#qi-hardware IRC log for Tuesday, 2011-05-17

DocScrutinizerwhy makes me every detail of that google stuff feel like  D-:00:54
wpwrakDocScrutinizer: not |-O***** ? ;-)00:58
DocScrutinizerindeed, just my graphical skills weren't sufficient00:59
wpwrakDocScrutinizer: maybe they're just trying to piss off intel ;-)00:59
DocScrutinizerthey can take their fakelocks and eat them01:00
DocScrutinizereven friggin nitdridiot on N900 NIT acts like windows 98 - "now I got control over the machine. let's make sure I don't have to share and never ever will lose it again"01:03
qi-bot[commit] Neil Stockbridge: There is now a package for SDL API man pages http://qi-hw.com/p/openwrt-packages/aca02e101:11
wpwrakDocScrutinizer: ah, i thought you were commenting on their reverse usb :)01:23
wpwrakDocScrutinizer: the locks are nice, yes. they already have quite a tradition. i think i kicked them out of the openmoko kernel, oh so many years ago :)01:25
qi-bot[commit] Xiangfu Liu: nanonote-files: data: config.full_system: enable 'man' in busybox http://qi-hw.com/p/openwrt-packages/860605302:33
uncloudedhey xiangfu, thanks for including "man".  it looks like busybox "man" expects to find page in /usr/man ( info thanks to strace being included by default!).  the pages I just uploaded are in /usr/share/man and busybox "man" doesn't find them by default02:48
uncloudedwhat the best way to make them work together?02:49
uncloudedI could move the libsdl pages to /usr/man, or /usr/man could be a link to /usr/share/man or the other way around, or maybe busybox can be configured to look in both locations?  what the most traditional way to do it?02:50
xiangfuunclouded: have you try this setup variable " $MANPATH"02:56
uncloudedxiangfu: that's the most elegant solution.  thanks!02:57
xiangfuunclouded: I still don't have one busybox that have 'man' so I can not test :(02:57
uncloudedxiangfu: it works great:  MANPATH=/usr/share/man man SDL_Delay02:58
xiangfudis MANPATH support like  /usr/share/man:/usr/man ?02:59
uncloudednext issue though: when I run that command, I get: sh: more: not found02:59
uncloudedsh: gtbl: not found02:59
xiangfuwe can export the MANPATH in '/etc/profile'02:59
uncloudedthat would be great.  looks like it doesn't support multiple paths though: MANPATH=/usr/man:/usr/share/man  man SDL_Delay03:00
uncloudedman: no manual entry for 'SDL_Delay'03:00
xiangfuok. then we just 'export MANPATH=/usr/share/man' in /etc/profile03:10
uncloudedok.  what about this difficulty "man" has in finding more and gtbl?03:11
xiangfuunclouded: I try to compile the busybox. with 'man' , downloading the SDK now. need 20 mis. 03:34
vladkorotnevhello everyone06:00
vladkorotnevI just started translating the Qi wiki to Russian. Already translated http://en.qi-hardware.com/wiki/FAQ :)06:02
vladkorotnevAnybody here?06:29
kyakxiangfu: having a look at buildhost, perhaps it is better to separate jobs for building the trunk and master07:42
kyakor maybe "separate more"07:42
kyakthere are just 8.5 hours between them, might be not sufficient07:43
Jay7why not to enqueue?07:45
kyakyea, maybe launch the jobs in sequence07:46
kyakand then maybe build it forever, i.e. trunk-master-trunk.. :)07:47
nomorebeerGuys, i have a notebook keyboard. I have the product number. Is there any way to find out what chip model number drives this keyboard ?11:00
dvdknomorebeer: wrong channel?  you might have been looking for #hardware11:01
nomorebeerthanks .11:01
vladkorotnevHello everyone :P13:50
scanfhow much is the nanonote15:26
scanfand what are the prospects for robot control and video control15:26
kristianpaulscanf: nanonote.cc 99USD last time i checked15:27
kristianpaulplus shipping etc..15:27
scanfit was $99 or the one last year15:27
scanfis it the same still15:27
kristianpaulbut be aware there are ditributors in some zones15:28
kristianpaulthe robot and video control15:28
kristianpaulwell, i dont know what you want to do specifically15:28
kristianpaulthe most common and used port in the nanonote is the microsd-like port15:28
scanfcontrol motors15:29
kristianpaularduino^ 15:29
kristianpauldunno motors as it depend on your driver etc.15:30
kristianpauli think at tuxbrain.com there is a post about nanonote and arduino15:30
kristianpaulalso please check http://en.qi-hardware.com/wiki/UBB15:30
kristianpaulUBB allow you to bitbang for I/O porpuses15:31
kristianpauland if you want to hack more,15:31
kristianpaulthere is a serial port in the nanoonte15:31
kristianpaulthat need solder some wires to get work propiertly but is another optio15:31
kristianpaulbut in theory usb port plus linux support can emulate a serial port too i think15:32
GNUtoohow does tuxbrain connect their arduino?15:32
kristianpaulscanf: http://en.qi-hardware.com/wiki/Serial_console15:33
GNUtooor do they use some other way15:33
GNUtoolike spi?15:33
kristianpauldont know15:33
kristianpaulspi is posible15:33
kristianpaulbut i think it was serial15:33
GNUtoome too since they flash it15:33
kristianpaulflash was acomplised by bitbanging and spi i think15:34
GNUtooah ok15:34
kristianpaulbut not sure, i'm not arduino fan15:34
kristianpaulbut sourcode usually dont lie http://projects.qi-hardware.com/index.php/p/ben-blinkenlights/15:35
scanfim trying to figure out if i want to go netbook or nanonote for an ROV brain15:37
ron_scanf are you talking about a computer on teh robot to use to control it, or a remote console that talks to various microcontrollers on the robot15:45
scanfcomputer on robot hooked to mcus15:46
scanfprobably wirelessly (wifi?) linking to yet another computer15:46
rjeffriesscanf Nanonote is a cute and clever little device. IMOI it is not a good match for what you need to do. very lkimited i/o ports15:48
rjeffriesabout a year ago there was an interesting project that married the COC used by Nanonote with an FPGA. removed keyboard, freeing a bunch of GPIOs15:48
rjeffriesit implemented handy things like I2C and Ethernet and oiptionally could have an LCD15:49
Jay7don't forget about MilkyMist SoC15:49
rjeffriesbut that project withered on the vine15:49
rjeffriesMilkymist is a possibility. it costs $50015:49
scanfi should probably just do arm cortex + linux netbook15:52
Jay7seems we are needed some board based on NN jz SoC15:53
kristianpaulrjeffries: Ethernet where?15:54
kristianpaulJay7: may be but is not a really demanded product i think15:55
Jay7kristianpaul: but it may be 'sellable' :)15:55
kristianpaulmay be :-)15:55
Action: Jay7 have bought Samsung TV..15:55
Jay7it have some ARM or MIPS board inside15:56
kristianpaulBut i realize mips is not part of the DIY prototyping worl15:56
Jay7and firmware is linux-based..15:56
kristianpaulmay be also a fpga15:56
Jay7but protected :(15:56
kristianpaulnot released at opensource.samsung?15:56
Jay7I want a free TV one :)15:56
kristianpaulBut a Analog one 15:56
Jay7hehe :)15:57
Jay7there is samygo.tv project about hacking that devices15:57
Jay7but I have latest firmware.. and not sure how hackable it is..15:58
wpwrakscanf: the ben is a good choice for tasks that require a small number of i/o signals, particularly if the device itself also has to be small. examples for this kinda of use would be user interfaces or prototyping of subsystems.16:12
wpwrakscanf: if you need more I/Os you would want a dedicated board or, if it gets really complex, a netbook plus usb-based peripherals16:13
wpwrakscanf: besides hw capabilities, you'll also want to pay attention to the porting effort involved in getting things to run on the device, and whether kernel drivers are properly mainlined16:19
wpwrakscanf: finally, anything running linux will have only limited real-time capabilities. the ben can do quite amazing things via the MMC port, including vga output, but if you continuously need operations with very precise timing and/or very short response time, you may want to consider adding some small microcontroller for this specific task.16:21
wpwrak(that's irrespective of what you use as the "brain" in the end - a netbook and even a prototyping board will have similar limitations)16:23
wpwrakscanf: in terms of hw-assisted peripherals, the ben can give you a configurable clock of up to 56 MHz, and via MMC limited-length sequences of clocked output signals (up to four of them), and probably also sequences of input signals (afaik, nobody has tried input yet, though). you can also access a UART if you don't mind a little soldering. if you have no or only limited use for the keyboard, you can use even more peripherals (SPI, etc16:26
wpwrak.). for the rest, you have to use gpio.16:26
wpwrakscanf: you may want to study the ben's schematics: http://en.qi-hardware.com/w/images/9/9c/Lb60_schematic.pdf16:27
rjeffries_wpwrak prolly a dumb question, but how general is the SPI driver in Ben? Could Ben perform as SPI slave (pure curiosity)20:33
rjeffries_this article got me thinking. actually, for general utility with an AVR/Arduino sensor node, two-way communication with Ben as Master would be great20:34
wpwrakrjeffries: no, the ben can't do SPI slave easily. its hardware SPI controller is master only. when doing it with bit-banging, you could theoretically implement slave, but you'd have to worry about timing. (i.e., the SPI clock would have to be quite slow)22:57
wpwrakrjeffries: however, since SPI data is bi-directional, there's usually no problem designing your protocol such that the Ben can be the master, and the MCU, with generally more predictable response time, can be slave22:59
wpwrakrjeffries: if the MCU needs to initiate a communication, it would use an interrupt line for this purpose23:00
rjeffrieswpwrak thanks for unconfusing me23:11
wpwrakDocScrutinizer: as a wild guess, when you'd take this board to SMT http://downloads.qi-hardware.com/people/werner/wpan/tmp/atben-110324-overview.png23:49
wpwrakDocScrutinizer: how many (very roughly) boards would you expect to need to figure out any SMT process issues (e.g., pick and place configuration and such) ?23:51
wpwrakDocScrutinizer: and, more importantly, once the initial problems have been overcome, what yield would you expect in a small run of 100 units ?23:51
wpwrakDocScrutinizer: my guesses would be something like 10 board for debugging the SMT process, maybe less. not sure about the yield. something in the ballpark of > 90%, < 99%.23:53
DocScrutinizerextremely dependant on the fab, and competence of operators. Your numbers looks not completely insane23:54
DocScrutinizerthere might be some issues like spilled header boards (reflow temp balance needs min of 10 boards to run thru until temp profile settled on 11 board which is first one expected to be "normally good"23:56
wpwrakDocScrutinizer: good ;-) yes, i know there are lots of unknowns in this equation. kinda like the drake equation :)23:56
wpwrakurgh. that would be nasty for a small run23:57
DocScrutinizerdunno, iirc at OM we had PV runs of 5023:58
DocScrutinizeror 100?23:58
DocScrutinizeryou'd need that much to find issues of the more statistical class23:59
wpwrakthis is a small production run of 100-130 units23:59
wpwraknobody around here has any money for real quantities ;-)23:59
DocScrutinizeryeah, I see23:59
wpwraknaw, no yield optimization23:59
--- Wed May 18 201100:00

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