#qi-hardware IRC log for Wednesday, 2010-12-15

andres-calderon I'm not sure about  the mechanical coupling between boards00:00
andres-calderonYou can see that each board has a different connector?00:01
andres-calderonnon a common block connector00:01
wolfspraulyes but they all seem mostly to be a random variation of P+ P- C D T00:02
wpwrakandres-calderon: if you still have these battery controllers, could you please take a high-resolution photo of each and upload it somewhere ? (qi-hardware.com wiki or such)00:02
wpwrakandres-calderon: that way, more people can try to find out how these thingies work00:02
andres-calderonwpwrak ok, I'll upload them.00:03
wpwrakthanks !00:03
wolfspraulunfortunately there is quite a bit of glue on some of them00:03
wolfspraulandres-calderon: given how the industry works, there is very little real innovation actually00:03
wpwraklet's hope for the best then :)00:03
wolfspraulso even though every company wants to have their own connector to sell batteries at a huge margin, electrically it's all the same00:03
wolfsprauljust for small variations in the mechanical alignment and shape of the connector (maybe patented :-))00:04
wolfspraulI would say that's true for anybody but Apple, maybe Sony.00:04
wolfspraulthe rest is all the same, no worries :-)00:04
wolfspraulonce you understand the circuit, that is00:04
andres-calderonjeje... I like the idea of using some of them.00:04
andres-calderonBut as I said, I worry about the mechanical coupling.00:05
wpwrakandres-calderon: what is actually the plan so far for the power supply of xue ? i haven't found any power connector.00:05
andres-calderonwpwrak design from scratch,  again..00:06
wolfspraulwpwrak: maybe you saw above, I asked about Sensor PSU and Andres said "copied from Aptina reference design unchecked"00:07
wolfspraulhe likes the idea of moving the sensor (and sensor psu) to a daughterboard00:07
andres-calderonI chose wrong ... none of the AiT component is available.00:07
wpwrakandres-calderon: hm. okay, there's one (non-technical) oddity: wolfgang mentioned that carlos told him xue was ready for production and carlos was complaining that wolfgang didn't move forward.00:07
wpwrakandres-calderon: i think that, from this, wolfgang got the idea that also you would consider xue to be "finished". and he was disappointed to find out that the reality is very different.00:08
andres-calderonxue was NOT ready for production00:08
wpwrakandres-calderon: so .. what happened ? :-)00:08
wpwrakandres-calderon: (NOT ready) good that we agree on this ;-)00:09
andres-calderonwpwrak Carlos is very enthusiastic.00:09
andres-calderonhas been a misunderstanding00:09
wpwrakwolfspraul: (2nd psu) yes, that's a different story. (having everything sensor-related on the same board makes perfect sense, of course)00:09
wpwrakandres-calderon: okay. is there any external pressure on the schedule ? e.g., have you or has carlos committed any schedule in the near future to any presentations, customers, or such ?00:11
andres-calderonwpwrak We have no external pressure.00:12
wpwrakandres-calderon: because if he or you feels in a rush for some reason, that's probably also something wolfgang should know. if only to correct unrealistic expectations :-) but sometimes, things can also be helped in other ways00:12
wpwrakokay, great. just a communication glitch then.00:12
andres-calderonYes, an stupid glitch :)00:13
andres-calderonjust noise in the channel00:13
wpwrakheh ;-)00:13
wpwrakandres-calderon: now .. now that it seems quite a few things will change, are there any other items in the current xue design that you'd rather change ?00:15
andres-calderonI feel lost with the design of the PSU. We not have experience with chinese and TW manufactures...00:15
wolfspraulandres-calderon: going forward, let's try to use only public communication to discuss the state and progress and plan for Xue00:15
wpwrakandres-calderon: you mean the AiT parts ? where does this design came from ?00:15
andres-calderonYes, AiT parts...00:16
wolfspraulfor example this channel is logged, so it helps avoid misunderstandings00:16
andres-calderonok, bookmarked.00:16
wolfspraulandres-calderon: how do you want to proceed now? where do you want to start?00:17
wolfspraulI can try to cleanup the schematics names today.00:17
wolfsprauland maybe add pdfs to BOOKSHELF little by little00:17
wolfspraulandres-calderon: you can consider parts we use in Milkymist One, see http://en.qi-hardware.com/wiki/Milkymist_One_RC2_BOM00:18
andres-calderonWe need to change of vision, design the product, not just a board.00:18
wolfspraulor parts used in the Ben00:18
wpwrakandres-calderon: what are the specific use cases you had in mind ? you mentioned automotive00:19
wolfsprauland finally, here is stuff I have in inventory: http://en.qi-hardware.com/wiki/Sharism_inventory00:19
andres-calderonnice componente, but very expensive : http://focus.ti.com/general/docs/lit/getliterature.tsp?genericPartNumber=pth04000w&fileType=pdf00:20
wolfspraulI can explain my vision :-)00:20
wolfspraulwe all see something, don't we :-)00:20
wolfspraulso first it's about another application of the Milkymist SoC00:20
wolfspraulthen it's about the KiCad process, including fped, boom00:20
wolfspraulthen it's mobile, battery powered should be possible00:21
wpwrakfor sourcing, one rule i found quite useful is that anything that digi-key don't stock is a potential problem :)00:21
wolfspraulthat's not always a good rule, but of course if digikey has something in 'real' quantities, that's a good indicator00:21
wolfspraulI like the idea of making the expansion header more central in the design, thus also moving the Aptina sensor there00:22
andres-calderonwpwrak:  I will use the digikey  heuristic00:22
wolfspraulso basically the Xue board provides basic digital interfaces, the fpga, usb, ethernet, microsd, nand, etc.00:22
wolfspraulwhereas the expansion boards provide analog/sensor stuff, image sensors, rf front-end boards, whatever00:23
andres-calderonmy vision: http://www.matrox.com/imaging/en/products/smart_cameras/iris_gt/00:23
wolfspraulbut the first application would just be the camera as discussed00:23
andres-calderonan artificial vision platform00:23
wolfspraulwe chose the APtina sensor already, and I have it in stock (bad decision, too early!)00:23
wolfspraulso that's roughly my vision00:24
wolfspraulon the current Xue, I would remove the AVR00:24
wolfspraulI would move the Aptina to a daugtherboard (just move the expansion header in a little)00:24
wolfspraulI would remove the ft2232d and provide the same headers as on milkymist one, for use with the jtag-serial daughterboards I am already making00:25
wolfspraulI would investigate the notebook battery controllers see whether that's a quick and dirty solution to hookup batteries to Xue00:25
wpwrakremove AVR -> probably remove the four FAN4010 too00:25
andres-calderonHDR is a nice feature. We will gain attention  for this feature.00:26
wolfspraulI would leave the spi-nand as is, a switch to nor flash would destroy too much value in Xue and I guess the technical challenge of spi-nand is under control00:26
wolfspraulsure remove the FAN00:26
wolfsprauloh my vision is also to keep the costs down00:26
wolfspraulthat's one reason why I like spi-nand00:26
wolfsprauland why I like to remove the ft2232, because later less and less people will need it, so daughterboard is better, especially since we have it already due to m100:27
andres-calderonI do not like the  jtag daugtherboard connector...  is enormous00:27
wolfspraulmaybe we can bring the cost of the xue base board under 100 USD00:27
wolfspraulthat would be cool00:27
wolfspraulit's smaller than all the chips you put on the board now00:27
wpwrakwolfspraul: sharing the ft2232 with MM1 also reduces the design risk00:27
wolfspraulif one design idea of xue is the expansion header, that means the board will gain height anyway00:28
wolfspraulso another (small) daughterboard somewhere will not make it more bulky overall00:28
wolfsprauljust fill space that would otherwise be air00:29
andres-calderonin fact, are 2 connectors...  but with the space released by the CIS boards.. may be will be a good choice00:29
wolfspraulthe Ethernet connector for example is really big, I doubt the jtag-serial daughterboard on xue would even be higher than that connector00:29
wpwrak(jtag) hmm, 2mm spacing, 14 pins of which ~9 are used00:30
wolfspraulhere's a picture of the little thing00:30
wpwrakthis is the connector: http://search.digikey.com/scripts/DkSearch/dksus.dll?lang=en&site=US&WT.z_homepage_link=hp_go_button&KeyWords=87832-1420&x=0&y=0&=go00:30
wpwrak(according to the MM1 schematics)00:31
wolfspraulI have the jtag-serial board already00:31
wpwrak16.65 x 6.3 mm00:31
wolfspraulso using the same board multiple times is good for me00:31
andres-calderonok, I think that  can work00:31
wolfspraulhave mercy with the poor manufacturer :-)00:32
wpwraki like the idea of sharing jtag00:32
wolfspraulwpwrak: it's 2 connectors, Sebastien's idea :-)00:32
wpwrakfewer boards to make00:32
andres-calderonwolfspraul I forgot!!, today arrived the M12 holders00:32
wolfspraulnice idea, creates a lot of problems00:32
wpwrak2 ?00:33
wolfspraulbut now we did it, so what00:33
wolfspraulyes, he had a NICE IDEA00:33
wpwrakJ6 and ...00:33
wolfsprauldon't we like nice ideas in hardware and mechanical?00:33
wolfspraulhe had the idea to put the jtag and serial connectors next to each other, at a certain distance00:33
wolfsprauland then have the board connect to both of them at once00:33
wolfspraulcool, eh/00:33
wolfspraulunfortunately we forgot how difficult it would be to find 2 matching pairs of connectors that come out at exactly the same height00:34
wpwrakwhy would that be difficult ? pick from the same series00:34
wolfspraulthere are 2 connectors00:34
wpwrakyou still have positioning tolerances and all that00:34
wolfspraulone is the 7*2 you pointed out00:35
wpwraksure, but same series = same nominal dimensions00:35
wolfspraulthe other one is a 4 pin serial connector00:35
andres-calderondifferent pitch?00:35
wpwrakJ5 then00:35
wolfspraulI am gladly sourcing them from you.00:35
wolfspraulmatching pairs please :-)00:35
wolfspraulthe problem is not the pitch, but the height00:35
wolfspraulyou have 2 boards, top and bottom00:35
wolfspraulbut they are connected to each other with 2 connectors00:35
wolfspraulso the 2 connectors (male and female) both need to have the same height00:35
wpwrak(same series) e.g., for 100 mil headers, you get all sorts of shapes, 1xN and 2xN with the same general shape within the same series00:36
andres-calderonThey share the same pitch?00:36
wolfspraulhere's the tool that helps us out right now :-)00:36
wpwrakMM1 schematics just say "Header 4" for J5. grrr.00:36
wolfspraulthe issue is HEIGHT00:37
wolfspraulwe are talking about 4 connectors in total, right?00:37
wolfspraul2 on the top, 2 on the bottom00:37
wolfsprauland they need to meet in the middle at the same height00:37
wpwrakwolfspraul: no problem so far :)00:37
wpwrakwolfspraul: no. only each pair has to have the same height.00:37
wolfspraulwe've made a few of these boards already, and every time it was a hassle with the height, well, unless for the help of the tool I just pointed to :-)00:38
wolfspraulbut anyway00:38
wpwrakwolfspraul: so take two board-to-board connectors, M and F. put M:1x4 and F:2x7 on one board and F:1x4 and M:2x7 on the other00:38
wolfspraulhow we have those jtag-serial boards already00:38
wpwrak(or make one all-male and the other all-female)00:38
wolfspraulyes and then most likely the two will not match00:39
wpwrakyou probably mixed series. that's when the problems start ;-)00:39
wolfspraulbut anyway, for Xue this is no issue00:39
wolfsprauljust use the same connectors as on m1, done00:39
wolfspraulpick from the m1 bom00:40
wpwrakas long as you stay within the same series, the only differences you get are component tolerances and positioning errors. can't avoid these. but they're usually not overly bad.00:40
wpwrakdo we have a picture of MM1 jtag ?00:40
wpwrakthe serial connector may get in the way00:40
andres-calderonwpwrak http://en.qi-hardware.com/wiki/File:Jtag-serial-board-on-m1.jpg00:41
wolfspraulwpwrak: serial connector in the way?00:42
wolfspraulmaybe we misunderstand00:42
wpwrakthanks ! do we also have the other side ?00:42
wolfspraulthey are _BOTH_ between the main pcb and the jtag-serial board00:42
wpwrakwolfspraul: i think xue only has jtag, no serial00:42
wolfspraulmaybe it should have serial as well?00:42
wpwrakwolfspraul: yes. so there has to be room for where serial goes. or that, yes00:42
wolfspraulwe have the jtag-serial board, and will get this supported out of the box in upstream tools00:43
wolfspraulsomething worthwhile reusing, I think00:43
wolfspraulthose are the 2 connectors from m1, if xue uses the same, all perfect for me00:43
wolfspraulthey need to be in the same distance to each other too, as on m100:44
andres-calderonI have not Altium. I need these measures.00:44
wolfspraulit's not in Altium00:44
wolfspraul100% kicad00:45
andres-calderonof course! jeje00:45
wolfspraulthat's the jtag-serial side, but it should exactly match the other side00:45
andres-calderoni know.. I'm almost asleep, sorry00:46
wolfspraulthen sleep a bit, we learnt a lot already :-)00:46
wolfspraulthanks for sharing your views!00:46
wolfspraulbtw, I'm currently making 100 of these critters, I think that's the first 'real' KiCad run in a while00:47
wolfspraulunfortunately no fped, didn't find the time to fpedize it00:47
andres-calderonwe need to seriously review the issue of the PSU. I listen carefully any idea.00:48
wolfspraulbut we use the KiCad generated gerbers for pcb, stencil, etc.00:48
wolfspraulandres-calderon: is the PSU on Ben NanoNote any good?00:48
wolfspraulI only have extremely high-level thoughts on the psu.00:49
andres-calderonI can  use the charger BEN00:49
wolfspraulprobably not very helpful00:49
wolfspraulwell the Ben has a battery00:49
wolfsprauland it's very robust00:49
wolfspraulyou can plug in or out the battery00:49
wolfspraulit will start charging00:49
wolfspraulit all seems to work00:49
wolfsprauland it's all done in a simple and cheap circuit, afaik00:49
wolfspraulXue has different, I understand higher, power requirements? So I'm not sure whether the Ben circuit applies.00:50
wpwrakben charger isn't too horrible. you can probably get rid of the FETs with a different chip, though, without picking anything overly exotic00:50
wolfspraulandres-calderon: did yousee the link to the Ben BOM?00:50
wolfspraulthe parts should be in there, let me check (it's already a while ago...)00:51
wpwrakwolfspraul: are all the ben parts considered easily available ?00:51
wolfspraulthat's a bit too broad00:51
wpwrakhehe ;-)00:51
wolfspraulbut definitely not a more bold statement than saying 'check digikey'00:51
andres-calderonwolfspraul: the BEN battery  charger can be enough00:52
wolfspraulso if 'check digikey' works, then 'ben parts can generally be considered widely available' also works00:52
wolfspraulwell, same as in digikey, there can be exceptions/surprises on individual chips00:52
wolfspraulbut for the most part, yes, sure, can be assumed to be widely available00:52
wpwraksimple but not quite correct rules are good. with them, you can impress people, because those simple rules show the clarity of how you see simplicity in the complexity. and you can impress people again when the rules don't work, because you wisely deviate from the beaten path where appropriate. you can't lose ;-)00:54
wolfspraulhmm, that ben .ods bom is not so great00:54
wolfspraulno links to datasheets00:54
wolfspraulwe have improved since then :-)00:54
wolfspraulyou need to cross-reference it with the schematics00:55
wpwraki like the inventory00:55
wolfspraulbut other than that it works00:55
wolfsprauloh, I want to get the inventory into boom standard :-)00:55
wolfspraulbut one by one00:55
andres-calderonI go to sleep. Bye00:55
wpwrakandres-calderon: cya !00:55
andres-calderonwpwrak: buenas noches00:56
wpwrakwith each digi-key parcel i get, i feel guilty for not tracking them in a boom-friendly format00:56
wolfspraulboom united01:01
wolfspraulamateurs of the world connect to form the largest inventory of strange stuff01:02
wpwrakheh. larger than digi-key's database ;-)01:02
wpwrakwell, i'd say the xue situation looks a lot friendlier than it did just two days ago. alas, also with a schedule that extends deeper into 2011.01:18
wolfspraulI enjoy walking my mind through it.01:19
wolfspraulit's scary to read how many things are supposedly based on my suggestions or decisions though01:20
wolfspraulvery scary01:20
wpwrakgood :)01:20
wolfspraulbut oh well01:20
wpwrakyes, indeed ;-)01:20
wpwrakwe found the implicit answer to the ownership question01:20
wolfspraulyes but if that's all, it means the board is doomed01:23
wolfspraulgood thing I think this is not all01:23
wpwrakyeah, that ownership needs to move again. you don't need it :-)01:27
wpwraklet's hope andres has enough time to do all the things that need doing01:27
wpwrakyes ! the fped fix fixed the weird build problem on sh4 ! :)01:32
wolfspraulwpwrak: sh4, that's where fped will take off01:46
wpwrakdefinitely ;-)01:50
wpwrakdebian on sh4 appears to have some odd shell. either a different kind of shell, or something much older or newer than the rest.01:51
wpwrakthat shell prints the "Aborted" when a program aborts not to that program's stderr but to the shell's stderr. this threw off my regression tests.01:51
qi-bot[commit] Werner Almesberger: cameo: migrated Excellon reader from drl2gp http://qi-hw.com/p/cae-tools/848eb2a04:38
qi-bot[commit] Werner Almesberger: cameo: adding support for dril/mill substitutions (in progress) http://qi-hw.com/p/cae-tools/b1652fc04:38
qi-bot[commit] Werner Almesberger: cameo: completed support for dril/mill conversions http://qi-hw.com/p/cae-tools/7a64e4204:38
qi-bot[commit] Werner Almesberger: cameo: added command "append", improved file output error checking http://qi-hw.com/p/cae-tools/b8f654504:38
qi-bot[commit] Werner Almesberger: cameo: "array" did not affect subsequently loaded files http://qi-hw.com/p/cae-tools/b76d17f04:38
qi-bot[commit] Werner Almesberger: cameo: apply translation after loading a file only to the new file http://qi-hw.com/p/cae-tools/7cb25c504:38
wpwrakhmm ... "cannot boot after fast power cycling" looks like quite the understatement. wasn't that "cannot boot ever after" yesterday ? ;-)04:41
wpwrakthe production/smt note must be from sebastien ;-)04:43
adamwangwpwrak, umm...i need write later04:43
adamwangha...no...i did in honest. :)04:44
adamwangi wrote what i thought they are not good enough..04:44
adamwangwe're not OM, pls..:)04:44
wpwrakadamwang: then his language rubs off ;-)04:45
adamwangi can see my parts here some of them had already pins lifted up04:46
wolfspraulwpwrak: no no :-)04:47
wolfspraulthe bootup issues are difficult, and probably multiple bugs masking each other04:47
wolfspraulbut if we ship out a board, it means we feel it's usable04:47
adamwangbulk parts are sort of management with One 'adam'...hehe .:)04:47
wolfspraulit's called 'early developer kit'04:47
wolfspraulthe power cycles we did were pretty drastic04:47
wolfspraulhundreds of cycles 200ms 100ms etc.04:47
wolfspraulmaybe the test was stupid04:48
wolfspraulif you use a hammer and smash down on the board you will also see a problem :-)04:48
wpwrakwolfspraul: still, that shouldn't kill the board04:48
wolfspraulit 'shouldn't' I agree, but even the setup is already special with a lab supply and a fast on-off switch04:48
wolfspraulyou cannot even try this with the supplied power supply04:48
wpwrakwolfspraul: poorly inserted connector ? :)04:49
wolfspraulmechanically you cannot be that fast, maybe if you have a switch further down somewhere04:49
wolfspraulwe can't resolve it here so let's see what further tests show04:49
adamwangeven if this test was stupid at all, it's still existed.04:49
wolfspraulmaybe Sebastien can find some root causes of flash corruption04:49
wolfspraulmaybe the Xilinx FAE can find some root causes in the fpga power-up sequence04:49
wolfspraulslowly we will understand it better04:49
wolfspraulall Xilinx boards show this too, btw04:50
wolfspraulSebastien is aware of this bug forever, he even thinks it might be a 'silicon bug' although I think we should not jump to such conclusions and just dig in step by step04:50
wolfspraulhe has seen this bug on the ml-401 xilinx board, etc.04:50
wolfspraulall under control :-)04:51
wpwrakbut flash corruption != the power cycle thing, right ?04:52
wpwrakthe latter is terminal, the former not04:52
wolfspraulthen there are 2 stages of 'terminal'04:53
wolfspraulflash corruption can be remedied with reflashing04:53
wpwrakokay, ICU :)04:53
wolfspraulthis bug is rare, so tracking it down is hard. there may be multiple bugs masking each other.04:53
wolfspraulif you are talking about something you have seen once, reading too much into it doesn't help.04:53
wolfspraulthis whole thing is unfortunate, without a doubt04:54
wolfspraulwe won't rest until it is fully understood and more robust04:54
wolfspraulwe had it on rc1 as well, btw, only that there were so few boards that it almost never showed up04:54
wolfspraulAdam has a pretty good attack plan.04:55
wolfsprauland I'm sure if Sebastien can reproduce it, every time he can he will try to peel off another layer from the onion, no? :-)04:55
wolfspraulalso I don't want to get completely side-tracked by this bug. that's the advantage of working with 40 boards first-hand for 2 weeks. I have a pretty good feeling on the severity of issues.04:56
wolfspraulthis one, even though it sounds scary, is not at the top04:56
wpwrakthe flash corruption doesn't sound so nasty. inconvenient for sure, and if it happens frequenty and under any operating condition, may disqualify the device for the intended use. but that's probably still a bit out.04:56
wolfspraulI highly doubt that.04:56
wolfspraulthere is a lot of alpha-level code in the stack04:57
wolfspraulkeep in mind how 'early' all this is04:57
wpwrakyes, that's why i say it still takes a while before this becomes the #1 blocker04:57
wolfspraulat those low levels hardware and software need to get really cozy with each other to become 'robust' to the higher levels04:57
wpwrakof course, for a VJ it would be rather embarrassing if the box all of a sudden doesn't boot ;-)04:57
wolfspraulmaybe the flash writing or reading routines need to become more forgiving04:58
wolfspraulit's hard to say04:58
wolfspraulalso I doubt sebastien will stop everything else and only hunt down this thing, it's too rare04:58
wolfspraulit's not the right priority04:58
wolfspraulput in some way to log/see what's going on, slowly make the codes more robust04:59
wolfspraulthat's the way to go04:59
wpwrakbut the fpga killer could be trouble even now. particularly if you don't know what exactly triggers it. could be that your test that needs hundreds of cycles is not the most "efficient" or common way to make it happen.04:59
wolfspraulmeanwhile Adam will go to the Xilinx FAE and that should be very nice too04:59
wolfspraulyeah we treated those 2 boards pretty badly, he he04:59
wolfspraulit's close to the hammer05:00
wolfspraulbut once I go into testing mode, well, I do05:00
wolfsprauladamwang: did I personally 'terminate' those 2, or you 1 and me 1? I forgot...05:00
wpwrakthey will haunt you in your dreams05:00
wolfspraulwell I took it sportive. I said 'let's see what's really going on in this power cycling thing' "let's stress test it a bit"05:01
wolfspraulthat was not a brilliant idea in hindsight05:01
adamwang0x2c you & me did, 0x07 you did. your hands were lucky.! :)05:01
wolfspraulso now I handed back to cooler heads05:01
wolfspraulI seriously forgot that it doesn't make sense to run a test that is totally removed from normal use patterns.05:02
wpwrakwolfspraul: get the process patented. then, when any other device fails in a remotely similar way, sue them ;-)05:02
wolfspraulat some point anything can be 'unforeseen' and damage the board.05:02
wolfspraulrun it in a sauna?05:02
wolfspraulit's also a bit stupid05:02
wolfspraulmaybe a VJ wants to perform in a Turkish steam bath?05:03
wolfspraulwe should test that05:03
adamwanghey, don't say in stupid way05:03
wolfspraulhumidity so high that there will be drops of water everywhere immediately05:03
wpwrakwater-proofing the box wouldn't be the worst possible idea05:03
wolfsprauladamwang: well I think it was a bit not-thought-through. so you were right :-)05:03
adamwangany software test plan you guys are also pretty sure they are perfect? That's the same story on h/w05:04
wolfspraulyou suggested to first study some more, then test. but I thought test first, collect data :-)05:04
wolfspraulnow we have data05:04
wolfsprauladamwang: yes but software tests normally don't leave destroyed stuff behind.05:04
wolfspraulthat's probably where my obsession comes from.05:04
wpwraknot only do clubs sometimes get humid, but there can also be spilled drinks if the VJ perimeter security isn't attentive enough05:04
wolfspraulso anyway05:04
wolfspraulI hold back a little.05:04
wolfspraulwe will improve this boot-bug thing.05:04
wolfspraulbut not out of proportion, no need to pass the stupid Wolfgang test. We should focus on really valuable use cases.05:05
wolfsprauladamwang: do you know roughly when you plan to go to the Xilinx FAE?05:06
adamwangtoday just work on wiki and others05:07
adamwangbefore I go  to there, I'd like to measure/scope all falling pulse after power down.05:08
wolfsprauladamwang: can you reproduce the .56A problem?05:08
adamwangi'd prefer i plot all power sequence and some control signals done before I go05:08
wolfspraulI think it's hard when you try to power-off for >2 seconds05:09
wolfspraulbut when you cycle too fast, maybe it's a separate issue then?05:09
adamwangright, i think i can reproduce easily, then scope all supplies (falling edges)..05:09
wolfspraulreproduce with fast power cycles, or with slow power cycles?05:10
adamwangthen compared to rising edges recorded first...05:10
adamwangi want to reproduce it by slow power cycles (say > 2s)05:10
wolfspraulok good05:11
adamwangthis step first to capture falling edges pulse then read it05:11
adamwangthere're 4 assembly:05:12
adamwangpower up (0.44A) rising edges05:12
adamwangpower down (0.44A) falling edges05:12
adamwangpower up (0.55A) rising edges05:12
adamwangpower down(0.55A) falling edges05:13
wolfspraulgot it05:13
wolfspraulpublish it all in a wiki page later, could be valuable for others too05:13
wolfspraulsebastien thinks he has seen power-up problems on every spartan-6 board, if I remember correctly05:13
adamwangsure, just to see boards itself behaviour first. I think we will all haunt in this. :)05:15
adamwangwell...so before that, i don't want to do more guess.05:16
wolfspraulI learnt my lesson05:17
wolfspraulsome peaceful study first05:17
wolfspraulmeanwhile development and use of Milkymist One can go in parallel05:17
lekernel_wolfspraul: can you bring one of the "terminated" boards with you to Berlin?05:46
wolfspraulno, they are in Taipei05:47
wolfspraulI want to give Adam some time now05:47
lekernel_ah, I thought you still were in Taipei...05:47
wolfspraulwhen he 'releases' them we can send one to you05:47
wolfspraulbut I really think we are better off slowing this down a little and not all jumping onto this one bug05:48
wolfspraulmaybe they can even be helpful with the Xilinx FAE, who knows05:48
wolfspraulgive them some time...05:48
zearhey guys, is there openssl port for the nanonote yet?06:15
wolfspraulis there not? I would hope so...06:16
zearno idea06:17
zearbut one guy contacted me asking about the nanonote06:17
wolfspraulthe packages are outdated, but there are a number of openssl packages there06:17
wolfspraulI don't know whether it's included in the latest image06:17
zearhe wants to know if nanonote can work as a password key06:17
wolfspraulyes definitely, it can06:18
zearso you plug it into a pc and it works as a usb disk mode06:18
wolfspraulout of the box maybe not though, more software work needed06:18
zearand, as an unrelated feature it has openssl for encrypting data06:18
zearwolfspraul, are you guys willing to help him if he has any problems with realizing his goals?06:19
zeari don't want to recommend some guy a 100EUR device if he wants to buy it only for one thing and this thing cannot be done with it06:20
zearso he can later blame me for wasting his money ;)06:20
wolfspraulhe, of course we are willing to help06:21
wolfspraulwe need use cases and demanding users06:21
wolfspraulon the other hand, he does need to be somewhat forgiving, realistically06:21
wolfspraulbut yeah, please. send him here or email or anything, and we work on it.06:22
zearhmm.. he seems to have no skills to realize his idea himself06:23
zearhe was hoping for a device that already has software for his needs06:23
lekernel_why use a complete computer for this?06:23
zearlekernel_, do you know any non-full computer devices that can change HIDs and are programmable?06:24
zearand are cheaper than a nanonote?06:24
zear*usb hids06:24
lekernel_there's the at90 usb key, but with no software skills it's going to be hard to get to work06:26
lekernel_same for nanonote anyway06:26
zearhe seems to have some skills06:26
zearyou know, the generic linux ones06:26
zearbut nothing to implement usb hid emulation06:27
zearhe says he needs an opensource platform, so usb keys are useless06:27
lekernel_I said at90 ... http://www.atmel.com/dyn/products/tools_card.asp?tool_id=387906:28
zearhe also says he needs at least 4 buttons and LCD06:28
zearso nanonote looks perfect for his needs06:28
andres-calderonI'm thinking of using the design of nano charger http://en.qi-hardware.com/w/images/9/9c/Lb60_schematic.pdf06:58
wolfspraulhi good morning!06:59
andres-calderonthe Nano charger need one ADC input (+VBAT signal),07:03
andres-calderonI worry a little with the  charger operation when the fpga is not programmed to control the charger.07:05
wolfspraulthe Jz4720 ADC is connected to that line? (not sure what you mean) if the jz4720 can control it, why shouldn't the fpga be able to control it?07:08
wolfspraulwell, I have no overview over the circuit :-)07:09
andres-calderonPBAT/ADIN0 pin 9207:10
andres-calderonfrom de 7z47xx IC07:10
andres-calderonI'm going to study better the design of the NN charger. Maybe I'm asking too early.07:14
wolfspraulyour concern is that the fpga cannot control that wire? I know too little to give meaningful feedback.07:15
wolfspraulof course I can study too and start learning :-)07:15
wolfspraulandres-calderon: ahh, you mentioned that you received the M12 lense holders?07:23
wolfspraulwhat do you think?07:23
wolfspraulnext time I will start to buy some M12 lenses07:23
wolfspraulfine out which lenses exist, different types (IR filter, no IR filter, telecentric, wide-angle, etc)07:24
andres-calderonwolfspraul the holders seem enough for the project.07:25
wolfsprauldid I only send you the plastic type, or also metal?07:25
wolfspraulnext time I send you metal as well07:25
wolfspraulplastic type costs 3 US cents07:25
wolfspraulmetal ca. 15 US cents07:25
wolfspraulone thing we have to find out is how to make the bottom side sit tight on the pcb07:26
wolfspraulso that no light gets in from the bottom07:26
wolfspraulnot sure, I know nothing about these things...07:26
andres-calderonI received 3 plastic type holders.07:32
andres-calderonseem designed to be fixed with screws.07:33
andres-calderonlike in the surveyor camera: http://www.surveyor.com/images/camera_hand.jpg07:35
wolfspraulyes screws07:36
wolfspraulbut the gap between the lense holder and the pcb - how can it be guaranteed to be completely shut?07:36
wolfspraullet's say you are standing in super bright sunlight outside...07:36
wolfspraulis the force from the screws pressing the thin (0.5mm?) plastic wall against the pcb silkscreen enough?07:37
wolfspraulI somehow doubt that07:37
wolfspraulI have seen some pictures where it seemed that a little black rubber piece was between the lense holder and PCB07:37
wolfspraulAdam has a friend who worked on cameras before, so maybe we get some advice from there. we try everywhere :-)07:37
wolfspraulandres-calderon: see here http://peauproductions.com/store/images/item_pictures/IMG_1760.JPG07:40
wolfspraulit's sitting on something07:40
wolfspraulsince that makes the connector higher, I would think it also affects the optics of the lense, maybe only slightly? don't know...07:40
wolfspraulwe don't have that little thing (rubber?) it sits on yet, sometimes those small details can be harder to get than expected...07:41
lekernelwpwrak: http://www.milkymist.org/flickernoise.html "Software stack": you'll be surprised how many components I have re-used :)07:41
lekernelGNU/Linux/X11 isn't the only alternative07:42
lekernel(and not even the best in many cases, I'd even say)07:42
andres-calderonwolfspraul: I hope not need rubber between holder and PCB.  Maybe the fact of the sensor is higher will be enough.07:47
wolfspraulandres-calderon: I doubt that. why no rubber?07:48
wolfspraulbest is to talk to someone who has already made cameras before...07:49
wolfspraulwe need to understand the pros and cons of it. same for metal vs. plastic holder.07:49
andres-calderonwolfspraul: for avoid a bit more complicated production process.07:50
wolfspraulthat's the wrong order, first we need to understand the pros and cons for camera quality.07:51
wolfspraulin parallel I will already start sourcing this, just in case because it will only cost pennies anyway07:51
wolfsprauladding that little spacer is no concern for 'production process' when only making a handful of boards07:52
wolfsprauleven when making hundreds or thousands07:52
wolfspraulfor now we should mostly worry about quality and functionality07:52
wolfspraulfor example - the spacer means that the lense sits higher. To which degree does that affect lense 'performance'?07:53
wolfspraulmaybe it's negligible, maybe not. don't know.07:53
wolfspraulunlike Andrey I think none of use can do some math to calculate the effect :-)07:53
andres-calderonwolfspraul: another design constrain to evaluate...   how to connect the sensor board: 1. block connector (like in the Surveyor) or 2. flat flexible cable (like in the elphel cam)07:57
wolfspraulI would make it a daughterboard using the expansion header we already have on Xue07:58
wolfspraulprobably that daughterboard sitting flat on top of Xue, back towards where the fpga is (basically where the current sensor is just a little higher)07:59
wolfspraulthen point the lense upwards as right now07:59
wolfspraulthose things are easily changed I guess. I would just start with the current expansion header.07:59
wolfspraulandres-calderon: can the fpc end directly into the expansion header?08:09
wolfspraulit's quite a bulky connector, for a high density fpc I'm not sure how easy it is. normally they go into those flat connectors.08:10
andres-calderonwolfspraul: i guess..08:10
andres-calderonwe can use the Ben flat connectors.08:13
andres-calderonhave enough lines08:13
wolfspraulthere is no need to use the Ben connectors only because I may have a few of them in stock08:14
wolfspraulwe can focus on the connectors that are best for Xue08:14
andres-calderongood news (congrats!)  http://lekernel.net/blog/?p=131408:22
wolfspraulyes :-)08:23
wolfspraulexcellent work from Sebastien (really)08:23
wolfspraulwe just screwed it together...08:23
lekernel:) thanks08:34
tuxbrain_awaylekernel: congrants lekernel btw can you also advice Tuxbrain will have some too ? :P09:42
lekernelsure, when you have them i'll put a link to tuxbrain09:43
tuxbrain_awaythanks :)09:43
wpwrakandres-calderon: (PBAT/ADIN0) i don't think this is part of the control loop09:44
wpwraklekernel: nice demos ! too bad there's no sound.09:53
wpwrakandres-calderon: (connector) another thing to consider is how the boards connect mechanically. e.g., you may want a hole for a spacer in each corner of the sensor board, and matching holes on the main pcb.09:55
lekernelwell those video suck, and don't even reflect the current state of the software09:55
lekernelI lack time and motivation to re-shoot some09:56
lekernelbut it's encouraging you find them nice already :)09:56
andres-calderonwpwrak: i agree, but its a nice feature can stimate  the battery level.09:57
wpwraklekernel: hehe ;) for sound, dunno if you also make music. there's a cute litte device made my Koss called the Kaossilater that's quite handy for producing nice-sounding loops, even for the musically challenged (like me ;-)09:57
lekerneloh, well09:57
wpwraklekernel: of course, i feel rather embarrased when a friend who does make music gets to play with it and produces something that sounds great in just a few minutes09:58
lekernelmid January I should have a milkymist workshop + party in a Berlin kind-of club/artist space09:58
lekernelI should have proper music there, and a nice setting09:58
wpwraklekernel: some "quick hacks" on the music side would help to have stuff you can publish without copyright issues09:58
wpwrakah yes, that ought to be great09:59
lekernelI should also pull of the video input there10:01
wpwrakandres-calderon: adc are somewhat unreliable for Li batteries, because of their discharge profile. but yes, it's better than nothing.10:01
lekernelI got myself a decent camera, and many of the visikord effects you've seen on youtube work on milkymist now10:01
wpwrakexcellent ! now all you need is a few pretty girls ;-)10:02
andres-calderonwpwrak: What do you think about this chip? http://focus.ti.com/lit/ds/symlink/bq24030.pdf11:14
andres-calderonwpwrak: one LiPo cell, dual input, thermal regulation, may be too expensive.11:14
andres-calderonwpwrak: this is the NANO charger: http://www.seawardinc.com.cn/Products_list.asp?id=198&action=new&language=zh-cn11:20
wpwrak(bq2403x) let's see ...11:22
wpwraklooks quite good. the chip takes care of a lot of little issues. pity that the LDO is extremely weak, only 20 mA11:26
wpwrak(bq) and it also passes the digi-key test ;-)11:29
wpwrak(bq price) not too bad. it doesn't need external FETs, takes care of USB current limiting, and even separates battery from output. do you need adapter and USB power ?11:34
kristianpaulIs not that SIE current missing feature?11:35
kristianpaul(do you need adapter and USB power ?)11:35
wpwrakmaybe, carlos' description wasn't quite clear11:36
kristianpaulI wonder why SIE dint use same power schema from nanonote, i guess power/current limits may be?11:37
wpwrakno idea. also the nanonote's charger isn't too great. doens't separate the current paths.11:39
kristianpaulah i see11:39
wpwrakandres-calderon: (NANO charger) oh, now i saw that you wrote NANO, not NAND ! i was already wondering ;-))11:39
andres-calderonwpwrak: Eso me pasa por escribir como Toro Sentado :)11:45
wpwrakif you want a cheaper chip, mcirochip MCP73837/8 may be an option. doesn't separate battery and system current, but also handled USB.11:47
andres-calderonwpwrak: MCP73837/8 may be a better choice than the exotic chinise Nano Ben charger IC.11:49
wpwrakoh, definitely. the one in the nano is overly simplistic.11:50
wpwrakyou can tell by the external FETs ;-)11:50
wpwrakah, what's the projected current consumption of Xue ?11:51
wpwrakhere's another one, similar to the MCP73837. this time from National Semi, not microchip:11:53
andres-calderonwpwrak: Difficult question when it related to FPGAs (between  5 and 10 Watts)11:55
wpwrakWOW ! 1-2 A at 5 V. crazy.11:56
wpwrakokay, let's look for the high-current chips then ...11:56
kristianpaul(5 and 10 Watts) why so much?11:57
andres-calderonkristianpaul: please see the TI power range for the XC6SLX45 FPGA11:59
kristianpaulsorry i'm not aware of that values, but i agree with wpwrak it is kind of crazy11:59
andres-calderonwpwrak: i like the LM3658 option.12:00
wpwrakLM3658 only supplies 1 A :-(12:00
andres-calderonPower Requirements  this is one of the dark sides of FPGAs12:01
wpwrakis current consumption also so high in real life ? or are these just extreme value, e.g., if all cells are switching at the maximum clock speed ?12:02
andres-calderonI have no real numbers ... estimate the consumption of the FPGA is a very difficult task12:03
andres-calderonSébastien should have this clear..12:04
andres-calderonplease see: http://en.qi-hardware.com/wiki/Milkymist_One_RC2_BOM12:05
andres-calderon3-A, 3.3/5-V Input Adjustable Switching Regulator with Auto-Track Sequencing12:06
Action: wpwrak quietly buries usb power12:07
lekernelFPGA is not the only thing on 2.5V12:07
wpwraklekernel: what real power consumptions have you measured on MM1 ? assuming no overly drainy peripherals12:09
wpwrakandres-calderon: there's that USB host. that could also suck 0.5 A if you let it12:09
lekernelwith current bitstream/software the complete board draws less than 5W12:10
wpwrakokay .. 1 A@5 V then ...12:12
wpwraklekernel: does it ever drop below, say, 2 W, while doing something useful (e.g., vi ;-) ? e.g., would it even make sense to consider a scenario with only USB power ? (for operation. may still make sense for charging and duty cycle < 100%)12:13
lekernelI don't know12:14
lekernelI haven't tried to optimize power at all, power doesn't matter for this design12:14
andres-calderonWe can forget  operation based on single USB port (dual port neither)12:14
andres-calderonwpwrak: For these power optimization of FPGA cores is that I had to put the AVR in the current design (seemed like a good idea  use the power metering  at different level).12:22
wpwraklekernel: sure, MM1 has other things to do :)12:25
wpwrakandres-calderon: thinking about the battery ... what characteristics do you have in mind ? e.g., how long should xue be aboe to run continuously on battery alone ?15:46
[Jc]Hi , i try to reflash the nanote whith the reflash_ben.sh but i have nothing after "try fetching .ubi.bz2 rootfs....15:58
[Jc]can you help me ?15:58
[Jc]strange, the flash is ok but very slow16:06
kristianpaul[Jc]: are you flahing now?16:31
kristianpaulWell that take some time, yes flash is kinf slow16:32
kristianpaulI may think is faster if is done from uSD-like memories16:59
kristianpauldamn taxes (wikireader)17:10
qi-bot[commit] David Kühling: tune Emacs' RAM usage (down to approx 10MB now).  Hide menu bar, http://qi-hw.com/p/openwrt-packages/ccd581818:46
qi-bot[commit] David Kühling: updated Emacs loadup.el http://qi-hw.com/p/openwrt-packages/5c645d118:49
qi-bot[commit] Werner Almesberger: cameo/gerber.c (gerber_read): store coordinates as floats after conversion http://qi-hw.com/p/cae-tools/995922420:43
qi-bot[commit] Werner Almesberger: cngt: tool change utility for MDX15/20 (in progress) http://qi-hw.com/p/cae-tools/154be2b20:43
qi-bot[commit] Werner Almesberger: cngt: some small fixes http://qi-hw.com/p/cae-tools/62152e220:43
wolfspraulhey just a reminder...20:58
wolfspraulwe need to purge 'SD' (as in Secure Digital) everywhere20:59
wolfspraulI see 'microSD' popping up all over the place20:59
wpwrakwhat shall we call it ?20:59
wolfspraulflash memory card?21:00
wolfspraulwe should also stay away from 'SDIO'21:00
wpwrakFMC - too general21:00
wolfspraulafaik the most we can go is microSD-like21:00
wpwrakmaybe micro5D ? ;-)21:01
wolfspraulalthough that is suspicious lawyer-speak :-)21:01
wolfspraulI like micro5D21:01
wolfspraulyou are almost Chinese now, Werner...21:01
wpwrakhaha ;-)21:01
wolfspraulbut seriously, the SD needs to go21:01
wolfspraulfor our own good, and the good of copyleft hardware and people who want to manufacture and sell devices21:01
wolfspraulit's a trademark, we have no rights to it21:01
wpwrakwe need to name the thing somehow. so, what shall it be ?21:01
wpwrakMicroFlash ? :)21:02
wpwrakuSD = SmallMicroFlash (SMF), with apologies to "Prince" :)21:02
wolfspraulas usual there are patents, copyright, trademark law21:03
wpwrakSDIO -> MFIO21:03
wolfspraulcopyright we should be OK, because we use the Linux kernel and if there is any problem then others have dealt with it before21:03
wolfspraulI think there was legal action over this years ago21:03
wolfspraulpatents are always nasty, but I doubt the SD-3C would directly use them, or whether we would actually infringe on any patent21:04
wolfspraulthat leaves the trademark21:04
wpwraki wonder what the status of the SD-card physical shape is21:05
wpwrakso, how to avoid the trademark ? 5D ? invent some name ? explain each time with a long sentence ?21:05
wolfspraulwell we need to start, I feel we don't pay attention to this and it will get nasty later21:06
kristianpauluSD-like is ok?21:06
wolfspraul(me included)21:06
wolfspraulthe -like thing is a legal safeguard, afaik21:06
wolfspraulI would prefer to only use it on the side though, to explain what the 'flash memory card' is21:07
wolfspraullike flash memory card (*)21:07
wolfsprauland then (*) micro-SD like flash card21:07
wolfspraulsomewhere far away :-)21:07
wpwrakhow do we reference it contexts that are not "official documents" ? irc, mailing list, etc.21:08
wpwraks/it/it in/21:08
kristianpaulas i understand micro is okay but not the "ese de", right?21:08
wpwrakthe channel could use an editbot :)21:08
wolfspraulthe SD is the core trademark, I think21:08
wolfspraulSecure Digital21:08
wolfspraul*SD and SD*21:08
wpwrakuEseDe ;-)21:08
wpwrakLibreDigital ;-)21:08
wolfspraulwell, my first concern is that the hardware we are making is clean21:09
wolfspraulbecause it should be the basis for 'copylefting' by others, right?21:09
wolfspraulI mean I can pay 4000 USD21:09
kristianpaulyou should :)21:09
kristianpaulshould not***21:09
wpwrakor we write it like we spell it. ED ;-)21:09
wolfspraulbut I just create a 4000 USD copy barrier, at least for people who want to be legal21:09
wpwrakuED, EDIO, etc.21:09
wolfspraulplus 1000-2000 / year I think21:09
kristianpauli like that21:09
wpwrakor the same with LD. uLD, LDIO, etc.21:10
kristianpaulbot will be funny21:10
wolfspraulso I want to increase the copyleftedness by making hardware that is as least likely to step onto the SD trademarks, copyrights, patents, as possible21:10
kristianpauls/n/N pops every where21:10
wolfspraulI mean right now we have 'microSD' right in the tooling of the Ben :-)21:10
wolfspraulcan't get any worse21:10
wolfspraulthat all needs to go21:11
wpwrakLibre Small Digital, LSD :)21:11
wolfspraullibre flash?21:11
kristianpauldont abouse libre word i think, flash is not libre at all (copyleft hw way)21:11
wolfspraulalso for Milkymist One microSD is popping up everywhere21:11
wolfspraulok fair enough21:11
kristianpaulhmm true21:11
wolfspraulSebastien proposed flash memory card21:11
wpwraklet them try to argue that LSD is likely to be mistaken for SD in court ;-)21:11
wolfspraulI like it21:12
wolfspraulI would write it out each time.21:12
wolfspraulno FMC21:12
wolfspraulwe can write flash memory card (*)21:12
kristianpauli prefer flash memory card, not resumed21:12
wolfsprauland then (*) microSD-like flash memory card21:12
kristianpaulhell ;)21:12
wpwrakwolfspraul: you're ignoring day to day communication. nobody will write a long name all the time21:12
wolfspraulor leave out the (*) alltogether21:13
wolfspraulactually I don't think it's that important21:13
wolfspraulthere are way too many of these cards21:13
wolfspraul(for regular people)21:13
wolfspraulrs-mmc, miniSD, microSD, full size, etc.21:14
wolfspraulanyway, I just wanted to post a reminder21:14
wolfspraul'SD' is a trademark, using it devalues our copyleft stuff21:14
wpwrakQIC - Qi Card :)21:14
wolfspraulit creates problems21:14
wolfspraulsooner or later21:14
wolfspraulthe SD people will not release their trademark into the public domain, etc.21:15
wpwrakwolfspraul: yes, you've repeated that SD is a problem about 50 times by now. i still get it ;-)21:15
wolfspraulwe cannot assume those things to happen21:15
wolfspraulthis will go on... SD to the moon! :-)21:15
wolfspraulwpwrak: for the archives :-)21:15
wpwrakwolfspraul: but you're dodging the question of a practical alternative :)21:15
wolfspraulflash memory card21:15
wolfspraulwhich you already said you don't like...21:15
kristianpaulexternal memory21:16
wpwraksure, in marketing material and such, you can write Flash Memory Card (*) Mycro-SD-alike21:16
wolfspraulthe first step is raising awareness that SD is a trademark we haven't licensed and shouldn't use21:16
wpwrakthat alone is not enough21:16
wpwrakwe still need to call it something21:16
wpwrakand we don't want people to fall back to "Flash" or "Card"21:16
wolfspraulwell there are several use cases. microSD cards (as in the mechanical size), SDIO protocol, others?21:17
wpwrakwhich is what will happen if you offer only an unwieldy name21:17
wpwrakso 1) SD is bad. avoid it. agreed.21:17
wolfspraulI do something practical, I start ripping out *SD* when I see it :-)21:17
wpwrak2) official documents, press statements, case engraving, whatever, can use something nice and complicated. your choice. fine with that.21:18
wpwrak3) what do we use for other communication ?21:18
wpwrakso, remove the hardware ?21:18
wpwrakthen we don't need to name it :)21:19
wpwrak /nick Salomon21:19
wpwrakSmall Removable Card ? SRC ?21:20
wolfspraulare you trying to find a name for the microSD mechanical shape?21:20
wolfspraulthe connector/height? (to the outside we are going pretty wild anyway... :-))21:21
wpwrakfor shape, the item itself, the connectors, etc.21:21
wpwrakprotocol, too.21:21
wolfspraulif your new names are not clear in meaning, people will fall back to the old ones similar to how you think they will do it with my 'flash memory card'21:21
wpwrakwell, we can probably talk about the SDIO protocol and so on and implicitly refer to out slot21:22
wpwraku5D then ;-) it's not our fault that people mistype the acronym of ultrasmall 5 dimensions all the time :)  (5D: 1) mechanical, 2) industry standard memory, 3) industry standard I/O, 4) GPIOs, 5) power)21:24
wpwrakor scrap mechanical and make it 1) clock21:24
wolfspraulif I compare with the headset connectors, we use 2.5mm, 3.5mm21:25
wolfspraulmaybe we say 2-pin, 3-pin, TRS, TRRS21:26
wpwrak5D FAQ: Q: Is u5D the same as uSD. A: No. u5D is a device and interface format specified [...bla...]. The SD Association has until date not sought certification of their product as u5D compatible.21:26
wolfspraulyes but if we do this the u5D thing will be seriously confusing21:26
wpwrak8card :)21:27
kristianpaulSPI Card21:28
wolfspraul8x8 flash memory card?21:29
wpwrakED - External Device. uED, EDIO, ... :)21:29
wolfspraul8 pin, 8 mm (not exactly, but for the sake of 8x8)21:29
wpwrakah :)21:29
wpwrak11 x 15 mm :)21:29
wolfspraulI am comparing with some other connectors21:29
wolfspraulnormally people 'describe' them21:29
wolfspraullike the 2.5mm headset thing21:30
wpwrak(11*15 = 165)21:30
wolfspraulthen of course often there is a trademark involved21:30
wolfspraulMIDI, DMX21:30
kristianpaulthose too?21:30
wpwrakMIDI too ? shit21:30
wolfsprauldon't know21:30
wolfspraulrelax guys21:30
wolfspraulhey Werner, let's make a fped view of the connector21:30
wolfspraulthen the fped text is the name of the connector21:31
wolfspraul(the xxx is the necessary source)21:31
wolfspraulmaybe I try 8x8 fmc for a while21:32
wolfspraulit's not more confusing the u5D21:32
wpwrak(fped) http://projects.qi-hardware.com/index.php/p/kicad-libs/source/tree/master/modules/usd-card.fpd21:33
wpwrakthat's the card/contact area shape21:34
wolfspraulof course it says 'SD' in there :-)21:34
wolfspraulyou trademark infringer!21:34
wpwrakof course :)21:34
kristianpaulrun !21:34
wolfspraulsteal other peoples work21:34
wpwrakwell, i've heard guns are cheap around here. so let the trademark/patent holder come ;-)21:35
wolfspraulI'm looking at a full-size SD card right now and I notice that the right-most pin (looking from the bottom) seems to be split into 2?21:35
wolfspraulso that makes it 9 pins21:35
wolfspraulit looks like this is some backward-compatibility thing, because the split is done in such a way that it would also fit a connector that had only 8 pins on the other side21:36
wpwraki think it's for card insertion detection21:37
wolfspraulmaybe we should move towards mechanical measurements in general?21:37
wolfspraulas an idea... (don't poke holes into it right away, Werner...)21:37
wolfspraulmechanical and math are patent and trademark free21:37
wolfspraulwe are still free to count21:38
wolfsprauland measure21:38
wpwrakor just number them ? card #1 :)21:38
wpwrak(math) hmm. mp3 algorithm ? :)21:38
wolfspraulfor example the trs trrs are called like that as 'tip ring sleeve'21:39
wolfspraulI like that21:39
wolfspraulno way this will ever be claimed by anybody21:39
wolfspraultrrs = tip ring ring sleeve21:39
wolfsprauljust describe the looks/shape of the connector21:39
wolfspraulwpwrak: no not the algorithm is patented, but the implementation :-) still a difference...21:40
wolfspraulguys until I hear protests, I will try 8x8 fmc21:40
wolfspraulmaybe someone can find something nice and descriptive along the lines of trs?21:40
wolfsprauljust use plain language, plain words, measurements/counts, etc.21:41
wpwrak8x8 does not reflect real dimensions21:41
wolfspraul8 pins, 8 millimeters21:41
wpwrak165 ? for the 165 square-millimeters a rectangular bounding box would cover ?21:41
wpwrakwhere are the 8 millimeters ?21:41
wolfspraulroughly :-)21:42
wolfspraulbecause 8x8 is easy to remember21:42
wpwrakthe card is 11x15 mm21:42
wolfspraulthe connector at the top is narrower21:42
wpwraksure. but then it's no longer a physical measurement21:42
wolfspraulthe pins are pretty much exactly 8mm wide21:42
wolfspraulalthough the pitch is not exactly 1mm21:43
wpwrak9.6 mm21:43
wpwrakpitch is 1.1 mm21:43
wpwrakjust look at the fped file. it's all in there :)21:43
wolfspraulthat's for the whole front, including left and right margin21:43
wpwrak(with fped - so you can see the dimensions)21:43
wolfspraulI'm not saying 8x8 fmc is perfect21:44
wolfspraulI'm also fine with 8x10, it's equally wrong/confusing I think21:44
wolfspraulbut my idea was something measured, or described21:44
wolfspraullike trs21:44
wolfspraulnot a play like u5D21:44
wolfspraulirony or humor doesn't work in such cases21:44
wolfspraulsomeone else will have the last laugh21:45
wpwrak8x10 is much better. 9.6 rounded :)21:45
wolfspraul8x10 fmc21:45
wpwrakjust "card" ?21:45
wolfspraulwell the 'x' makes it look like a 2D measurement21:45
wolfspraulbut one is number of pins, the other is mm21:45
wolfspraul8p10mm, argh21:45
wpwrakmaybe 8/10 or 8:1021:45
wolfspraulI like the : better21:46
wolfspraul8:10 flash card21:46
wpwrakdoesn't have to be flash21:46
kristianpaul4bit card21:46
wpwrakis Flash trademarked ? :)21:46
wolfspraulI don't think so21:46
wpwrakF-card :)21:46
wolfspraulthat's suspicious21:46
wpwrakespecially when it doens't work :)21:46
wolfspraul8:10 card21:47
wolfspraulmaybe 8:10 card holder21:47
wolfspraulyes I think that's nice21:47
wolfspraulI can live with that21:47
wpwraknow, the protocol ? we have SD and SDIO to rename.21:47
wolfspraul8:10 is pretty cool, maybe we can find something better even so it still works when the next nanoSD is introduced...21:48
wpwrakit would then be 8:6 or such21:48
wolfspraulbut number of pins is quite specific, and rounded mm at the front too21:48
wolfspraulunless they introduce a round one21:48
wpwrakspecific is good in this case21:48
wpwrakafter all, this should be an alias for uSD21:49
wpwrakfull-siz SD would be 9:2421:49
wolfspraulman I hope Sebastien doesn't kill us over this21:49
wolfspraulSD and SDIO21:49
wolfspraulwhen do we use this?21:49
wpwrakminiSD something like 11:1921:50
wolfspraulI'm not using it, I think21:50
wolfspraulare the Linux kernel sourcing talking about SDIO actually?21:50
wpwrakwhen talking about the protocol21:50
wpwrakof course they do21:50
wpwrakperhaps we can use SD/SDIO when talking about the protocol. after all, we're not identifying the card per se21:51
kristianpaultrade mark !21:51
kristianpaulno matter card is there i guess21:51
wolfspraulalso I don't think this will show up anywhere in or on the product, or marketing21:51
kristianpaulah ok21:51
wpwrakit should appear on some occasions in official documentation, to clarify that the interface is IO-capable21:52
wpwrakbut maybe it's possible to treat this separate from technical discussion21:53
wolfspraulwhen I think about Milkymist One, I cannot imagine SD or SDIO to show up anywhere21:54
wolfspraulwhy should it?21:54
wolfspraulif the specs list the connector as '8:10 card holder' then that's it21:54
wolfspraulof course nobody will know what a 8:10 card holder is21:54
wolfspraulso we may need a (*) microSD-like somewhere21:55
wolfspraulmaybe even (*) 8 pins : 10 mm connector, microSD-like21:55
wpwrakMM1: dunno. depends on the intended role for the 8:10 card21:56
wpwrakbut think of a general-purpose device like the ben. there you want to know whether it can talk to SDIO peripherals21:56
wolfspraulon m1, it cannot find nice cards like ben-wpan, because it's in the middle of the board and no space, and lying flat on the pcb as well21:56
wpwrakof course, you can write something like "for memory and I/O cards"21:57
wolfspraulfor this to have any practical effect, you would have to list the tested and supported cards anyway21:57
wolfspraulone by one, with brand, model, etc.21:57
wpwraksure. it's a first overview. what you print on the box.21:57
wolfspraul(I'm talking about 'practical effect' for a normal end user)21:58
wpwrakmaybe you already know whether card X is linux-friendly or not21:58
kristianpaulis MMC trademarktoo?21:58
wolfspraulthere is no way someone will buy 2 products because both say 'SDIO' and he thinks they should work together21:58
wpwraki wouldn't be so sure ;-)21:58
wolfspraulno way21:58
wolfspraulmaybe if the sales rep says they will work21:58
wolfspraulbut definitely not based on seeing the same trademark21:59
wolfspraulkristianpaul: I think mmc, if even a trademark, is not owned/managed by the SD-3C21:59
wpwraksometimes you have to rely on guesses. you go to the shop with a list of possible devices. they have none of these but something else. do you want to return home and do more research or buy the damn thing for USD 10 and hope for the best ?22:00
wolfsprauljust looking at uspto.gov22:01
wolfspraulso many mmc22:01
wolfspraulone is a 'foldable pet ramp'22:01
wpwrakthere's also quite a lot of people who simply don't know how to find out beforehand about such problems. to them, it may just be technology that sometimes works and sometimes doesn't. at some point in time, you may also reach this part of the market ;-)22:01
wolfspraulmmc mobile...22:02
wpwraki mean, even technically competent people would, for example, have a hard time buying a home theatre where all the elements work properly for HD with all the copy-protection and such :)22:02
wolfspraulregistered by the multimedia card association, but now owned by jedec solid state technology association22:02
wolfspra1lgot disconnected, did it get through?22:07
wpwrakmmc -> jedec, aye22:07
wolfspra1lok I just added that we also need to look at trademarks case by case22:08
wolfspra1lif one is licensed in terms that are acceptable, then using it may be better for everybody22:08
wolfspra1llike Linux22:08
wolfspra1lso I don't know about MMC22:08
wolfspra1lbut I am pretty sure about *SD*22:08
wolfspra1lthat one needs to go, or sooner or later everybody who manufactures 'SD host devices' will have to stay away from large markets like the US, or pay 4000 USD + 2000 USD/yr22:08
wolfspra1lI don't know about MIDI, DMX, and many others22:09
wolfspra1lcase by case, our resources are limited and opportunity costs are high...22:09
wolfspra1lso we got a solution for now?22:09
wolfspra1lthe card is: 8:10 card22:09
wolfspra1lthe protocol is: memory and I/O22:09
wpwrakconsidering how profitable the drug business is, perhaps making the device illegal in the US wouldn't be such a bad idea ;-)22:10
wolfspra1ldon't you think that 8:10 flash card is better?22:11
wolfspra1lhmm, maybe not22:11
wolfspra1lit's just a desperate attempt to ring a bell :-)22:11
wpwrak(protocol) i'd rename only in "marketing" communication. for descriptive use in technical communication, we ought to be safe.22:12
wpwrake.g., "implement the SDIO protocol". or "the SDIO protocol [says this or that]"22:13
wolfspra1lshould be OK, and this shouldn't make it through to the product or marketing material anyway22:13
wpwrakhmm ......... they also use the x:y format for bible references ... googling it, there's a nice one: http://bible.cc/luke/8-10.htm22:15
wpwrakalludes to the obscurity we're after ;-)22:16
wolfspra1lkristianpaul: for trademarks, you can normally use a quick search at uspto.gov22:16
wolfspra1lin technology, the US is basically the gold standard, or at least it is what we are talking about here22:16
kristianpaulwolfspra1l: looking now, good to know :)22:16
wolfspra1lif someone registers a trademark in Kazachstan, and rips off some importers, ok that's a separate issue22:17
wolfspra1lit's a big world22:17
wpwrak("gold" standard) high time for china to take over :)22:17
wolfspra1lwell yeah that will be a big mess22:17
wolfspra1lit's probably already happening22:17
wolfspra1lkristianpaul: also, for a trademark to be enforced, it doesn't necessarily need to be registered22:17
kristianpaul(take over) dont do it already?22:17
wolfspra1lsometimes you see this (tm) and (r)22:17
wolfspra1l(tm) is a unregistered trademark22:18
wpwraknow all we need is to make patent production and trafficking a capital offense and the world will be a much better place :)22:18
wolfspra1lin the end any company can bring forward a trademark infringement claim, at any time, and for any reason22:18
wolfspra1lthe question is whether it can stand in court22:18
wolfspra1lso it's not as easy as looking in uspto.gov and then you know what's going on22:18
wolfspra1land some big corps are known to trample over this system anyway22:19
wolfspra1lthey just don't care22:19
wolfspra1lso even if you have a trademark registered for x years, if you are in the way they will attack you22:19
wolfspra1land you better give in to their demands, whatever it is :-)22:19
kristianpaulso fair ;^)22:19
wolfspra1lso similar to patents, I approach this in a case by case style22:20
wolfspra1lwatch out for aggressive people, stay away from them22:20
wolfspra1ldon't pay a ransom, because then others will have to pay too in the future22:20
wolfspra1lI will probably stop registering trademarks, and use unregistered instead22:21
wolfspra1ljust (tm)22:21
wolfspra1llike NanoNote, Milkymist22:21
wolfspra1lthe registration has questionable value anyway, in my practical experience22:22
wolfspra1lyou are just feeding lawyers, parasites22:22
wolfspra1lin the end they come trample over you anyway, so you better hold fire and still have a few bucks when they come attack :-)22:22
wolfspra1lwpwrak: hey wanna hear a China story?22:27
wpwrakalways :)22:28
wolfspra1lno rumors, all first hand22:28
wolfspra1lI walked around many offices of IC design firms in China, and always wondered why I never saw IC design software running on any computer screen...22:28
wolfspra1l(I'm quite observing looking at what is running on screens when I walk around new offices)22:29
wolfspra1lfinally I found the answer :-)22:29
wpwrakthey don't have licenses ?22:29
wolfspra1lthe Chinese government has setup special small companies to buy 1 license of expensive US software, install it all properly etc.22:29
wolfspra1loh no22:29
wolfspra1lmuch smarter22:29
wpwrakand only run the stuff when there are no visitors22:29
wolfspra1land then they use whatever they can to 'share' this22:29
wolfspra1lpeople can go 'visit' there22:29
wolfspra1lfinally a government that truly helps its people, no?22:30
kristianpaulsharing  is good ! :)22:30
wolfspra1lthey call this 'IC center' etc.22:30
wolfspra1la center it is, indeed :-)22:30
kristianpaulbus is it time shared?22:30
wolfspra1lwho knows, whatever works22:30
wpwrakthen there must also be a lot of leakage between companies - someone forgetting files, etc. ;-)22:31
wpwrakall very good, of course. more sharing :)22:31
wolfspra1lprobably, but it's a lot of copy/paste anyway22:31
wolfspra1lbut this is a true story, organized by the Chinese government22:31
wpwrak.cn - copy nation :)22:31
wolfspra1lnice service22:31
wpwrakexcellent service :)22:31
wpwrakof course, not so good from the perspective of furthering freedom. they should give the money they save to people who develop free alternatives :)22:32
wolfspra1lwatch out they have their own definition of freedom22:33
kristianpaulI always want to ask that, how is "free alternives" fit to china rules?22:33
kristianpaulah that explain better ( own definition of freedom)22:33
wolfspra1lsure not at all22:33
wolfspra1lif anything 'free' means under 100% CHinese govt control22:33
wolfspra1lif only 99% control, then not entirely free yet22:34
wolfspra1lbut yes, I'm sure they work on those 'free alternatives' elsewhere22:34
kristianpaulAre you aware of Lemote company and chinese gov?22:34
wolfspra1lwhat do you mean with aware?22:35
wolfspra1lI've heard of Lemote22:35
kristianpauls/aware/what do you know22:35
kristianpaulIs Ingenic some how linked/parner with china gov?22:37
kristianpaulwell i dont know if parner is a right word there22:37
wolfspra1lChina is a dictatorship. every chinese company, everything located on chinese soil is under 100% control of the Chinese government.22:39
wolfspra1lone phone call is enough :-)22:39
kristianpauli see22:39
wolfspra1lthe CEO of any Chinese company can be called to show up for a 'government conference' within hours22:39
wolfspra1land all of them will go22:40
kristianpaulThat includes Sharism? :)22:40
wolfspra1lI'm not making this up, this is how it is.22:40
wolfspra1lI would never ever register a company in mainland China.22:40
wolfspra1lSharism Ltd. is registered in Hong Kong, which still benefits from British traditions.22:40
kristianpaulah ok22:40
wolfspra1lfor the next 20-30 years we should be safe there22:40
kristianpaulgood :)22:41
wolfspra1lalthough they are slowly trying to undermine it22:41
wolfspra1lI made the mistake of registering some .cn domains, I am giving them up now22:41
wolfspra1loh btw, I think it's crazy that Mozilla and others install SSl root certificates controlled by the Chinese government22:41
wolfspra1lthat gives them a chance to run nice man-in-the-middle attacks without you ever noticing22:42
wolfspra1land you bet they already work on this, have the capability, and use it when appropriate22:42
kristianpaulnext year will be the year to dont trust SSL root ceriticates22:42
wolfspra1l100% guaranteed22:42
wolfspra1lseriously, this is serious stuff22:43
wolfspra1lthose root certificates need to be removed asap22:43
kristianpaulThat root-trust is scary22:43
wpwrakthe whole concept is broken22:44
wolfspra1lwpwrak: yeah but come on. a root certificate from the chinese government, preinstalled in the whole western world. hello???22:44
wolfspra1lone phone call22:44
wolfspra1lwhen they call, the caller ID is 0000000022:45
wolfspra1lno name, just when and where to go22:45
wpwrak(0000000) like telemarketing :)22:46
wolfspra1lkristianpaul: .tw is cool, of course.22:47
kristianpaulargg what i was thinking my data bus is just 8 bits :/22:50
wpwrakkristianpaul: simpler :)22:57
wpwrakkristianpaul: where are you actually stuck ? a good while ago, it looked as if you had pretty much everything in place23:01
kristianpaulwpwrak: i'm implementing what we talked yday23:01
wpwrakkristianpaul: are you cooking up some fancy handshake protocol ?23:01
kristianpaulnot fancy23:01
kristianpaulas simple as i can make it23:01
wpwrakkristianpaul: should be quite straightforward. particularly if you just make an 8 bit shifter23:01
kristianpaulfor now just a 8 bit shifter, plus a 4 bit counter in order tow adress a sram arragement (buffer)23:02
wpwrakkristianpaul: 1) extend the old 4 bit to 8 bit. 2) divide SYNC by two and output it23:02
wpwrakkristianpaul: ah, you're adding a buffer. i see.23:02
kristianpauli MUST, is no somethin i wish23:03
kristianpaulbut SIE desing push me to do it that way23:03
wpwrakhow come ?23:03
kristianpaulwas something i dint realize until i crasshed CPU, (shared bus with other memories)23:03
kristianpaulJZ4725 processor has a peripheral in charge of controlling extern memories, this controller allows to handle static, NOR, NAND and SDRAM memories. SIE uses the CS2 signal which belongs to the second pool of static memory and is activated (low active) whenever an address in between 0x1400|0000 to 0x17FF|FFFF is accessed. Next picture shows reading and writing cycles for this processor.23:04
kristianpaulhttp://en.qi-hardware.com/wiki/Xburst-FPGA_Interface ^23:04
wpwrakhmm. are you saying that you don't have 9-10 dedicated I/Os between FGPA and CPU ?23:04
kristianpaulsadly no :'(23:05
kristianpaulreally i tought i had23:05
kristianpauli just (as always) ignore some documetation and discover it by trial/error23:05
wpwrakinteresting. do you have DMA handshake signals ?23:05
kristianpauli'm not really aware of DMA handshake now, i must read first a bit more of this topic23:06
wpwrakmay be easier to go directly to DMA if you don't have a nice and simple interface23:06
wpwrakyou need a request line. maybe also the acknowledgement line. documentation seems a bit spotty.23:07
kristianpaulIf i see this get messy (i donk think, just in case) i 'll end wiring the spi-to-parallel module to the SDIO of my ben23:07
wpwrakif the FPGA appears like a memory device, you'd implement your 8 bit register just there and then request one DMA cycle. if this is handled quickly enough, you win :)23:07
kristianpaulwell i already have that23:08
wpwrakin any case, DMA is what you want. anything else is still some sort of polling and probably will be messy23:08
kristianpaulso i'mm missing DMA23:08
wpwraklet's see what they write about dma ...23:08
kristianpaulhmm seems is worth read more about DMA23:09
wpwrakgz4725_ds .. 3.9.3. ah, that helps .. NOT :)23:09
kristianpaulin SIE fpga have only I/O to real world afaik :/23:09
wpwrakmaybe you can loop something back :)23:11
GWEARIi have a simple but kinda tricky question23:13
kristianpaulDrop it :)23:14
GWEARImy old computer had a folder with over 3000 picture i took, its stuck in an endless cycle when trying to boot. i was able to install my old computers internal HD in another one, the problem is that the folder i had made was no longer in the desktop. yet it kept my other backed up cell phone pics in the My Pictures area (i had put that folder there.. is there any way to retrieve the folder i had created on the desktop?23:16
wpwrak*hmm* there's no external DMA ?!?23:20
wpwrakamazing. seems they really left it out.23:22
wpwrakGWEARI: on the XBurst JZ4740/20 and similar23:23
kristianpaulpoll poll poll23:23
wpwraksuck suck suck :-(23:23
wpwrakwell, you could send stuff over the UART ;-) that's about the only interface that's slave-capable. okay, and USB.23:24
kristianpauli was thinking ignore xburst-shared-bus and wire to SDIO23:25
kristianpaulwich is actually IO23:25
wpwrakplan B: maybe the bus is multi-master-capable, so the FPGA could become bus master23:25
kristianpaulargg thats is just 2bit, i touch that before23:26
kristianpaulbus i dont know, i need dig more on it23:26
wpwraksdio is 4 bits23:26
kristianpaulnot in SIE ;-)23:27
kristianpaullet me confirm (kicad)23:27
kristianpaulah yes23:27
kristianpaulyou right23:27
kristianpaulhmm i'm confused but Carlos send a patch for a 2 bit thing on SIE, maybe i'm wrong23:28
wpwrakphew :)23:28
kristianpaulah zoom tricky me23:28
wpwrakbus probably isn't master-capable. sdio would be a possibility. but the protocol is a little tricky23:29
GWEARIr u still referring to my question or something else23:30
kristianpaulwpwrak: http://kristianpaul.org/~paul/tmp/SIESDIO.png23:31
wpwrakGWEARI: we're trying to figure out a reasonable way to shovel data from an FPGA into the CPU (be it on SIE or ben)23:31
kristianpaulGWEARI: something else, actually your question is not too much related to copyleft hardware btw :)23:31
wpwrakkewl. one bit :-(23:32
kristianpaulah thats it23:32
wpwrakwhy efficient if you can make it horribly slow ? :)23:32
kristianpauli think that EMC bus sharing in SIE is there because a reason23:33
wpwraki wonder what the reason for this is. ran out of GPIOs ?23:33
kristianpaul(i hope)23:33
kristianpaulwpwrak: yes23:33
wpwrakit better be ;-)23:33
GWEARIu don't know nething on the subject then?23:33
kristianpauli tought it was same cpu as ben  (hoping re-use kerboard pins) but this differ23:34
wpwraknot sure if you can do SDIO with that configuration23:34
wpwrakisn't it the 4740 ?23:34
wpwrakor maybe i'm confused :)23:34
kristianpauli was too23:34
wpwrakaah, 4725. right.23:34
wpwrakhehe. that cpu indeed has only one pin there. wow. that takes courage ;-)23:36
GWEARIwhere would you recomend i go with this question?23:39
kristianpaulGWEARI: #electronics may guide you better i think23:40
kristianpaulwpwrak: you think implement sync/2 in a register bit then data in the next address, and read from there is too much pain ?23:46
kristianpaulany way, i'll try it that way23:47
kristianpauli'm off bed now, read you soon23:47
wpwrak(too much pain) dunno. but seems the next logical step23:49
--- Thu Dec 16 201000:00

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