#milkymist IRC log for Tuesday, 2012-01-31

cladamwwpwrak, (usb) for now, will ext. = max(4, 1) or min. = max(0, 1)? seems that haven't determined yet.02:00
wpwraki'd say ext = 4, int = 2. keep it simple :)02:01
wpwrakint = 1 would be a highly unusual configuration anyway02:02
cladamwmmm...current existing usb extendable cable seems not easy to find a one.02:03
wpwrakvirtually all the internal USB port of PCs are of that kind. they use 100 mil headers in a 2x5-1 configuration02:05
cladamwalso if 5 for rest, the footprints for two right-angle 1*2 parts plus another 1*1 usb connector are also not good for arrangement.02:05
wpwrak2x5-1 = a 2x5 pattern but with one pin cut02:05
wpwrakwolfgang seems to know some magic cables for internal use. i don't know what they are. what i know are the 2x5-1 connectors used in pcs for the last 10 years or so :)02:07
cladamwi would still leave this whole 6 ports to let house generating placements to see if they are okay at some time, or we decide it at that time, how do you think?02:07
wpwraki think it's okay if they choose the placement of the connector for the internal ports02:08
cladamwokay...later i asked what magic cables (int.) are. ;-)02:08
wpwrakremember, the internal ports would be a 100 mil header, not a USB A connector02:08
wpwrakyeah :)02:08
wpwrakthe placement would depend a bit on what cables wolfgang can find02:09
cladamwyeah... a one 2*5 pins header ( precisely said that it's a housing header in pitch 2.54mm) ;-)02:09
wpwrakif he can find a nice cable, the there should be some distance between the header and the location of the usb device02:09
wpwrakif he can't, and we have to use an adapter board, then they should be close02:10
wpwrak(housing) you mean shrouded ?02:10
cladamwyes02:10
wpwrakthat's not required02:11
cladamwwe can see if later what real cable we select. of course we want 2.54mm for 2*502:11
Action: kristianpaul interested on migen but will try later02:11
wpwrakif we connect some 4x1 plus, we bypass any keying anyway02:11
cladamwsure02:12
wpwrakif we connect a pc-type cable, it should be 2x5-102:12
kristianpauldont read all log, but please dont make M1 users to convert their videos to "specialiced" formats..02:12
wpwrakkristianpaul: we probably will :)02:12
wpwrakkristianpaul: at leat in layman's terms02:13
cladamwkristianpaul, ;-)02:13
wpwraklayman: just drop an .avi, .wmv, .mp3. or .mkv there, and it'll work02:13
wpwrakus: well, .mpg at 160x8002:14
kristianpaulwow nocarrier twitvid is AWESOME02:14
kristianpaulwpwrak: srop and work sounds nice to me :)02:14
wpwraksrop ?02:14
kristianpauls/srop/drop02:14
wpwrakah yes :)02:15
wpwrakbut it probably won't work that way :)02:15
kristianpauli  know :)02:15
wpwrakbut that's an integration issue02:15
kristianpaulwow nocarrier video is _awesome_, i wached twice i dint get bored :)02:17
GitHub192[scripts] xiangfu pushed 1 new commit to master: http://git.io/6e3RwA04:05
GitHub192[scripts/master] reflash_m1.sh cleanup help message - Xiangfu Liu04:05
qi-botThe Firmware build was successfull, see images here: http://fidelio.qi-hardware.com/~xiangfu/build-milkymist/milkymist-firmware-20120131-0507/05:56
lekernel_larsc: at the moment the expressions are dumbly translated into verilog, and it's verilog which takes care of expressions widths/signedness08:00
lekernel_larsc: there is a way, but you need to reimplement Verilog's rules08:01
lekernel_wpwrak: FFMPEG can read all sorts of formats... I'm not worried09:02
wpwraklekernel_: we'll have to be careful to remove the encumbered ones, though10:26
wpwrakso that would mean MPEG2, MPEG4, H.264, Quicktime at least >= 2, etc.10:34
wpwrakfor the rationale, see also http://en.qi-hardware.com/wiki/Patent_Policy_Public_Face10:37
lekernel_aren't those the most popular ones?10:41
Fallenousure10:43
Fallenoubut if you support MPEG-2 you have to pay something AFAIK10:43
lekernel_no, you don't have to :)10:44
wpwrakMPEG-LA claims you have to. see also http://en.wikipedia.org/wiki/Mpeg2#Patent_holders10:46
wpwrakmost of the patents should expire around 2014, though10:47
wpwrak(and some earlier)10:47
lekernel_they would claim anything that helps them make money10:47
wpwrakand yes, _of_course_ the encumbered formats are the popular ones :)10:47
lekernel_great. so let's just include them, we're too small for them to care.10:48
lekernel_they would have to do things like buy a M1, analyze the software, ....10:49
cladamw(upward10:49
cladamw emission of bottom-mounted side-facing LEDs) FR4 allows probably10:49
cladamwsome light to pass at a steeper angle, but how translucent it will, i don't know. at least side-facing led is good idea.10:49
lekernel_just don't mention anything about the supported formats10:49
wpwraklekernel_: no, don't do that. remember openmoko ? we got bitten, too. and yes, they did make the effort.10:50
lekernel_http://www.google.com/trends/?q=openmoko,milkymist10:50
wpwraklekernel_: FLAC should be safe. maybe MPEG1, too10:50
wpwrakit's not about size ...10:50
FallenouWell I heard recently from a lawyer that MPEG-2 makes you pay10:52
Fallenouand pay even more than MPEG-410:52
Fallenoueven if it's not logical10:52
Fallenouit's like $0.50 per unit10:53
Fallenouor $0.2010:53
cladamwwpwrak, APA1606SURCK is currently proposed for leds? ;-)10:53
Fallenouoh no it's $2.00 for MPEG-2 since january 201010:54
Fallenouhell it's expensive10:54
wpwrakcladamw: yeah, nice and bright. do you like them ? :)10:54
cladamwwpwrak, yeah pretty cool. seems that we had have side-facing led in stock, need to find it out. it's used in avt2.10:55
cladamwwpwrak, also red color then. well...i need to check.10:56
cladamw(U5 p/n content) just set visibled.10:58
FallenouMPEG-4 : 0 to 50 000 decoders sold per year => 0 royalties11:00
Fallenouafter that it's $0.25 per decoder11:00
lekernel_Fallenou: but to properly get into the "0 royalties" category, I guess you have to sign something not so nice, right?11:01
Fallenouhttp://www.mpegla.com/main/programs/M4V/Documents/m4vweb.pdf11:01
FallenouI don't know if the license agreement (la) is like global11:02
Fallenouor if you have to actually sign it11:02
wpwrakmpeg4 is a minefield because there's not just one place you have to get licenses from: http://www.m4if.org/patents/11:02
wpwraklekernel_: presumably, yes11:03
Fallenouhum ok, i'm not an expert anyway about this kind of stuff11:04
Fallenousounds scary11:04
wpwrakand since M1 is a product targeted at professional use, it may not even fall into the royalty-free category11:05
lekernel_seems FFMPEG can already read PNG/JPG11:05
lekernel_http://ffmpeg.org/general.html11:05
lekernel_http://ffmpeg.org/general.html#Image-Formats11:05
wpwrakFallenou: it's scary and unjust11:05
lekernel_so I guess we can simply base all the image/video reading stuff on FFMPEG11:08
cladamwwpwrak, http://www.kingbright.com/manager/upload/pdf/KPA-3210SURCK(Ver1189146070.9)11:12
wpwrakas long as it's not too hard to keep the ugly bits out, why not11:13
cladamwwpwrak, it's 3.2*1.0mm, yours APA1606SURCK is 1.6*0.6mm,11:13
lekernelwpwrak: what about distributing a package consisting of all the FN object files, a compiler/linker, and some instructions to build and link FFMPEG?11:14
wpwrakcladamw: yeah. the one you have is HUGE :)11:15
lekernelalso the web updater should have an option to choose where to download stuff from11:15
cladamwwpwrak, ;-) but from your led demo, you used 0603 package. the APA one is smaller.11:16
cladamwwpwrak, yeah...so forget about my finding. ;-)11:17
wpwraklekernel: as long as it's not the 3rd bullet of http://en.qi-hardware.com/wiki/Patent_Policy_Public_Face#Quick_summary11:17
wpwrakor, for that matter bullet 211:18
wpwrakthe laws are not friendly, and neither is the way the courts rule. most of what's on that page is based on court rulings11:19
wpwrakalso, the deeper you enter the grey zone, the more likely it is that you could invite unwanted attention. even a lawsuit you could win can kill you financially.11:20
wpwrakcladamw: the APA... shuold be 0605, no ? checking ...11:21
wpwrakyes, 060511:22
wpwrakso it's a little bit larger than the 0603 is used11:22
wpwraks/is used/i used/11:24
wpwrakwolfspraul: your opinion on using ffmpeg ?11:25
cladamwwpwrak, yes, since the distance between board edge and switches and ir is small, so yes, stick on this APA... ;-)11:26
lekernelhaha http://hardware.slashdot.org/comments.pl?sid=1690774&cid=3261278615:38
Fallenoulekernel: do you really need python 3.2 for migen ?20:21
FallenouI have python 3.1 on debian squeeze :x20:21
lekernelit seems larsc is using 3.1 successfully20:24
Fallenouok nice !20:24
Fallenouwill try to kick out uart and norflash for simulation20:25
Fallenouand just keep lm32/wishbone/sram20:25
larscFallenou: python 3.1 on squeeze works, but you have to install some dependencies manually, since most of the python libraries that come with squeeze aren't 3.x compatible20:31
Fallenouit worked without any special dependency here20:33
larscif you want to use migen.flow you need networkx20:33
Fallenoulekernel: if there is only one slave, migen generates things like sel[0] which xst refuses to synthetise20:46
Fallenoubecause it's notan array20:46
Action: Fallenou corrected by hand20:46
Fallenouworks great anyway20:47
lekernelhm, ok20:49
lekernelgrmbl, verilog is annoying20:49
lekernelbtw if you want to load code in the memory there's the "init=" parameter20:50
lekerneland I don't think you should need the interconnect component... the i-bus should have the instruction memory only, and the d-bus, something to display accesses20:51
lekernelI can make you a point-to-point bus interconnect component for those cases.20:52
--- Wed Feb 1 201200:00

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