#qi-hardware IRC log for Monday, 2011-08-08

kristianpaulha, i found the root cause for the accum int signal not asserted... i did a bad asumption about ignoring bus cycle signal from  avalon to wishbone port...01:44
kristianpauloh well01:45
kristianpauldamn, can be? in namuru i need to read/qeite some registers  to clean some flags...02:10
kristianpauli cant see where is it pointed in the documentation.. but02:10
Action: kristianpaul cross fingers and hope Artyom to confirm this02:10
kristianpaulwell, if this is true i can implemnent id to fit milkymist csr bus workflow, but is tricky to cath it from avalon as current it is..02:12
kristianpaul"real soon now " :-)04:56
kristianpaul300 more and i can buy a ephel ;)04:57
wolfspraulkristianpaul: should we remove the "real soon now"?04:59
wolfspraulI mean the whole line?04:59
wolfspraulI didn't wnat it there but Sebastien added it...05:00
wolfspraulif there's something I don't like, it's announcements05:00
kristianpaulwell, the soon now is kinda odd imho05:00
wolfspraulthanks, that helps05:01
kristianpaullike, is here but not can touch it !!, just wait a bit ;-)05:01
kristianpaulalso i asume you had a BIG annoucment for the this rc3?05:02
wolfspraulI first want to actually have it05:02
kristianpaulconsidering new stuff on the box and all software improvements since the EDK era :)05:02
kristianpaulnot to much effort i think all is already wrritten just need to be organized.05:03
kristianpaulthose videos from xiangfu are a must to include,05:04
kristianpaulfrom all my mm1 demostration people always get amazed with the effects and the music just playing in the background :)05:04
kristianpaul(no complaing for the resolution or dunno what else)05:05
kristianpaulone it run people like. so the preview videos should keep that sensation05:05
kristianpaulnow i consider again the must word :)05:06
kristianpaulrjeffries_: \.? ;-)05:06
wolfspraulkristianpaul: I don't understand you. What is your point?05:07
kristianpaulwolfspraul: better quality on the preview videos for patches :)05:08
wolfspraulyou mean the 4 new ones?05:08
kristianpaulwell i'm not saying current are bad, just keep improving :)05:08
wolfspraulnot good enough?05:08
wolfspraulwhat needs to be improved?05:08
wolfspraulfor the 4 new ones, what needs improvement?05:08
wolfspraulprobably that needs improvement on the m1 side05:09
wolfspraulthose videos are taken with a camera mounted on a tripod filming an LCM :-)05:09
kristianpaullooks like those movies ripped from cinema , in the good senseof my comment05:09
kristianpaulhe, yes that !05:09
wolfspraulyes, but that's the only practical way I have right now05:10
wolfspraulI'm not sure whether the resolution actually can be improved on the recording side05:10
wolfspraulthe camera certainly has a high resolution, and then there's video compression05:10
kristianpaulsure i understand just dropping some comments for what i think can call atention for rc3 annoucement05:10
wolfspraulI don't know how to quickly increase the quality of recordings05:11
wolfspraulvga grabber _might_ help but I don't have one and it's expensive and may need Windows etc.05:11
wolfsprauland it may very well end up in the same quality05:11
kristianpaulhe i was about to point that (vga grabber)05:11
wolfspraulwhat would help is real-time video encoding on the m1, or higher m1 resolution05:11
wolfsprauluntil then I think the camera from lcm approach is not bad05:12
kristianpaulif you say so :)05:12
wolfspraulkristianpaul: you think it's too bad?05:13
kristianpaulno no05:13
wolfspraulit's just a practical idea now that works, and we can record a lot of patches fast05:13
wolfspraulI want to record some with different kinds of music05:13
kristianpauljust "rip" a video from another video camera is not always good, sure is practical05:13
wolfspraulyou try with a vga grabber then we compare05:14
wolfspraulit's slow and expensive05:14
wolfspraulI am thinking practical05:14
wolfspraulnow we have those 4 videos, today05:14
wolfspraulin the wiki05:14
kristianpaulsure, i just complaing...05:14
wolfsprauland I can easily make another 10 or 2005:14
wolfspraulI will not go down the vga grabber direction05:14
wolfspraulI checked a little, not good05:14
wolfspraulif someone else does - great :-)05:14
wolfspraulwiki uploads work for anybody...05:15
wolfspraulso better videos from me will only come if either real-time m1 encoding/streaming works, or if the m1 resolution goes up05:15
wolfspraulboth of which is not to be expected soon05:15
wolfspraulmeaning that, unless someone else steps up and beats me to better quality, we will get more of this current quality level you see there05:15
wolfspraulit's the best I can pull off right now05:15
wolfspraulI think it's better than the ones we had before, a step in the right direction05:16
kristianpauloh, sure05:17
kristianpaulas i said, keep improving :)05:17
kristianpauljust after you watched several times, details pop-up05:17
kristianpaulno more :)05:17
wolfspraulI am just describing what I did so far and why, and what next steps are planned.05:20
wolfspraulSebastien tried to work on 1024x768 resolution, but could only get it to about 4 fps at first05:21
wolfspraulthen he worked on several speed improvements like a TMU05:21
wolfspraulwhich got it up but not enough05:21
wolfspraulso as a side effect we have faster memory now, and smoother 640x480 rendering :-)05:21
wolfspraulbut more work is needed to squeeze out more performance on the memory side05:22
wolfspraulhe mentioned a complete redesign of the memory controller, which would require serious work and research he said05:22
wolfspraulwon't happen soon05:22
wolfsprauland/or that Xilinx code snippet, which unfortunately comes without source redistribution rights05:22
wolfspraulwe'll get there eventually, but more people need to join and help ramp up the speed05:23
wolfspraulat least Sebastien already saw the rendering in 1024x768 and said it looked noticably better :-)05:23
wolfspraulthat's great05:23
wolfspraulhe said 800x600 doesn't look much better than the current 640x480 (also related to some texture sizes I think)05:24
wolfspraulso I guess we can skip 800x60005:25
wolfsprauleither 1024x768 or stay at 640x480 and make that better05:25
wolfspraulthat's all I know05:25
kristianpaul640x480 looks nice05:26
kristianpaulyou really take a while to realize the real resolution and bits per pixed, if you are not told 05:27
kristianpaulactually i started the concern after playing with the video-in, before that i even  thought about it05:28
kristianpaulArtyom, how do you clean status_read flag? i guess as writing to STATUS reg isnt?05:33
kristianpaulthat will explain also that i need to re-load some values per channel, after a accumulator interrupt05:34
kristianpauli poiting it because i think that was my problem with zeroes accumulators, i just implemented wrong all the flags for enable and registers and internal flags in namuru..05:35
kristianpaulso once written they keep asserted hihg.. ;-)05:35
kristianpaulanyway, fixing  now i think that was my problem :), waiting for your feedback05:36
kristianpaulwolfspraul: can you check please sige.com i dont see GPS in products anymore...05:55
kristianpaulah yes it is ;)05:55
kristianpaulsorry for the noise05:55
kristianpaulthat Search box is not very helpfull05:56
kristianpauland also http://www.skyworksinc.com/Products_SiGe_Mobile_Devices.aspx05:57
wolfspraulat the level you are working on, we will be able to move to different front-ends later I'm sure, with relative ease05:58
wolfspraulthat's the whole point of copyleft hardware after all, to push the software level to the point that every chip we source can easily be replaced with another chip from a second, independent, manufacturer05:59
wolfspraulI am not checking with SiGe on the business side, because I have so little to show to them / talk about05:59
kristianpaulhe, sure, just wpwrak was concern about the acquired i wanted to make sure current status :)05:59
wolfspraulbut once I do, I will definitely reach out to them05:59
wolfspraulremember they sent us this EVB of a (back then) unreleased chip for free, to support our project06:00
wolfspraulI owe them a report back06:00
kristianpaulsure :)06:00
kristianpaulcurrent report signal detected no tracking, i'm working on it.. :)06:00
wolfspraulsure, no rush06:00
kristianpaulok off bed now tomorrow i have boring meeting ;)06:01
rjeffries_wpwrak trying to recall when I advocated "open loop design" 06:08
rjeffries_was that the time I suggested Argintina should become a US Territoty? I can't remember. ;)06:32
rjeffries_"Even a mistake may turn out to be the one thing necessary to a worthwhile achievement." - Henry Ford06:44
xiangfuqwebirc73004, hi07:06
aw0x50: C154 re-soldered its two pads, the impedance err disappears. phew~I took out L8/U11/C136/U13 sequentially then found C154. ;-)07:10
wolfspraulaw: not bad07:12
wolfspraulanother board recovered?07:12
wolfspraulwas the cold soldering visually detectable? (would have been cool if you had taken a microscope picture before the resoldering)07:12
awdidn't take microscope. :( but I would like to change 0805's footprint to avoid of this. the gap between two pads seems too closer. 07:14
wolfspraulsure sounds good. take a pic now to document it.07:15
wolfspraulwould have been interesting to know whether it was detectable before, visually07:15
awcold soldering can't cause "short" in C15407:15
wolfsprauloh sorry, yes07:15
wolfspraulwrong word07:15
wolfspraulI was just curious how it looked before07:16
wolfspraulbut no pic, no problem07:16
awit must be haven a "short" condition under two soldering pads to cause this. but bad that I forgot to take mircoscope, it may still not easily to discover by microscope top view though.07:17
wolfspraultop view not, maybe from the side?07:17
wolfspraulwell, now it's too late07:18
wolfspraulbut a good discovery, let's see how the rest of the board works...07:18
awthis happens of course if solder paste is too much then get this.07:18
wolfspraulchange footprint sounds good07:18
wolfspraulif that helps07:18
awbut I would also narrow the gap between 0805 footprint. ;-)07:19
awnow~go to soldering those parts back and see. :)07:21
wolfspraulwpwrak: yes, I really like Milkymist Hardware Family, but one more word missing07:51
wolfspraulHardware Family Requirements?07:51
wolfspraulwhich family has 'requirements'...07:51
wolfspraulso in that style it needs to be something friendlier then07:51
wolfsprauldescription? too meaningless07:52
wolfspraulcharacterizations? particulars?08:00
wolfspraulMilkymist Hardware Family Particulars? MHFP :-)08:00
wolfspraulor just Milkymist Hardware Family, nothing else08:01
LunaVoraxHi everyone !09:20
DocScrutinizershort probably caused by either to much solder paster, or poor alignment/stencil-layout, or too much pressure when placing components into solder paste bed09:29
DocScrutinizerdang, I bet this time it will take a week til OM box chandra is up again09:30
DocScrutinizerI have a pretty clear idea where "our" sysops are right now and what they're doing09:31
wolfspraulLunaVorax: hi09:37
LunaVoraxwolfspraul, :)09:39
LunaVoraxGot the newsletter09:39
LunaVoraxI have ONE complain09:39
LunaVoraxSome people have computers that cannot play HD videos correctly09:39
LunaVorax(in this case, me)09:40
LunaVoraxSo I cannot enjoy some of them09:40
wolfspraulwhich one?09:40
LunaVoraxThe hello world video09:40
LunaVoraxAnd probably the "How to boot the Nanonote into USB mode" also09:41
LunaVoraxI would have rencoded them in half the size but for some reason avidemux doesn't want to open them09:41
awDocScrutinizer, yes, since I currently found one 0805 for short condition: although surely the stencil-layout and much solder pastes or much pressure will cause this, I'd still prefer to narrow the gap between 0805's two pads. ;-) surely the a smaller stencil's aperture can help this. the right effort should not be at smt vendor side though. ;-)09:42
wolfspraulLunaVorax: can you explain more?09:42
wolfspraulwhat system do you have? how do you watch videos?09:43
wolfsprauldo you use the javascript viewer?09:43
wolfspraulor download the .ogv file?09:43
wolfsprauloh wow, indeed. the original of those two videos was done at 1920x1080 pixels - cool09:44
DocScrutinizeraw: narrowing gap? please elaborate09:44
wolfspraulthat will still play nice on the 4000x4000 pixel screens we will have in a few years :-)09:44
LunaVoraxwolfspraul, I am using a 1.3ghz (mono core) computer (with an old gefore mx400 64mb card running the open source driver)09:44
LunaVoraxwolfspraul, and I download the .ogv files on the computer and play it with mplayer09:45
wolfspraulhave you tried using the javascript viewer?09:45
LunaVoraxwolfspraul, not sure it will go faster but let me try09:45
LunaVoraxI browse the web with JS off09:45
wolfspraulI don't know technically where it's scaled down, but the view is definitely much smaller than 1920x1080 there09:45
LunaVoraxOh ok09:46
wolfspraulwhat I can do in the future is for any video above a certain resolution, to upload a second lower resolution one09:46
wolfspraulbut it's all work and I need to understand where the problem is first09:46
wolfspraulyour mplayer cannot scale down the 1920x1080 video?09:46
LunaVoraxHmm no, no good, twice as sluggish09:46
wolfspraulit may be scaled down locally09:46
LunaVoraxwolfspraul, i tried with no success but maybe I'm using the wrong parametters09:47
wolfspraulotherwise I don't know who could do the computing intensive scaling09:47
awDocScrutinizer, the distance between two pads of 0805 I'll expand it, currently it's too close. sorry that I used wrong word. ;-)09:47
wolfspraulI guess MediaWiki is not that good yet in auto-scaling videos09:47
wolfspraulfor pictures it's great09:47
wolfspraulso I can easily upload the absolute highest-res picture I have, because in any place I use it MediaWiki will scale it down on the server, and cache the scaled down version09:48
wolfspraulbut for videos that's not the case (yet, I would hope)09:48
wolfspraulso if I upload a super high-res 4000x4000 video, that'd be bad09:48
wolfspraulpeople would have to download all that data, only to have it scaled down after download - not good09:49
wolfspraulthat's my current understanding09:49
wolfspraulso I guess I need to manually create lower-res versions, until MediaWiki is better with videos09:49
wolfspraulLunaVorax: which resolution is best for you?09:49
awDocScrutinizer, C154/0805 is surrounding U13: http://en.qi-hardware.com/w/images/1/1f/M1_rc3_front_gerber_view_image.png09:49
DocScrutinizeraw: look, neither the PCB nor the component has any surface that would make the molten solder migrate and close the gap between the two pads. The pads shall be no closer to each other than the contacts of the component. There is no way the solder shorts the gap unless either your solderpaste blobs merged into each other due to too much paste or pressure, or you apply pessure when solder is molten so it will spread under component09:49
LunaVoraxwoakas, for video playback, 640x480 is good (in know it's 2011)09:50
LunaVorax-in + I09:50
wolfspraulthe size of the window on the news will not be bigger than that anyway09:50
wolfspraulit's just maintenance overhead09:50
wolfspraulfirst a lower-res video needs to be created manually, then uploaded09:50
wolfspraulthen there are 2 uploads of the same video09:51
wolfsprauland some people may prefer the high-res one, but even if you link to both resolutions, that's not very user friendly09:51
wolfspraulbest would be if the scaling would be more automatic and built-in...09:51
DocScrutinizeraw: general rule: the amount of solder paste applied never can be too small09:51
wolfspraulLunaVorax: thanks for the feedback! I will look into this09:51
wolfspraulI noticed that on Wikimedia Commons many videos are uploaded in multiple resolutions09:52
LunaVoraxwolfspraul, thank you :)09:52
LunaVoraxwolfspraul, i'll try to scale down the video using ffmpeg09:52
wolfspraula lot of work09:52
wolfspraulthe server should do that for us09:52
wolfspraulour time is valuable, and there are better things to do...09:52
LunaVoraxOh ok09:53
LunaVoraxLet the server do the work then :P09:53
wolfspraulfeel free, feel free...09:53
wolfspraulbest would be if MediaWiki could do video scaling (and caching) like they do for images09:53
wolfspraulbut it's a lot of work throughout the whole system09:53
wolfspraulI'm sure09:53
wolfspraulI'll think about it09:53
LunaVoraxIf you can tweak that also, try to make the server use the svn version of theora, it has really been improved and makes videos looks better09:54
awDocScrutinizer, the stencil built from smt vendor, the stencil's 0805 size will never be the same as equivalent to design solder paste layer size, I agreed your idea though. but I would change our design not to ask smt vendor to request if their thickness of solder paste is under spec. or not. ;-) The design from ours should be producible everywhere to let any smt vendor can mount/assembly boards. ;-)09:54
DocScrutinizeraw: otoh increasing the gap between pads to more than the distance of contact to contact distance on component is generally not a good thing either09:54
LunaVoraxwoakas, (even better than wp8, hum hum)09:55
awDocScrutinizer, :-)09:55
DocScrutinizeraw: what you sugest can not work. The SMT population and reflow process has to be tuned anyway, and there's no such thing like a universal layout09:56
DocScrutinizerIOW when smt fab is doing nonsense, there's nothing you can do against it on layout side09:57
DocScrutinizerbut maybe the footprint is actually sub-optimal, I haven't checked it09:58
DocScrutinizergetting good footprints seems the holy grail of smt designers09:58
wolfspraulLunaVorax: the server is not using theora (what exactly?)10:00
wolfspraulthe only thing the server does is the thumbnails from videos10:00
wolfspraulusing oggthumb now because ffmpeg caused too many problems10:00
wolfspraulbut the server is not doing video scaling, at all10:00
wolfspraulthat's the thing10:00
wolfspraulso a video uploaded as 1920x1080 will be downloaded like that, full resolution10:01
wolfsprauland then scaled down locally, I'm pretty sure10:01
LunaVoraxwolfspraul, hum ok10:01
LunaVoraxtheora = the ogg video codec, is that what you were asking ?10:01
DocScrutinizeraw: anyway doublecheck your pad size vs size of component contacts. I'd think they have to exactly match, otherwise you're asking for trouble10:01
DocScrutinizerSMT components get pulled into position by solder cohesion and adhesion10:02
DocScrutinizeronly if footprints match exactly - otherwise they get pulled *out of pos*, they move in unpleasant ways10:03
wolfspraulLunaVorax: I do not believe the server touches (encodes or decodes) the ogv file10:03
wolfspraulfrom my understanding, the only thing the server does with uploaded ogv files is to extract thumbnails with oggthumb10:03
wolfspraulthat's it10:03
wolfspraulthe rest is on the client10:03
wolfspraulthat's my understanding, I need to look into it10:03
wolfspraulnow that I understand your problem I think we will find a solution :-)10:04
wolfspraulthanks A LOT for the feedback10:04
wolfspraulthat's the kind of stuff we need to get better...10:04
LunaVoraxI'm sorry I'm a bit confused with all of this10:04
wolfspraulwell I am guessing too10:04
LunaVoraxWell, thank you... I though i was only one single grumpy user10:04
wolfspraulI need to do some study on it...10:04
wolfsprauleven if so, so what10:04
wolfspraulthen we improve it for you10:04
wolfspraulindividuals always have the purest perspective on things, imho10:05
wolfspraulonce you start to ask groups/polling, you are in all sorts of social experiments and surprises10:05
wolfspraulhard to understand what 'the group' really meant and wanted :-)10:05
wolfspraulso I rather go with individual feedback10:05
awDocScrutinizer, pad size and size of components are not match well though I just realized when I see it. ;-) this is primary, others surely is secondary idea. but glad to see here you gave us great inputs.10:05
wolfsprauland then either you are happy or not, I can understand that feedback10:05
DocScrutinizeraw: bottom line: make sure your footprint is *correct*, there's only one correct padsize for footprints. And you can do nothing if SMT fab applies too much solder paste10:05
LunaVoraxAnyway, about this tutorial, I think I'll have to work on it and write a wiki page about it, cause the guys is asuming that the user already installed the GCC toolchain. That's pretty discouraging if you're a noob.10:06
wolfspraulDocScrutinizer: keep in mind the statistics too. If we have a problem with 1 of tens of thousands of parts, that needs to be considered as well.10:06
wolfspraulso a solution like "too much solder paste" may well lead to problems with a few hundred parts in the next run, instead of 1 :-)10:07
wolfspraulthis was a run of 90 boards, and Adam now foudn this one instance of this once capacitor on one board10:07
wolfspraulthat's a pretty important data point to consider in any counter measure10:08
DocScrutinizerapplying solder paste is a critical process and prone to have glitches every now and then10:08
DocScrutinizerbut uneven printing of solder paste is a problem of SMT manuf10:09
DocScrutinizeronly thing designer can do is make sure the FP is _correct_10:10
DocScrutinizera _correct_ footprint is already optimized to avoid issues with too much solder paste - there's no ueber-optimization especially to avoid shorts10:11
DocScrutinizeras you got a distance of contacts on component as well and that won't change no matter which footprint. Ideally footprint and component match 100%10:12
DocScrutinizeryou *may* optimize solder paste stencil though10:13
DocScrutinizerand make sure your solder stop mask printing is OK10:14
DocScrutinizer(the usually green stuff)10:14
wolfspraulLunaVorax: there's some good information here http://commons.wikimedia.org/wiki/Help:Converting_video#General_conversion_tips10:38
wolfspraulI will do that for the 2 tutorial videos10:39
LunaVoraxPerfect ! I didn't knew wikimedia what hosting such helping instructions10:40
DocScrutinizerwolfspraul: do you have administrative access to openmoko chandra?10:41
wolfspraulLunaVorax: more here http://en.wikipedia.org/wiki/Wikipedia:Creation_and_usage_of_media_files#Video10:44
wolfspraulDocScrutinizer: I doubt it, sorry10:44
LunaVoraxwolfspraul, the red square probably is the most valuable information, it also talk about bandwidth10:45
DocScrutinizerwolfspraul: who's gismo's "boss"?10:48
DocScrutinizerit's a NOGO those 2 dudes keep exclusive admin of OM infra10:48
DocScrutinizerthey damn share credentials to sb not living in Berlin, pleae10:49
DocScrutinizerpreferrably sb who knows about admin stuff, e.g mrmoku10:50
LunaVoraxwolfspraul, I'm encoding the video you'll need for the wiki10:58
LunaVoraxit'll be 400x225 the exact size of the thumbnail10:59
LunaVoraxwolfspraul, ok got the right settings, I'll send you the video when it's ready11:15
wolfspraulLunaVorax: he, nice! :-)11:24
wolfspraulsend me?11:29
wolfsprauljust upload them11:29
wolfsprauland then we can link from the embedded video and caption11:29
wolfspraulI'll do that once it's uploaded11:29
kristianpaulbah, 10 minutes later :(11:50
qi-botThe build was successfull, see images here: http://fidelio.qi-hardware.com/~xiangfu/compile-log/openwrt-xburst.full_system-08072011-0759/11:52
wolfspraulLunaVorax: you there? did you run into problems? I can finish it too if you like...12:24
LunaVoraxwolfspraul, sorry I just came back eating12:29
LunaVoraxI'm finished12:29
LunaVoraxDon't worry, I have a long experience playing with ffmpeg :P12:29
LunaVoraxWhere can I upload the files ? Can I do it with my account on the wiki ?12:30
wolfspraulset the license to cc-by-sa, copyright status and source just fill in your own name, if you want to be perfect you can link to the higher-resolution original somewhere12:31
wolfspraulmaybe in source12:31
LunaVoraxok 12:31
LunaVoraxwolfspraul, uploaded and newsletter fixed12:44
wolfsprauloh, let me check12:45
wolfspraulok, the thumbnails look different already, strange :-)12:47
LunaVoraxYeah I saw that12:47
LunaVoraxI don't know if that's very important12:47
LunaVoraxAnyway, I think it teached us a valuable lesson, video embedded in the newsletter should follow some standards like max height = 360px and max bitrate = 220kbits/s12:47
LunaVoraxThe question is now, how to configure the server to do the job for us every time. Maybe a simple bash script should do the trick12:48
LunaVoraxwolfspraul, does the reencoded video suits your standards ?12:54
wolfspraulthe thumbnail is very important12:55
wolfspraulbecause that's the impression people get from looking over the page12:55
wolfspraulthe thumbnail will make a _huge_ difference as to whether someone will watch the video or not12:55
wolfspraulit's a bit tedious to find a good one, and unfortunately you can only pick a second, not a frame. so it's a hit and miss.12:56
wolfspraulI think I found 2 acceptable ones for the 2 new videos now, check it out12:56
wolfspraulLunaVorax: ok check it, now I changed the download link to download medium/high13:02
wolfspraulit's not nice but I guess we have to do this because we cannot sense which resolution make sense for the client13:02
wolfsprauloh, and the size is also wrong :-)13:02
LunaVoraxwhich size ?13:03
wolfspraulaw_: how can we find the root cause of boards like 0x34 ?13:03
wolfspraulLunaVorax: look under the video13:04
wolfspraulwait let me update this13:04
LunaVoraxwolfspraul, I thing that what would make sense for the client would be something they see every day. Therefore (if the video were encoded following a pixel height rule) using terms like (360p, 720p, 1080p) instead of (low, medium, high) which can be considered somewhat vague13:06
wolfspraulyou will find out that there are as many resolutions as there are videos13:08
wolfspraulcheck now, updated again13:08
wolfspraulI think now it's good13:08
wolfspraulI removed the medium/high, sure that is vague but we will always be pushing the problem around13:08
aw_wolfspraul, not only 0x34 which stopped at 'Bitstream length: 1484404', few boards have been also stopped at that err, it seems that i would say(but right now I can't approve it. ;-) ) that 1). s/w codes can't fit h/w tolerance caused by parts, then corrupted somewhere then can't restore/recover. 2) what's exactly part caused shift easily by part itself tolerance?13:08
wolfspraulyes but which h/w tolerance? :-)13:09
wolfspraulwe can send 0x34 to Sebastien, that's one option13:09
LunaVoraxwolfspraul, your edit is perfect :)13:09
wolfspraulor you try to find out the hardware difference, by comparing until you find something suspicious. or maybe by looking for certain things? (if sebastien has ideas)13:09
wolfspraulthe good news is that it seems we can reproduce the problem on 0x34, that is it happens every time13:10
wolfspraulthat's better than a problem that appears only rarely13:10
wolfspraulLunaVorax: so basically what we do in the future is this:13:10
wolfspraula video should always first be created and uploaded in the highest resolution possible13:10
wolfspraulno limits13:10
wolfspraulthe higher the better13:10
aw_wolfspraul, don't know yet just guess (without hard data), just to be verified later like: taking working board and that issued board, taking scopes again. but it could be hard to take many signals in parallel from my current equipment.13:11
wolfspraulthen, if it's more than 1024x768 or so, we will create a scaled down version that matches the window size we will have for the embedded viewing13:11
wolfspraulin that case, instead of one 'download' link, we will have 2 links linking to the smaller and larger video by stating their file sizes (for people who want to download)13:11
wolfspraulaw_: understand, hard. maybe sebastien can give us some ideas, or we send 0x34 to Sebastien.13:12
wolfspraulwe'll see13:12
wolfspraulwe do have a few of those boards, so at some point we probably need to investigate13:12
wolfspraulLunaVorax: sounds like a plan?13:13
LunaVoraxwolfspraul, be also sure when embedding a video in a webpage that your server can deliver the video fast enough so users don't have lags because of their/your connection speed13:13
LunaVoraxwolfspraul, perfect but I have two suggestion13:13
LunaVoraxEmbedded videos = 220kbits/s13:13
LunaVoraxI suggest a limit of 480pixels in height instead of 1024x76813:13
wolfspraulhe, ok :-)13:14
LunaVoraxsorry if I'm too picky :P13:14
aw_wolfspraul, regards to which h/w part I don't know now only once we deeply digging into probe sequencies of several signals to see if any secrets behind.13:14
wolfspraulwe are 'wasting' some download capacity and client load when embedding larger videos and letting the client scale down13:14
wolfspraulbut you were the first one to notice and complain13:14
LunaVoraxMaybe because I'm the first to have weak hardware13:15
wolfspraulit's a tradeoff between maintenance overhead and our actual users13:15
wolfspraulwell the rule is for 1 person to speak up, there's 100 with the same problem who will not speak up13:15
wolfspraulso it's a good thing you brought up there, I think13:15
wolfspraulbut, you will admit, this creates more manual work13:15
wolfspraulso I'll see on which videos I do this extra work :-)13:15
wolfspraulit's a wiki, anybody can help13:15
wolfspraulalso this can be done retroactively13:16
LunaVoraxwolfspraul, that's why I was wondering how to solve that with a script13:16
wolfspraulas long as we first always upload the highest-resolution video first, everything else is optimization and can be done at any time13:16
wolfspraulnah, too much work13:16
LunaVoraxYou sure ?13:16
wolfspraulif even Wikimedia Commons is doing this manually now, I wait until their pain is high enough that they implement it server side13:16
wolfspraulQi is a copyleft hardware project13:16
wolfspraulit's not that I'm dwiddling my thumbs looking for ways to kill my time13:17
wolfspraulpush the Mediawiki guys :-)13:17
wolfspraulor even help them13:17
wolfspraulbut no custom scripts on the qi server, which will add more maintenance work for me13:17
LunaVoraxThat sounds better indeed13:17
LunaVoraxYep I understand perfectly13:17
wolfsprauland plus, this won't be easy13:17
wolfspraulI would extend the pretty neat picture thumbnailing infrastructure they have already13:18
wolfspraulthat just needs to be extended to video thumbnailing13:18
wolfspraulI guess it's not high enough in their huge priority list13:18
wolfspraulso video thumbnailing needs to be done manually right now13:18
wolfspraulwhich is only really an issue for high-res videos, and people on slow links or lower mhz machines13:19
LunaVoraxwolfspraul, I'm asking the problem on #mediawiki :)13:20
LunaVoraxWe'll see if they have a solution13:20
wolfspraullet's see whether you get an answer13:20
wolfspraulthat channel is pretty crowded13:20
wolfspraulyou have a good chance when the full-time san francisco staff are online, which is probably a bit too early now13:20
wolfspraul6 am there now, I think13:21
wolfsprauleven the hackers are sleeping then, except if they have an American roh there... :-)13:21
LunaVoraxOh darn I forgot about that13:22
wolfspraulmediawiki is global, just wait and see. but in my experience in that channel it helps when the americans are awake...13:23
wolfspraulI've fixed a few bugs over the years and I like the project in general. they really do well.13:24
wpwrakrjeffries: (vga tester) heh, that would work ;-)14:15
LunaVoraxWould it be possible to display a little animation when the Nanonote is booting ?14:18
wolfspraulLunaVorax: surely possible but someone needs to implement it14:39
wolfsprauland not everybody would like it probably, for example I just want it to boot as fast as possible14:40
LunaVoraxI was just wondering14:40
wolfspraulbut you can experiment in that area :-)14:40
LunaVoraxwolfspraul, I think my knowledge is too poor for now14:40
LunaVoraxBut sure14:40
Aylahi guys15:15
Aylaone question about gmenu2x:15:15
Aylaon nanonote, the icons of the buttons are shown as "A", "B", etc?15:17
AylaI mean, does the the graphic of the "accept" button on gmenu2x is a button with the "A" or "B" letter? If so, would you want to change that?15:17
AylaI believe you don't use the "A" letter on your keyboard to launch apps15:18
xiangfuAyla, it's only shows 'A' 'B', it should be : cancel=keyboard,104#NanoNote H15:24
xiangfuaccept=keyboard,13#NanoNote Enter15:24
xiangfuthe new gmenu2x input system changed them to cancel and accept I think.15:24
AylaI'm speaking about button icons15:24
Aylathe .png files still show "A" and "B" buttons15:24
xiangfuAyla, oh. sorry. yes. 15:26
xiangfuyes. change to a better one will be great.15:26
rjeffries_wpwrak someone on G+ commented thusly re qi-hardware newsletter I pimped today:19:25
rjeffries_[1 of 2] 19:25
rjeffries_Guess what everyone will notice. Milkymist stuff? Open hardware progress? Nope, nyan-cat. (Of course)19:25
rjeffries_[2 of 2] Ah yes, and this slide is one of the best, indeed: http://en.qi-hardware.com/w/index.php?title=File:Fisl12.pdf&page=2219:26
wpwrak(working the backlog once more. you guys are killing me. go out for the evening and you have a day's worth of stuff piling up on the channel)19:34
wpwrakrjeffries_: (open loop design) that was in reference to your question about the cost of doing purely the "design" (without making any hardware) for a potential ya19:35
wpwrakrjeffries_: i think you underestimate just how unpredictable making hardware can be. e.g., it's extremely easy to misunderstand how a chip works, and fixing the misunderstanding can be difficult. then, a chip's documentation may claim this or that, but the chip behaves differently.19:37
wpwrakrjeffries_: also, sometimes all you can understand from the data sheet is that you have to be careful about certain things, but you need to combine this with hands-on experience to be sure just how careful, and what happens if you aren't (e.g., because doing what they suggests would conflict with other design goals)19:38
wpwrakrjeffries_: and so on. if you're doing a design with components that you're not already very familiar with, you have no chance if you don't make prototypes and check them.19:39
wpwrakrjeffries_: we also have the issue that our component and footprint libraries are not very complete yet. so this is an additional risk factor. again, it's all well and under control if you actually try things.19:40
wpwrakrjeffries_: (G+ responses) heh ;-)19:42
wpwrakwolfspraul: (hw family) characteristics ? but i think it can stand without "characteristics". if you talk about a family, it's kinda implied that it'll be about some of the characteristics of its members19:43
wpwrakDocScrutinizer: (I have a pretty clear idea where "our" sysops are right now and what they're doing) in the tent where all the raucous singing and drinking happens till the early dawn ? ;)19:46
DocScrutinizeractually I guess they are crawling some cable tunnel or sth, at a location next to berlin19:47
wpwrak(M1 demo videos) one problem are the washed-out colors and the blur. when you see the demo videos, they look "nice" but not very captivating. they look a LOT better in real life. kinda like going from VHS-to-YouTube to a DVD.19:48
Artyomkristianpaul: yes, when you read data from couple of registers their value is cleared automatically.20:19
qi-bot[commit] Werner Almesberger: ircstat/: IRC traffic analysis (master) http://qi-hw.com/p/wernermisc/db8c6d520:28
qi-bot[commit] Ayla: Shorted a too long caption on the contextual menu. (master) http://qi-hw.com/p/gmenu2x/5b49bf021:37
qi-bot[commit] Ayla: Added a default config file for the Dingux port. (master) http://qi-hw.com/p/gmenu2x/c5d4cfb21:37
qi-bot[commit] Ayla: Shorted a too long caption on the contextual menu. (2) (master) http://qi-hw.com/p/gmenu2x/7f08e8521:37
LunaFrizzleBtw way21:57
LunaFrizzleI couldn't NOT see you were interested by marcan's OpenLase project21:57
LunaFrizzleIf Qi makes a device that is OpenLase compatible, you can have ALL of my damn money21:58
wpwrakLunaFrizzle: i hope wolfgang records this statement of yours and one day when you're rich and famous, he'll bring it out again ;-)22:03
LunaFrizzlewpwrak, I'll prostitute myself just to get this device ;P22:04
LunaFrizzleOr sell one of my kidneys... sounds familiar ? ;)22:04
wpwrakvery good. i think we could put that to good use :)22:04
wpwrakand the kidney, too :)22:05
LunaFrizzleBack to the serious thing, I still have troubles compiling SDL things on the Ben22:05
LunaFrizzleI need to talk about it to xiangfu22:06
LunaFrizzleI though about a neat thing to do to advertise the Nanonote, I should try to code a demo in the same spirit of the first Macintosh show off by Apple22:07
LunaFrizzleYou know, the one with the Chariots of Fire music :P22:07
LunaFrizzleThe bad side is, in order to show how capable the device it, I should integrate some 3D in it... too complex for me22:08
wpwrakare the SDL problems caused by the build environment ? or are you uncertain how to use SDL ?22:09
LunaFrizzlewpwrak, the code is a simple C hello world with only #include <SDL/SDL.h>. I'm sure there's something wrong with the build environment22:10
wpwrakwhat error do you get ?22:12
marcanLunaFrizzle: there's a hacked up branch of openlase with ASCII art output instead of laser output22:16
marcanI wrote that while on more than a couple drinks during the 27c3 afterparty22:16
marcanto put 3D ascii art on a little PC they had there ;)22:16
wpwrakLunaFrizzle: you should send him a few bottles of cognac ;-)22:18
LunaFrizzleOh ! I didn't saw you were there marcan :o22:27
LunaFrizzleI'm a big fan of your work !22:27
marcan:) thanks!22:27
LunaFrizzlewpwrak, I think I fixed it. I need to compile more complex code to be sure22:32
LunaFrizzlewpwrak, so far I just installed the uclibcxx package22:32
LunaFrizzlemarcan, a few month ago we were still laughing with a college professor about Sony's mistake with the encryption key when using a random number that wasn't random at all22:33
marcanlol :D22:33
LunaFrizzleHe suggested me that it was a trainee's mistake 'cause he couldn't see any other alternative22:34
marcanit's not just Sony, Nintendo botched the Wii signature verification because they strncmp()ed the hash22:34
marcaninstead of memcmp()ing it22:34
LunaFrizzleYeah I'm sure they made that kind of thing also22:35
marcanSony is worse though, Nintendo had good design and horrible implementation. Sony has horrible design and horrible implementation22:35
LunaFrizzleStill it was more impressive since the PS3 has been still untouched for a very long time22:35
marcannobody cared because of OtherOS22:35
marcanI'm entirely serious, none of the hackers that I know did anything until OtherOS was pulled22:35
marcanOtherOS is the only _real_, working PS3 security system, by accident or not :)22:36
marcanor, rather, was22:36
AylaotherOS was a good idea against piracy...22:36
LunaFrizzlemarcan, wow but I wa still hearing people not happy about the fact that they couldn't get full power from the console. That's what Geohot tried to fix initially, right ?22:36
marcannot really, geohot started once they had already pulled it from the slim22:37
LunaFrizzleAh yeah I forgot about that22:37
marcanand OtherOS wasn't really _that_ limiting, the _only_ thing it didn't have was 3D graphics22:37
wpwrakAyla: and pulling it just got everyone very determined to take revenge. very smooth planning on sony's side :)22:37
LunaFrizzleYeah that's it22:37
marcannow we know that GameOS is exactly the same thing with exactly the same interfaces for everything, except the 3D reference to the GPU is nulled out22:37
marcanwpwrak: when they retroactively pulled OtherOS from the Fat they basically sealed their fate hard22:38
LunaFrizzleI might buy a second-hand PS3 when people will thing PS3 is lame because of PS422:38
marcanif there's something worse than not allowing homebrew, it's allowing homebrew and then illegally removing the feature retroactively22:38
LunaFrizzle-thing +think22:38
wpwrakmarcan: yeah, i think we know the answer to whether it is worse to have loved and lost or to never have loved ;-)22:40
wpwrakmarcan: at least as the geeky kind of love of technology is concerned ;-)22:40
marcanoh, I love it when companies do evil stuff, it gives me the motivation to hack their products :P22:41
LunaFrizzlemarcan, fail0verfow is still active or did Sony just pushed you down ?22:43
LunaFrizzleI have to admit I didnot quite understood what really happenned and how are the things now22:44
LunaFrizzle(if I can ask...)22:44
marcanfail0verflow was never about anything in particular, just a group of people doing stuff22:44
marcanthe lawsuit made us go into stealth mode regarding the PS3 and then we got bored of the PS322:45
LunaFrizzleno code release then ?22:45
marcanwe already released everything :22:45
LunaFrizzleWhat ?22:45
marcanthe ps3 tools, asbestos, noralizer22:45
LunaFrizzleDammit I'm always the last one informed :(22:45
marcanthat stuff got released like a week after 27c322:45
LunaFrizzleOh ok22:45
marcanit got pulled later due to the lawsuit, but there are tons of mirrors22:46
LunaFrizzleWhen I heard you saying "It still need a lot of cleaning" I was like "oh shit, not after 32months or so"22:46
marcanwell, it did still need a lot of cleaning, but then the lawsuit happened and we got bored22:46
LunaFrizzleOk, good to know marcan :)22:46
LunaFrizzleOk I understand22:46
marcanand in the meantime lots of other people started going about their own way with the PS322:46
marcanbut basically everything we had/used for the demo was published22:47
marcanand it was in fact significantly cleaner, some of the hacks we used for the demo were... hilariously horrid.22:47
marcanpicture brute-forcing zlib lengths to match the compressed size because we didn't have a working self creator yet22:47
marcanthe tools did come with a working self packer afterwards ;)22:48
LunaFrizzleGood to know :)22:49
LunaFrizzleGood thing I'm broke or that would have make me buy a ps322:49
marcanalso, dhewg (Andre Heider) is working on getting some clean PS3 Linux patches upstream22:50
marcan(to support all the new GameOS stuff)22:50
marcanincluding cleaning up my ugly patches ;)22:50
marcanbut yeah, fail0verflow is just a bunch of hackers doing random stuff, mostly an offshoot of the iPhone Dev Team and Team Twiizers (wii)22:51
marcanthere are other projects in the works, probably nowhere near as media-attention-grabbing as the PS3 stuff, but still interesting22:51
marcan(probably nowhere near as lawsuit-inducing too :P)22:52
LunaFrizzlefound some mirros22:53
LunaFrizzlethere's really few code in fact !22:53
LunaFrizzle(or maybe my mirror is clumsy)22:53
marcanwell, asbestos is public at http://git.marcansoft.com/?p=asbestos.git;a=summary , I never took that down entirely, just filtered out some ancient PS3 exploit stuff that nobody had used anyway22:54
Aylais the PS3 fully open now?22:56
marcanthen there's noralizer, which is just a fairly generic nor reader/writer FPGA thing, and the tools to work with PS3 formats/crypto22:56
LunaFrizzleok :)22:56
Aylait's easy to run homebrew?22:56
marcanAyla: there are many gradations of "open", I don't think anything is "fully open" :P22:56
LunaFrizzleThanks marcan22:56
marcanAyla: if you're running an older firmware, yes. if you're running a newer firmware, no, but that's entirely predictable22:57
marcanevery existing PS3 (except for maybe the most recent batches) can be flashed via hardware to run homebrew though, that's what we "broke"22:57
AylaI thought the official signing key was found, thus it was possible to sign homebrew?22:57
marcanyes, but obviously they changed the key for newer firmwares (and just keep a whitelist of existing games)22:58
marcanthere are certain "absolutes" that you can hit in security22:58
marcanit's nice when there's, say, a vulnerable ROM that lets you exploit existing systems forever in software, but the PS3 has no such thing22:58
LunaFrizzleHehehe people are all asking PS3's firmware on second-hand websites22:59
marcanthe absolute that we did hit is that you can hardware-flash all PS3s to run homebrew22:59
marcanand you can run homebrew on PS3s with (now) old firmware22:59
mthtaking over the firmware update proceess would be the only way to keep control, I guess22:59
mthbut then you'd have to patch every new firmware release22:59
marcanwell, sony hid the new firmware's keys inside by wraping them in bootldr, which is the single last unbroken thing in the PS323:00
mthah, so a patched firmware won't even boot then23:00
marcanyou can't decrypt the latest firmwares (or any games released for them)23:00
marcanit's unbroken because the people who want to break it want to do so for piracy and the people who want piracy tend not to be very good at breaking things in practice23:00
LunaFrizzlemarcan, so if someone break it, Sony won't have anymore places to hide ?23:00
Aylaheh, that's actually true23:01
marcan(bootldr is vulnerable to the same issues, and the keys have been used with the same non-random numbers, but for practical reasons you need hardware to exploit it and not that many people know what they're doing)23:01
marcanLunaFrizzle: correct23:01
LunaFrizzleend of the world or something for Sony then23:01
marcanif someone gets the bootldr keys then Sony cannot hide new firmwares and games from people23:01
marcanbasically they'd entirely lose any chance at protecting newer games23:02
marcansince all the PS3's cryptographic secrets would be divulged, i.e. by necessity anything they do will have to be decryptable by existing hardware and we'd have all the keys23:02
LunaFrizzlecan't say I wouldn't like to see such a thing happen 23:02
marcanwhat they did right now is a hack, btw23:02
marcanup until recently bootldr was only used early on in the boot chain and was largely irrelevant23:02
LunaFrizzleEven if I won't use it, just to see SOny's reaction23:02
marcanbut since it's the only bit left, they wrapped everything else inside the bootldr crypto23:02
marcanit's funny, because we theorized they could do this, and then geohot claimed if they hired him he'd help them "fix" things (he was definitely referring to this idea too)23:03
marcanand of course that's what they ended up doing23:04
marcanit really is their last chance though23:04
mththey're probably hoping it will last until the PS4 launch; it doesn't have to last forever23:04
Aylathe PS4 has been announced?23:04
mthI haven't heard anything about the PS4 yet though, so it's probably not close to launch23:04
marcanthe PS4 will be... interesting23:04
LunaFrizzleAyla, they said they'll put less money in it so far23:04
marcanI highly doubt it will be backwards-compatible23:04
AylaI'm not really following video games news23:05
LunaFrizzleAyla, I don't think it's anymore "video games" at this state. Business maybe23:05
marcanI wonder if they've fired their engineering team or if they're still going to design a hilariously cost-unoptimized design again23:05
marcanthis should amuse the people in here, some gems about the original PS3 hardware design23:05
AylaLunaFrizzle: that's sad, but true :(23:06
marcanthey put WiFi in the PS3, fair enough23:06
marcanbut instead of using a USB or SDIO device like everyone else, they decided to... put a wireless router inside the PS3 and have it act as a WiFi to Ethernet bridge23:06
marcancomplete with its own marvell ARM SoC running eCos23:06
LunaFrizzlewow wtf ?23:06
marcanand since the chipset only has one GMII port, they put an 8-port gigabit switch inside the PS323:07
LunaFrizzleit really is overengineered23:07
marcanone gigE port for the external plug,one GMII port for the chipset23:07
marcanone 10/100 port for the wlan23:07
marcanthe rest, unused23:07
marcanthey use VLANs to separate the traffic23:07
marcanspeaking of the chipset, it has more die area than the Cell23:07
mthsounds like they took a prototype and did a 1:1 translation to make it a product23:07
Aylamarcan: awesome :D23:08
marcanit has a built-in graphics and DRAM subsystem, PCI Express, a MIPS processor, media accelerators, ...23:08
marcanall entirely unused23:08
marcan(it's meant for STBs and TVs)23:08
marcanthe original chipset only had IDE though, no SATA, so they used an IDE to SATA converter for the HDD23:09
marcanthen they came out with a new revision of the chipset, with SATA23:09
marcanbut since it didn't have IDE, and their Blu-Ray drive was IDE, they kept the IDE to SATA bridge... in reverse23:09
Aylalet me guess23:09
Aylaah, ok23:09
mththey had a plan once to use the PS3 as an STB; said the Cell was powerful enough to demultiplex transport streams; but I don't think it ever materialized23:09
LunaFrizzlemarcan, no wonder why they lost so much money in the PS3 (?)23:11
marcanmore funsies, in older PS3s with 4 USB ports they use a built-in USB hub, since the chipset doesn't have enough ports23:11
marcanin newer ones with two ports, they do have enough ports... but they kept using the hub23:11
marcanLunaFrizzle: exactly.23:11
marcanit's by far the worst cost-optimization (or lack thereof) I've ever seen23:11
marcanthey finally got their act together with the slim and removed the switch, the vlans, the wifi router (now it's just usb), the hub, the SATA...23:11
marcanbut that's after losing a LOT of money on the Fats23:12
LunaFrizzlemarcan, you mean the slim is a fixed ps3 ?23:12
marcandesign-wise, it's what the PS3 should've been from the get go23:12
Aylayou bet it's slim now :p23:14
LunaFrizzlemarcan, sorry to have bothered you with the ps3, maybe you wanted to talk about something else23:14
marcanthe funny part is it's not that much smaller than the original PS3 :P23:14
marcanin fact it's deeper23:14
marcanLunaFrizzle: nah, I'm bored :)23:15
Aylasoooo they'll build a 'discount' PS4?23:25
LunaFrizzle"Doesn't play games, just pretend you have a functionnal PS4 to your friends, only 899$"23:26
Ayladon't forget the Blu-Ray2 reader23:27
mthand 3D... or is that hype already past?23:28
wolfspraulwpwrak: ok, Milkymist Hardware Family Characteristics it is - thanks!23:28
LunaFrizzlemth, intra-neuronal connection will be the next big thing... wait no23:28
mthLunaFrizzle: I'll wait a while until they got all the bugs out... ;)23:29
wolfspraulmarcan: did you know that Andrew Zonenberg will work on a MEMS mirror next?23:29
marcanugh, nvidia's blob crashed, did I miss anything? (besides what wolfspraul just said, I got that :)23:30
LunaFrizzleJust some trolling23:30
marcanwolfspraul: a MEMS mirror for?23:31
wolfspraulyou tell me :-) for a laser?23:41
wolfspraulAndrew's project is a home MEMS and CMOS fab23:42
wolfspraulhis process can only produce one die at a time, but still... maybe something useful comes out of it one day...23:42
marcanwolfspraul: huh, wow23:44
marcanthat's amazing23:46
wolfspraulyes it is :-) next steps are comb drive and mirror. no idea where it goes but we try to find any form of use case asap.23:58
wolfspraulif you want some of that stuff for a laser project, let me know and I can try to recreate his fab process and manufacture some :-)23:58
--- Tue Aug 9 201100:00

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