#qi-hardware IRC log for Saturday, 2012-01-28

lindi-_and yeah, schedule does not seem to be called ever00:03
lindi-_I wish openocd supported single stepping...00:07
Action: DocScrutinizer would toss a Lauterbach over to lindi, if it wasn't such an incredibly expensive and closed and dongled tool that lindi-_ wouldn't like it anyway00:10
DocScrutinizerit's really convenient to click the "load" button which implicitly resets the attached DUT, and a few moments later you see the first opcode of ROM-BL highlighted in the listing window, and a "stopped" in the status line00:14
DocScrutinizereven more convenient is to set a breakpoint to the interesting line of your patched code, and when it stops again, click on the two parameters to strcmp() to find you passed a pointer to "LLDEBUG" while the compared item in the list is "LLDebug"00:16
DocScrutinizerprobably saved me like 2h of code analysis today00:17
DocScrutinizernow you could say "duh, use printf() - just does the same for you" - ust LLDebug was the name of the UART where printf() prints to ;-D00:18
DocScrutinizerso you could say I patched printf(), and it'S not a nice idea to use printf(9 inside printf() to debug it00:19
DocScrutinizerbtw Lauterbach trace32 software can do this on ARM core SoC via JTAG00:22
DocScrutinizerso if it wasn't for the dongle, you probably wouldn't even need the Lauterbach hardware00:24
lindi-_seems it is never exiting the interrupt handler00:24
DocScrutinizeronly thing that doesn't work when you use JTAG only is the 512MB backtrace buffer in Lauterbach hw, that allows you to run the code "backwards"00:25
DocScrutinizerat least that's not possible in realtime via JTAG ;-D00:26
lindi-_yeah nice non-free tools indeed :/00:28
DocScrutinizerI thought you'd appreciate to hear what's possible via JTAG if only the free tools were smart enough00:28
lindi-_now I'd like to know how the cpu gets to execute instruction at 0xc02975a400:28
lindi-_it's not executing the instruction before that so something is branching there00:29
lindi-_but according to source there isn't any kind of loop :/00:29
DocScrutinizera clear case for backtrace buffer00:30
DocScrutinizeror auto-singlestepped mode00:30
DocScrutinizerthough *terribly* slow, it should be able to create a backtrace as well00:31
lindi-_hmm, single stepping seems to work after some openocd magic :/00:33
whitequarkDocScrutinizer: gdb has some support for reversible debugging, doesn't it?00:36
DocScrutinizerI admit I'm not a gdb wizard00:36
whitequarkI'll just try it for my arm now00:37
whitequark... only to find out that I do not have it atm.00:37
whitequarkDocScrutinizer: actually, if your arch has a proper gdbserver, then you have half of those features for certain, and maybe have reversible debugging, if it's actually working. so things are not that bad00:39
DocScrutinizerI never said things were bad00:40
whitequarkok, then I misunderstood you00:40
DocScrutinizerjust debugging kernel on arm seems wasn't possible a year ago, since no free (or no at all) version of kdbg available for that00:41
DocScrutinizerlindi-_ is debugging suspend/resume issues on GTA02, not exactly gdb's primary domain00:42
DocScrutinizerand I think I should get a beer now at my pub, so I'll see sth else than LCD screens all day and all night00:48
wpwrakyou could try to find a CRT for home use :)00:49
DocScrutinizermaybe I could invent some goggles/glasses that are made of some special material energized by washing with liquid air from Himalaya, to remove that nasty mind corrupting vibrations from that light that goes thru the liquid crystal. Might be a commercial success story on next esoteric convention00:52
wpwraksounds like a plan00:53
wpwrakthat was a quick beer01:25
DocScrutinizerno, that was the slowest waitress I've ever seen, paired with amnesia and a bunch of drunken assholes 01:26
DocScrutinizerI probably would die from thirst in there01:26
DocScrutinizerthe reward for trying to do weird things which are not related to looking at LCD01:27
wpwrakpoetic justice, swiftly served by the agent on duty, disguised as a waitress01:29
DocScrutinizerdon't forget her troops! a really nice bunch of weird drunken people, one sitting left to me and arguing with the one right to me, face to face. Probably the waitress didn'T even dare to bring a beer for me, no way to place it anywhere01:33
Action: DocScrutinizer gors washing his face01:34
wpwrakyou could have asked one of them to swap places with you :)01:36
rjeffrieswhitequark Jean Claude Whippler of JeeLabs does a lot of home automation stuff, all very open. AVRs are ok/fine for the individual sensor nodes, need something beefier for central coordinator. He alswo recently worked through a design to steal a small amount of energy form 230V mains. is not selling due to liability.05:05
rjeffriesHe uses the HopeRF RFM12B radio. cheap, and it works. I think that's also the device ?? mirko ?? uses for his project 05:07
qi-bot[commit] Maarten ter Huurne: MIPS: A320: Fine-tune PWM backlight settings. (jz-3.2) http://qi-hw.com/p/qi-kernel/0bc374e11:38
qi-bot[commit] Paul Cercueil: MIPS: JZ4740: enable the OpenDingux logo only when targetting the Dingoo. (jz-3.2) http://qi-hw.com/p/qi-kernel/268810111:45
qi-bot[commit] Paul Cercueil: MIPS: A320: Updated defconfig to match Kconfig changes. (jz-3.2) http://qi-hw.com/p/qi-kernel/9c420ed11:45
qi-bot[commit] Paul Cercueil: MIPS: A320: defconfig: Tweak scheduling options to get better performance. (jz-3.2) http://qi-hw.com/p/qi-kernel/d9e5b2b11:45
qi-bot[commit] Paul Cercueil: OpenDingux: initrd: added 'splashkill' program. (jz-3.2) http://qi-hw.com/p/qi-kernel/c9b9fe012:07
wpwrakwolfspra1l: hmm, this could be an interesting product for qi-hw. should have a nice margin when produced in china. (page in german) http://www.bpes.de/de/boxentransformer.html12:52
wolfspraulwpwrak: thanks for the link! the text sounds a little esoteric at first read, but I'll read again13:59
wolfspraulI am in no way worried about finding product opportunities13:59
wpwrak(esoteric) i think it's highly concentrated snake oil :)14:00
wolfsprauland I also don't think there will be anything specific about any country in the manufacturing process14:00
wolfspraulno second read needed?14:00
wpwrakbut you can't beat the margin on a few pieces of what looks like plastic sold for EUR 2100 :)14:00
wpwrak2nd read only for entertainment value14:01
wolfsprauloh that doesn't surprise me. I can't put a percentage figure but A LOT of businesses only survive because at some underlying point, they have these kinds of margins14:01
wolfspraulmy gut feeling says 50% of businesses at least :-)14:01
wpwrakyou can also look at their product catalog. they also have some energy varnish for violins and such :)14:02
wolfspraulok then no 2nd read, I thought it's something serious14:02
wolfspraulthere's a lot of innovation in speakers indeed, that's why I even read it in the first place14:02
wolfspraulbecause I have come across a lot of initially exotic sounding speaker stuff lately14:02
wpwrakah ! sorry :) speakers sounded like just credible enough at first to make it more funny14:02
wolfspraulmaybe I become to open-minded towards 'creative' people nowadays14:03
wolfspraulabout margins, I went through a number of computer/electronics retailers in Germany in recent weeks, and wow, they are changing a lot14:04
wpwrakyeah, though these show more the kind of creativity that gets you executed in china ;-)14:04
wolfspraulall I see are piles of very high-margin accessories14:04
wolfspraulit's like people are stuck with their computers, and try to keep them alive14:04
wpwrakall the rest is mail order ?14:04
wolfspraulso they sell all these crufty little things, extensions, expansions, cables, etc.14:05
wolfspraulthe computers are also there, but they are clearly loss leaders14:05
wolfspraulthe store makes money with accessories, when people come back because something doesn't work :-)14:06
wpwrakyeah, cables have traditionally had interesting pricing14:07
wpwrakdid you have any luck pimping M1 to shops ? (if you got to doing that ... you were once talking about it)14:10
wpwrakthey shuold like it better now that is has midi. more accessories to sell ;-))14:10
wpwrak(accessories) maybe we could offer a special usb-midi cable :)14:11
wpwrak(i.e., USB A for the M1, USB mini B for the MIDI device. perhaps USD 50. after all, it is something a little exclusive that's hard to find elsewhere)14:12
whitequark... wait what14:15
Action: whitequark had to read it three times before he finally understood14:16
wpwrakimportant properties could include: 1) low transmission delay (< 10 ns); 2) ultra-low delay variation; 3) electromagnetic shielding14:16
whitequark4) asbestos-free14:17
wpwrakoh right, GREEN !! :)14:17
whitequarkyou need to add nanotechnologies somewhere14:18
wpwrakmaybe it could come in a sealed bag, sterilized14:18
wolfspraulno pitching yet, for that I have to travel around to visit music/dj shops14:18
whitequarkeverything is better (from the "get money from govt" point of view) with some nanotechnologies14:18
wpwrakwolfspraul: train network down ? :)14:19
Action: DocScrutinizer puts on his frozen-air activated rock-crystal goggles before continuing to read http://www.bpes.de/de/boxentransformer.html on his evil LCD screen14:22
wolfspraulfor the archive records: that link is a hoax14:23
wpwraki still kinda like the usb-midi cable idea, though. you could probably make good money with this. and you wouldn't even have to lie :)14:36
wpwrakas in "tell a lie". lies of omission are a different topic14:37
wolfspraulwhat cable? size-A to mini-B is the most common type14:39
wolfspraulfaderfox comes with a cable, I forgot whether one was included in the icreativ14:40
wolfspraulI will focus on the product, accessories are a distraction now.14:41
wpwrakiCreativ comes with A to mini-B. so does the nanoKONTROL2 (and probably a ton of midi for iGadget devices)14:41
wolfspraulfirst step is to finish the remote kbd idea, that's really nice14:41
wolfsprauland why would that not work with the M1?14:42
wolfspraulthe connector on M1 is the same as the most common connector on notebooks14:42
wolfspraulmini-b into icreativ, the other size into m114:43
wpwrakno no ... i'm not saying that any random usb cable wouldn't work. that would be a lie. but one could still offer a usb-midi cable, no ?14:44
wpwraklike you can get some fancy ethernet cables for audio14:44
wpwrakof course, that would be preying on the technically weak, who may be led to think (once you tell them that such a product exists) that you actually NEED a special usb-midi cable for usb-midi.14:45
wpwrakor maybe they're just not sure and prefer to err on the safe side.14:46
wolfspraulit won't sell14:53
wolfspraulI think you underestimate the amount of work needed to actually sell such cables14:53
wolfspraullet's make a strong core product first, then accessories14:54
wolfsprauland btw, the #1 best accessory for the Ben NanoNote already exists - jane's pouches14:54
rohwpwrak: do you know of a cable which has usb-midi on one end (host) and midi on the other?14:55
rohto use usb-midi devices with stuff that likes real midi14:55
wolfspraulwe've only found the other way round14:55
rohhehe... i know the other way round exists ;)14:56
wolfspraulto connect old-style MIDI devices to a notebook, that exists for 5 USD or so14:56
wolfspraulwe have not found what you mean14:56
rohi was just thinking if i could trigger wpwrak into hacking mode to brew something up *ducks*14:56
wpwrakyou can use a pc ;-)14:56
rohwouldnt that be a job for your 8051?14:57
wpwrakwith a pc, it works great. plug in all the usb-midi you want, add a usb-to-midi dongle, connect them all together with qjackctl14:57
wpwrak(my 8051) it doesn't do usb host :)14:59
wolfspraulabout the led mails... it was so much that I was overwhelmed in replying14:59
wolfspraulI think all is good now - design verified14:59
wolfspraulwhen Adam is back we can proceed to actual schematic entry?14:59
wolfspraulthe only question was the size of grid?14:59
wolfspraulI have no overview over free pins and routing15:00
wolfspraulotherwise you can just pick what you like15:00
wolfspraulif we do well, I can imagine other uses for the led array later15:00
wpwrakyeah. the only things missing are 1) see if there is led vs. ir interference. but i'm sufficiently unworried about that that i think i'll postpone such testing until a day i feel really totally and utterly bored :)15:00
wolfspraulnot needed15:00
wolfspraulworst case that led has to stay off, so what15:01
wpwrak2) pick a location of the leds in the matrix. i'd leave that to adam / the layout guys.15:01
wpwrak3) size of matrix, yes. that can simply be use 3+3 pins, reserve one more15:02
wolfspraullet's have one asking size15:02
wolfspraulotherwise we just spread uncertainty to more people15:02
wpwrak4) picking pins would also be adam / layout guys. we can adapt to anything they pick within ~30 minutes, 25 of which are the xilinx tools doing their thing15:03
wolfspraulthat asking size can come with a note *) can be smaller if needed15:03
wpwrakfor the size, i'd go with 3 x 4. it's not an excessive number of pins and it allows for up to 24 leds. depending on how you populate the matrix, only a 2x4 or 3x3 area of it would be used, i.e., one pin could be unconnected15:04
wpwrak(there would be no physical matrix anyway. just an equivalent topology)15:05
wpwrak5) decide on the exact layout rules for the leds. such as leaving an area for translucency, distance from connector edge, distance from board edge, any exceptions to the rules, etc.15:06
wpwrak(ir led) yes, that worst case gives me great comfort :)15:08
wolfsprauldon't make 5) too complicated15:12
wolfspraulbut yes, sure. I saw your drawing :-)15:12
wpwrakit's one more thing for adam to decide :)15:15
wolfspraulok, so 3*4 is the preferred one15:17
wolfspraulgreat, seems led is exhaustively covered15:21
wolfspraulif we are fiddling with layout, a wish I want to add is to move the memcard or jtag-serial so that the memcard can be fully opened15:21
wolfsprauljust move it a bit, no big change15:21
wolfspraulalso the jtag-serial has one corner cut out slightly, I am wondering whether that is still needed (there was a 2nd batch of jtag-serial board with minor improvements)15:22
wpwrak3*4 also happens to be the maximum size we can support with 6 mA per LED :) anything larger would get a bit more demanding (well, a little. we could go to vast quantities of LEDs without working up a sweat, if we really must :)15:22
wolfspraulif it's still there, that's another thing we can fix15:22
wpwrakyeah, saw the corner. very cute ;-)15:22
wpwrakwhy is this an issue ?15:23
wolfspraulfinally the jtag-serial coult be 1mm or so further away from the acrylic, that would make inserting the cable a tad easier15:23
wolfspraulnot an issue, just a small annoyance15:23
wpwraki would make a hole in the acrylic and make both the cable and its insertion even easier ;-))15:23
wolfspraulthat's another idea15:23
wpwrak"live hinge" anyone ? :)15:24
wolfspraulso I just want to move jtag-serial and memcard a little (1-2mm) to solve those 3 issues15:24
wolfspraulmemcard open, 1mm more clearance to wall, no need for cut corner15:24
wolfspraulsure we can make bigger changes, but that's a separate idea then15:24
wolfspraulI like having the jtag-serial separate now, it's good15:24
wpwrakagain, why is the cut corner an issue ? once you know you need it, it should be trivial to have15:25
wolfsprauleven though for the foreseeable future all boards will have it inside by default15:25
wolfspraulit's a detail that needs to be remembered, and really just a leftover from lack of mechanical overview early on15:25
wpwrakif you want new jtag boards you make backward-compatible, you need to keep the corner anyway15:25
wolfspraulit may actually already be fixed in the 2nd-gen jtag-serial, I don't know15:26
wolfspraulno need because all m1 have one already and I have plenty in stock15:26
wolfspraulI just want to cleanup the board a little, that's al15:26
wolfspraulI'm talking about 1-2mm moves15:26
wpwrakmy M1r4 has an "RC2" jtab board. that has (and needs) the corner15:26
wolfspraulif it's easy, layout should just do it15:26
wolfspraulwe already have two versions of jtag boards15:27
wolfsprauland plenty in stock15:27
wolfspraulthe point is to cleanup the files going forward15:27
wolfspraulhas to start at some point...15:27
wolfspraulsame for memcard blockage15:27
wpwrakthis looks alike a case of fighting yesterday's battles :)15:28
wolfspraulif those little things won't get done, no big deal either15:28
wpwrakthe memcard i get15:28
wpwrakthe jtag has already sailed15:28
wpwrakif you move it and anyone makes cards that don't have the corner cut out, they won't work in rc2/rc3 boards15:28
wolfspraulI would move the headers on m1 a little so that the board doesn't need the cut corner15:28
wolfsprauldoesn't matter, since all rc2/rc3 boards have jtag-serial already15:29
wpwrakif you keep on cutting out the corner forever anyway, why bother moving ? :)15:29
wolfspraulthat's what I want to cleanup, the cut corner need15:29
wolfspraulgee, small detail :-)15:29
wpwrakthe jtag board could need replacing due to loss, defect (real or imagined)15:29
wolfspraulit should take layout less time to move this by 1mm than our discussion here :-)15:29
wolfspraulI have plenty in stock15:29
wolfsprauland I don't even plan another run without cutting the corner15:30
wpwrakyou're trying to solve a problem you've already solved :)15:30
wolfspraulbut I want to move the headers by 1mm :-)15:30
wolfspraulso you don't make the same argument when we are at R515:30
wolfspraulif routing says it's easy, i will ask them to move15:30
wolfspraulalso memcard opening15:30
wolfsprauland also 1mm extra clearance towards wall15:30
wolfspraula bigger thing would be whether to move the memcard to the side15:31
wolfspraulI was worried you bring that up :-)15:31
wolfspraulbut I think we should not do that now15:31
wpwrakwell, probably no harm in it. just totally pointless :)15:31
wolfspraulnot in R415:31
wpwrakkinda like the endless renames in kicad ;-))15:31
wolfspraulyou agree about the memcard?15:32
wolfspraulthat's highest value of those 315:32
wolfspraulsecond is 1mm extra clearance to wall15:32
wpwrakthe memcard is a disaster. i don't think you can make it worse ;-)15:32
wolfspraulthird is remove need to cut corner15:32
wolfspraulI think we agree actually, on everything. *if* they are moving the jtag-serial headers already, they can also consider that corner thing.15:32
wpwrakso any change to the memcard is good :)15:33
wolfspraulbecause you bet that was not intentional...15:33
wolfspraulit has very small practical value now, only cleanup15:33
wolfspraulbut that most minor thing is what we discuss about15:33
wolfspraulif routing says it's difficult in that area, or the power stuff is too close, or whatever, then I wouldn't touch any of this15:33
wolfspraulI think the entire current situation is not bad, including memcard15:33
wolfspraulI see the memcard as some sort of internal storage15:33
wolfsprauland I ship all products with 2gb formatted and installed there15:34
wpwrakwhile changes to jtag don't seem to do more than change for the sake of changing. kinda like when you go to the supermarket and find your favourite butter with a slightly different logo and somewhere in the corner they printed "NEW LOGO !"15:34
wpwrak(i'm not kidding. i actually saw something like this. though it wasn't butter)15:34
wolfspraulthe corner cut was not intentional15:34
wolfsprauljust believe me15:35
wolfspraulanyway this is indeed the most minor of those 3 small mechanical annoyances15:35
wpwrakoh, i get that :) just, now that it's there and done, it has no further cost15:35
wpwrakthe pcb-cutting cnc machine won't mind that little turn :)15:36
wolfsprauland you have no problem discussing whitespace in sources...15:36
whitequarkwpwrak: how do you think, how long will it take to make M1 Rlast? and how long will it be until it will be obsolete?15:36
wpwrakRlast ?15:36
wpwrakah, final :)15:36
wolfsprauldon't understand your question15:36
whitequarkR1, R2, ..., R5, Release.15:36
whitequarkyeah, final revision15:36
wolfspraulthe more it sells, the more revisions15:37
whitequarkah, so that M1 revision isn't like release candidate. my bad.15:37
wolfsprauleven after a Milkymist Two or some other product that replaces M1 comes out, there may still be new M1 revisions15:37
wpwrakit's like cars. you make a new model every year. you stop the line only when it stops selling.15:37
wolfspraulwhitequark: yes, but you noticed an important detail :-)15:37
wolfspraulthe 'candidate' was indeed confusing15:37
wolfspraulwe realized that recently15:38
wolfspraulso we stopped saying RC1, RC2, RC315:38
wolfsprauland instead now it's just revision, or R415:38
wpwrakhave we actually announced the name change on the list ? i'd still consider that our "official" channel15:38
whitequarkI'd personally suggest V4, or V1.415:38
whitequarklike router vendors do15:38
whitequarkthat's way less confusing15:38
wolfspraulwell, we've just settled on R415:39
wolfspraulbut maybe V5 in honor of whitequark ? :-)15:39
wolfspraulthen W6 in honor of Werner15:39
wolfspraulhow about that?15:39
wpwrakit was M1rc3 (confusing). wolfgang suggested M1r4, which is better. i think i kinda like the idea of a M1.4. been thinking about that for a while, too.15:39
wpwrakeach time i type "M1r4", it feels a little wrong15:39
wolfsprauloh god, version number discussion15:40
wolfspraulI am relaxed about this15:40
wolfspraulif it makes werner happy, I am also ok with M1.415:40
wolfspraulthe main thing on the sales side is product literature, manual etc.15:40
whitequarkyeah, M1.4 is fine too15:40
wolfsprauland they all talk about 'milkymist one' only anyway15:40
wolfspraulthe rest is 'internal' (open for Qi of course), so we can communicate about improvements15:40
wolfspraulso as long as the numbers make sense, including the old ones, I don't care which words or characters there are15:41
wolfspraulmeaning that... M1R4 can also be M1.415:41
wpwrakwe have another thing, just "Milkymist". that would be the whole category. e.g., i could imagine an M2 with a >= 7" panel15:41
wolfspraulthat's another discussion15:42
wolfspraulthe Milkymist brand is confusing, overloaded a bit15:42
wolfspraulbut I wouldn't change that right now, the brand needs more visibility first15:42
wolfspraulso I tend to say Milkymist SoC or Milkymist One15:42
wolfspraulor sometimes I just say Milkymist when talking to new people, since I mainly want to catch their attention15:42
wolfspraulwpwrak: btw, about the 1.4 thing15:43
wolfspraulthe current silkscreen says Milkymist One RC3 <datecode>15:43
wpwrakdatecodes are great to have15:43
wolfsprauloh definitely15:43
wolfspraulso you suggest Milkymist One M1.4 <datecode> ?15:43
wolfspraulyou want to repeat the "M1"?15:43
wpwrakhmm, dunno15:44
whitequark"Milkymist One V1.4 <datecode>" ?15:44
wolfspraulwith R4 the flow is better Milkymist One R4 <datecode>15:44
whitequarkor V415:44
wolfspraulsince we are deep in opinion land, I think V is geeky and cheap15:44
wpwrakyes, for the long form, R4 looks nicer15:45
wolfspraulI believe if it's R, the product will take off! :-)15:45
wolfspraulthere's a story why the founder of Roland chose that name15:45
wolfspraulone reason supposedly was that there were very few/no other companies starting with 'R', and that would give Roland a very visible spot in alphabetical listings at tradeshows and exhibitions15:46
wolfspraulthat's the legend...15:46
wpwrak(roland) before, they called themselves "Eyjafjallajökull", but there were never any orders for them in the mail ?  :)15:46
wpwrakheh :)15:47
wolfspraulI continue with "R4" unless some other active contributor insists on something else15:47
wolfspraulI never type M1R4 anyway15:47
wpwrakthere's also the eternal battle for having the most "A" at the beginning of the name15:47
wolfspraulbecause the context is clear15:47
wolfspraulyeah but that's cheap15:47
wpwrakvery :)15:47
wolfspraulbut I can see the point of being the only one under 'R'15:47
wolfspraulif the story is true15:47
wolfspraulso about R4 - leds are settled15:47
wolfspraulwpwrak: infinite thanks, really!15:48
wolfspraulI hope this turns into a multi-billion dollar LED spinoff, one day15:48
whitequarker, what?15:48
wpwrak(M1R4) you should type M1r4 anyway ;-)15:48
whitequarkcan I see the history of led thing?15:48
wolfspraulit's mostly in #milkymist and the milkymist devel list15:48
wpwrakwhitequark: are you on the #milkymist list ?15:48
whitequarkwpwrak: nope15:48
wolfsprauljust have a bit here today so the Qi crowd doesn't feel excluded15:48
wpwrakmilkymist list, without #15:48
wolfspraulmost likely there would have been more feedback on the Qi list...15:49
wolfspraulso leds are settled15:49
wolfspraulI will add my wish item of moving the jtag-serial header slightly to address the issues above - if easy15:49
wpwrakwhitequark: for some visual impressions and random stuff: http://downloads.qi-hardware.com/people/werner/m1/leds/15:50
wolfspraulthen we have the expansion system15:50
wpwrakwhitequark: to make sense of it, you'll have to read the list, though :)15:50
wpwrak(qi crowd excluded) boosting the traffic for the monthly statistics ? :)15:51
wpwrakyup. still needs more specs. but ... the part adam needs should be pretty much covered15:51
wolfspraulquality stuff will find its way15:51
wolfspraulyou mean the leds?15:51
wpwrakthe expansion header15:51
wpwrakor header(s)15:52
wolfsprauloh, we settled on the pins for the second header?15:52
wpwraki'm fine with the pins adam uses15:52
wpwrakthere are two clock pins too (i think), so in case we need any of these, we'll have them15:52
wpwraknobody mentioned any other special pin features to take into account, so maybe there are none we should care about15:53
wpwrakthe ones we do take into account are: voltage (all 3.3 V), differential pairs, and having > 1 clock pins15:54
wpwrak(up from 0 clock pins in M1rc3 :)15:54
wpwrakwhat i'm not so happy about is that J21 is upside down. but that's too late to fix.15:55
wpwrakand we should probably keep J22 like this too, for consistency15:55
wpwrakbetter consistently bad than inconsistent :)15:55
wolfspraulupside down?15:56
wolfspraul>1 you mean >=2 ?15:56
wolfspraulpeople don't mention special pins because almost nobody is using the header now15:57
wolfspraulchicken & egg problem, like so many on m115:57
wolfspraulbut I agree, the above looks right15:58
wolfspraulwe do factor in what we've learnt from kpaul and others - clock & differential pairs15:58
wolfsprauland if more people learn more things, we will factor them in in the future...15:58
wolfspraulha ha, I like how you reply to the html mail in this ugly way15:58
wolfspraulspan span span span15:58
wolfsprauldiv div div15:59
wolfspraulspan span15:59
wolfsprauldiv div div15:59
wolfspraul(the "yes" was my actual message)15:59
wolfspraulknowing you I'm pretty sure you did this on purpose :-)15:59
wpwrak(special pins) well, also people who know that there special pins can mention them. they don't have to actually use them. but since nobody said anything, i suppose it's fine. else, we'll eventually find out ;-)16:01
wolfspraulyes, fully agree16:01
wolfspraulR4 looks like a huge step up from R316:01
wpwrak(html mail) obviously ;-)16:02
wpwrak(j21 upside down) in the schematics, it's shown with ground at the bottom, which is a common configuration (of the connector i mean, not just of the presentation)16:04
wpwrakin the board, you notice that the ground pins are towards the center of the main pcb. if you work on such a board, you'll probably consider the edge of the M1 as your "bottom". thus, the connector will appear reversed.16:05
wolfspraulwhy not rotate the header?16:05
wpwrakwe can ease the pain by rotating it in the schematics or in the supporting documentation16:06
wolfspraulor on the board16:06
wpwrakif you rotate now, it'll be incompatible with rc2/rc316:06
wolfspraulshow the layout folks some respect16:06
wolfspraulwouldn't it just be upside down?16:06
wolfsprauland plus, we are going from male to female anyway16:07
wolfspraulwithout having found an easy-to-point-to adapter cable or piece yet16:07
wpwrakwell, ledm would still fit, even if rotated. not as elegantly, but yes. so 100% of my existing boards would be compatible ;-)16:08
wpwrak(adapter) just take a header16:08
wpwrakthe typical header has ~5 mm mating pin on the top and ~2.5 mm pin for soldering at the bottom. those 2.5 mm are just enough to mate with a female socket. it won't be great mechanically, but you get a contact16:10
wpwrakof course, this gives you a bit of a stack16:10
wolfspraulbut if you rotate your board, it would fit even on rc2/3, no?16:10
wolfspraulI think we can go for reversal of the physical header, no problem16:11
wpwrakledm can be rotated, yes. it'll sit on the dvi connector, but that's okay. there's nothing on the bottom16:12
wolfspraulok, so let's rotate the header?16:12
wpwrakof course, future boards may be more difficult16:12
wpwrakif we do it, it's now or never :)16:13
wolfspraulnot really because we already know that most <R4 boards are in the hands of developers or people that are very strong technically16:13
wolfsprauland I have no problem at all offering ugprade options to the few that are not, if this should ever become an issue16:13
wolfspraulI think any argument for not rotating the header is quite hypothetical16:14
wolfspraulwith so many changes we are making already16:14
wpwrak(future boards) i was thinking of making some experimental board and being able to try it out on an rc3. as the rc3 ages, it also becomes attractive for more risky experiments ;-)16:14
wolfspraulsure but that is still possible, even with rotation, no?16:14
wolfspraulif that makes you feel uneasy or slow down, then let's not rotate16:15
wpwraklet's float the idea on #milkymist or the list ?16:15
wolfspraulyou can but there will be 0 responses16:15
wpwrak(possible) yes, but you may have to design specifically for rc316:15
wolfspraulwe can also leave the header unrotated16:15
wolfspraulI can see arguments either way16:15
wolfspraulbut as you said, no more changes after R416:16
wolfspraulotherwise we really completely screw up any expansion idea16:16
wpwrakyeah :)16:17
wolfspraulso maybe no rotate?16:17
wolfspraulyour pick16:17
wolfspraulyou are *by far* the one who gave the expansion system the most thought and practical usage16:18
wolfsprauland yes, if we leave J21 the way it is now, then j22 should be the same and we update schematics and documentation16:18
wpwrakno rotating on the pcb is the safer choice. sometimes, cowardice is the better part of valor ;-) but we can still float it on the list16:21
wpwrakit's also more democratic ;-)16:21
wolfspraulsure, fine by me. do I understand this right: if list is silent we leave as-is, and update documentation?16:22
wpwrakok :)16:25
wolfspraulcool, so that was all for R4?16:36
wpwrakhow about connecting 5V on J21 via a header ?16:37
wpwrakwith jumper16:37
wpwraki.e., a cheap "switch"16:38
wolfsprauldon't understand. you mean a switch between 3.3V and 5V?16:38
wpwrakno, 3V3 would always be there. but we'd get the ability to disable 5V.16:39
wolfspraulsome pins are 5V? wait, checking16:40
wolfspraulah yes, 1 and 216:41
wolfspraulI have no idea. You hear people saying they want 5V, others are saying they want lower voltages, even below 3.3V16:41
wolfspraulthe default is to leave the jumper open?16:42
wolfspraulso 5V only after closing the jumper?16:42
DocScrutinizermoo wolfspraul wpwrak 16:43
wpwraki'm writing a mail with the rationale ...16:43
wolfspraulhi there :-)16:43
wpwrakgotta boost the list traffic stats ;-)16:43
wpwrakhmm, what's the sound bats make ? chirp ? :)16:43
DocScrutinizerwhat you're messing up right here? ;-D16:43
DocScrutinizerwpwrak: depends on converter, here they ROAR16:44
wpwrak(messing) keeping 5V away from the FPGA's 3.3V domain :)16:45
DocScrutinizersounds like a good idea :-D16:45
wpwrak(roar) naw, our bats are small16:45
wolfspraulyou mean if someone shorts 1 and 3, the fpga and board are toast?16:45
wpwrakthat could happen, too16:46
wpwrakor bring 5V to one of the I/O pins16:46
wolfspraulhow about bringing out lower voltages like 1.8, 1.2 ?16:46
wpwraknot sure what happens if you short the 5V and 3V3 rails. maybe that's actually survivable because everything is then at a higher level16:47
wolfspraulI somehow doubt that, but ok16:47
DocScrutinizernot for long16:47
wpwraki wouldn't bother with that16:47
wolfspraulwell I have nothing against a jumper (default off) to disable 5V16:47
wolfspraulof course it's more work16:47
wpwrakperhaps we can have some use for the rc2 wolfgang is replacing with newer boards ;-)16:47
DocScrutinizerif you're concerned, you usually use crowbar circuits16:47
DocScrutinizerand fuses16:48
wpwrakDocScrutinizer: are you on the milkymist mailing list ?16:48
wolfspraulwhat's the proposed pin structure for J22 ?16:48
wolfspraulanother 2 gnd, 2 3V3, 2 5V ?16:48
DocScrutinizernope, and I'm happy with it16:48
wpwrak(j22) 2 x GND, 2 x 3V3, the rest i/o16:48
DocScrutinizerI got just a few lists with several 1000 unread mails already16:48
wolfspraulQi is far bigger than Milkymist and has a more diverse set of people on the list16:49
wpwrakwell, you know the schematics. the issue in question is the to of the expansion header J21 (page 3)16:49
DocScrutinizerlinkie to schem pdf pls16:49
wpwrakthere we have the 5 V rail, right next to it the 3.3 V rail, and then a lot of 3.3 V FPGA I/Os16:49
wpwrakbookmark this :) http://milkymist.org/mmone/rc3_schematics.pdf16:49
DocScrutinizermeh maybe I got it somewhare under my 587 bookmarks already ;-)16:50
Action: DocScrutinizer missing a search in bookmarks, heard FF got sth like that16:50
wpwrakyeah, bookmarks are not all that useful :)16:53
DocScrutinizerJ21 has no 5V?16:53
wpwraki have a little HTML page i keep on editing. would actually be easy to turn this into a smarter bookmark page.16:53
wpwrakpins 1 and 216:53
DocScrutinizerooh it has16:53
DocScrutinizerprety please consider changing VDD symbols from upside-down GND symbol to a circle symbol!16:54
Action: DocScrutinizer notices the "GND" symbols are actually PE symbols there16:56
wpwrakwrong tree ;-)16:56
wpwrakthe plan is to migrate the schematics to kicad anyway. soon. like next week.16:56
wpwrakon that occasion, a number of perversions can die :)16:57
DocScrutinizerI'd happily help killing them :-D16:57
wolfspraulnext week?16:57
wpwrakexcellent. we have our first volunteer ! :)16:58
wpwraktoo soon ? basically once adam has processed the backlog16:58
DocScrutinizerI just siign the execution orders though :-)16:58
wolfsprauloh sure16:58
wpwrakyou;ll make a fine military governor ;-)16:58
wpwrakbut yes, perhaps next week is to ambitious. there is a bit of a pile awaiting him already :)16:59
DocScrutinizerI already got upset about illegible GTA04 schematics, and about goldelico not even offering a way I might edit this mess17:00
DocScrutinizerthey got no GND symbols at all, rather wire up GND like any other trace17:00
wpwrakexcellent ! ;-)17:01
DocScrutinizerconfuses the living hell out of me17:01
wpwraklearn from the rookiest of rookies ;-)17:01
wolfspraulDocScrutinizer: you want to help with the kicad schematics?17:01
wolfspraulI'll pray 5 times a day towards any direction of the sky you are telling me to :-)17:02
DocScrutinizerwolfspraul: I'd maybe take some fun in doing so, alas I am afraid kicad is no fun to work with, esp on a laptop with ~300MB free space (actually less than the 5% reserved for root-only)17:03
wolfspraulmy installed size is 102 MB17:05
DocScrutinizernew laptop is on my ToBuy list for next week17:05
wpwrakkicad isn't commercial bloatware ;-)17:05
DocScrutinizer(J21) so what's wrong with it?17:05
DocScrutinizerfor sure my choice to place the GND, 5V, 3V3 would have been slightly different17:06
wpwrakthe risk of accidently bringing those 5V somewhere they shouldn't go17:06
wpwrake.g., short to the neighbouring 3V3. or to any of the 3.3 V I/Os17:07
wpwrakjust described it here: http://lists.milkymist.org/pipermail/devel-milkymist.org/2012-January/002655.html17:07
DocScrutinizerthe latter seems unlikely to happen, the first I'd tackle by placing GND *between* 5V and 3V317:08
wpwrakit's not just shorts by metal objects, but could be traces on the board17:08
wpwrakthis is for experimental boards. so they can have errors. or they could also simply be mis-inserted.17:09
DocScrutinizershorts on traces caused by faulty PCB can never be handled in any sane way by design17:09
wpwrakcorrect. that's why i'm suggesting to keep 5 V away if we don't need it17:09
DocScrutinizerfor mis-inserting plugs, you usually cut one pin and plug one "hole"17:10
wpwraki'm not worried about shorting anything there to GND or to 3V3. that's survivable, often indefinitely.17:10
wpwrakthat assumes you use proper connectors :)17:10
DocScrutinizeralso a common best practice is to place GND pins so the plug's short between all the GND contacts would short VDD to GND when plug inserted wrong way round (180° rotated)17:11
wpwrakfor DIY/prototypes, improvisation is quite likely, so i wouldn't count on people to have the "right" connectors17:12
DocScrutinizerfor that you can't count on anything, so as well just forget about the whole issue17:12
DocScrutinizerapart from separating 5V from 3V3 by one positions so shorting the pins by metal opbjects gets unlikely17:13
wpwrakyes, short-on-rotation is good. seems to be a bit tricky here, though. or not ?17:13
DocScrutinizer5V, GND, 3V3, IO [...]17:14
DocScrutinizer(tricky) not that much if you'd use 4 GND and only one for 5V and for 3V3if 17:14
DocScrutinizerer, where from is that trailing "if"?17:15
whitequarkDocScrutinizer: sorry what? why is it a best practice to short VDD to GND?17:16
DocScrutinizerread context17:16
wpwrakwe'd rather not reassign pins. since that would break compatibility with rc2 and rc3 boards already in the field17:16
wpwrakwhitequark: if you short the rails, the short-circuit protection of the regulator will kick in and you just power down17:16
DocScrutinizerget a crowbar on 3V317:16
wpwrakhow to do this without reassigning pins ? :)17:17
whitequarkwpwrak: ah ok, that's how it should work theoretically then17:17
wpwrakwhitequark: got some cases where it didn't ? ;-)17:17
DocScrutinizerso same short-circuit shutdown would kick in when shorting 5V to 3V317:17
whitequarkolimex did it with their jtag. plug the connector wrong way and their LDO explodes17:17
whitequarkalmost literally17:17
DocScrutinizerwpwrak: electronic crowbar circuit on 3V3 rail17:18
wpwrakdesigned according to military standards, including the self-destruct :)17:18
wpwrak(crowbar) but that still wouldn't help with 5V reaching the I/Os17:19
wpwrakdon't you like the simplicity of being able to just disable 5V with a jumper ?17:20
DocScrutinizerwpwrak: it might, when IO got proper clamp diodes to 3V317:20
wpwrakit's the clamp diodes i'm worried about :)17:20
wpwraktiny little clamp diodes17:20
DocScrutinizeryeah sure17:20
DocScrutinizerthey should survive a surge to charge up 3V3 buffer C to the point where crowbar kicks in and ters down both 3V3 *and* 5V *directly*17:21
wpwraksounds like a major redesign of the power supply circuit17:22
DocScrutinizerjust add a synced crowbar circuit on both 5V and 3V3 rail17:23
wpwrakcrowbar .... "No records match your search criteria" says digi-key :-(17:23
DocScrutinizerno crowbar chips out there17:24
DocScrutinizerthyristor, Zener, R17:24
DocScrutinizerthyristor, Zener, 2 R, C17:25
wpwrakhmm, the 2nd circuit looks scary. the 1st isn't so bad. but i notice that it works with 12 V. that may explain some of the simplicity.17:27
DocScrutinizeryou'd need some sync as well to trigger 5V crowbar when 3V3 engages17:27
wpwrakphew. does sound major.17:27
DocScrutinizergood EE isn't for free on BOM17:28
wpwraki'm more worried about EE-time :)17:28
wpwrakand design risk17:28
DocScrutinizerbetter get some positions in BOM for *that* rather than for useless inductors etc17:28
wpwrakbut for an imperfect solution, a jumper on 5 V doesn't sound so bad, does it ?17:29
DocScrutinizerdesign risk?17:29
wpwrakyeah, these inductors where evil ;-)17:29
wpwrakrisk of a fuckup. wouldn't be the first :)17:29
DocScrutinizeryou can remove the whole circuit without any impact on normal operation, so what's the "risk"?17:29
wpwrakrework :)17:30
DocScrutinizerjust make whole block "NC" and risk killed ;-P17:30
whitequarkwpwrak: what's with those inductors?17:30
wpwraknaw, while i think the whole power supply circuit needs some overhaul, r4 may not be a good moment for it17:31
wpwrakwe don't want to delay that thing forever. and opinions on the importance of the expansion header are already divided17:31
wpwrakso making this the source for another week of discussion and experiment wouldn't be popular17:32
DocScrutinizera "meh" for me. You asked me, I give a proposal. Up to you to deal with orga crap17:32
wpwrakhence my proposal for a jumper :)17:32
DocScrutinizerjumper for sure can't hurt17:32
wpwrakOVP crowbar idea duly noted :) that's a good item to have on the requirements for the future proper power supply circuit. ideally already as part of the chip. but we'll see.17:33
wpwrakwhitequark: some ground planes are connected with inductors. e.g., the audio section with the main part of the circuit.17:34
wpwrakor, rather, were :)17:34
wpwraknow these inductors have yielded to a significant amount of tin17:35
wpwrakkewl, already two votes for the jumper ;)17:35
whitequarkwpwrak: hm. I heard some guidelines recommend that. What was wrong with them?17:36
wpwrakDocScrutinizer: and you agree that 5 V reaching the 3.3 V rail or the I/O pins in the 3.3 V domain is generally bad news. i.e., it's not just some theoretical risk i'm imagining17:36
wpwrakwhitequark: i'm glad you asked :-) connect audio equipment. a little voltage surge goes through ground (they're at different "floating" potentials). the inductor won't let the surge through. so it goes through the audio codec in one way or another.17:38
whitequarksounds like fubar17:39
wpwrakwhitequark: audio codec hates that and crashes. luckily, it takes no permanent damage. system clock happens to be provided by audio codec. thus whole systems hangs. VJ is not happy ;-)17:39
whitequark>no permanent damage17:40
whitequarkI guess it depends on the surge energy...17:40
wpwrakyeah, i was lucky :)17:40
wpwraki would have kinda hated to burn my brand new M1 already on the first day17:41
DocScrutinizerwpwrak: yes, I agree on 5V is a bad thing on 3V3 domain, and should never happen17:41
wpwrakkewl, thanks !17:42
DocScrutinizerand risk to get shorts on expansion cables would cause exactly that to happen when 5V are on a connector that has lots of 3V3 stuff is not negligible17:42
DocScrutinizerideally your 3V3 IO is protected against such incidents, but yes... I know17:44
whitequarkDocScrutinizer: I heard of a cheap way to do overvoltage protection. Basically, to protect a 3.3V max input you place a 3.3V (or slightly bigger to account for imprecision) zener on it. Zeners always fail short, and when it fails short, your power supply protection kicks in. Afterwards, you replace failed zeners.17:48
whitequarkDoes that work?17:48
DocScrutinizerusually yes18:05
wolfspraulso the end result is we do the jumper, and we leave it open by default (but jumper seated there)18:06
DocScrutinizerbut Zener are no ideal perfect components, so you might see overvoltage nevertheless18:06
whitequarkDocScrutinizer: thanks18:07
wpwrakwolfspraul: perfect18:07
wolfspraulyeah well, work work work. but what wouldn't we do for THE FUTURE18:11
wpwrakit's probably the easiest of all the changes in r4 ;-))18:11
wolfspraullet's see18:13
DocScrutinizerthe reluctance to do changes on new hw rev basically defeats the whole purpose of getting a new hw rev at all18:16
DocScrutinizerthis old OM notion of "well, this maybe is a bug, but until now it didn't cause major issues so let's keep it this way it's now, rather than introducing new untested stuff into next hw rev" somehow always bewildered me18:18
wpwraknot at all. it's a question of what gets improvements to people as efficiently as possible18:19
wpwraknobody benefits if we have the perfect design but it takes us ten years to finish it18:19
DocScrutinizerit's a question of self esteem and professional experience whether you think you could fix a bug, or are incompetent to a degree where your fix would rather make things worse than better18:20
wpwrakoh, i'm confident we can do anything :)18:20
wpwrakgiven enough time and money :)18:21
DocScrutinizerworkload for *implementing* an existing improvement mustn't be a major concern, otherwise you again picked the wrong profession18:21
wpwraksee, there is only a finite number of M1rc3. and they're selling. when wolfgang runs out of them, would it be better to sell M1r4 with some improvements or make more M1rc3 without any of them ?18:21
DocScrutinizerdepends on the workload vs benefit from a new revision, Often it's better to produce some more of the old rev, and get a real improved version later on18:23
DocScrutinizera major fail in procedures at OM18:23
wpwrakwell, rc3 had a number of nightmarish issues :)18:23
wpwrakhere, i prefer the "release often" approach over the "it's done when it's done" approach. making hardware is slow enough as it is.18:24
DocScrutinizerhaha, it only gets much slower by this release often approach18:25
wpwrakcorrect. but you're making progress.18:25
DocScrutinizerit's not like software where you simply hit a button to release18:25
DocScrutinizerthis misconception was another reason why OM hw release policy and procedures been so therribly fsckdup18:26
wpwrakalso, in case you really mess up, you have a previous version to fall back to that's reasonably close, not something already completely prehistoric18:26
wpwrakOM also suffered from the "this is our last chance to do it" syndrome18:27
DocScrutinizerwhich is somehow identical18:27
wpwrakif you just accept that there will be, say, two hw revisions per year, you can be a lot more relaxed18:27
wpwrakif it doesn't make it this time, you'll only have to wait a few moremonths18:28
DocScrutinizerat OM Sean more thought of two *products* per year18:28
wpwraknot like several years. or until your next life :)18:28
wpwraklet's just say that he had a great many ideas ;-)18:29
wolfspraulwow your voting works better than I thought20:25
wolfspraulmehr Demokratie wagen :-)20:26
DocScrutinizerthere's been any votes lately?20:28
wpwrakand unanimously ! ;-)20:41
wolfspraulthat was about whether we should rotate J21 or not20:50
wolfspraula minor issue and I suggested Werner just decides what he likes better, but Werner wanted to ask the list (where I thought nobody would care)20:50
wolfsprauland I was wrong, already 3 answers by now, all favoring to keep the layout as-is and fix the schematics/docs20:51
wpwrakin general, the simpler the issue, the more people will offer their opinion :)20:54
Action: mstevens attempts to reflash a long neglected nanonote20:54
Action: lindi-_ still struggles with suspend bug :/20:55
lindi-_currently it seems that an irq occurs even though it has been masked and __irq_svc does not cope with the unexpected situation and instead loop forever and never returns from the interrupt handler20:56
mstevensooh you guys improved things so much since last time I flashed21:30
wolfspraulmstevens: thanks! :-)21:56
wolfsprauland I think we got 10% of what the Ben could do out of it21:56
wolfspraulthat bugs me, but well, just steady improvement is fine21:56
wolfspraulthe Ben could be a really great notebook companion, on a day when lugging around the big thing is a bad idea21:57
mstevenswolfspraul: there are like actual applications now21:57
GNUtoohi kristianpaul 23:26
GNUtoodid you already heard of MEIF?23:27
kristianpaulGNUtoo: hey there23:27
GNUtoobasically it's a GPS protocol present on some phones23:27
GNUtoothe CPU has to compute the fix23:27
GNUtoosome people are interested about reversing it23:28
GNUtoolike the new main replicant developer23:28
kristianpaulis not another network to improve gps accuracy?23:28
GNUtoo+ another person in #openmoko-cdevel(morphis)23:28
kristianpauloh interesting (reverse)23:28
kristianpaulwhat hardware does it need to work?23:29
GNUtooso we might share work and/or knowledge23:29
kristianpauli mean receiver part23:29
kristianpaulbut dont you want to implement a gps sofyware receiver from scratch? dont they?23:29
GNUtooI don't know, MEIF is present at least on some phones like some samsung ones or the nexus S(our target)23:29
kristianpaulif you have more information about hw part, IF, etc..23:30
GNUtoobut the interesting part is that it get the raw data, and compute the fix on the phone's CPU23:30
kristianpaulyes this sounds very famiiar to RRLP23:30
GNUtoowe don't have something working yet but as soon as we have the init sequence we should share infos/code23:30
kristianpaulwhat about hw?23:31
kristianpaulwhat chip does it use?23:31
GNUtooah I remember now23:31
GNUtoobroadcom something23:31
kristianpaulhmm, still hard to guess ;)23:31
kristianpaullets see23:31
GNUtoobcm4751 it seem23:32
kristianpaulwich phone have it?23:32
GNUtoothere is a datasheet floating on the net23:32
GNUtoosome samsung phones23:32
GNUtoolike the nexus S 23:32
GNUtooor other phones23:32
GNUtoolook for MEIF on xda23:32
kristianpaulargh, cant find anythoing23:35
kristianpaulfirst time look at that xda23:35
GNUtoodo you know xda?23:35
GNUtoolet me look23:36
kristianpaul"The GPS software also provides native support for Broadcom's Long Term Orbit (LTO) extended ephemeris service" what on hell is that??23:37
kristianpaulwhy they need extend ehepmeris..23:38
GNUtoono idea23:38
kristianpauloh oh, you're going to replace firmware inside  bcm4751 ?23:39
GNUtooit's da_G23:39
GNUtoowe only want to talk to ir right now23:40
GNUtooda_G is the author of the posts if I remember well23:40
kristianpauli was looking a datasheet, seems baseband is just another blackbox23:41
GNUtoohowever it's way better than qualcomm phones (for now)23:41
kristianpaulso nah, nothing i could get interesting it, as the idea is to get a libre baseband SoC and firmware23:41
kristianpaulsure sure23:41
GNUtoobasically we should share the :23:42
GNUtoo"how to compute GPS fix on a CPU"23:42
kristianpaulthat dont sound right to me,23:42
GNUtooah? why?23:43
GNUtoonot free enough?23:43
GNUtooI lack context in your don't sound right to me23:43
kristianpauli'm reading this http://wenku.baidu.com/view/0128a23467ec102de2bd8973.html23:44
kristianpaulindeed is RRLP related23:44
GNUtooI've downloaded the datasheet if you want23:45
kristianpaulwikisend it! thanks23:45
kristianpaulso far your project look liks this to me http://gps.psas.pdx.edu/23:46
kristianpauli never tought found a "GPS SoC" this days nice !23:47
kristianpaulGPS receiver23:47
kristianpaulyou plan get to the arm cortex later correct?23:48
GNUtooI'll explain23:48
GNUtooreplicant is the project which replaces non-free software and libraries on android phones23:49
kristianpaulbut you meant, calculate fix on tye cpu23:49
GNUtoowe started with the htcdream and qualcomm devices23:49
GNUtooand we found out that they were very bad for freedom since the modem controlled:23:50
GNUtoo*the GPS23:50
GNUtoo*the sound card23:50
GNUtoo*could read/write in the ram of the main CPU23:50
GNUtoobut then we found the nexus S23:50
GNUtooanother dev which is now the main dev reversed the modem protocol of it23:51
GNUtoowe still need to reverse the GPS protocol23:51
GNUtoowhich is MEIF23:51
kristianpaulMEIF is like NMEA ?23:51
GNUtoonot at all23:51
GNUtooMEIF is like raw mesurement of the GPS sent to the main CPU(the arm CPU running android)23:52
GNUtoowhich is why I talked to you about it23:52
GNUtoothe fix and everything must be calculated from the main CPU(running android)23:52
kristianpaulsounds good23:52
GNUtooso if you have hints on how to calculate fixes with free software we are very interested23:52
kristianpaulyes i DO !23:53
kristianpaulwell, lindi-_ had some experiences on field too23:53
kristianpaulfrom ublox chips23:53
kristianpaulthat provides navigation data23:53
GNUtooI'd love to have free software GPS running on my freerunner btw23:53
kristianpaulbut i dont see any hope if you cant hack the arm cortex soc from the broadcon chip23:53
kristianpaulwich is that gpl-gps link above is about23:54
kristianpaulGNUtoo: if you have more datasheets perhaps i coould help or point something more accurate to do23:55
lindi-_GNUtoo: currently I think it's a major milestone if we can extract raw data from non-free firmware and then recalculate the results from that23:55
kristianpaulbut resumiing if you can get navigation data from the soc, well is a early first step23:56
lindi-_GNUtoo: replacing the non-free firmware on the chip sounds currently too hard23:56
lindi-_GNUtoo: got a spec for this MEIF?23:56
GNUtoothat's the problem23:56
kristianpaulotherwise, trying to reverse eng closed source protocols like that MEIF is a waste of time i think23:56
GNUtoonokia offered it under NDA but they do not anymore23:56
lindi-_kristianpaul: it can't be that complex23:57
kristianpaulhack the arm cortex inside the bcm chip !!23:57
kristianpaullearn how internals correlator works23:57
kristianpaulGPS protocol as you called it not secret, specificaions are public23:58
GNUtooyou mean we hack it and reimplement something like NMEA instead?23:58
lindi-_nmea is not very nice format23:58
GNUtooI know23:58
kristianpauli said you cant ingnore this bcm chip run a firmware23:58
GNUtoolike AT23:58
kristianpaulhack that23:58
lindi-_ubx is actually quite nice23:58
kristianpaulno learn how talk the chip, no just that23:58
lindi-_compact, simple23:58
kristianpaulGNUtoo: or help us port osgps to milkymist soc !23:59
kristianpaulwe are working on a free gps baseband and receiver here23:59
GNUtoolindi-_, what's the chip in the openmoko?23:59
--- Sun Jan 29 201200:00

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