
emeband the chip can tune over the whole band?01:51
wpwrakit's like wifi01:52
wpwrak2405 - 2480 MHz01:52
wpwrakchannel spacing is 5 MHz01:53
emebwow - more than I thought...01:53
wpwrakdata rate is 250 kbps01:53
Action: DocScrutinizer muses idly what WPAN might be at all01:54
emebso there's a lot of dead air between channels.01:54
wpwrakchip rate 2 Mchips/s01:54
DocScrutinizeremeb: for sure not01:54
emeb?? didn't realize it was dsss01:54
wpwrakwireless personal area network. basically a bubble arund you with ~10 m radius01:54
emebguess I was still thinking the FM stuff that HopeRF was using...01:55
DocScrutinizerwhat's it worth for?01:55
wpwraknot sure about modulation details yet. that's still on mt reading list01:55
wpwrakemeb: all the sub-ghz stuff is basically unusable for us. way too many regulatory restrictions.01:56
wpwrakemeb: such as duty cycle. 1 s per hour or so, in some of the more extreme cases.01:57
emebheh - not really useful01:57
emebwow - those at86rf230 chips are cheap - $2.27 in 100qty from Digi-key02:06
wpwrakthe cc2520 is also not bad. according to ti, as low as $1,95/100002:07
wpwrakso both are interesting02:07
DocScrutinizerwpwrak: I strongly suggest to also review the documentation of CC2511 reference design, schematics and all. Also to find out about mystery of that "upper GNDplane"02:08
wpwrakti's has a few more features, such as more buffers. atmel has onlt one, shared between rx/tx/host. kinda like the 3c501 ethernet :)02:08
emebso are you still undecided which chip to use?02:08
wpwrakyup, good idea. there are also a few mystery components next to the antenna02:09
wpwrakemeb: they both look attractive02:09
DocScrutinizerI'd expect to find some words about "making antenna shorter/longer" in the included readme02:10
emebhmm - Digi-Key's 100pc price for the CC2511 is $6.1002:10
emeb$4.73 for the CC252002:11
wpwrak2511 is mcu + usb + rf02:12
wpwrak2520 is just rf, like the at86rf23002:12
emebatmel datasheet is pretty sparse on RF details02:14
wpwrakemeb: then try to find the impedance in ti's ...02:17
emebsection 9.1 of the cc2520 ds says "high impedance". What more do you want?02:18
wpwrakgood luck finding the rest :)02:20
emebnice chatting - gn02:25
wpwrakDocScrutinizer: any other "red flags" in my design ? the whole mcu side and the crystal will go away when putting all this into the ben02:26
DocScrutinizerwpwrak: I missed the point of WPAN advantages over more established protocols like e.g BT PAN02:26
DocScrutinizer(red flag) I gather the crappy SMT only non-thruhole USB receptacle will go away as well02:27
wpwrakopenness. basically everything WLAN or even BT is under NDA. and sometimes you can't even buy the chips/modules. or they're excessively expensive.02:27
wpwrakhey, everybody uses that receptacle ;-) and i soldered it well :)02:28
DocScrutinizerwhat an epic fail02:28
wpwrakand yes, it'll go away. the idea is to basiclly cut at the two v-shaped dents. and of course remove the crystal. didn't know for sure what to with it when i designed the board02:29
DocScrutinizer(USB) yeah, and N900 returns/repairs/MIAs are legion, due to micro-USB rcptcl coming off02:29
wpwrakso far, _my_ usb receptacles are holding :)02:29
wpwrakbut it's nice to hear that not only OM screwed up their usb :)02:30
DocScrutinizerindustry should specify drop-tests to be done -mandatory - with cables attached02:30
DocScrutinizerthe whole USB receptacle and plugs shit is a botch, at best02:31
wpwrakbtw, the receptacle i used in idbg is pretty much indestructible. it's mounted inside the board plane, so there's no lever on the pull/push forces. as long as the pcb holds, that thing stays there.02:31
DocScrutinizergenearlly receptacles must be mounted to solid case, and connected to PCB by flexible  / detachable connection02:32
wpwrakhere;s the critter: http://www.almesberger.net/misc/ben/idbg-v1-pcb.jpg02:32
DocScrutinizeryup, that's another viable way to avoid it breaking on your first sneezing02:33
wpwrakalso note the six (!) solder pads. i think nokia wouldn't have much trouble if they had used this one :)02:33
DocScrutinizerforget it! no matter what breaks, whether solder points or clue holding copper traces to base epoxy or the whole PCBA cracks. It's a flimsy design - by design02:34
wpwrakyeah, pcb cracking is nasty02:35
wpwrakthe proble is that you can't produce case-mounted connectors nearly as cheaply as pcb-mounted ones02:35
DocScrutinizernever will stand bending/torque forces >200Ncm02:35
DocScrutinizererr 2Ncm?02:36
wpwrak200 Ncm is a lot :)02:36
DocScrutinizer1Newton == 100g02:36
DocScrutinizerso make that 2Ncm02:36
wpwrak2 Ncm should be fine02:37
DocScrutinizeri'd have used pcm but that's quite uncommon and ambiguous02:37
wpwrakthat's one chocolate suspended from the end of a usb connector02:37
DocScrutinizerand you'll exceed that easily, in a lot of everyday situations02:38
DocScrutinizerlike dropping device on connector, or just putting it on sofy and then sitting on it02:38
wpwrakdon't sit on your computers ! :)02:38
DocScrutinizerI bet a lot of less coordinated people will exceed that on mere unplugging02:39
wpwrakmaybe a sofa-compatible computer would need a big capacitative sensor, to detect the approaching of the owner's arse ;-)02:39
DocScrutinizeror trip over the wire of charger - bang, dead02:40
wpwrakyes, i think plug/unplug forces can be quite large02:40
wpwrakmost things don't break quite so easily02:40
DocScrutinizerUSB receptacles do, esp the SMT type02:40
wpwrakhmm, i have stuff fall down, too. nothing broke so far. granted, i don't tie them to my car and drag them across town ...02:42
DocScrutinizerto make matters worse, it seems a lot of micro-USB come with excessive plug forces, by design02:42
wpwrakyup. micro-usb wants a good push02:42
DocScrutinizerhonestly, micro-USB rcptcl should have its own mounting PCB, fixed to case, and any sort of wire or spring contacting to main PCB02:44
wpwrakgood luck with that proposal :-)02:44
DocScrutinizerlook at N810 power connector (2mm barrel) or all the Nxxx series AV/headset connector. Those are nice plastic things with a number of goldplated springs to contact to pads on PCB02:46
DocScrutinizercase has a recession to hold the thing mechanically, a 'bay'02:46
wpwrakah, they're spring-connected ? nice02:47
DocScrutinizercan't see fundamental obstacles to implement USB that way02:47
wpwrakcontact spacing perhaps ?02:48
DocScrutinizerlook there, the golden thing at right (6 contacts): http://people.openmoko.org/joerg/n900/images/IMG_1939.JPG02:48
DocScrutinizerAV connector02:48
DocScrutinizer(sorry for blurr)02:49
wpwrakah, not bad02:49
wpwrakstill bigger than usb, i think, but smaller than i would have expected02:49
DocScrutinizeroops N810 has SMT AV02:50
wpwrakthere goes the neighbourhood :)02:51
DocScrutinizerhttp://people.openmoko.org/joerg/N810-disassembly/n810_disassembly/snc00297_jpg.jpeg.html for power barrel spring type02:51
wpwrakkinda biggish, if considering usb size02:52
DocScrutinizerit's not like those devices are <2% air inside02:52
wpwrakthe usb receptacle does look a bit fragile, though02:53
wpwrakonly half the pads large, the other(s) already quite small02:54
DocScrutinizerflimsy crap02:54
DocScrutinizernever can work decently02:54
DocScrutinizerI mean, thru hole would give at least a minimum of ruggedness02:55
DocScrutinizerspring load contacts and fixed in a bay would be optimum02:55
DocScrutinizeryou literally had to break the case to ruin the receptacle, and even then you can just get a spare case which is supposed to be much cheaper than spare main PCBA02:57
wpwraki thnk large enough pads should make it work okay for smt too02:57
wpwrakjust tiny pads don't work. well, ironically, often it's even the solder than breaks, not the pad02:57
wpwrakyes :)02:58
DocScrutinizerit's a botch, and it stays that way02:58
DocScrutinizercould be done better02:59
DocScrutinizerjust like most pushbutton designs02:59
DocScrutinizerfull load of force to the switching element... no wonder it breaks on first hard push02:59
DocScrutinizerwhy not have a latch or something where the plastic knob movement is limited hard, and pushing the switch with a "spring" designed into the knob plastic03:00
DocScrutinizerthis way you never can break the switch, unless you poke *through* the plastic knob03:01
wpwrakhmm. patent it :)03:02
DocScrutinizerI'd rather love to see it *used* than patented03:03
wpwrakyou could replace the spring with some silicone03:03
DocScrutinizerthough this is harder to assemble, one component more03:04
wpwrakgive the button four little feet. then you can use an off-the shelf dome.03:04
wpwrakyou'd just fill the keycap with the silicone03:04
wpwrakeasier to do then placing a spring03:04
wpwrakwell, scratch the silicone03:05
DocScrutinizerI thought of making the 'spring' just a thin lever - part of the plastic of knob03:05
wpwrakjust a button with feet to terminate the movement will do03:05
DocScrutinizercan't manufacture that exact enough03:06
wpwrakthe dome also allows for some tolerance03:06
DocScrutinizerheadroom from 'click point' to maximum push in of a smt switch is <0.05 mm03:07
DocScrutinizera dome type spring switch has no such headroom at all03:07
wpwrakhmm. lemme check this ...03:08
DocScrutinizeryou need to push it down all the way until the metal dome touches the PCB pad. In that moment contact and limiting of movement are same thing03:09
DocScrutinizerI was thinking of side force operated type smt switches, like in FR03:09
DocScrutinizerthough the silicone thing + feet works for dome type as well03:10
DocScrutinizerhttp://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=P10850SCT-ND FR AUX button03:11
DocScrutinizersure like hell will come off, just needs enough operation force03:12
wpwrakhmm, dome data sheet doesn't say how far it can travel once active. there must be some tolerance.03:13
DocScrutinizerthere isn't by design03:14
DocScrutinizer(well axcept for how much you can compress the metal of the dome :-P)03:14
DocScrutinizeror the PCB and pad03:15
wpwrakyup. that's what i was thinking of03:17
DocScrutinizerand that are, incidentally, exactly the two failure patterns you see on dome sheet kbds like the one on N810: puncture the pad by excessive force, so there's no more contact when dome pressed down to the plain epoxy - and deformation of the dome spring, so it loses its properties03:17
wpwrakboth should need quite a bit of force ...03:18
DocScrutinizeroften less than you might think03:18
wpwrakyou also have the dome's base that may be flexible03:18
wpwrakthis one has a PUR ring03:19
wpwrakah, but the metal is inside, so, doesn't help03:19
DocScrutinizerif patenting would help to propagate it into industry, I'd for sure do it. But I'm sure it'd be rather contraproductive03:20
DocScrutinizeranyway, 9'o saturday morning. The time they run their tests on TV how far the idiocy may be pushed on the customers03:22
DocScrutinizermustn't miss that :-P03:22
DocScrutinizeralso *yawn*03:23
wpwrak;-)) yeah, soon 4:30 here. may be a good moment for a nap.03:24
wpwrakthanks a lot for your rf suggestions !03:24
wpwrakthat gives me hope the next version will perform a bit less crappy than the current one :)03:25
DocScrutinizeralways a please, yw03:27
DocScrutinizerpleasure even03:28
wpwraki'll try to have more difficult questions the next time :)03:40
rohabout connectors and so.. if you want to know how to _really_ do that.. check out a lenovo T60 power plug03:57
rohthey put the socket into a 'cave' inside the case which encloses it from all sides, a hole to the back which it cant fall through (rim) and multiple wires (4 in total, 2 for each pin in parallel since its power) go to the board and are connected with a 4 pin minimolex or so03:58
rohmeans you can a) exchange the power-socket (big barrel type) seperately, and it can 'wobble' a bit. make breaking out of the socket basically impossible, and also breaking it off a mainboard or so.03:59
rohgreat design.03:59
rohi guess one could mount a usb that way. put it on a pcb which is only 2-3mm wider than the socket, use 4 wires to the main pcb, put the whole thing in a 'recessed' indention in one of the case parts and fixate it by putting the other casing half over it. no more broken pcbs04:01
kyakunclouded: nightsky 20100827 is working fine!04:37
kyakis a great piece of software :)04:38
kyakunclouded: how do you zoom in/out? it's bound to F11/F1204:39
uncloudedkyak: it says F11 and F12 but its the volume up and down keys in SDL05:27
kyakunclouded: hm, you mean Vol Up and Down06:06
kyakyeah, works great :) thanks!06:06
uncloudedyes.  they turn up as F11 and F12 in SDL.  I should adjust the help text for the NanoNote though06:06
kyakunclouded: do you know how i can see the Sky in specific date/time?06:06
kyakwithout changin system date/time06:06
kyakunclouded: maybe the F1 help text in source code? Definitely should be patched06:07
uncloudednot without editing the source06:07
kyakF11/F12 are bound very conveniently .. i don't know if this is just a conicidence06:08
kyakunclouded: would be better if after pressing F1 it would indicate that zoom in/out keys are VolUp and VolDown06:08
uncloudedno coincidence.  it was written for the NanoNote06:09
kyakheh, great06:09
uncloudedif you wanted to edit the source to show the sky at different times you want the lines that looks like: model.when_updated = time( NULL); in the update_main() method in main.c06:10
uncloudedtime(NULL) returns the time now.  using a different value will show the sky at the time represented by that different value06:11
kyaki see.. thanks for the hint06:11
kyakmaybe for starters it can read this from config file06:11
kyakreading from user input would be harder06:12
kyakunclouded: i see you've updated version on SF06:14
uncloudedthe latest version only adds the ability to resize the window and build on Mac OS X, nothing new for the NN06:17
kyakdo you plan to put more effort into nightsky? for example, implement possibility to search objects?06:20
uncloudednot sure yet06:22
kyakwhat exists now is already great though :)06:24
uncloudedto type stuff in it would have to have a primitive text field: type letters, backspace, delete, cursor, etc. but otherwise searching for an object would be a useful feature06:26
kyakyeah, it would be.. searching for objects and setting arbitrary dates/times06:27
uncloudedalthough out of the 10,000 odd brightest objects, only 230 have common names06:45
kristianpaulnice screenshot14:51
wpwrakhmm. needs a hierarchy by topic. cutest on top. there should also be an "install me right now" link :)14:56
wpwrakbut yes, screenshots are important. we're very visual animals. that won't go away anytime soon :)14:57
wpwrakascii art ski jumping. oh dead ;-)14:57
wpwrakdeaR even14:57
virickristianpaul: through that I discovered 'asc'! I go building it14:59
kristianpaulwpwrak: indeed topic lacking15:00
wpwrakgrr. git log --follow  follows renames only into the past, not into the future :-(17:26
wpwrakthis is going to be messy ...17:36
rafa__are news around?18:30
wpwrakrafa__: everybody seems to have a relaxing saturday:)20:37
viricI just discovered ufoai :)20:38
wolfspraulrafa__: you online? where are you?20:38
wpwrakhe should still be in barcelona20:42
virichm I live near Barcelona20:46
viric100km away20:46
wolfspraulwpwrak: nice, so he made it through immigration :-)20:48
wpwrakwolfspraul: yeah. apparently they only had a quick look at the passport. didn't even want to see all the stuff they had prepared.20:49
wolfspraulthat's how it should be20:50
wpwrakyes, indeed20:51
rafa__wolfspraul: hey!20:54
rafa__yes, we are in barcelona!20:55
rafa__we also met tuxbrain and we had some nice relaxing afternoonn talking20:55
wpwrakwolfspraul: will be interesting to find out what chip hoperf picked. it has a few unique magic numbers that only appear in their data sheet, according to google20:55
wpwrakthings like 0xD99E8621 and 0xFFFFFEF7CF20810408204120:56
rafa__wolfspraul: the airport polices just check my face against passport. and i looks like and good england boy right now.. without beard and with a reallyy short haiir20:56
rafa__and i look like a good..20:57
wpwrakrafa: ah, you disguised yourself as skinhead ;-)20:58
rafa__let me show you a photo working with tuxbrain20:58
rafa__wpwrak: haha... nono. like a good boy.. skinhead is a little bad boy i guess20:58
rafa__well, i have beard again.. that is fast20:59
rafa__wpwrak: anyway, i do not rely in the spanish airport police yet.. aaand i feel the same when i need to get in to spain again after jlime meeting21:02
rafa__when i was there.. the police man was so tired he seemed... but i was seeing that other people started to have problems with other polices21:04
rafa__wpwrak: perhaps the strategy is to check which police man is wanting to leave the office :))21:06
rafa__hola gustavo ;)21:08
wpwrakrafa__: so another shave before returning to spain ? :) anyway, you'll return from a EU country. so there shouldn't be much in terms of border controls.21:09
rafa__ah yes? i do not know well the papers there.. if it asks about the "from" country  maybe yes, it should be like you say21:12
wolfspraulrafa__: wow great you are with David?21:14
wpwrakrafa__: maybe keep boarding passes and hotel receipts. that way, you can always prove that you're indeed on an fairly extensive trip.21:14
wpwrakrafa__: i don't quite remember how the airports deal with intra-EU immigration. in theory, there should be no control. but if passengers from EU can be mixed with passengers connecting from non-EU flights, then there would be controls.21:16
rafa__wolfspraul: it was great.. he and victor are really nice guys and very funny guys :D21:18
rafa__wolfspraul: wpwrak : http://fz.hobby-site.org/hp660lx/nn/working_at_tuxbrain.jpg21:19
rafa__i wrote the link manually, let me know if it works :P (i am chating with my hp palmtop)21:19
wpwrakempty beer glasses ?!?21:21
rafa__wpwrak: that was before to leave21:21
wolfsprauljoshcryer: welcome to our copyleft hardware project :-)21:33
wolfspraulI wish one day we have a page like that Milkymist One SMT/DIP flow for a nice little ASIC manufacturing job...21:33
wolfsprauljoshcryer: lekernel is the main force behind Milkymist, the channel is #milkymist on freenode21:34
joshcryerwolfspraul, thank you for the warm welcome. :)21:50
kristianpauljoshcryer: why milkymist is not free sofware?22:29
kristianpaulis code as we can see it, sure thee is a synteshis process22:30
kristianpaulsorry ask again i really want read your opinion22:30
joshcryers/free software/free hardware/22:31
joshcryeruses Spartan-6 FPGA22:31
joshcryerWas not bashing, just thought it was not, exactly, what wolfspraul was talking about.22:31
joshcryer(with regards to freedombox)22:31
kristianpaulyeah sure the fpga is theere but at least we can implement free designs on it22:32
joshcryerVery true.22:32
kristianpaulhardware is not free by default ever22:32
joshcryerIt's closer than FreedomBox.22:32
joshcryerFree as in freedom kristianpaul.22:32
joshcryerIn English you do not say "Freedom Hardware."22:32
kristianpaulwell you here thats good :)22:33
wolfsprauljoshcryer: what is closer than FreedomBox?22:33
wpwrakdon't "freedom fries" count ? ;-)22:33
joshcryerwolfspraul, Milkymist...22:33
wolfspraulMilkymist the IC design?22:33
joshcryerwolfspraul, yes, it incorporates non-free(dom) hardware.22:34
joshcryerLike FreedomBox does.22:34
kristianpaulis other level22:35
wolfspraulMilkymist is just an IC design, which on top of being entirely copyleft or BSD licensed, does go to great lengths to not build dependencies to proprietary vendor-specific FPGA extensions22:35
wpwrakjoshcryer: so are you saying fpgas in general are out ? or just the spartan-6 ?22:35
kristianpaulwe cant see hardware as just a big squared box and thats it22:36
joshcryerwpwrak, nah, wolfspraul even wants to move on to ASIC.22:36
kristianpaulthere are leves we can achive freedon goals in some of then22:36
kristianpauljoshcryer: Having an ASIC doent mean we're done thereare propietary stuff required to make an ASIC22:37
kristianpaulso level by level...22:37
joshcryerwolfspraul, that's very fair enough. FreedomBox main competition is PlugComputer.22:37
wpwrakjoshcryer: yup, but for now we're at fpgas. kinda more affordable :)22:38
joshcryerI did say that it was closer.22:38
joshcryer<joshcryer> It's closer than FreedomBox.22:38
kristianpauland for what we really need "practical" freedom22:38
joshcryerBut this discussion really split when I suggested that wolfspraul make himself an option for FreedomBox.22:38
joshcryerSo FreedomBox 1.0 is PlugComputer or SheevaPlug, eventually it can become Milkymist-derived.22:39
wpwrakhmm, the mill has been enjoying more than a day off. time to give it something to do ... let's see ... the topmost plastic looks easy :) kinda useless, but all the rest i've left will be hell to mount.22:39
wolfsprauljoshcryer: oh I misunderstood you then. you meant 'closer to freedom' not 'more closed' :-)22:40
kristianpauljoshcryer: so lets call it "free that non-free box"22:41
kristianpauljoshcryer: as we do when run free sofware on out computers22:41
joshcryerwolfspraul, apologies, English is a very difficult language, even for native speakers.22:41
joshcryerwolfspraul, I do sincerely hope you haven't ruled out being a FreedomBox platform some day though, as I think you have the right idea.22:42
joshcryer(I'm maybe a bit more optimistic / naive than most.)22:42
wolfspraulhard to compete with Marvell22:42
joshcryerShame it's not atheros (packet injection). >:)22:43
wolfspraulthey will integrate the CPU and Wi-Fi stuff, price will come down. maybe you will get that thing for 20 USD soon... who knows22:43
joshcryerSo what, if you design ASIC and IC and we design PCB and cap/res. components you build the thing at materials cost (very very cheap).22:44
joshcryerwolfspraul, ShivaPlug is $125, V3.0 will likely be in same price range.22:46
joshcryerThey should use Fonera but it's very limited (I am IRCing via my Fonera).22:46
joshcryer(And locked.)22:46
kristianpaulwolfspraul: why not Xburst + ethernet lilke a freedom box?22:47
kristianpaulXburst chips are cheap hopefully can run server apps22:48
wpwraki think a shivaplug-like device could make a useful platform for bringing MM to the masses. similar to a MM-based ben. basically take a design people are familiar with and totally change one key item.22:51
wpwrakhowever, a shivaplug-alike would also face regulatory issues. you may find it surprisingly difficult to ship such a thing to a lot of places.22:52
joshcryerIndeed, especially if it ran FreedomBox social networking cloud apps.22:52
joshcryerYou destory facebook and fund free hardware designs, in one stroke.22:52
wpwrakwell, the shipping will be fine. but then there are customs at the end, watching over local regulations.22:53
wpwrakjoshcryer: "world domination, fast and on all fronts at once !" ;-)22:53
joshcryerThat's my optimism speaking, btw. :)22:54
joshcryerWhat started it all was my mentioning my admiration of Moglen and someone posting a FreedomBox link and I'm afraid that wolfspraul thought we were saying it was better, but that was not the intention at all.22:54
joshcryerI quite like Qi-hardware now that I discovered it. :)22:54
wpwrakwolfspraul: btw, you'll notice that my "schematics diffs" often don't pick up a project from its beginning (happens in ben-wpan/atrf and xue). that's because not all the relevant top-level files (i.e., .pro and .sch) were committed from the start. so schhist2web starts looking at things beginning with the first commit but doesn't find anything it can use until later.23:00
kristianpaulwell, i'mm off bed see ya :)23:01
joshcryerLater kristianpaul.23:02
qi-bot[commit] Andres Calderon: 5V out DC-DC added http://qi-hw.com/p/xue/d9b9c8423:02
wolfspraulwpwrak: well but that sounds like a feature not a bug, right?23:03
wolfspraulmy main concern is that the scripts are robust23:03
wolfspraulnever hang, never end in loops spamming the server disk space, those things23:03
wpwrakwolfspraul: it's a feature in the sense that the project was indeed not really usable at that point. it's a bug in the sense that there may be a history of changes we're missing. but i'd be a bit afraid of adding heuristics that move up in the history to cherry-pick files that may be missing. maybe add some mechanism for manual fixups in the future.23:05
wpwrakwolfspraul: (robust) time will tell ;-)23:05
wpwrakit would be good to run them in some sort of restricted environment. so that, even if something goes wrong, nothing horrible happens.23:06
wpwrake.g., there are a few rm -rf in there. would be ugly if their arguments turn into a bare / by some twist of fate and the uid running all this actually can delete something valuable.23:07
wpwrakif someone was looking for a malicious exploit, there will almost certainly be something to find. after all, there are not only the shell scripts (shell being a fragile language when it comes to metacharacter exploits), but also git and eeschema. git has some interesting parsing heuristics that add a whole new set of meta-characters.23:09
wpwrakif we ignore malicious exploits, things should be a bit less rough. but i'd still take precautions.23:10
wpwrake.g., if wouldn't feel comfortable creating a new unprivileged user and running an rm -rf / as that user, then you probably don't want such a beast run wild23:11
wolfspraulwe can ignore truly malicious exploits, but the 'restricted environment' still needs to be easily setup and maintainable23:11
wolfspraulI'll look into it23:11
wolfspraulwpwrak: would you feel comfortable running the script as an unprivileged user?23:12
wpwrakthere are some forms of sandboxes that may be sufficient. package build processes often face the same challenges.23:12
wpwrakif your system is set up in a way that you wouldn't be afraid of your unprivileged user running rm -rf /, then yes23:12
wpwrakalso, be warned that it can produce quite a lot of temporary output. e.g., xue has a cache of 3.7 GB23:15
wpwrakben-wpan/atrf has 357 MB. ben-wpan/cntr only 32 MB (but the latter may get a bit bigger. there's still some bug somewhere hiding some things)23:17
wpwrakthe cache is so big because it generates ppm files for all the schematics in all revisions. i can add an optimization for things that don't change later, though.23:18
wpwrakactually .. i could just use links ... let's see ...23:20
wpwrakfor the things running crazy, you could just give it a generous cpu limit. maybe combined with a disk quota if you have that kind of stuff enabled.23:23
wpwrakfor kicad, it also need some sort of X display. that can of course be some X server variant that doesn't need a head. vnc or such.23:24
wpwrakwithvnc, you could also easily check what's wrong in case it hangs at some error dialog.23:25
wolfspraulok got it23:31
wpwrakoh, and it can run in a pretty isolated environment if you want. no need to be able to write the repository or any other "valuable" data. at the end, you can just copy what it has generated from the output directory.23:33
wpwrakso even a VM that you scrap after use would be fine.23:33
wpwrakthe only issue with recycling after use is that, if you lose the cache, it takes a while to process all the old commits. e.g., it needs about one minute for ben-wpan/atrf from a "cold" cache. after that, it's ~5 seconds if nothing has changed.23:36
wpwrakit's worse with xue. xue with a cold cache takes ~12 minutes. 16 seconds from up to date cache.23:36
wpwrakmaybe i can make all this a bit faster when eliminating some of the rituals for unchanged files. unfortunately, it's hard to predict when a file may change or not. e.g., if the .sch is the same but a .lib or even the .pro has changed, the output could be different.23:38
wolfsprauldon't go crazy, I think I need to move on the server now23:38
wolfspraulfor Xue, the #1 thing that stood out last time still were the sheets that had no images at all, just an empty column23:39
wpwrakyup. that's the name tracking problem. it's a bit harder than i thought, because git won't tell me about renames in the past->present direction, only present->past.23:40
qi-bot[commit] Xiangfu Liu: test-memory-card.sh: cleanup the output message, modify by Adam http://qi-hw.com/p/nn-usb-fpga/b6f32d523:42
wolfspraulxiangfu: the 8 and 16 gb cards still don't boot23:43
wpwrakthere's also the problem that xue uses a3 while i expect a4. i need to poke around inside the .sch for this. the .ps generated by eeschema always claims it's a4 :-(23:43
wolfspraulcan you send me a u-boot with some more on-screen logging? maybe also just some number at the end (shortly before the error message so it's still visible), so that I know for sure that I am booting with the right u-boot version23:44
wolfspraulor maybe I need to send you my sd cards...23:44
xiangfuwolfspraul: yes. I saw your email. then I fixed the last bug that I know.23:44
xiangfuwolfspraul: it's hard to output log to screen.23:44
xiangfuwolfspraul: send me the sd cards is better.23:45
wpwrakcut out the man in the middle :)23:45
wolfsprauldoes anybody want to adopt the qihardware twitter and identi.ca accounts? there is also a nanonote group on identi.ca23:47
wolfspraulall pretty badly neglected, so I want to clean up a bit23:47
wolfspraulif nobody wants to use them, I will just delete them23:47
wolfspraulthere are maybe 150 or so followers on identi.ca, maybe 300 on twitter. still a while until it catches up with Lady Gaga :-)23:49
wpwrakmaybe also kill the forum while you're at it ? ;-)23:49
wolfspraulI knew you would say that Werner :-)23:49
wolfspraulI think the forum is fixable, at least I will try myself.23:50
wpwrakthat problem with these "fad" platforms is that the accounts there tend to get orphaned quickly. e.g., just look at all the placed on myspace that wikipedia still often links to. meanwhile, the people have long moved to facebook or such.23:50
wpwrakthe forum would be great if it connected to the mailing list23:50
wpwrakbut if it's just a parallel universe, it just creates diversion23:51
wolfspraulI have no problem with twitter and identi.ca it's just that I only have so much time, so I think either someone steps up and brings some life into those accounts, or we do everybody a better service by closing them.23:52
wpwrakit's of course easy enough to ignore if you already know that the mailing list is where things are really happening. but it can trap newcomers. also, people who just want to see what's going on in that strange project find a seemingly abandoned site. not good PR.23:52
wpwrak(twitter etc.) they need constant attention to be useful, yes. you could perhaps automate some tweets or so. but i'm not sure if it's worth the trouble.23:54
wpwrak(automate) e.g., tweet is a thread on the list exceeds a certain size/message frequency. or pick commits, maybe even randomly.23:55
wpwraknew project creation could also go there. or maybe after a few commits, so that people don't only see that something will (maybe) happen, but that they come visiting when there's some meat already.23:56
wolfspraulnice ideas. let's see whether someone wants to step up around these accounts, and if not they are deleted :-)23:57
wpwrakif you want to experiment with automated news, you could send them to a list created for that purpose. if the concept works, you could add interfaces list->twitter etc.23:58
--- Sun Aug 29 201000:00

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