#milkymist IRC log for Tuesday, 2011-11-15

wolfspraulwpwrak: ok I read the new calculation thanks to Joerg... but - what does it mean?01:08
wolfspraulthe proposed usb power switch would help?01:09
wolfspraulthe 4.0v reset ic?01:09
wpwrakwolfspraul: i'd go for both :)01:16
wpwrakthe new calculations only show the problem more clearly from the theoretical side. we already know independently (from the experiments) that the drop is too low.01:16
wolfsprauland with those 2 changes, what does it look like?01:17
wolfsprauladam is going for x-ray and reworks on the remaining 14 boards this week and next week if needed01:18
wolfspraulin parallel he will talk to the layout people in case they have any advice about dvi-i routing, also this week or next I think01:18
wolfspraulthen we need to source a bunch of things for the design verification, including the new reset ic and usb power switch01:18
wolfspraulbut for that we need a final proposal :-)01:19
wolfsprauljust braindumping to sync everybody01:19
wolfspraulkristianpaul: what about this clock signal you keep talking about?01:21
wolfspraulsince we cannot get enough energy behind a big rethinking of the expansion header, maybe we just run a wire to a new test point in rc4? what exactly do you need?01:22
wpwrakokay, plenty of time :)01:23
wpwrak(what does it look like) you mean the functional prognosis ? or the schematics delta ?01:23
wpwrakanyway, we have a new problem at the moment. full-speed usb mysteriously doesn't work, even though we think it should (after the latest code changes). not sure yet what's amiss. hopefully just software.01:25
wolfsprauloh sure, I am 100% aware :-)01:25
sh4rm4can someone recommend a "power flow" simulator ?01:26
wolfspraulthis project is so much fun, I feel every day my mind is just blowing out or something :-)01:26
wpwrak;-))01:26
wolfspraulalso how we can tackle this hardware business thing01:26
wolfspraulwhere is the continuity, where the costs, where the investments01:26
wolfspraulhow to get to great products01:26
wolfsprauland so on01:26
wolfspraulamazing01:26
wolfspraulbut if I mentally zoom back a year or two or three, and apply what I know today, I couldn't have gone any other way01:27
wpwrakfinding some investment that will allow to bring the system to a point where it's truly customer-ready would be a good first step :)01:27
wolfsprauljust nothing can replace actual experience in this field, I think01:27
wpwrakyeah, experience is critical01:27
wpwrakand tools :)01:27
wolfspraulI'm eternal optimist01:27
wolfspraulwe will find out what's wrong with full-speed usb01:28
wpwrakof course we will01:28
wolfsprauland worst case, everything up to rc3 is only low-speed :-)01:28
wolfspraulone thought i had last night was whether some of the non-working keyboards might be related to whatever unknown usb issues as well01:29
wolfspraulmany months ago when I tried to get m1 into a package that sort-of works out of the box, I started to buy a keyboard01:30
wolfspraulfirst very innocent, just got one at a corner shop01:30
wolfsprauldidn't work01:30
wolfspraulsecond one01:30
wolfsprauldidn't work01:30
wolfspraulthird one01:30
wolfsprauldidn't work01:30
wolfspraulall for different reasons01:30
wpwraki have one non-working keyboard left and that's the HHKB. M1 fails in two ways: 1) VBUS collapse (work-around: connect keyboard before booting) 2) keyboard is full-speed, which we don't support yet01:30
wolfspraulI threw away all non-working keyboards later, or maybe gave to xiangfu (forgot)01:31
kristianpaulwolfspraul: what i need seems to be very specific perhaps, a secondary system clock signal, i think dvi-i should be enought to fullfill this need at the end01:31
wpwrakif my other hhkb hadn't disappeared in the openmoko move, i'd also have one that doesn't work because it has an internal USB hub :)01:31
wolfspraulin hindsight we could have been a little more suspicious back then, but it's easy to say that with m1 having so many loose ends everywhere01:31
wolfspraulmost apple keyboards have built-in usb hubs01:32
wolfspraulthey all don't work01:32
wpwrakhub can't work without a hub driver01:32
wolfsprauldon't even know about low vs. full-speed01:32
wolfspraulyes sure01:32
wpwrakbeyond that, there's probably nothing overly magic. well, we couldn't hard-code device addresses in softusb, but that's not difficult to solve01:32
wolfsprauleven when we started sourcing those silicone keyboards we include now, we ran into quite a few that wouldn't work :-)01:33
wolfspraulhttp://en.qi-hardware.com/wiki/Milkymist_One_accessories#keyboard01:33
wolfspraulsorry the notes are messy01:33
wpwrakprobably full-speed and/or hub :)01:33
wolfspraullots of loose ends left lying around01:33
wolfspraulnah01:33
wolfspraulI did a lot of keyboard testing01:33
wpwrakfor mice, you have a similar set of issues:01:33
wolfspraulthere are multiple bugs/issues01:33
wolfspraulsome maybe clean-cut like usb-hub01:34
wolfspraulbut also others like maybe full-speed working 'sometimes' (see sebastien's comments), or even timing issues like we also found on the remote control side01:34
wpwrak1) possibly inrush current (not an issue in any M1rc3 but mine)01:34
wolfsprauland from the backlog I understand we are already fixing some usb issues, good...01:34
wpwrak2) full-speed. surprisingly, i haven't found any with that yet, but they must exist.01:35
wolfspraulif I source 10 different keyboards in China, I can guarantee you 5 will have different sorts of problems01:35
wolfspraulafter I was done, I discarded the whole pile of keyboards I went through, or maybe gave some to xiangfu, adding to his huge todo list01:35
wpwrak3) unexpected report format. there, we get 3a) old mice without wheel, 3b) mice high resolution, 3c) any other unusual report configuration01:36
wolfspraul"will sometimes issue multiple characters, or random characters (for example <l><return> when pressing only <return>), or spontaneous characters without any key being pressed "01:36
xiangfuthose keyboards under my desktop.01:36
wolfspraul:-)01:36
wolfspraulxiangfu: you also have that one with keyboard + mouse together, right?01:36
xiangfuyes.01:36
wolfspraulimplemented as 2 endpoints of 1 usb interface...01:37
wolfspraulI think Werner has something similar01:37
wpwrakthe report configuration is basically encoded in a little script the mouse sends to the host. softusb ignores that completely and instead uses a hard-coded layout. i added two special cases to that hard-coded layout to support more devices. it's not any less dirty than it was before, but at least more things work ;-)01:37
wolfspraulxiangfu: can you post the lsusb of it?01:37
wolfspraulwpwrak: after my testing, I am sure we have several bugs, not just one. But good we are unraveling it now.01:38
wpwrak(multiple chars) that sounds weird indeed01:38
wolfspraulyes01:38
wolfspraulwe should have investigated01:38
wolfspraulshould should01:38
wolfspraulxiangfu may still have it01:38
wpwrak(kbd+mouse) it's one EP per interface but two interfaces per configuration. "interface" and "configuration" are USB terms ;)01:38
wolfspraulxiangfu has one as well, let's see whether his setup is similar01:39
wolfspraulxiangfu: can you connect that kbd+mouse combo and post the lsusb?01:39
wpwraki changed softusb to simply parse all the interface descriptors and collect the data. before, it exited after the first interface descriptor, using only whatever it found there01:39
wolfspraulone reason we stopped trying to source a combo was that it wouldn't work anyway, and I was exhausted pointing to too many missing features01:39
wpwrakoh, i forgot some more reasons for trouble:01:40
xiangfuhttp://downloads.qi-hardware.com/people/xiangfu/tmp/lsusb.keyboard.mouse.together.txt01:40
wolfspraulperfect, thanks a lot!01:40
wolfspraulxiangfu: can you add it to this table? http://en.qi-hardware.com/wiki/Milkymist_One_accessories#keyboard01:41
wpwrak4) EP different from 1. 5) composite device01:41
wolfspraulwe can also add the one we include now01:41
wpwrakxiangfu: it may work now01:41
xiangfuwpwrak, I will try.01:42
wpwrakxiangfu: that is, unless it's full-speed. lsusb -t will tell you01:42
xiangfu    |__ Port 1: Dev 4, If 0, Class=HID, Driver=usbhid, 1.5M01:43
xiangfu    |__ Port 1: Dev 4, If 1, Class=HID, Driver=usbhid, 1.5M01:43
wolfspraulxiangfu: can you check whether the greenasia ic keyboard comes out at full speed?01:44
wpwraklow-speed. perfect.01:45
wolfspraulwpwrak: thanks a lot, I added your mice comments http://en.qi-hardware.com/wiki/Milkymist_One_accessories#mouse01:48
wolfspraulso what we had there before was too optimistic "all mice should world"01:48
wolfspraulwork01:48
wolfspraulwell, I think we were subconsciously clear that if we would do serious testing with lots of mice bought in the market, we would find some troublemakers01:49
wolfspraulguess you kill our dreams...01:49
wolfspraulno need to mention the in-rush case because it's temporary and only on your board01:49
wpwrakyup, i have a privileged configuration ;-)01:51
xiangfuwe can use the 'meta' key to emulate mouse under keyboard (in case someone don't  know that)01:51
xiangfusorry. it's the 'Win' key in most keyboard.01:52
xiangfuWin + Enter = left click01:52
xiangfuWin + L/R/U/D = mouse move01:52
wpwrakthat "win" key isn't the best possible choice for compact keyboards. there, it may be Fn+something and conflict with Fn+cursor being something else01:53
wpwrake.g., my Rii RF keyboard does just that01:53
wpwrakbut anyway, minor nuisance. particularly now that the pad on that same keyboard works ;-)01:54
wolfspraulxiangfu: can you find the greenasia ic keyboard and check whether it's full-speed or low-speed?01:57
wolfspraulthis other one that is sputtering out keys and spontaneously typing may also be interesting, but a little hard for you to debug without monitor equipment01:58
wpwraki still have one mouse that doesn't move smoothly. it's that old mechanical mouse. under linux, it works fine.01:58
xiangfuwolfspraul, I am not at home now. so I will do that later afternoon or tonight. and update the wiki.01:58
wolfspraulah sure, great01:58
wolfspraulno rush01:58
wolfspraulRome wasn't built in a day01:58
wolfsprauland neither is Milkymist :-)01:58
wpwrakmaybe we should have a page for keyboards and for mice each ?01:58
wolfspraulxiangfu: that's a saying "Rome wasn't built in a day"01:59
wolfspraulha ha, there's some stuff here for Werner... http://en.wikipedia.org/wiki/Rome_wasn%27t_built_in_a_day02:01
xiangfuwolfspraul, one issue.02:01
wolfspraulsuch as ... "Haste maketh waste."02:01
xiangfuwolfspraul, the qi-hardware.com run out of space.02:01
xiangfudev/vda1             49445508  44571324   2362424  95% /02:01
wolfsprauloops02:01
wolfspraulI have some server upgrades scheduled, but maybe we find a quick fix first02:02
wolfspraullet me check some logs...02:02
xiangfuwolfspraul, I try to copy new ben images from fidelio to en.qi-hardware.com.02:02
wpwrak(heywood) a busy man ;-)02:03
wolfspraulHaste maketh waste02:04
wolfspraulnot bad, eh?02:04
wolfspraulI can see you saying this with your typical nice smile!02:04
wolfspraulxiangfu: I deleted some logs, 700 MB02:04
wpwraki have some 2+ GB of .MOVies02:06
wolfspraulall fine02:08
wolfspraulthe job is on my side to upgrade asap02:08
wpwrakif you need more room, those .MOVs are my M1 demos and the uncut ben-wpan video02:12
wolfspraulno02:13
wolfspraulyour stuff are the crown jewels02:13
wpwrak;-)02:13
wolfspraulxiangfu: ok I made a little more space02:29
wolfspraul4.5 GB avail now02:30
wolfspraulthanks for the notice02:30
xiangfuI will copy another 1GB file later(this week). :)02:31
xiangfuwolfspraul,  great. thanks.02:32
Action: xiangfu update the m1 now by click 'webupdate'02:41
wolfspraulhe02:46
wolfspraulxiangfu: and? :-)02:46
wpwrakwolfspraul: btw, a PR idea: get M1 covered as "new concept" by the music press. not as a customer-ready product. and mention that the project is looking for investors to cover the work towards "polishing" it.02:51
wpwrakwolfspraul: the current PR suffers from being in a hurry for reaching profitability. this means that you present it as something it isn't yet, and even sink a lot of energy into giving it a glossy appearance (which may still be a worthwhile exercise for getting experience). haste makes waste :-)02:53
wolfspraulhmm02:56
wolfspraulyes I agree, but I think we do that already02:56
wolfspraulthe problem is that the world is full of people that talk about new 'concepts'02:56
wpwrakyes, but most of them don't have a nice prototype to show :)02:57
wolfspraulmeaning that they think they are the greatest and deserve more recognition for the genius they think they are02:57
wolfspraulyes02:57
wolfspraulso in order to break through this group, we chose to make a functioning and fun product ourselves, as a first step02:57
wolfspraulwhich is what we do02:57
wpwrakM1 is pretty advanced for a prototype, of course. but it still falls short of what most customers will expect.02:57
wolfspraulso now, either we like this product, and the fact that it's so good drives us forward, or we don't like it ourselves and then it deserves to die and everybody who ignored it until now made the right decision and saved themselves a lot of time from another crazy doomed project02:58
wpwrakand current PR is centered on placing it as a product, not as a call for investors02:58
wpwrakthat fact that it's good isn't enough to drive it forward02:59
wolfspraulif we don't like it ourselves, nobody will join02:59
wolfspraulI think most people are very sensitive to not join a dead-end effort03:00
wpwrakit still needs a lot of work before it's nice and rounded. take my experience in the last weeks. basically anything i tried to do with it, that was supposed to "work", failed miserably, with each experiment producing half a dozen patches03:00
wolfsprauland even though we look at one problem after another, I think the stuff that is out on the market and fights for customers is also not perfect03:00
wpwraknow, i can fix things that get in my way. but most customers will not even be good at reporting them.03:01
wolfsprauloh sure, I know03:01
wolfspraulthat's why most sales disappear into a black hole imho03:01
wolfspraulbecause the truth is they run into things that don't work for them pretty fast, and give up03:01
wpwrakperhaps it's that, yes03:01
wolfspraulwhere's nocarrier? :-)03:01
wolfspraulso yes, I agree with what you say about 'concept', just which concept exactly, and why people should believe in it. we have to proove that, nobody else will.03:02
wpwrakpressed for time, you also picked an odd use case as one of the main directions, namely the "automatic VJ". it took me a while to understand that M1 doesn't work like that.03:02
wpwrakalso sebastien doesn't use it like that03:03
wolfspraulyep our marketing is also evolving03:03
wolfspraulthe 'video tool' is much better03:03
wolfspraulonly that you have created a setup with host redirection etc. that is beyond the reach of most people03:03
wpwraki'd call it an instrument :) like a saxophone03:03
wolfspraulyes03:04
wolfspraulso the output of the instrument is a video signal?03:04
wpwrakyup03:04
wpwrak(midi et al.) it's not only missing functionality but also making what is there work better. sebastien's approach is to work on a feature until it scratches his itch but rarely more than that. of course, that makes sense for the proof of concept, because there's a LOT of territory to cover.03:07
wpwrakbut when you bring it to customers, you need to allow for variations. you can't just ship a box that contains M1 plus the whole world surrounding it.03:07
wpwrakeven though you're partially doing that ;-)03:07
wolfspraulhere you see some of our wild attempts at headlines... http://en.qi-hardware.com/wiki/Milkymist_One_marketing#Headlines03:08
wolfspraulthe 3rd one is "instrument"03:08
wolfspraulnot a good line though03:08
wpwrakas you broaden the coverage of scenarios for the existing functionality you converge towards covering a large part of real-life environments03:08
wolfsprauloh not whole world, definitely not03:09
wolfspraulyou are right on that too - the accessories have to relate to the main marketing message03:09
wolfsprauland have to create value, make the product more usable03:09
wpwrak"from music to VJing in 3 minutes" is bad. all the things that emphasize that it's no effort to use it well are silly.03:10
wpwrak"visual power unleashed" is better ;-)03:10
wpwrakthe ""instrument" doesn't have to be in the headline. but it has to be clear in the context03:10
wpwrakotherwise you get those "USD 500 screensafer" comments ;-)03:10
wpwrakthat one was actually great. that guy understood the (all wrong) marketing message perfectly ;-)03:11
wpwrak"cool animations out of the box" is also bad. again, it suggests a triviality that isn't there.03:12
wpwrak"programming high performance visuals" and "the highest performance visuals, accessible in an easy programming language" position it towards sw development. that's probably not what VJs want to hear first03:13
wpwrak"no need to be a VJ to get it work" oh dead. talk about alienating your customers ;-)03:14
wpwraks/dead/dear/ # freud :)03:14
wolfspraulbtw, steve's good old Creative Labs dabbled in this too, called 'lava' http://connect.creativelabs.com/emu/lava/lava.aspx03:14
wolfspraulah no, you are too strict there. think about "VJing for dummies"03:14
wpwrak"hardware acceleration ala carte" is lame. everybody has hw acceleration :)03:14
wolfspraulwe would get sued...03:14
wolfspraulbecause it's such a successful franchise to call something "for dummies" and people will love it03:15
wpwrakwell, "for dummies" may work because the franchise downplays the dumbness aspect03:15
wolfspraulhttp://www.mycreativefansite.com/2008/10/some-history-lava-music-in-motion.html03:16
wolfspraulxiangfu: how did the update go?03:17
xiangfuwolfspraul, one moment. I use a un-stable release. just reflash to rc3. now try to update.03:17
wpwrak(lava) "the emm pee three industry" ;)03:18
wolfspraulI only gave you those links as an example how other people before us tried to market/talk about this kind of features.03:21
wolfspraulit didn't work...03:21
wolfspraul:-)03:21
wolfspraullight instrument03:21
wolfspraulactually I've heard a couple of times how people like the effect of m1 in a dark room where m1 is the only light source in the room03:22
wpwrakthey lacked the itunes store03:22
wolfspraulI meant those texts, how they tried to market it.03:23
wpwrakyes yes. a bit similar to M103:23
wpwrakbut all virtual03:23
wpwrakso there's a difference. do aspiring rock musicians get some iGuitar app or do they buy the real thing ?03:24
xiangfu(Failed to download version information)03:24
wpwrakwolfspraul: anyway, the marketing message is just one thing. the other aspect is that M1 needs more work. first of all, to make the existing features real-world usable. and then also to create a proper workflow or set of workflows. like the control architecture i described.03:28
wpwrakat some point in time, you'd also have to take this to real VJs (or go VJing) to see if it holds up in practice03:28
wpwrakbasically get the nasty bits out of the way before you try to reach regular customers and reviewers03:29
wolfspraulxiangfu: [failed to download version info] :-) and now?03:29
wpwrakif you go to reviewers now and present it as a finished product, you just confuse them.03:30
wolfsprauldid dhcp work? you got an ip address?03:31
xiangfuwithout VPN open the version file under bowers needs ~4 seconds.03:31
xiangfuyes. I got the correct IP address.03:31
wpwrakIMHO, the whole "connect it to some idle screen" meme has to die. it's like advertizing some yamaha synthesizer on the strength of its automatic percussion03:31
wolfsprauland still not working?03:31
wolfspraulyes agreed, but we are just at the very beginning of making it work with controllers03:32
wolfspraulhave to have some patience sometimes03:32
wolfspraulthe only thing that works today is what you know works :-)03:32
wolfspraulreally works03:32
wolfspraulwhether you like the power of that or not03:32
wpwrakexactly. so ... it needs more work before it's ready.03:32
wolfspraulwith the controllers we are on a very good path to reach the next level03:33
wpwrakhowever, you're marketing it as if it was ready now. even go to reviewers with that premise.03:33
wolfspraulas soon as I hear some positive news from you or Sebastien, I will buy a bunch more to accelerate this03:33
xiangfuthe curl not work well when Internet unstable. we have to improve the download code to fit all Internet-environment03:33
wpwrakdon't get me wrong. i think we are on a good path. what i'm saying is that this path goes on for a bit.03:33
wolfspraulyes03:34
wolfspraulxiangfu: ok, not robust enough03:34
wolfspraulgot it03:34
xiangfuI will test later and report. I never test the 'webupdate' outside.(always test that at home)03:34
wpwrakand then the question is how the project will stay on that path. right now, there are extremely few people working on it and none are on a particularly sustainable path.03:35
wpwrakthat's where external investment would help. i know you don't like the idea. but would you rather have the project fade and die ?03:36
wolfspraulno 'don't like'03:37
wolfspraulJon and I are working on it03:37
wpwrakdevelopment needs a certain minimum speed to work. if things drop to the level of sebastien spending a few hours every other weekend, M1 will never be ready.03:37
wolfspraulagree03:37
wolfspraulthen it will be added to the pile of failed 'open hardware' projects :-)03:37
wpwrakyeah03:38
wolfspraulI think Sebastien plans or is already working on serious polishing though03:38
wolfspraulok gotta run, bbiab03:39
wpwrakyes, i think he has now returned to work full-time on some features03:39
wpwrakbut i suspect that the next "real work" will have to happen sooner or later. well, maybe he's got a pot of gold somewhere that sustains him economically :)03:40
xiangfu(webupdate) not robust enough.04:26
wpwrakwhat trips it ?04:30
wpwrakback to USB. incoming message is now [~SOF15F][~SETUP0/0][~DATA0:00.05.01.00.00.00.00.00]. so far, so good.04:42
wpwrakcomparison with PC: [~SOF180][~+ also good  (the ~ are differential ones before the beginning of the SYNC pattern)04:49
wolfspraulback :-) translating to action mode... first I will reorder the headlines as per what we like best now05:08
wolfspraulwpwrak: btw I think you see m1 too 'glass half empty' now, too negative. I think it's a good and worthwhile start and we can attract new people and don't need to be afraid of the invitational effort.05:10
wpwraki think you misunderstand the nature of my criticism05:10
wolfspraulin your feedback of the headlines, all but one are negative. so I will move those lower, and the one you liked a little to the top.05:11
wpwraki agree that it's a good and worthwhile start. in fact, it's amazing what it can do already.05:11
wolfspraulok, now we need to find new and more friends05:11
wolfsprauland some rich friends too, ideally :-)05:11
wpwrakbut you're pushing it too aggressively now. you're pretending it's more finished than it is, pushing it to customers or reviewers that will review it as if ready for customers.05:12
wpwrakyes, exactly ;-)05:12
wpwraki think that's at the root of the misrepresentation - the hurried need to make money05:13
wolfspraulok but don't babysit people either. there's a lot that perfectly know what they get, and like to experiment with new stuff.05:13
wolfspraulI moved 'visual power unleashed' to the top05:13
wolfspraulnow, I try to email heise about that, let's see :-)05:13
wpwrakand i think you should make it clear that the project is looking for investment05:14
wpwrakinvestors interested in that segment will probably read the news. so if you can get some "future tech pre-review", this reaches them.05:15
wolfspraulI want a customer who orders 1000 m1, not an investor.05:27
wpwrakyes, and i want a pink unicorn that breathes fire ;-)05:28
wolfspraulthe difference is subtle, but when I go around I look for customers... I think that's the way that can work. Maybe Jon is a little more focused on plain 'investors', he has to speak up about the exact nature of whom he tries to find.05:28
wolfspraulI believe we will not find an investor unless that person also buy and believes in the sellability.05:28
wolfspraulanybody who studies the project will determine the stage of it, and understand that to lift it to the next level as a product is the same as 'investing'05:29
wpwrakyour customer segment is very limited. there aren't a lot of people who will buy something that's still "under construction"05:29
wolfspraulbut in order for this to become real, it has to be a strong buyer or distributor who has the financial ability to order (and sell)05:29
wpwrakthey'll just wait until you finish05:29
wolfspraulthat's rare enough05:29
wolfsprauloh no, they would have the product first, they can have it exclusive, etc.05:29
wpwrakthat's where investors enter the picture. they finance the phase from great idea to sellable product05:30
wolfspraulthe problem is that a lot of buyers are not that involved, in fact they would expect financing and risk to be taken elsewhere05:30
wolfspraulI don't think the type of investor you have in mind exists.05:30
wpwrakno distibutor will pick it up if it still needs half a year of development05:30
wpwrakmaybe what you're looking for is a distributor who thinks and acts like an investor05:31
wolfsprauldon't get hung up on 'distributor'05:31
wolfspraulthere are many businesses around05:31
wpwrakbut then, why not just look for an investor ? ;-)05:31
wolfspraulbut I'm saying unless the 'investor' we are looking for can imagine to suddenly sit on 1000 m1, he will not invest05:31
wolfspraulI will give him m1 for his dollars05:31
wolfspraulif he doesn't like that, he will not invest anyway05:31
wpwrakthe investor will probably have several steps in mind: certain R&D objectives. then, pre-production evaluation. if the device passes, you get the money to produce these 1000 units. or maybe 200 first and then re-evaluation. whatever :)05:33
wolfspraulwe will have found our investor once we have found a customer who wants to order 1k or more units05:36
wolfspraulthat's the same person05:36
wolfspraulone thing I foudn recently in terms of investing was the latest OpenPandora moves05:37
wolfspraulthe OP team continues to amaze me :-)05:37
wpwrakno. the customer will want a finished device. but you don't have that.05:37
wolfspraulthey have ripped off their customers several times, and are now raising money from small private guys on a contractual basis05:37
wolfspraulbypassing the typical VC or banking world05:37
wpwrakand you won't have a finished device without more money.05:37
wolfsprauland it seems to work, they think they can collect ca. 350k EUR that way05:37
wolfspraulno wrong05:37
wolfspraulevery large customer knows very well how much money needs to be invested05:38
wpwraksebastien will only be able to work full-time for so long on it. and since he'll be working almost alone, it will be hard to motivate him.05:38
wolfspraulor do you think they fall for any of our "it's complete" bs :-)05:38
wpwrakwell, what is the product you're trying to sell ? M1 ? or the project ? the latter could work. that's something "sellable", even if still in progress05:39
wolfspraulhttp://boards.openpandora.org/index.php?/topic/5860-contracts-sent-out-parts-hopefully-moving-in-2011-10-26/05:39
wolfspraulif that works, it's something to learn from (not the 'rip off your customers' part, but the fact that you can bypass professional investors by going the community route)05:39
wolfspraulno the question is what they want to buy05:40
wolfspraulI propose m1 right now, sure05:40
wolfspraulwe can make an increasingly good pitch for that05:40
wpwrak(community) well, why not :) some people probably have heaps of money laying around and they may like the idea better to invest into a worthwhile project than giving it to their banksters05:40
wpwrakfine. so they'll listen to your increasingly good pitch. then they'll go home and wait for how good your pitch will be in three months :)05:41
wolfspraulthat's also a way to combine people who believe in the product with those that help finance and produce it05:41
wpwrakif you die until then, well, it's just one lost opportunity05:42
wolfspraulevery investor can do that, what's your point?05:42
wolfsprauleither they think it's an interesting product that helps their business, or not05:42
wpwrakdistributor = buys finished things. investor = buys into projects05:43
wolfspraulcash by definition is the most liquid form of value, meaning you think hardest moving out of it :-)05:43
wolfspraulonce you are out of it, getting back to cash is hard enough...05:43
wolfspraulI am looking for a customer who orders 1k or more m1, I think it's the easier route than looking for someone to just send money (invest) with whatever ultimate goal in mind05:44
wpwrakan investor could in theory also be a distributor, but that's an uncommon combination. they have different fields of expertise.05:44
wolfspraulcustomer, better05:45
wolfspraulmy 2c05:45
wpwrakbut you do understand that M1 is not in a shape where a reasonable customer will buy 1k units ?05:45
wolfspraulthere are tons of successful businesses, I don't think what you mean with 'distributor' is what I mean05:45
wolfspraulI think we are close05:46
wolfsprauland that customer would understand that the delivery time is 3 months or so and we will spend part of his money on continuing bugfixes05:46
wpwrakwould you have said the same 2 weeks ago ? :)05:46
wpwraki don't think your "customer" exists :)05:47
wpwrakwhy not call him "investor", because he'd have to behave exactly like one ;-)05:47
wpwrakincluding trying to get some control over the development process05:47
wpwraki think that's what you're afraid of. having to give up some control05:48
wpwrakmaybe you can find a customer who will pre-order 1k units. sign a MoU. pay on delivery. does that help you ?05:49
wpwrakmaybe you can take that MoU to the bank and get a loan05:49
wolfspraulno they pay 50% when ordering, and 50% before shipping05:50
wpwrakwhy should they ?05:50
wpwrakthey may do that if the product is finished and all you need to do is the manufacturing. low risk.05:51
wpwrakbut it's not there yet. you can't show the finished product yet.05:51
wpwrak(low risk) well, lower risk. still messy, as we know.05:51
wolfspraulone by one05:51
wolfspraulbusiness is about risk05:52
wolfsprauleverybody takes risks05:52
wpwrakagreed05:52
wolfspraulthat's why any serious customer first wants to see the product, test it05:52
wolfspraulANY05:52
wolfspraulguaranteed05:52
wpwrakprecisely05:52
wolfspraulwhy do they accept 50/50 terms - because those are our terms, unnegotiable05:52
wolfspraulwe are a manufacturer, not a bank05:52
wolfspraulthe entire odm/oem industry works on such terms05:53
wolfspraulsomeone is doing just the manufacturing, that company is paid partially when ordering, partially before delivery05:53
wolfspraulsomeone else is doing financing05:53
wolfspraulthat company carries the return risk too, they are usually paid by the retailer 30 or even 60 or 90 days *after* the actual sale in the store :-)05:54
wolfspraulminus returns, minus fees, minus whatever05:54
wolfspraulI cannot and do not want to play that role.05:54
wpwrakso far, so good05:54
wpwrakbut that's manufacturing05:54
wolfspraulyes, I can manufacture this thing05:54
wolfspraulthat's all05:54
wolfspraulfor normal manufacturing terms05:54
wpwrakbut you can't design it05:54
wolfspraulkeep in mind all files are open, so if someone doesn't like this - manufacture yourself!05:54
wpwrakhah ! straight from the linux kernel:05:57
wpwrak /* TRSTRCY = 10 ms; plus some extra */05:57
wpwrakmsleep(10 + 40);05:57
wpwrakweird. now i just managed to hang navre, it seems06:11
wpwrakhmm. how big can the navre code get ?06:14
wpwrakah. 11 address bits, each for a 2 byte word. and "size" says we're at 4 kB.06:16
wolfspraulwpwrak: what control do you believe we (or me) retain that may be bad for the product/project?06:29
wpwrakwhat i meant is that i suspect you don't want to find an investor because you're afraid of losing some control over the project06:31
wpwrakregarding jon's search for investors, i think he's going in a totally different direction. nothing directly related to technical development, more related to the very big picture. M1 is a mere drop in that ocean.06:35
wpwrakthat doesn't mean it couldn't work, but it's something where the strength of M1 is of limited use06:36
wpwrakwhen looking for R&D investment on the other hand, you have something very convincing to show right now06:36
wolfspraulloosing which control?06:40
wolfspraulbtw, we should also not underestimate the value of having produced over 130 units by now06:40
wolfspraula run of 500 or 1k is by far not completely crazy06:40
wpwrakover what you do what commitments you make, etc.06:40
wolfsprauldon't understand06:41
wolfspraulI don't like to receive investments based on over-promising, yes, true.06:41
wpwrak(130 units) oh, sure06:41
wolfspraulso I need to understand the upside potential06:41
wolfspraulback to those 130 units, they have given us incredible insights into the technology, and also users06:42
wolfsprauleven if we are still dealing with a lot of silence06:42
wolfspraulthere was and is no other path06:42
wolfspraul136 to be precice, phew06:42
wolfspraulprecise06:42
wolfspraultoo hard :-) hardware is too hard...06:42
wpwrakmaybe. perhaps a few more eyes could have helped. as well as some more hands. not a lot. just a little.06:43
wpwrakbut in any case, what's important is not the past but the future06:44
wpwrakwhat you have is a "half-finished" product. and you have demonstrable production experience. you have or know people who you believe can bring the product to completion.06:46
rohwpwrak: i think what we have is 'a plan' aswell as enough experience for an educated guess about the short to medium future possibilities of getting stuff done06:47
rohanything else can change at any time (availability on nonstocked parts, people walking away due to financial pressure etc)06:48
wpwrakroh: yes, the plan is there, too. probably needs writing down, with numbers06:48
rohmaybe. i see the key in customers. make the numbers go up, the cost will go down, profit margin rises a bit... -> maybe one could pay people doing more work then ;)06:49
wpwrakyes, but that's not product development. then it's just a hobby06:49
rohinvestors are also just 'weird banks' .. they want to see something in return. so calling people who get hardware in return customers fits better06:50
wpwrakthat's wolfgang's strategy. where's the revenue ? :)06:50
rohselling the hw.06:50
rohthats why i would also call it customer.06:50
wpwrakbut he's not selling nearly enough hardware06:50
rohwell.. its very weird special case hw06:51
wpwrakand he doesn't have the resources to improve the hardware06:51
wpwraksee the ben. abandoned.06:51
rohmaybe we need some more simple products to make the brand known? well.. thats guesswork. i am not a salesman ;)06:51
wpwrakand yes, of course investors want something in return06:52
wpwrakthey want a share of the profits when the thing finally starts selling. probably a big share.06:53
wpwrakit's obviously better if you don't need external financing (as long as you succeed in the end)06:53
wpwrak(simple) arduinos ;-)06:53
rohtrue06:54
rohabout ben.. i see the major problem on the mechanical side (costly to develop, not much 'inhouse' expertise) as well as the cost06:54
wpwrakthere's nothing in the ben we wouldn't want to be able to master. it's even a simple device, also mechanically06:55
wpwrakand yes, it's costly to develop. anything is costly to develop :)06:56
wpwrakM1 has several sebastien-years of R&D in it. that's already a large value. even if we discount the student years, that's still 1.5 calendar years of work. maybe he'd make EUR 100k per year doing similar work in industry. so 150 kEUR, just opportunity cost on his side.06:59
rohtrue. quite a lot investment in there already, especially in lifetime of the developers07:00
wolfspraulben is not abandoned07:00
wolfspraulI know because I keep paying the bills07:00
wpwrakthen we have adam working pretty much full-time on it. xiangfu part-time. wolfgang, yi, jon, lars, mwalle, me a bit, various other contributors07:00
wolfsprauland very much like to do that07:01
rohand with simple i meant somethign much simpler as the ben. maybe without case07:01
wolfspraulben is off-topic in this channel, hope Sebastien doesn't mind :-)07:01
wolfspraulso far we have been unable to involve Ingenic in any serious way07:01
wpwrakroh: then the cost of prototype development07:01
wolfspraulI will continue to work on that07:01
wolfspraulI agree with roh that the ben mechanical is a blocker for further progress, that's my #1 thing now07:01
rohwpwrak: sure. but for simple devices that sinks enormously07:01
wpwrakroh: i mean on the investment side of M107:02
wolfspraulthe value of the Ben software is high and increasing imho, because I think for most manufacturers and products, Android is a bad trap07:02
rohe.g. for a device without case (thing the arduino-zoo and add on boards for an example) you only need a few bucks... similar development like atben/atusb07:02
wpwrakroh: M1 probably already has the equivalent of maybe 2 MUSD of R&D value. maybe more.07:03
rohmeans one or only a few persons complete development in one location, including dvt runs, and then the pvt etc happens whereever the real volumes are made.07:03
wpwrakwolfspraul: it's deprecating because it's tied to a dying platform07:03
wolfspraulthe notebook market has zero money left for innovation07:03
wolfspraultablet still has money, but will be just the mainstream consumption device, especially if Amazon can successfully reduce the tablet price to 199 USD (to be decided next year)07:04
rohwolfspraul: true. and devices are dead-ends on their latest vendor supported kernels etc. no fixes, no security fixes07:04
wolfspraulwpwrak: really not true. the openwrt basis is great.07:04
wolfspraul'dying platform' = what? Ingenic?07:04
wolfspraulyes, with Ingenic I will continue to push for a sizable investment actually07:05
Action: roh is developing home-automation stuff atm. avr based. redneck home solding compatible technology. most of it even single sided.07:05
wpwrakwolfspraul: the ben. how popular is the ben openwrt on other devices ? in terms of people using it ?07:05
wolfspraulother devices?07:05
wpwrakwolfspraul: other ingenic devices07:05
wolfspraulshould we move the ben chat over to #qi-hardware ? people there will miss it07:06
wolfsprauloh I'm on that07:06
rohto be fair... my main blocker with the ben (motivational wise) are its limits for my daily use. i simply dont have one to be earnest07:06
wolfspraulbut Ingenic has extremely few brain cells to think about software, or even understand what's going on :-)07:06
rohwolfspraul: ok07:06
wpwrakalas, it still doesn't seem to work08:13
wolfspraulwhat doesn't work?08:14
wpwrakfull-speed USB08:16
wpwraki'm still not getting a response08:16
wolfspraulah, the delay08:26
wpwrakno no, no delay08:30
wpwrakthe usb device simply doesn't anser08:30
aw_http://en.qi-hardware.com/wiki/Milkymist_One_run_3_schedule#BGA_Balls_Inspection08:31
aw_just went to smt factory there to do some works for bga inspections. ;-) phew...08:31
wpwrakaw_: sucks, eh ? ;-)08:32
aw_retrieve two m1 boards back. ;-)08:32
aw_i'll continue working on this. not finished all.08:33
aw_still have some pictures not all uploaded. will do soon.08:34
wolfspraulexcellent08:36
wolfspraulwpwrak: sorry yes, I meant the delay you found in the Linux sources didn't make the problem go away on m108:38
wpwrakah yes. that's the slightly disappointing result. well, it'll make things easier in the future :)08:39
wpwrakthe next step will be to examine what the device really receives. but .. first a nap :)08:42
lekernelkristianpaul: if it's just a clock input and you don't have anything to synchronize it to, clock_dedicated_route=false has no side effect except waste a negligible amount of the FPGA routing resource09:04
lekernelwpwrak: does the device answer after that 50ms delay (FYI: I had implemented something like that at the beginning, but I removed it because the spec also says the devices go to suspend mode after no activity has been detected for a few ms... wtf?)09:05
stekernwhat 50ms delay are you referring to?09:20
lekernelhttp://lists.milkymist.org/pipermail/devel-milkymist.org/2011-November/002145.html09:23
stekernaah, ok, thanks. but surely that "go to suspend mode" is not valid during a bus reset?10:00
lekernelthat's immediately after the bus reset and before the first transaction10:14
lekernelmaybe we should send SOFs/keepalive during that period?10:14
stekernI would assume so10:18
wpwrak(idle) ah, i wasn't sure about that. well, nothing easier to try than that :)11:51
wpwraklekernel: and no, the device doesn't answer. but ... my device doesn't implement suspend anyway11:52
wpwrakwolfspraul: (c't) nice headline11:57
wpwrakwolfspraul: hmm, faderfox's M1 has run into a "shipment exception". never saw those "details". each country its own weird procedures :)12:00
rohheh, our hackspace is listed in the recent ct hardware hacks (special issue)12:02
wolfspraulshipment exception, let's see12:07
wolfspraulI had doubts about the low value, German customs are bean counters12:08
wolfspraulhey, lies mal die antwort von heise :-)12:09
wolfspraulsorry, German12:09
wolfspraullemme reply to that as well :-)12:09
wolfspraulI think this reply is very interesting12:09
wolfspraulforwarded to you12:09
wolfspraulwhat do you think?12:10
wpwraklemme check ...12:11
wpwrakkaraoke machine. hehe ;-)12:11
wpwraki guess the marketing message this far may have left some impressions. with M1 being positioned as a device with only multimedia input but no significant controls of its own.12:17
wpwrakand yes, the low video in and out specs are coming back to haunt us once more.12:18
wpwrakthey're also one of the reasons why i think we should go slowly with approaching customers. because if we decide to fix the video in/out quality, it will take time.12:19
wolfspraulwait wait12:20
wolfspraulfirst of all he spends his time and tries to understand what this is12:20
wolfspraulwhich is great12:20
wpwrakunfortunately, it's not only the specs. you can see the problems quite clearly. maybe we can hide some of the video out issues, without upgrading overall specs. lekernel's plan sounds promising. but it's still a bit of work.12:20
wolfsprauland he is immediately misled12:20
wpwrakis he ?12:20
wpwrakmisled i mean12:20
wolfspraulfrom our perspective, sure12:20
wolfspraulif the end result is that a notebook is better, then we misled people that looked at our stuff in good faith :-)12:21
wpwrakokay, the karaoke machine is a little odd :)12:21
wolfspraulbecause of course we cannot win the notebook comparison12:21
wolfspraulhe tries to understand what this strange box is12:21
wolfspraulhe just cannot imagine that someone comes out with a new machine with these specs in 201112:21
wolfspraulso he thinks it must be something completely crazy out of Japan :-)12:21
wolfspraulI think that's very understandable and funny12:22
wpwrakyes, but part of the specs is not just poor specs. it's poor results. let's not pretend everything is perfect.12:22
wolfspraulso which way do we go now in explaining?12:22
wolfspraulthe "oh, but it's all gpl" route?12:22
wpwrake.g., i have yet to see a usable camera image12:22
wolfspraulah, tell me more about camera issues12:22
wpwrak(part of that may be equipment, lighting, etc., though)12:22
wolfspraulyou think we have a problem somewhere?12:23
wolfspraulfirst time I hear that from you12:23
wolfsprauland for the heise guy - how to answer his questions?12:23
wpwrakcamera images have low resolution and bad color contrast12:23
wolfspraulhe is open-minded to carrying this thing, that's why he writes back12:23
wolfspraulbut right now he doesn't get it :-)12:23
wpwrakthe color contrast may be an equipment/lighting issue, though. if i get people to stand still long enough to shine some strong light into their faces, the images get better :)12:24
wolfsprauldo you want to try to reply?12:24
wolfspraulyour way to answer is often breadth-taking, maybe he likes it :-)12:24
rohhm. i havent seen/played with the camera now packed.. is it that bad?12:24
wolfspraulnah, no way12:25
wolfspraulI did a 1-month long camera comparison, must have compared 20+ cameras12:25
wolfspraulfor an analog camera, what we include is a really nice one, 1/3'' Sony CCD sensor etc.12:25
wpwrakthe low camera resolution is part low specs, part the soc only using one of the two half-frames, and maybe also due to poor color contrast (which acts a blur. perhaps also caused by composite video. that's why i once asked about using r/g/b split)12:25
rohhm.12:25
wolfspraulof course it cannot beat a HD/hdmi/whatever cam12:26
wpwraki tried two video sources: the bundled camera and my canon still image camera (with composite out)12:26
rohok.. the only one field issue is evil.. that halves vertical resolution. which may already be only ~400 lines12:26
wpwrakroh: yes ;-)12:26
wpwrakof course, merging the two fields is hard, as everyone knows. and afaik there's no progressive composite. so it comes down to composite being the limiting factor.12:27
wolfspraulyou can also hook up the cam to a TV if you have one, see whether that comes out better or not12:28
wpwraknot sure if separate rgb is also interlaced or not12:28
rohtrue. simply said.. if its analog video and not hd, its progressive (99.9% of the time)12:28
wolfsprauland what did the comparison with your canon show you?12:28
wpwrakcanon: image quality still not convincing12:28
rohwpwrak: also interlaced, usually. i know very few devices which can work with progressive content at all12:28
lekernelwhat? milkymist is on c't?12:29
wpwrakbut i didn't go too deeply yet. just noticed that it's bad12:29
wolfspraullekernel: no it's not on c't, but almost ;-)12:29
wpwraklekernel: not yet :) we're trying to shape the message12:29
rohhm. is the video adc too simple? or whats the issue?12:29
wolfspraul(though I don't think we can pass that barrier, to be honest. not right now...)12:29
rohi mean.. there isnt much schematic from input to chip to do something wrong.... ;)12:29
wolfspraulI have fairly good contacts at c't, well, at least they listen to what I have to say and they know I'm a serious guy12:29
wpwrakroh: the adc does everything right. but you still need to piece the two fields together12:29
rohmaybe only setting proper adc register values will fix it?12:30
wolfspraulbut no way I can make them write what I think they should write ;-) (and that's perfectly fine)12:30
lekernelthe 7181c may improve color btw12:30
wolfspraulin fact I highly value their feedback, since they deal with news all the time, if we believe being carried there is a big win for us (which I am not so sure about), then we have to pass their checks...12:30
wolfspraulwpwrak: do you want to reply?12:31
wpwraklekernel: how's your experience with video in so far ? does it look "good" to you ? or more like "good enough to leave it alone for a while and rush to the next fire that needs fighting" ?12:31
wolfspraulI'll contemplate for half an hour or so :-)12:31
lekernelas you said, color is often not enough12:32
lekernelthat's the main problem I have with it, and I hope the 7181C would improve a bit12:32
wpwrakwolfspraul: yo're better at smooth writing. i'll be happy to reply with proof-reading, though :)12:32
wolfsprauland I love your arguments so much12:32
wolfspraulI think you have some way to write to him that leaves him no chance but to feature Milkymist all over their publications for a month!12:33
lekernelresolution issues, not really. it doesn't matter that much when it goes through the average beamer...12:33
wpwrakmmh, 240 lines on an XXXXx768 or better unit ? that's almost 1:4. you ought to notice, even if it's blurred and such12:34
lekernelI don't want to write a deinterlacer. if I want to spend time on something pesky, I'd rather design a HDMI input.12:35
lekernelboth are equally annoying, with the HDMI input giving a far better result in the end :)12:36
rohsure. but which cams do have live hdmi out?12:37
rohwell.. in the 'reasonable price' range12:38
wpwrakmaybe we can have composite and hdmi in ? :)12:39
wpwrakwhat kind of output DO cameras actually have ? i mean the not-too-expensive consumer cameras12:39
rohcvbs and maybe svideo12:40
rohsvideo already helps a lot12:40
lekernelI think there are a few that have HDMI, no?12:40
rohnothing below a few hundred bucks12:41
rohhdmi is designed for the living room. from pc/gaming console/set-top-box to tv.. not really for 'inputs'12:42
rohalso you have that sucky hdcp issue...do you?12:42
lekernelso, what do you recommend for high quality live digital video from cameras?12:42
lekernelHDCP has been broken, no?12:42
lekerneland I don't think it's enabled for your personal videos12:43
rohlekernel: its not a technical but a licensing issue afaik12:43
lekernelwe did not sign anything12:43
wpwrakcan HDMI work without HDCP ? i.e., can you simply not encrypt and the receiver will accept it ?12:43
lekerneland it's not even HDMI, it's digital video input12:44
rohsure. but maybe you cannot buy the needed chips if you don't or so12:44
lekernelthe FPGA takes HDMI signals directly12:44
rohhdmi as output to a beamer is more common. but i havent seen it on consumer stuff for live camera video. maybe some high-end dslr do that12:44
rohprofessional video is all sdi12:45
lekernelSDI cameras are several thousand euros, no?12:45
rohone bnc socket. digital, uncompressed serial bitstream.12:45
rohyes.12:45
lekernelHDMI looks like a nice in-between CVBS/S-video and SDI ...12:46
rohsure.12:47
rohi'd recommend to first see whats wrong with the video. make svideo work with an mini-din to rca adapter.. add that to the package12:48
rohshouldnt be magic. maybe you can 'solve' it by some input switching logic and some static presets like more color gain or so12:49
rohfound this http://www.seiwaamerica.com/products/hdmi.html12:49
rohand http://reviews.cnet.com/digital-camera-reviews/?filter=501414_12818333_12:49
wolfspraulif we have a dvi-i connector, can we use vga for output and hdmi for input? (over that same connector)12:49
rohbut the latter doesnt say if thats a live video without osd or what is coming out of the hdmi12:49
wolfsprauloh, and I totally think we should bugfix/improve/support the hw we have now, rather than thinking about which other hw to switch to. first the former, then the latter.12:50
wolfspraulotherwise we burn our early adopters/customers, and I cannot think of any tech/brand that would have survived that12:50
wpwrak(make the most with the existing hw) agreed. but also keep on looking for low-hanging fruits for future M1rcX. like dvi-i out.12:53
wolfspraulif we have a dvi-i connector, can it also serve as video input?12:54
rohnope. not really the idea of that connector12:54
lekernelit's a hack, but it should work12:54
wolfspraulok, thanks12:54
wolfsprauljust curious12:55
rohelectrically.. well.. if you add a weird adapter cable.. but thats fishy12:55
wolfspraulnot weird, dvi-i adapter cables are very common12:55
wolfspraulmaybe not that one? I don't know12:55
wolfspraulanyway, my question is already answered, I understand it's a hack12:55
wolfspraulif we currently only decode 1/2 of the lines of our cam, I'd say that sounds like an easy way to improve a lot :-) (I may underestimate the amount of work it takes to make this work, I admit)12:56
wolfspraulthe uptick to 7181c is planned, so that's good12:56
wolfspraulwe can try to source a cam with higher output specs, but I understand that's not even the bottleneck today, so why bother... (right now)12:56
lekerneldeinterlacing is a massive mess12:57
wolfspraulI got the feeling above12:57
lekernelwill use at least 1/4 of the FPGA just for that, not to count man-hours12:57
wolfspraulto me cam quality is not a high-priority problem12:58
wolfspraulin fact I hear it the first time today from Werner12:58
wolfspraulwhat I hear a lot is still-pics, usb-midi12:58
lekernelno, it's not. and within cam quality problems, color is more important than resolution.12:58
wpwrakyeah, i think color is what you notice first. images look washed out.13:06
wpwrakadd to this the low color depth on the digital side and the quick propagation of rounding errors it causes, and the overall effect is much less than what should be possible13:07
ab5tractwolfspraul, i would like the ability to stream video from video files (which i guess is along similar principles to static images)13:12
ab5tractmixing live stream and pre-recorded13:13
ab5tractso, stream mixing (all inputs should be equiv: live, static, and AVI)13:13
ab5tractand by avi i mean MPEG-4, which i think the mm1 could handle decoding?13:14
lekernelhttp://opencores.org/forum,OC%20H.264%20project ahem...13:15
lekernelI've tried compiled ffmpeg for rtems - it does compile, but I don't know if it works and at what speed13:16
wolfspraulab5tract: realistically that's a tough development job and may take months13:17
wolfspraulI'm lowering expectations first :-)13:17
wolfspraulnow that they are lowered, let's find hackers!13:18
ab5tracthehe13:18
ab5tractwell, static images is a nice start13:18
wolfspraulm1 is definitely one of the most interesting computing platforms right now, I think that's fair to say13:18
ab5tractim assuming the best documentation is sebastian's thesis?13:18
wolfspraulmaybe you can go around and motivate some developers to have a look?13:18
ab5tractthough it occurs to me that it might be in french13:18
ab5tractyeah i can obviously try13:19
ab5tractand i will :)13:19
wolfspraulab5tract: thanks!13:24
wolfspraulwe can make a todo list that would make any computer science prof proud :-)13:24
ab5tractwolfspraul, that is an interesting angle actually13:29
ab5tractthere might be compsci departments interested in teaching on such a platform13:29
ab5tractthough i expect documentation would be an issue there (no textbooks, not sure what the reference manual for the processor looks like a la my question about sebastiens thesis above)13:30
wolfspraulI think documentation is not that bad, just start with the sources13:31
wolfspraulthe advantage of us working on an actual functioning product is huge pull/focus towards something that works overall13:32
wolfspraulif I were to learn something, I would want to start with something that works, and then tweak it13:32
wolfspraulwpwrak: do you have a TV you can connect the composite out to for quick quality comparison?13:40
wolfspraulbefore we blame composite or the cam...13:40
lekernelab5tract: we have proper English technical documentation http://www.milkymist.org/mmsoc.html13:45
lekerneland some more13:46
wpwrakwolfspraul: of course. don't you remember watching argentina's shame before germany ? :)13:46
wolfspraulah yes, of course. great game13:46
wolfspraul4:013:47
wpwrakgrmbl13:47
wpwrakat least the city gave a rather convincing "28 days later" experience afterwards13:49
wpwrakab5tract: technical documentation is pretty good. some things may need a bit of updating, but overall, it tells you want you want to know.13:55
wpwrakab5tract: user documentation is so-so. the reference material gives you a starting point for doing your own experiments.13:56
wpwrakab5tract: things get more grim when it comes to introductions/tutorials (for users). this is an area that needs more work. maybe also by combining them with the reference material in the FN handbook (which is the general direction of that manual anyway)13:57
xiangfuwpwrak, is the fullspeed show 12M under lsusb -t?14:35
xiangfuwolfspraul, the GreenAsia is lowspeed.14:35
wpwrak(full-speed) yes. in the case of atusb, i even know that it can't be low-speed, no matter how hard you try ;-) (some devices can be both)14:37
wolfspraulxiangfu: how about the one with sputtering output?14:38
wolfspraulxiangfu_: brand of that one was "Yaoguangqishi" - do you still have it?14:39
wolfspraulmy guess would be low-speed as well, but now that we know this distinction we can check real quick...14:39
xiangfu_yes.14:39
xiangfu_just checked. that one low-speed as well14:40
xiangfu_but I have a high speed mouse.14:40
wolfsprauldoes it work?14:41
xiangfu_no.14:41
wolfspraulok great, please add to the wiki14:41
wolfspraulwith vid/pid/brand etc.14:41
wolfspraulhttp://en.qi-hardware.com/wiki/Milkymist_One_accessories#mouse14:42
wolfspraulfirst sentence is wrong14:42
wolfsprauland that was partially ironic when I wrote it! :-)14:42
wolfspraul"good news - we haven't found a USB mouse yet that would not work with Milkymist One."14:42
wolfspraulnotice the 'yet' :-)14:42
wolfspraulI was frustrated and worn out from keyboard testing at that time, so I was hoping writing this there would somehow be a good defense. didn't work.14:43
wolfspraulI never did serious mice testing either, that also didn't help14:44
wolfsprauloh well14:44
wolfspraulanyway now we found our first troublemaker, that's good to track and hopefully eventually fix this as well14:45
wolfspraulgood catch!14:45
xiangfu_(webupdate) at home. works fine. it only works fine when Internet connect is stable and fast.14:49
xiangfu_my home is 2M ADSL. download speed ~270KB/s14:49
xiangfu_200KB/s ~ 300KB/s14:49
xiangfu_Kb14:49
wolfspraulwe need to track down what exactly it needs in terms of 'stable and fast'14:50
wolfsprauland then make the update more robust14:50
wolfspraulit should be able to support and survive any interruption, packet loss, cable unplugged in the middle, dc jack unplugged, etc.14:51
xiangfu_yes.14:52
wolfspraulwpwrak: faderfox package looks good, no worry14:53
wolfspraulthat looks more like an internal fedex hickup to me, a little slower than expected basically14:53
wolfspraulfrom my understanding he will only get the m1 as a casual/interested user anyway14:54
wolfspraulhe wants to know what this is, how it works14:54
wolfspraulbut very good for us to get this feedback since he is deeply involved in the controller scene for 7 years at least, and most likely in that field even before then, I would think14:55
wpwrak(package) yes, took me a while to decode it. looks good then :)14:55
wolfspraulso I'm looking forward to his feedback :-)14:55
xiangfu_mouse updated: http://en.qi-hardware.com/wiki/Milkymist_One_accessories#mouse15:09
xiangfu_after webupdate. I re-plug the power. the NOR-corruption happen again. now I am read the standby.bin back.15:10
xiangfu_there is no NOR-corruption. maybe new issue. what I do is. do webupdate. then re-plug the power cable.15:19
xiangfu_press the middle button can not boot m1.15:19
lekernelvery nice. like I didn't test that stuff 20 times or so already15:20
xiangfu_read the standby.fpg back. show all fine. no NOR-corruption.15:20
xiangfu_but now it can  boot again.15:20
wpwrakxiangfu_: what do the NOR lock bits look like now ?15:21
lekernelwith the new version?15:21
xiangfu_lekernel, OS still show: RTEMS 4.10.99.0.15:21
lekernelthe OS version number is meaningless15:21
lekernelwhat are the others?15:21
xiangfu_lekernel, other fine. I take some screenshot. will send a quick feedback now (to mailing list)15:21
wpwraklekernel: btw, regarding navre progmem size. which way should it go ? like my patch does, try to make things fit in 4 kB (i did it this way because that lets me avoid touching verilog and all that entails),or the other way around, grow the navre progmem in the SOC to 8 kB ?15:23
lekernelno, grow the soc program15:25
rohgrow soc.15:25
rohmake it 16k or so. or is that memory thats 'precious'?15:27
wpwrakit's fpga memory. so yes, my preciousss :)15:28
rohmeh15:28
wpwraklekernel: so ... would you mind growing that soc for me ? :) so that i can chase the usb full-speed mystery some more today instead of spending the next days figuring out what failure modes exist between the xilinx tools and wine ?15:30
lekernelthere's a linux version of the xilinx tools15:30
lekerneland yes, anyway15:30
wpwrakthanks ! :)15:33
wpwrak(isn't the "linux version" just a win binaries plus wine bundle ?)15:33
wpwrak(thought i had once read something like this. not sure if it was about these tools, though)15:34
lekernelno, it's not15:37
lekernelthey have questionable "portability" toolkits, but most of them aren't as bad as wine15:37
wpwrakcool. one of these days then15:39
GitHub20[milkymist] sbourdeauducq pushed 1 new commit to master: http://git.io/qKe3Ig15:53
GitHub20[milkymist/master] softusb: increase program memory size to 8K - Sebastien Bourdeauducq15:53
xiangfu_wpwrak, lockbit all correct.15:55
xiangfu_wpwrak, maybe I do something wrong.15:55
wpwrakxiangfu_: or maybe you do something very right ;-)15:56
xiangfu_wpwrak, there is no 'unlock' code under flickernoise or rtems driver. so webupdate will never touch the lock bit (if there is no other bug :)15:56
xiangfu_wpwrak, when I plug power cable to m1. sometimes just press middle button cann't boot m1. I have to replug the power cable again.15:57
xiangfu_but it rare happen.15:57
xiangfu_wpwrak, I have two combine keyboard and mouse devices. both only keyboard working. will testing your new usb code when the buildhost finish the build15:59
xiangfu_it still building: http://fidelio.qi-hardware.com/~xiangfu/build-milkymist/curr.build.firmware.log15:59
xiangfu_this build stuck at 'cvs update' for very long time. I just fix that start a new build.16:00
lekernelxiangfu_: this sounds like our old friend, the reset/norruption bug, no?16:01
Action: xiangfu_ will fix a small thing. display 'English' even m1 select France language. I just found all cellphone language setting is like that.16:05
xiangfu_small thing, I will fix that and commit tomorrow. :)16:05
wpwrakxiangfu_: (webupdate) okay. it doesn't need to unlock those other blocks. so i guess all is well then.16:05
xiangfu_wpwrak, (webupdate)yes. it only touch regular partitions.16:05
xiangfu_lekernel, yes. it reset bug. but don't remember if we focus on this before. :(16:06
xiangfu_s/don't remember/I don't remember/16:06
wpwraklekernel: hmm ... what good will the 1.1 soc do me if it doesn't work with rtems ? :)16:07
lekernelwpwrak: use the silly demo firmware :)16:07
wpwrakurgh16:07
lekernelbut i'm planning to add rtems support soon anyway...16:07
lekernelmaybe tonight16:07
wpwrakcould you synthesize a soc that's rtems compatible ?16:08
wpwrak(with the 8 kB navre) or is that too messy ?16:08
lekernelit's a bit messy, but why don't you use the demo firmware until I add support in rtems?16:08
lekernelmuch easier16:08
wpwrakit's not a blocker for me at the moment anyway, but would be nice to have16:08
lekernelalso, the demo image is smaller and faster to netboot16:08
lekernel(not to mention compile)16:09
wpwraki'm trying to pick my battles ;-) there's already an awful lot of moving parts ...16:09
wpwrakbah, compilation is fast16:09
lekernelthat's not a moving part. ./build_demo.sh && cp software/demo/boot.bin ...16:10
lekernelthat's all16:10
lekernelas a little extra, build_demo.sh also rebuilds the navre firmware and automatically integrates it in the final boot.bin16:11
wpwrakdoes the demo replace the bios ? or does it replace flickernoise ?16:15
lekernelit replaces flickernoise16:15
lekerneland you load it from the bios16:15
lekernele.g. with tftp16:16
lekernelatm I have deleted flickernoise from my M1 so the BIOS automatically netboots16:16
lekernelI just have to reboot the M1 and it loads the new software from the PC...16:16
lekernelif you insist on flashing, you can do it too, but it's slower16:16
wpwrak File not found: BOOT.BIN16:17
wpwrakthen it went on to serial boot16:17
wpwrakheh, serial boot works ;-)16:18
lekernelwell, your tftp server has to handle /boot.bin16:19
wpwrakyou make it sound as if anyone just had a tftp server sitting around. most be a hundred years since the last time i did that ;-)16:20
wpwraklet's see if anything works ...16:23
wpwrakmhh. of course, it doens't remember the network settings ...16:23
xiangfu_wpwrak,  you have the 'make load' under flickernoise.git :)16:27
wpwraktftpd-hpa doesn't like the path name. wants an absolute path.16:32
wpwraklet's see what happens with original tftpd ...16:33
wpwrakInvalid flash boot image length16:34
wpwrakno idea what it tried to load16:34
xiangfu_wpwrak, 'if((length < 32) || (length > 4*1024*1024)) {  '16:35
xiangfu_what is the size of boot.bin of the demo.16:35
wpwrakand now it seems to have died (tftpd)16:36
xiangfu_so a hundred years software not working now. :D16:37
xiangfu_that is why we keep update/upstream everywhere. :)16:37
lekernelwpwrak: you need to add a .fbi header to boot.bin if you want to flash it16:38
lekernel(with length and CRC)16:38
lekernelbut netboot is easier and faster, really16:38
wpwrakphew. with aftpd it finally worked16:41
lekernelhttp://www.jacquardcomputing.com/roccc/17:26
qi-botThe Firmware build was successfull, see images here: http://fidelio.qi-hardware.com/~xiangfu/build-milkymist/milkymist-firmware-11152011-1653/17:59
GitHub7[flickernoise] sbourdeauducq pushed 1 new commit to master: http://git.io/-_SyEw18:16
GitHub7[flickernoise/master] Rename Werner's patch - Sebastien Bourdeauducq18:16
lars_is it even possible to built current milkymist with 12.4 without timing problems?19:21
kristianpaulwpwrak, ab5tract, and intros tutorials for developers, like  Artyom the other day asking how to get into milkymist soc19:42
kristianpaulwpwrak: tftpd works for me19:55
kristianpaultftpd-hpa actually19:55
kristianpaulmy /etc/default/tftpd-hpa //pastebin.com/BGTWriWC19:56
kristianpauland i can confirm is the most pleasent way boot linux :)19:57
kristianpaulafter giveup with memcard ;)19:57
kristianpaullars_: try? :) actually i think 12.3 worked fine for me then i jumped to 13.x20:00
lars_kristianpaul: well, i'm getting timing errors20:01
rohi also use the hpa one20:04
kristianpaul:|20:04
kristianpaullars_: i think thats why i wait until 13.x :)20:04
rohtftp that is.not for mm.. but in general. tftp is icky sometimes20:04
lekernel12.x is full of bugs, install 13.320:04
kristianpaulroh: what you think if or mm? memcard boot? or mm it self implementing its own tftp server?20:06
roh?20:06
kristianpauli understand you disagree tftp is not for mm and mm + milkymist?20:07
kristianpauls/+/=20:07
kristianpaulroccc,cool :)20:08
rohah.. no. i meant hat i agree that tftp in general (nothing to do with the mm1) is tricky and not that easy to get working. and i got good results using tftpd-hpa as server for embedded devices20:08
kristianpaul:)20:09
kristianpaultricky indeed20:09
qi-botThe Firmware build was successfull, see images here: http://fidelio.qi-hardware.com/~xiangfu/build-milkymist/milkymist-firmware-11152011-1916/20:22
wpwrakkristianpaul: ah, maybe the --secure is what does the trick. anyway, atftpd seems to work okay.20:24
GitHub88[milkymist] sbourdeauducq pushed 3 new commits to master: http://git.io/1VJJDg20:35
GitHub88[milkymist/master] softusb: remove compiler warning about unused variable - Werner Almesberger20:35
GitHub88[milkymist/master] softusb: implement the reset recovery interval - Werner Almesberger20:35
GitHub88[milkymist/master] USB: send SOFs and keepalives on both ports and immediately after reset - Sebastien Bourdeauducq20:35
lekernel...but it still gets ignored20:36
lekernelthis is amazing. I get the exact same sequence from the PC and M1: 50 SOFs + get device descriptor at address 0, byte for byte21:02
lekernelthe M1 gets ignored, the PC does not21:02
lekernelI wonder what the flying fuck is going on there again21:02
wpwrakinteresting, isn't it ? :)21:32
wpwraki also had a brief look at one of the analog signals and it didn't look unusually ugly. but then, my equipment is not ideal for analyzing such things.21:36
wpwrakUSB is known to be rather finicky about the analog domain.21:36
lekernelis the atusb recognized by the M1?21:38
lekernelI mean, does the "set address" or "get device descriptor" works?21:38
lekernelmaybe inspecting what the atusb receives on the other side would help?21:39
wpwrakno, no response. still need to start tweaking it, though. at least there, i can control both sides21:39
wpwrakyes, that's the idea21:39
wpwrakhmm, shouldn't the demo fw print usb debug messages ?21:55
wpwrakah wait. need to update the soc ..21:56
lekerneli'm nearly done with the new uart driver, don't worry :p22:00
wpwrakgood. for now i've reached nirvana. updated everything. and nothing works anymore ;-)22:02
lekernelsounds like you forgot to update the bios?22:03
wpwraki did that as well. before i got at least one hopeful newline when booting. afterwards, not even that :-(22:03
wpwrakreverting to a known good state now ...22:04
wpwrakbtw, for banners it may be useful to include the build date and time. that way, you notice if by some accident something wasn't updated22:05
wpwrakthese tend to be the nastiest "bugs" to fix. because everything will be perfect in the code ... :)22:05
wpwrakgood. the old order has been restored :)22:10
wpwrakin honor of this, i shall call my next patch modifications "reactionary mix" ;-)22:11
wpwraknow let's see if i can beat some sense into this old atusb ... the tricky bit is that i need to replace the boot loader. else, that one will fight with the m1 and i'll never get to the "interesting" code22:12
lekernelworks22:14
lekernelnow, on to the fun of formatting a patch with that CVS crap22:14
lekernelgrmbl22:14
lekernelsomething that can be done with "git diff" in < 1s takes several minutes with that retardware (including rants)22:15
wpwrakone should support for efficient ranting an essential feature of any non-trivial piece of software22:16
wpwraks/should/should consider/22:20
wpwrakoh great. this is the board where the reset pad with trace came off. no in-circuit programming with the fellow ...22:28
wpwrakit never sees the SETUP22:33
wpwraklet's see if it receives at least SOFs22:40
qi-botThe Firmware build was successfull, see images here: http://fidelio.qi-hardware.com/~xiangfu/build-milkymist/milkymist-firmware-11152011-2136/22:43
wpwrakis sees something. hmm. now .. is this SOFs or was that a setup ? :)22:46
mwalleah usb debug action :)22:46
wpwrakweird. seems that it hit a panic. that would be the one i planted next to setup22:48
wpwrakit's one of those moments when the cursor is in front of the computer ...22:54
wpwrak"cursor" from "to curse"22:54
wpwraklet's see what these SOF really contain ...22:55
wpwrakhmm. atusb claims to see SOF changes at some 2 Hz (guess from visual observation). intriguing.22:58
wpwrakoh wait. that's the upper 3 bits ! heh, sense has been restored :)22:59
wpwraklet's have a good long look at the bit timing23:01
wpwrakhmm, looks roughly correct, to the extent i can tell23:08
wpwrakwell. the sun has set. time for breakfast.23:25
--- Wed Nov 16 201100:00

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