#qi-hardware IRC log for Saturday, 2011-09-10

GitHub199[milkymist] kristianpaul pushed 2 new commits to gps-sdr-testing: https://github.com/kristianpaul/milkymist/compare/d4b6326...3a4218c00:30
GitHub199[milkymist/gps-sdr-testing] missing enable and bad out asigment for code generator - Cristian Paul PeƱaranda Rojas00:30
GitHub199[milkymist/gps-sdr-testing] duplicated enable bit removed - Cristian Paul PeƱaranda Rojas00:30
qi-botThe build has FAILED, see log here: http://fidelio.qi-hardware.com/~xiangfu/compile-log/openwrt-xburst.full_system-09092011-2229/06:57
JasmineAA in games isn't really needed at high res right07:51
Guest83143awh my name changed07:51
kyakxiangfu: hi09:49
kyakxiangfu: it is hard to track down build errors on build host with -j4, the log file is just a mess. How do you think, what could be done?09:50
kyakespecially, the last 100 lines are of no use09:51
virickyak: look at more than 100 lines10:08
virickyak: 'task spooler'! :)10:08
virickyak: with it, "ts make -j4", "ts -c | less"10:29
viricthen /error, ...10:29
kyakviric: hm, interesting..10:54
kyakviric: how i can kill the last job?11:26
kyakviric: nah, task spooler doesn't really help. The output of make -j4 is still intermixed11:43
Jay7kyak: imho, nothing may be done here11:57
kyakJay7: other than running make in a single thread, that is :)12:03
Jay7or patch make ;)12:04
Jay7but I'm unsure it's possible to do12:04
Jay7well.. it's possible but I'm sure this would be hard task :)12:05
Jay7but even just prefixing every line with task no would be enough..12:10
Jay7kyak: anyway, store full log or do grep -B 100 error12:10
kyakJay7: the full log is already stored, but it was handy to just have a look at the last 100 lines of output to see the error right away12:15
Jay7in OE is possible to run builds of multiple packages 12:17
Jay7so there is possible to set bitbake threads to high no and set make threads to 1 :)12:18
Jay7but anyway this would be slower12:18
virickyak: kill -- -`ts -p`13:16
viric(kills the job and childs)13:16
viricah, against intermixed it does nothing, of course :)13:16
viricbut it easily stores the output :)13:16
kyakbtw does it store both stout and stderr?13:17
virickyak: yes, mixed13:17
virickyak: you can do easy tests, like "ts ls asdfasjdf"13:17
kyakyeah, i can :)13:17
viric'-t' runs as if you made a tail -f of it13:18
viricand you can do that from any terminal13:18
viric(the job gets unlinked from the terminal)13:18
kyakwhat is your use case for ts?13:18
kyaki can think of something like start-stop-daemon :)13:18
kyaki don't really need task spoolinh13:19
viricWell, sometime I enqueue downloads...13:26
viricnothing related to start-stop-daemon13:26
viricI enqueue some reencodings of audio, video...13:27
viricnetwork synchronisations of directories (rsync, unison, ...)13:27
viricsystem or package updates...13:29
viricsometimes I need spooling, sometimes I only need easy storage of the output or detachment of terminal13:30
virickyak: it can also spool multiple jobs at once (-S), and sometimes this way I can feed easily a multicore computer13:30
viricmany uses! :)13:30
kyakwhat would happen if you ts an interactive program?13:31
viricit starts programs with closed stdin13:31
kyakyeah, that's many uses :)13:31
viricfor interactive programs you may use 'tm'!13:31
kyaktask manager, right? :)13:31
viricno ;)13:31
viricterminal mixer13:32
kyaki'll have a look later13:32
viric'ts' can also mail the output of the programs :)13:33
viricanyway... I also leave 2 3>AB813:33
virickyak: have good vacations :)13:34
rjeffriesChumby NeTV looks interesting. Inckludes Spartan 6 FPGA. http://www.linuxfordevices.com/c/a/News/Chumby-NeTV/ Not sure about degree of open-ness. Others can evaluat ethat dimension better than moi.13:39
kristianpaul"that most app developers need to get their hands dirty with FPGA programming." nice :-)13:42
kristianpaulwe can introduce milkymist to those developers too ;-D13:43
lekernelrjeffries, do you also send news about qi projects into chumby channels? if not, please do.13:43
kristianpaulyes please :-)13:43
rjeffrieslekernel and kristianpaul look. I assume people here are smart and interested in a wide range of geeky things. I am not following Chumby but this is an interesting device.13:44
kristianpaulsure, but you can spread the word about copyleft hardware projects to 13:45
rjeffriesand I have done so on Twitter and G+ in teh past. ;)413:45
lekernelI have received messages about this netv 4 times or so already; unfortunately, our own PR doesn't work this good13:47
rjeffriesthis gadget overlaying character generator on a tV stram is a clever hack. now the question is what it will cost. MM's FPGA costs $40 as I recall. But maybe there are lower cost versions of Spartan 613:47
rjeffriesabsent a price, this is but a curiosity.13:47
kristianpaulhttp://wiki.chumby.com/index.php/NeTV_FPGA_architecture  arch coming soon :/13:48
kristianpaulah,  nice lekernel , so you know what are they doing with the fpga? looks related to a video acelaration core13:49
rjeffriesthey said they use it to overlay character generation on the video stream. lekernel does the same, naturally13:50
kristianpaulyes HDMI13:50
lekernelwithout framebuffer, which makes it a lot more easy13:50
kyakviric: thanks, and you, too :)13:51
rjeffriesHDMI is A Good Thing. It's waht all TVs and monitors support these days, at least in the developed world.13:51
kristianpauland the fees :-)13:51
lekernelrjeffries, hdmi = patents, licensing fees, DRM, etc.13:52
rjeffrieslekerne;l understood.13:52
wpwrakrjeffries: m1 actually does a lot more than overlay the video ;-) sometimes, things that look easy are nothing like that if you peek under the hood :)14:02
rjeffrieswpwrak I am curious whatr sell price  they will arrive at. FPGAs are not exactly low cost.14:03
kristianpaulXC6SLX9-2TQG144C at digikey14:07
wpwrakrjeffries: depends a bit. some aren't too bad. and you have CPLDs continuing at the lower range. it's all a question how what you really need. of course, a relatively static image/overlay needs a lot less than generating things at a high frame rate14:08
wpwraklekernel: if it stripped DRM transparently, that would be a hit ;-)14:09
kristianpaulhe i tought sparna6 dint come in TQFP14:09
wpwrakkristianpaul: real men don't use TQFP :)14:09
Action: wpwrak quickly hides his pictures of labsw14:09
wpwrakhmm, now .. what to do with that live action video of labsw torturing M1. i somehow don't feel very motivated to spend hours editing it. and unedited it's worse than watching paint dry. besides being pretty incomprehensible if you don't know M1 already.14:13
kristianpaulwhy not just publish it as raw?14:14
kristianpaulah ok14:16
wpwrak470 MB of boredom ? :)14:17
wpwrakit's 6+ minutes. you could cut it down to perhaps one. i'm uploading the monster to http://thisismynext.com/2011/09/08/chumby-netv-smart-tv/14:21
wpwrakETA ~2 hours14:21
wpwrakmake that http://downloads.qi-hardware.com/people/werner/labsw/raw/14:22
rjeffrieswpwrak as that nasty case of time travel cleare dup for you? that could be pretty scary!15:39
wpwrakrjeffries: yeah, it was easy to fix. i think the dinosaurs died out a few more times again, but it's always them who get to suffer the anomalies. kind of a running joke among time travelers :)15:40
wpwrakrjeffries: speaking of time travel, have you seen this one yet ? http://www.abyssandapex.com/200710-wikihistory.html15:48
qi-bot[commit] Werner Almesberger: m1rc3/norruption/: loop5 test: cut power while in standby (master) http://qi-hw.com/p/wernermisc/db7ae9419:34
qi-bot[commit] Werner Almesberger: m1rc3/norruption/: update of on-going testing; new test loop7 simulating "false start" (master) http://qi-hw.com/p/wernermisc/231046619:34
wpwrakif anyone is interested, i've uploaded the raw video material (uncut) of labsw power cycling M1: http://downloads.qi-hardware.com/people/werner/labsw/raw/MVI_1633.MOV19:35
kristianpaul45m to go..19:36
wpwrakDocScrutinizer: i have a number of digital inputs that come from places where they can be exposed to EMI. counter-measures i have in mind is deglitching in software (basically debouncing) and then increasing the pull-up current (tens of kOhm -> 1 kOhm). i'm also contemplating to bypassing these lines to ground by a bit (10-100 nF). do you think these caps would do any good ? or would that just be voodoo engineering ? :)21:25
DocScrutinizershould work to some degree, not for ESD prot though21:26
DocScrutinizeryou probably want to increase source Z by inserting e.g a 50R into the line from that C to exposure stage21:27
DocScrutinizer100nF sound fat21:28
wpwrakESD shouldn't be an issue there21:28
DocScrutinizerand also 100nF don't work for EMI21:28
DocScrutinizeryou'd prefer something like 100pF for that21:28
wpwrakthe EMI disturbances may last for some milliseconds21:29
DocScrutinizer100nF have way too much parasitic inductive Z21:29
DocScrutinizerlol, so it's EMI of 20Hz?21:29
wpwrakit's a relay closing and who knows what flowing through that relay :)21:29
DocScrutinizerit's all about a RC lowpass filter21:29
wpwrakyeah, i get the 50 Ohm + cap structure. darn. more components.21:30
DocScrutinizerwhere R is that 50R I suggested21:30
wpwraksoon, i'll switch even this critter to 0402 ...21:30
DocScrutinizerfor RF EMI you need low-ES"R" Cs, read 10..500pF21:31
DocScrutinizer100nF is like a bead for RF21:31
wpwrakno no, relays are LF :)21:31
DocScrutinizerEMI never is LF21:31
wpwrakwhat else do you all it then ? EMP ? :)21:32
DocScrutinizerthat's called ESD then ;-P21:32
wpwraknaw, nothing electrostatic there :)21:32
DocScrutinizerusually associated to sparks etc21:32
wpwrakyes, sparks fit21:33
DocScrutinizerso how is your several ms of *DC* supposed to enter the trace otherwise?21:33
wpwrakelectromagnetic field, i suppose21:33
wpwrakthe relay is isolated21:34
DocScrutinizerfor several ms of RF you still want a 50pF, not a 100nF21:34
DocScrutinizeryou can't induce a 20Hz full sine wave into a trace on a PCB21:34
wpwrakno no, it's not a full size. it's some upset - i haven't actually seen it - that makes the system think buttons have been pressed21:35
DocScrutinizeryou have 3 problems to handle: RFI, ESD, and OV21:35
wpwraki know that it's not a brown-out or such. that would look different21:36
DocScrutinizerRFI: 50pF21:36
DocScrutinizerESD: tranzorb or series Z with that 50pF21:36
wpwraki think it's induction from wires on which a high current is switched quickly21:36
DocScrutinizerOV: tranzorb21:37
DocScrutinizeryeah, that's inductive coupled spikes21:37
DocScrutinizerduration <1ms usually21:37
wpwrakthe whole episode takes up to ~10 ms. the spikes i've seen (on other lines) are much shorter, though21:38
DocScrutinizeranyway, you'll be fine with series-R 50R plus a 10nF I guess21:38
DocScrutinizerget a higher series-R if you can afford, and spend a 50pF in parallel to the 10nF21:39
DocScrutinizerand clamping-diodes or a tranzorb!21:39
wpwraki'll take 100 R then. nicer number :)21:39
wpwrakthe component count is getting out of hand. i'm already going from an internal pull-up and nothing on the outside to external pull-up plus filter.21:43
DocScrutinizerforget the external pullup21:43
wpwraki need a pull-up. the signal is a button that shorts to ground. well, i can put the pull-up on the "far" side. not sure if that is a good idea, though.21:45
DocScrutinizerdefiitely a good idea21:45
DocScrutinizeryou said you got an internal 10k pullup21:46
wpwrakno, the 10k would be external21:46
wpwraki have something like 50 k internal21:46
wpwrakbut i want to turn off the internal pull-up. causes problems elsewhere21:46
wpwrak(the chip has only a global pull-up enable/disable)21:47
DocScrutinizerand the wire to the switch is like 15m running in parallel to 220V switched line?21:48
wpwraknaw, it's only about 5-10 cm. may get close to relay, lines in/out of relay. relay is currently switch DC. that DC comes from switching power supply, goes into milkymist one. there's probably a lot of inrush current for the caps.21:50
DocScrutinizermeh, don't worry too much, get a 50pF GPIO->GND, a GPIO->50R->1kR-pullup, and your wire to switch from 50R*1kR21:53
DocScrutinizerif you're still worried, place a 100nF from 50R*1kR ->GND21:54
DocScrutinizerfor debouncing21:55
DocScrutinizerand eating spikes21:55
wpwraki'll debounce in sw :)21:55
DocScrutinizermake that 100R21:56
DocScrutinizeror even several kR, on a second thought21:56
wpwraki'm just worried about coupling persistent critters when going to AC. but of course, there i wouldn't have such rapid field changes all the time either. mains doesn't run at kHz ;-)21:56
DocScrutinizerinput Z of GPIO is high I guess21:56
wpwrakvery, yes21:57
wpwrak> 1 MOhm21:57
DocScrutinizerso s/50R/50kR/21:57
DocScrutinizeror sth21:57
DocScrutinizermaybe 22kR21:57
wpwrakhmm. 10 k ? i like my powers of ten :)21:58
DocScrutinizermakes a nice OVP and ESD and EMI together with the 50pF21:58
DocScrutinizerpullup of 100R..1k seems sane21:59
DocScrutinizer100R cleaning switch's contacts ;-)21:59
lekernelkristianpaul, I don't know how they implement stuff in the FPGA. as a matter of fact, I'm not impressed with an FPGA overlaying some text over video.21:59
wpwrakeek. don't have the current budget for that. 1 k would be the absolute limit22:00
lekernelwhat is remarkable, however, is that this project is yet another one with a publicity/coolness ratio much higher than that of milkymist. this is galling and demotivating, and the main reason why i'm not doing much technical work atm. instead of that I'm trying to do some PR and getting wolfspra1l to configure the webshop ...22:01
wpwraklekernel: scantly clad dancing girls !22:02
wpwrakDocScrutinizer: any change the filter would still work if i go from 50 pF to 1 nF ?22:08
wpwrak(better in-house availability of components :)22:08
DocScrutinizer1nF is close to the edge for RFI22:08
wpwrakabout 14 kHz for 10 kR + 1 nF22:09
DocScrutinizerif your device starts to act funny on American Forces Radio blasting out @ 14MHz you know what's up22:09
wpwrakso it'll wire it to my anti-aircraft installation, to set defcon 1 ? :)22:10
DocScrutinizer1nF starts to have too high a parasitic "choke" built in22:10
wpwraki'm not too worried about radio. that shouldn't really produce the power levels that could trip this anyway. (under sane usage conditions.)22:11
wpwrakit's just what happens in and around the relay that concerns me22:11
DocScrutinizerthat's why you frequently see a 10pF, 10nF and 100uF in parallel for really buffwring and smoothing VDD22:12
wpwrakand then you get the other EE crowd that says the whole thing is much worse than just any single cap ;-)22:13
DocScrutinizeryou'll not get anything near a virtual 0R for 100MHz across a 10nF22:13
wpwraki'm so unconcerned about FM ;-)22:13
DocScrutinizerit's more like a virtual break or NC for 100MHz22:13
wpwrakit's a digital input after all. a few tens of mV of FM noise floor or even a few hundred in this case don't matter.22:14
DocScrutinizerhaha, you'll easily see a few volts if bad luck adds to full moon22:14
wpwrakof course, i kinda miss the involuntary RF lab i had at my old place. with several microwave antenna clusters basically right in front of my nose (~50 m away at most, at eye level)22:15
DocScrutinizeryou got a non-linear high-Z input after all22:15
wpwrak(volts) i don't believe in astrology :)22:16
DocScrutinizerusually good setting for an unintentional high quality AM receiver ;-D22:16
kristianpaullekernel:(impressed) FN rsswall ;-)22:17
wpwrakyeah. add the microwave clusters and you have pretty much everything you could wish for ;-)22:17
DocScrutinizerplace your smartphone next to this contraption and watch it switch in the rythm of the 20Hz GSM data packages22:17
wpwrakyou need a metal case then anyway22:18
DocScrutinizerhmm, you remember my epic fail regarding my efforts to use a microwave oven for shielding GSM RF?22:18
wpwraki'm not sure microwave ovens are all that RF-tight anyway ...22:19
DocScrutinizerthe dam FR worked inside the (not operating) microwave better than outside22:19
DocScrutinizerthat's my point22:19
DocScrutinizerno metal case is22:19
wpwraktheir tightness may be more on the scale of "don't let more than 1 kW leak in any direction"22:20
wpwrak(at least not in the certification device :)22:20
DocScrutinizerno, it's just highly tuned to the magnetron's RF freq22:20
DocScrutinizerprobably for WLAN it would've worked just fine22:21
wpwrak(tuned) that would make sense22:21
DocScrutinizernot though for GSM22:21
wpwrakhm yes, gsm is pretty far off22:21
DocScrutinizerlambda/2 seals22:21
wpwrakyeah. and a window.22:21
DocScrutinizerthe whole magic is in the small step in the door frame plus the actual dimension of the gap between door and case, it's of a certain distance from "inside" to that small step22:23
DocScrutinizerso reflections cancel all RF that might want to go thru that gap22:23
DocScrutinizerjust for one freq that this door is designed for22:24
DocScrutinizerI also tested all-metal cookie boxes with a metal lid, phone worked inside just fine22:27
wpwrakpity. scarytech.com is already taken :)22:27
wpwrakyeah, the cookie boxes are worrying22:28
Jay7hehe.. another kexec user in ML22:44
wpwrakJay7: it's unstoppable :)22:54
Jay7yeah :)22:55
Jay7seems I should do another try of kexec and kexecboot on NN22:55
wpwrakYES ! :)22:57
kristianpaulweee :-)23:07
Jay7this may need to bring NN support into new OE (oe-core + meta-handhelds)23:09
kristianpaulwhat is that meta-handhelds?23:18
Jay7OE layer 23:19
Jay7support of Sharp Zaurus/HP iPAQ now is there23:19
Action: Jay7 -> sleep23:27
qi-botThe build has FAILED, see log here: http://fidelio.qi-hardware.com/~xiangfu/compile-log/openwrt-xburst.full_system-09102011-1134/23:32
qi-bot[commit] Werner Almesberger: labsw/: power bypassing is now the way SiLabs recommend (master) http://qi-hw.com/p/wernermisc/5d32ade23:42
qi-bot[commit] Werner Almesberger: labsw/ changed internal pull-ups to external to remove sneak current path (master) http://qi-hw.com/p/wernermisc/6190dbe23:42
qi-bot[commit] Werner Almesberger: labsw/: added low-pass filters to button input (with help from Joerg Reisenweber) (master) http://qi-hw.com/p/wernermisc/bdaff5123:42
qi-bot[commit] Werner Almesberger: labsw/: added hardware revision indication (master) http://qi-hw.com/p/wernermisc/761eae123:42
wpwrakkristianpaul: so .. how did you like my boring movie ? :)23:45
kristianpaulwpwrak: he23:46
kristianpaulbut gets the idea 23:47
wpwrakprecisely :)23:47
wpwrakof course, now imagine adam doing all that manually. hundreds of times.23:47
kristianpaulno no23:48
wpwrakat least i get to sleep while all the boredom unfolds :)23:50
kristianpaullabsw give you some feedback? i mean how you know if meanshile sleeping mm1 poweron but dont get render?23:52
kristianpaulwhere is the log? :-)23:54
wpwraki didn't add that feedback yet. i think i could, though. not in labsw itself but on the PC. there's a pair of registers in the FPGA that contains the status of the latest two attempts to reconfigure. these should tell me what happened. haven't tried that yet, though23:54
wpwrakso what i do is that i simply check visually from time to time and write down the result23:54
kristianpaulha! so not full sleep 23:55
wpwrakhere's the log :) http://projects.qi-hardware.com/index.php/p/wernermisc/source/tree/master/m1rc3/norruption/LOG23:55
wpwrakcatnaps :)23:55
kristianpaulwho/what generate that log?23:56
kristianpaullooks very human23:56
wpwrakbut yes, it's a bit annoying that labsw is busy so long. i still have a few improvements i want to test but i don't want to upset is before these experiments are done. alas, already the last time i took it out was enough to make my setup fail to trigger the NOR corruption anymore. very annoying.23:57
wpwrakyes, that's me. haven't taught my scripts enough AI yet :)23:57
wpwraki also have a serial console log. but that redefines boredom, even if after you've seen the slow video ...23:58
--- Sun Sep 11 201100:00

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