#milkymist IRC log for Saturday, 2013-05-04

lekernelbhamilton, btw you can ask questions here too14:24
bhamiltongreat :)14:24
GitHub87[migen] sbourdeauducq pushed 2 new commits to master: http://git.io/9iw8-A15:40
GitHub87migen/master fd089b1 Sebastien Bourdeauducq: actorlib/dma_asmi/OOOWriter: fix tag offset15:40
GitHub87migen/master 7a74dae Sebastien Bourdeauducq: actorlib/spi: add DMAWriteController15:40
lekernelfound the problem :)16:10
lekernelyay!!!16:11
lekernelpairs of 2 bits were swapped ...16:11
lekernelgosh this sort of bug can be so annoying ...16:11
lekernelnow the picture looks perfect16:12
larschm, yea that's what i expected, just couldn't figure out which two16:15
larscwhich two were swapped?16:16
lekernelall pairs16:16
lekernelall 5 pairs of 2 bits in the 10-bit words16:17
larscha, I stopped at trying to swap at a 4 bit boundary...16:17
lekernelhttp://milkymist.org/dvi_working.png16:17
lekerneland no more strange behavior like lines with the wrong size etc. except at beginning and end16:18
larscthat was due to the vsync character being wrong and it thought that was data16:19
larscyea16:19
larscactually are you sure that this is all?16:21
larscthe the vsync bit seems to be flipped16:23
lekernelupdated the data16:24
larscwhen I count the control character I get this16:25
larsc   1237 FOUND 2 0b101010100 0b101010101116:25
larsc   1236 FOUND 2 0b1010101011 0b10101010016:25
larsc      4 FOUND 2 0b10101011 0b110101010016:25
larsc      6 FOUND 2 0b1101010100 0b1010101116:25
larscwhile I expect the numbers to be the other way around16:26
larscwell not count the control characters, but count the transistions of hsync on <-> hsync off16:26
lekernelmh?16:27
larscIf vsync is off the transistion of hsync off to on should be the transition of c=0 to c=116:28
larscright?16:28
lekernelI think so16:28
lekernelwhat do you see instead?16:28
larscwell a transition from c=3 to c=216:29
larscI think16:29
lekernelhsync and vsync can have negative polarity - that's defined in EDID and enabled by default iirc16:29
lekernelmaybe it's just that16:29
larscah16:29
larscthat makes sense then16:29
larscuhm yea, 640x480 has negative polarity for both hsync and vsync16:30
lekernelare there any modes that have positive polarity? (ie do I need to care about that, or just hardcode negative polarity?)16:37
larscI think 720p, 1080p has positive polarity16:40
larscnegative polarity is mostly used for legacy modes, or something16:40
larschttp://lxr.free-electrons.com/source/drivers/gpu/drm/drm_edid.c#L13416:41
lekernelgrmbl16:48
lekernelhow is the gateware supposed to autodetect that?16:48
larsccheck which character comes after de goes from high to low16:50
larscor count the characters16:51
larsciirc vsync and hsync polarity are actually supposed to help to distiguish between modes16:51
lekernelok, so when de goes from high to low, it's usually hsync, and sometimes vsync+hsync ?16:54
lekernelit can never be vsync alone16:54
lekernelright?16:54
larscthe pattern is de nothing hsync nothing de16:55
larscor vsync hsync+vsync vsync16:55
larschttp://www.epanorama.net/documents/pc/multisync_modedetect.html so its legacy crap16:56
larscthe first symbol after de goes low will always be both hsync and vsync off16:56
lekernelok so if c=3 immediately after de=0 then polarity is negative?17:02
larscI'd just take the c after de=0 and xor all other c's with it17:03
lekernelgood idea17:03
GitHub45[milkymist-ng] sbourdeauducq pushed 1 new commit to master: http://git.io/nsChhw18:40
GitHub45milkymist-ng/master 6307331 Sebastien Bourdeauducq: dvisampler/datacapture: swap bit pairs18:40
GitHub91[milkymist-ng] sbourdeauducq pushed 1 new commit to master: http://git.io/qiTZQQ18:42
GitHub91milkymist-ng/master 4259699 Sebastien Bourdeauducq: dvisampler: add RawDVISampler18:42
Fallenoucongratz :)21:30
--- Sun May 5 201300:00

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