#qi-hardware IRC log for Thursday, 2011-12-08

kristianpaulmirko: is the rfm12 driver upstream ?01:27
kristianpaulat least qi kernel, will be nice to get support on it01:30
akiwiguyhello, everyone04:59
mirkokristianpaul: what's your definition of upstream?11:14
mirkokristianpaul: it's public :)11:14
kuribasIs there work being done for a successor of the nanonote?11:46
kuribasI love the idea of the nanonote, but I am taken back by the lack of connectivity.11:47
wolfspraulkuribas: work yes, but don't expect new hardware that can replace the nanonote anytime soon12:01
wolfspraulthe Ben I mean12:01
kuribasBecause of production costs?12:02
kuribasIt would be interesting if the device could offer something that isn't offered by any competitors.12:13
kuribasFor example, the ability to interface directly with the chip.12:14
wolfspraulkuribas: to interface directly with what chip?12:31
wolfsprauldo you have a Ben right now? what do you want to do?12:31
kuribasNot yet.  For example interfacing with an audio a/d converter trought I2S.12:33
kuribasOr interfacing with microcontrollers through a serial connector.12:34
kuribasI own a zipit (it was quite cheap), but I don't like the keyboard on the zipit.12:35
kuribasBy interfacing I mean having some kind of dock that connects directly with the processor pins.12:40
wpwrakkuribas: have you seen this yet ? http://en.qi-hardware.com/wiki/UBB12:41
wpwrakkuribas: it doesn't give your a lot of signals, but it's very very convenient12:42
kuribasAh, I see :)12:44
kuribasBut no usb host?12:44
kuribasThe zipit has a 32 pin connector, which exposes quite a lot of pins (for usb, camera, etc...).12:48
wpwrakno, no USB host. the chip would have it but the wires are not bonded :-( so close, yet so far ...13:06
kuribasWould it be exponsive to change the pcb?13:22
wpwraki think so, yes13:26
wolfspraulkuribas: on the Ben side, we work mostly on the software13:26
wolfspraulthen there is atben/atusb13:27
wolfspraulthen we work (or should work) on tools like KiCad and boom (sourcing), to help with future manufacturing runs13:27
wolfspraulthen we should work on mechanical, but don't right now, not much at least13:28
wpwrakkuribas: another issue is that we'd also need some way to bring the signal out. UBB isn't very nice for this. and adding any other kind of connector would mean a change of the case13:28
wolfspraulthen there is Milkymist (the SoC) and Milkymist One, the video synthesizer/instrument13:28
wolfspraulI don't want to list all open items there ;-)13:28
wpwrakkuribas: a case change would cost in the order of USD 100'000 (new molds). so that would have sell quite a lot of bens :)13:29
wolfspraulwe should make more mechanical experiments like we did with Milkymist One, not just follow old paths13:29
wolfsprauland no other than Werner did all sorts of steps into that direction as well...13:29
wpwrakagreed. having practical know-how is good to have13:29
wpwrakhmm, makes me remember that i still have to get you some counterweights :)13:30
wejpbut even without changing the case, routing some of the SoCs currently unused pins to some test points on the PCB would be very nice for those people who would like to tinker with the hardware and don't mind soldering13:30
wolfspraulon the Milkymist side, there is some slow but steady GPS hacking going on13:30
wejpwould it really be that expensive/difficult to just change the pcb a little?13:30
wpwrakdunno. the fab may require a certain MOQ to make such a change. perhaps in the order of ~ 2-3 times the number of units already sold, maybe more13:32
wejpmh :/13:32
wpwrakwolfspraul: did i get this estimate right ? based on: units sold so far = about 50% of the units you initially promised to have made. and for a pcb change, they'd want an order of comparable size.13:33
wolfspraulah, that's so hard to explain :-)13:34
wolfspraulin manufacturing there are tons of moving parts13:34
wolfspraulthat's why even serious 'industry analysts' are publishing numbers about the Amazon Fire 'BOM' that I can just laugh about13:34
wejpi'm just asking this, because probably most people interested in the ben are hackers in one way or another and as such even such a rather small change could make the device much more attractive to those people13:34
wolfspraulsame for any other 'component breakdown analysis'13:34
wolfspraulall rubbish13:34
wolfspraulI do not see a path ('path') to make that kind of Ben13:35
wolfspraulthe Ben is the Ben and done13:35
wolfsprauland selling (just sold one today, and a few the last days :-))13:35
wolfspraulwe have the name for the next one already, that's cool... Ya NanoNote!13:35
wpwrak(component breakdown) well, it establishes a lower bound (if they have realistic component prices)13:35
wolfsprauland when and how exactly that magic beast will come out I don't know13:35
wolfspraulall wrong13:35
wejpok, but is there anything wrong with creating a slightly updated Ben 1.1, if it wasn't too expensive?13:35
wolfspraulno, nothing wrong, in fact it's an honor that people want more13:36
wolfspraulthat's motivating13:36
wolfsprauland we are on this, yes13:36
wolfspraulbut the path there is not easy to explain unless you watch and understand the many moving parts in manufacturing in general13:36
wolfspraulsure we could jump in and make a Ya, as spec'ed13:36
wolfspraulthat would probably cost 500k USD or more13:36
wpwrakwolfspraul: so how much money would wejp have to put on the table for you to go to the manufacturer and order a batch with a modified pcb ? and how many bens would wejp get for that ?13:36
wolfspraulby now we have by far surpassed the knowledge level of the original device manufacturer13:37
wolfspraulthey invited us to show them some of the 'features' (they mean software) we have on the Ben13:37
wolfspraulthey said "our device only has 1x features, but yours has 8x"13:38
wolfspraulbla bla13:38
kuribasWhat about a funding system like kickstarter?13:38
wolfspraulof course it also went nowhere because our system and their system are fundamentally incompatible13:38
wolfsprauleverything, I don't even know where to start13:38
wolfspraulos any modified PCB we just make ourselves, like we did AVT213:38
wolfspraulthe avt2 is a 64mb + usb host board that fits into the Ben case (with some extra holes)13:39
wolfspraulit exists already13:39
wolfspraulall files are there13:39
wejpwolfspraul, i know that building the Ya right now would be expensive, which is why i was wondering if instead doing just a rather subtle change to the pcb would be much less expensive? if it is still very expensive,that's ok, but if it is not and if you guys consider producing anither batch of Ben's once the current batch has been sold, it might be worth considering13:39
wolfspraulyou can make it in batches of 10 or 20 more or less economically13:39
wolfspraulwhat do you mean by "doing a change to the pcb"13:39
wolfspraulyou can change design files and upload them somewhere, but then what?13:39
wolfsprauldesign is only worth so much in manufacturing, over-valued imho13:40
wolfspraulwe have AVT213:40
kuribasI am making a nice UI for the zipit and nanonote (based on X windows and EFL), that could make the device attractive for more people.13:40
wejpwolfspraul, as i said, routing some of the currently unused I/O pins of the SoC to the outsinde to some test points13:40
wejpso they are accessible for hackers13:40
wolfspraulalready done, called AVT213:40
wolfsprauldone and produced and tested and works13:40
wolfspraulkuribas: great! :-)13:40
wolfspraulI don't think any more AVT2 will be produced13:40
wpwrakso how much would it cost to make, say, 100 more AVT2 boards ?13:40
wpwrakor has the ability for this been lost ?13:41
wolfspraulyou want to know the famous "BOM"? :-)13:41
wolfspraulmaybe 50 USD or so, * 100 = 5000 USD13:41
wejpok, will that updated pcb ever be sold in a possible future batch of the Ben?13:41
wpwrak(bom) no, money wolfgang-in, 100 AVT2-PCBA wolfgang-out :)13:41
wolfspraullet's add some pcb+smt+wire bonding overhead, 10k13:41
wolfsprauldisplay will be hard to get nowadays13:41
wolfspraulI just sold some of my remaining inventory to Carlos for a local SAKC production I think, or derivative13:42
wpwrakjust the PCBA. people can get a ben and swap it themselves13:42
wolfspraulyeah so if you leave out all the work that goes into sourcing, yield, testing, maybe 80-100 USD / board13:42
wejpoh, yeah, just selling those pcb would be nice too i think13:42
wolfsprauldon't know about sourcing of 4720 dies13:43
wolfspraulAVT2 hasn't been touched in 2 years or so13:43
wpwrakwejp: so there's your bill: invest 8-10k into qi-hw and you can probably get those boards made13:43
wejpi see13:44
wejpwell when those boards have to be sold at about 100 USD, this probably won't work, as the ben costs about the same. probably not worth it then13:44
wolfspraulthe low prices kill innovation13:44
wolfspraulfor anybody13:44
wolfspraulso users/buyers have to decide what they want13:45
wejpit was just a suggestion13:45
wolfspraulyes, it's good. we are working on this, but with some detours.13:46
wolfspraulthat's why I say: boom, mechanical experiments/processes/materials, Milkymist SoC13:46
wolfspraulyes I know a little far-fetched to see how this all connects together, but well, I can only act upon what I learn as I move forward13:47
wejpyeah, sure13:47
wpwrakthat's what we need H5N1. a world population of only the 1% the current size wuoldn't offer the same economy of scale -> higher prices -> more innovation ;-)13:47
wolfspraulI am 100% confident that we will be able to continue to build the Ben NanoNote software platform and move it forward into new hardware products13:47
wolfspraulotherwise why would I continue to invest in it?13:47
wolfspraulwhen I started the Ben NanoNote, I was hoping to build a mutual win-win relationship with Ingenic13:48
wolfspraul(the SoC maker)13:48
wolfspraulafter I realized from my endeavor before that what we wanted to do was impossible with Samsung13:48
wolfspraulafter 2 years, I finally can say that the chances that this win-win will ever happen slowly approach 0%13:48
wolfspraulthat's ok13:48
wolfspraulthe Samsung experiment before also took me 2 years :-)13:48
wolfspraulnow... the good news is - free software has our own SoC coming13:49
wolfspraulit's called Milkymist SoC13:49
wejpyeah, that's nice13:49
wolfsprauland I saw this as a hedge against inability to really kick Ingenic alive for a long time, so the path is long but hey, it's not that we are unprepared13:49
wolfspraulin parallel I still try to work with Ingenic, of course13:49
wejpso will the Milymist get an MMU at some point?13:49
wolfspraulthey have lots of products popping up here and there13:50
wolfspraullike that race-to-bottom tablet Ron was pointing out the other day13:50
wolfspraulso it would be possible to bring some openess quality to some of that, by marrying with what we do with the millions of USD wasted on the side of Ingenic and Ingenic's customers13:50
wolfspraulalas, until today, that marriage hasn't happened13:50
wolfspraulbut Ingenic is proud how many millions they have from their IPO :-)13:51
wolfspraulhappy spending! :-)13:51
wejpheh, i see13:51
wolfspraulso the Ben->Ya path has 2 options13:51
wolfspraulone with Ingenic, one with Milkymist13:51
wolfspraulunless a third one pops up, but I'm not aware of that right now13:51
wolfspraulI'm working on both, but from my time maybe only 2% on the Ingenic side, 98% on Milkymist13:52
wolfspraulthat's for the Ya hardware13:52
wolfspraulnot much I can do on the Ingenic side unless we find a way to carry forward what we did with the Ben (hardware)13:52
wolfspraulwe cleaned up and opened up that whole platform once, for free, for 2 years. if they don't see a value in that, well, then I work on the Milkymist SoC for a while.13:52
wolfspraulMeanwhile Ingenic is so clueless that they are already considering switching from MIPS to ARM. that tells you something about the depth of their software strategy :-)13:53
wejpa Ya with an open SoC would be really nice, but if i see it right, currently the milkymist does not have a MMU, will that be changed in the future?13:53
wolfspraulhave some spare time? :-)13:53
wolfspraulthat's the interesting part, it's all open13:53
wolfspraulreally open13:53
wolfspraulwe have some plans if you are looking for a starting point :-)13:54
wolfspraulso yes, a MMU is definitely one of the highest wish-list items for the Milkymist SoC now13:55
wejpi don't have very much time right now, but on the other hand, i am very interested in such things, so a starting point would be nice. it is really difficult to get started with such rather complicated things13:55
wolfspraulit's on the official roadmap (but not at the top for Sebastien), there were plans and discussions on the milkymist devel list13:55
wolfspraulyes maybe, nobody really knows until we go there and try13:55
wolfspraulI also don't know13:55
wolfsprauland I was a long-term kernel hacker (not Linux though, mostly Windows 9X and Windows NT)13:55
wolfspraulbut I have no clue about HDL, how to write a MMU13:56
wolfspraulthat's the chance to go back to school! :-)13:56
wolfspraulautodidact school, of course13:56
wejphehe, yeah, i'm doing that all the time ;)13:56
wolfspraulwejp: check this out http://lists.milkymist.org/pipermail/devel-milkymist.org/2011-April/001418.html13:57
wolfspraulthe second one 'almost' has the full design already :-)13:58
wolfspraulonly some coding needed now13:58
wolfspraulso what will that be? 500 lines? 1000?13:58
wolfspraulVerilog is short13:58
wolfspraulif only we had more hacker capable in that kind of stuff...13:58
kristianpaulmirko: he, sure i mean linux upstream but i got the message :)14:11
kristianpaul make milkymist slim than now put a screen on it there is next ya ;)14:23
kristianpaulkeyboard came inside the same case, you just take it out of it14:23
wolfspraulyes something like that14:24
wolfsprauland integrated gps :-)14:24
kristianpaulyes :-)14:24
pabs3interesting discussion14:59
wpwrakwolfspraul: why would it be clueless if they switch from MIPS to ARM ?15:24
mirkokristianpaul: it won't get upstream, too hackish18:55
kristianpaulmirko: do you have a user space vesion of it?18:58
mirkokristianpaul: take a look at the webpage and ask again :)19:00
mirkokristianpaul: http://switchsmart.org/wiki/index.php?title=SwitchSmart_model19:00
Action: kristianpaul rtfm19:00
kristianpauli just skimed and noticed, but sure is there... well at the end i want my router and nanonote to talk, yo yes i _must_  read more19:01
kristianpaulmirko: mom, i saw that before, but the switchsmart library still need a driver on the kernel right?19:02
kristianpaul(let me know if i'm bothering you with all my questions please)19:02
kristianpaulmirko: ?19:15
mirkokristianpaul: just busy, you're right - it needs a kernel module20:59
mirkothat's because of timing and performance20:59
mirkotoggling gpio's must be very accurate and causes heavy cpu load at high frequencies20:59
wolfspraulwpwrak: you are right. not 'cluelesss' depending on how they see themselves. It's just amazing to see how they are unable to tell the difference between good and bad software, working and non-working product.22:14
wolfspraulthe only way they can find out the software they spend a lot of money on doesn't work is from sales numbers, I'd say that's quite late.22:14
wpwrakah ... so they think the root of all their worries is MIPS vs. ARM, while it's pretty irrelevant ?22:14
wpwrakthere are easier and harders ways of doing QA :-)22:15
wolfspraulthey spend a lot of money on software that would all need to be written off when switching to ARM22:16
wpwrakah, not the linux way then: just recompile :)22:16
wolfspraullike I said. I now realize what the price and consequences of making 2 USD SoCs are, compared to the 15-20 USD SoCs from TI, Samsung, Freescale, etc.22:17
wpwrakbtw, interesting article (in german): http://www.heise.de/mobil/meldung/Smartphone-Nutzung-in-Deutschland-Nokia-vor-Apple-1391860.html22:17
wolfsprauland the consequence is that you are flying a large airplane nearly blindly22:17
wolfspraulbecause you don't have any quality staff on board, zero, nobody22:17
wolfspraulthat's a problem sooner or later22:17
wolfspraulbut it's also the prerequisite for selling 2 USD SoCs22:17
wpwrakthe part about customer loyalty due to the cost of migration is interesting. that's a strong argument in favour of open designs (sw and hw), and it's good to see that also regular users realize that this is an issue.22:18
wolfspraulso they are not 'clueless' in a way, they continue their path22:18
wolfspraulit's interesting to see how software-dependant the investents of fabless SoC makers are22:19
wolfspraulthat makes me feel good about Milkymist SoC22:19
wpwrak(clueless) yeah, i guess it's okay while it works. do you expect to eventually find their iceberg ? or does this look like a model that's long-term sustainable ?22:19
wolfspraulcould be sustainable, depends on what they want22:20
wolfspraulI give you another example in 5 min, brb22:20
wolfspraul22:16 < wpwrak> ah, not the linux way then: just recompile :)22:20
wolfspraulwpwrak: back. so we talk about the Amazon Fire and why I believe it's the first iPad competitor...22:37
wolfsprauland what chip is inside... TI OMAP4. and how we can work in such a way that the next time a really great product comes out an Ingenic chip is inside, even if in 2-3 years.22:38
wolfspraulthe answer? :-)22:38
wolfspraul"we sent our chip to Quanta already, that's how we get it into the Fire. All fine there."22:38
wolfspraulha ha :-)22:38
kristianpaularm better supported in Linux?22:38
wolfspraulat that point I realized the downside of 2 USD SoCs22:38
wolfspraulit's just a piece of plastic to them22:39
wpwrakwww.quantatw.com ?22:39
wolfspraulit's the ODM that makes the Amazon Fire22:39
wolfspraulas if Quanta would have *any* authority to change the SoC inside the Amazon Fire!!!22:39
kristianpaulalso XO laptop for OLPC time ago22:39
wolfspraulit's just unbelievable22:39
wpwrakhmm yes, i see22:39
wolfspraulmaybe they have a strong input on changing some power ICs, maybe22:39
wolfspraulbut the SoC???22:39
wolfsprauloh my god22:39
wpwrakbut it most work from time to time, no ? or else they wouldn't do it22:40
wolfspraulso I say 'clueless' from a software perspective22:40
wolfsprauloh sure, they send their chips to anywhere where manufacturing is happening, I guess22:40
wolfspraullike a good salesman should...22:40
wolfspraulbut while I talk about Linaro, he talks about sending samples to Quanta22:40
wolfspraulbeing so proud to know that Quanta 'makes' the Fire!22:41
wpwrakwhat do quanta do ? manufactoring ? design ? chips ?22:41
wolfspraulwho knows, I'd think they are the typical ODM there22:41
wolfspraulbut changing the SoC from a TI OMAP to an Ingenic would rip the entire software stack apart22:42
wolfspraulyou know that22:42
wolfspraulthat's *years* of investments on Amazon's side22:42
wpwrakah, so you don't even know if quanta may be the wrong tree or not :)22:42
wolfspraulhow on earth can they not see this?22:42
wolfsprauloh for sure I know22:42
wpwrakmaybe they're not targetting the kindle fire per se22:42
wolfspraulwe talked about the Kindle Fire22:42
wolfspraulI'm just giving you this as an example of how far the vision goes.22:43
wpwrakbut if people go to quanta and ask them to make them a kindle fire clone, maybe then the ingenic comes out of the hat ?22:43
wolfspraulyou think complicated :-)22:43
wolfspraulIngenic makes SoCs but has zero visibility into software22:43
wolfspraulthat's difficult22:43
kristianpaulif i understand correcly Ingenic dont want to make SoCs anymore?22:44
wolfsprauloops, no :-)22:44
wolfspraulWerner asked about 'clueless'22:44
wolfspraulthat depends on perspective22:44
wolfspraulthey are marching forward on their roadmap, successful IPO in Shenzhen, flush in cash, etc.22:44
wpwrakwell,if there are others who do the software for ingenic chips ...22:44
wolfspraulbut you cannot buy brain easily :-)22:45
wolfspraulyes but can you work with an SoC maker that doesn't understand software at all?22:45
wolfspraulimagine this in the harshest way possible22:45
wolfsprauldon't you think that's a bit risky for their software partner? :-)22:45
kristianpaulah okay so its about software that works, so mips doest arm do? (sorry repeat this again)22:45
wolfspraulthat's why the Letux 400 changed CPUs 3 times in a year (!)22:46
wolfspraulso while Schaller was still struggling to get a 2.6 Linux to even boot, their manufacturer had already changed the CPU multiple times22:46
wolfspraulrotating through whatever they could get in random fights against 'bad sales'22:46
wolfspraulI have a feeling that independent SoC makers will go away or be consolidated into just 1 or 222:50
wpwrak(letux) nice :)22:51
wolfspraulnvidia already mostly works with very few large customers only, and Marvell is moving there22:51
wolfspraulSoC depends on software and embedded specs22:51
wolfspraulwill be interesting to see where Ingenic goes22:52
wpwrakwhat OS does the kindle use ? linux ?22:52
wolfspraulmaybe they only have the option of bottom fishing22:52
wolfspraulheavily modified Android, yes22:52
wolfspraulthere's a large market each year of super low price products that don't work22:53
wpwrakhas android been ported to mips ?22:53
wolfspraulIngenic's way is the only way to compete there22:53
wolfspraulah that's a long story22:53
wolfspraultoo long for our happy open hardware team, that's all legacy stuff imho22:53
wpwrakif the customer has a good sw strategy, then the SoC maker doesn't need any. e.g., it would be relatively painless for someone like us to switch from a MIPS-based ingenic 4xxx to an ARM-based 5xxx22:54
wpwrakor an lm32-based milky-ya, for that matter ;-)22:55
wolfspraulwhy does the SoC maker exist then? who designs the SoC and decides what goes in and what doesn't?22:55
wpwraklet them. they do whatever they think is good. then you choose :)22:55
wolfspraulan SoC maker without software visibility sits in a tough spot22:55
wolfspraulbut if they like it there... :-)22:56
wolfspraulmaybe Ron will buy one of their tablets :-)22:56
wpwraki see that more important with users who are themselves weak in sw development22:56
wolfspraulyes but most people are, software is expensive22:56
wpwrakof course, the typical "let's just fork the vendor BSP" approach leads straight into lifelong dependency22:57
wolfspraulthat's why the idea that Amazon will allow Quanta to change the SoC from a TI OMAP to an Ingenic MIPS is so absurd22:57
wolfspraulit's as if TI's 1000+ Linux team doesn't even exist22:57
wolfsprauland Amazon's work over the last 5 years22:57
wolfsprauland Linaro22:57
wolfspraulall of that doesn't exist in that thinking22:57
wolfspraulit's irrelevant22:57
wolfspraulbut that's why the TI OMAP costs 15 USD, and the Ingenic chip 5 USD22:57
wolfspraulremove some brain22:58
wolfspraulwhich chip is in the Amazon Fire, today and tomorrow?22:58
wpwrakso what you're saying is that there's no linaro for MIPS. that's the bottom line of the problem ?22:58
wolfspraulI don't know how Ingenic can make it there. We could aim for the next Fire, and I will continue to propose that.22:58
wolfspraulit goes higher to Amazon22:58
wejpTI's linux support sucks pretty much22:58
wolfspraulthey have made a lot of software investments22:58
wolfsprauland if you made software investments, you will not like someone switching SoCs under you22:59
wolfspraulunless you are really just doing html5 apps22:59
wpwrakwell, we have good ingenic support in the ben and even if we assume all the work was paid at industry rates, we'd probably still only be a fraction of amazon's kindle team alone22:59
wolfspraulyes we should use the Ben as the basis for future products22:59
wejpso if ti really has a linux team with more than 1000 people, they should probably fire them22:59
wolfspraulbut I will not do that with Ingenic being completely agnostic to it23:00
wolfspraulthen I rather spend more time on our own free SoC23:00
wpwrakso i don't see a need for the soc maker to specifically have good sw skills. but someone in the stack needs them. you run into trouble of none of them does.23:01
wpwraki think what you're saying is that you tried to "sell" qi-hw to ingenic and they didn't find the idea interesting. and now you're unhappy. fair enough :)23:01
wpwrakyou could also go to quanta and "sell" qi-hw as the enabler for using those cheap ingenic chips. or to amazon.23:02
wpwrakin fact, the higher up in that stack, the better, because you can just ignore the mistakes at the layers below23:03
wpwrake.g., if ingenic provide some closed BSP, you simply don't use it23:03
wpwrakif quanta give you a 2.4 linux kernel, you smile and bin it23:04
wolfspraulyes fully agree about 'the higher the better'23:05
wolfspraulthat's why I focus on products right away23:05
wolfsprauland make the ben and m1 work better each day23:05
wpwrakon the other hand, if you're at ingenic, provide the coolest up to date and perfectly mainlined linux support, it all is useless if, say, quanta make a dead-end fork and amazon builds upon that23:05
wolfspraulnot unhappy about Ingenic, we keep talking but no need to start something that fails23:05
wolfspraulno Amazon controls the software stack23:06
wolfspraulI like your logic of "sitting higher allows to compensate for mistakes of lower levels" :-)23:06
wpwrakgood for them :)23:06
wolfspraulgotta run, bbiab23:06
--- Fri Dec 9 201100:00

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