#milkymist IRC log for Thursday, 2011-03-17

manukoI'm interested on jtag for mico3203:51
kristianpaulxiangfu: hi !!04:02
kristianpaulokay lest try something here04:06
Action: kristianpaul missi mm1 early bios fast booting..04:06
Action: kristianpaul ccd + mm104:11
kristianpaulxiangfu: will be cool be able to get snanshots from the vide in04:13
kristianpaulor at least from the previw that is visualized in the video input settings04:13
xiangfukristianpaul: yes.04:31
kristianpaulhe, just telling :-)04:36
kristianpaulha, copy and paste could be usefull in the patch editor too04:49
lekernelkristianpaul: it works... ctrl c/v06:01
kristianpaullekernel: morning ! :-)06:06
Fallenoukristianpaul: I don't understand quite well what's on your screen on your video10:19
Fallenouis this your keyboard that your filming and feeding into the video input ?10:20
kristianpaulFallenou: heh10:41
kristianpaulwas a test but seems my camera is not good fore details10:41
kristianpaulFallenou: but yes it is :-)10:42
Fallenouoh ok !10:42
FallenouI recognized something looking like a keyboard but wasn't sure10:42
Fallenousince it's distorted by M1 (which is normal I would say it does it's job :p)10:42
Fallenouwell nice we need more video !10:42
FallenouThere is no video of the current version of Milkymist / Flickernoise10:43
lekernelwolfspraul: do you know any good manufacturer of ac97 codecs?10:47
Fallenoulekernel: maybe have a look on a PC sound card :o10:48
kristianpaulFallenou: more video, sure, !! but first i have some todos with you and they list10:48
wolfspraulI'm not sure changing the codec is the right solution for the line-out noise anyway. don't you think we should track down the real issue first?10:48
Fallenoukristianpaul: yes sure :)10:48
kristianpaulFallenou: btw i already forked lekernel milkymist repo10:48
Fallenoukristianpaul: good !! :)10:48
kristianpaulFallenou: i hope upload sistensis log today..10:48
Fallenouthank you10:48
wolfspraulkristianpaul: why is the camera not good on details? what details? it should be pretty good, especially considering that in that video, you can hardly see anything :-)10:48
lekernelwolfspraul: the ml401 has the same codec and is noticeably noisy too. so...10:48
kristianpaulwolfspraul: my camera not the ccd one10:48
kristianpaulit have details10:49
kristianpaulreally nice ones i think10:49
lekernelwolfspraul: also, we should be able to find a pin-compatible chip. just drop-in replacement, no pcb layout change10:49
lekernelso it's an easy test and fix10:49
wolfspraulsure, if that is possible let's do it10:49
wolfspraulwill you reply to the guy on the list?10:50
lekerneli'll try it on a rc1 board...10:50
Fallenouyeah maybe just source different ac 97 pinout compatible and test them since it's not that much work10:50
wolfspraulpcb layout change is no problem, unless the change starts to spread into other components etc.10:50
kristianpaulok, but why and what is problem is supposed to solve the new chip?10:50
wolfspraulif it's just a little different pad size/location, it shouldn't matter because we are doing rc3 anyway10:50
kristianpaulare we talking about audio-out noise?10:50
kristianpaulor other no so know problem?10:51
lekernelI could order samples from NS, but I'd like to try a different manufacturer :)10:51
lekernelkristianpaul: yes, audio out noise10:51
lekernelit's a minor issue, but everyone's complaining about that10:51
kristianpaullekernel: it was present on the ml401 as well?10:51
wolfspraulit's not minor. once the feature is there it should work.10:51
kristianpaulagree :-)10:51
wolfsprauland plus - connecting a headphone, even just for fun, is something a lot of people will try10:51
wolfspraulthen they hear a lot of noise, and they think 'what a crap'10:52
wolfspraulit's that easy10:52
lekernelyeah, I see10:52
lekernelthat's why i'll look a bit more into that10:52
kristianpaulyou asked xilinx with the ml401 a posible root case?10:52
wolfspraulif we leave it as is on rc3 (which I can live with), then I will start saying everywhere that line-out is basically not working. "forget that it exists"10:53
wolfspraulthat's better than leaving it there and letting people discover the noise10:53
wolfspraulit's called 'expectation management'10:53
wolfspraulwe can definitely do that - scratch line-out on the marketing side ("does not exist")10:53
lekernelwolfspraul: any sourcing experience from VIA?10:54
kristianpaulhmm via. :-)10:54
lekernelyeah, I know, we can also simply remove the jack10:54
wolfspraulthen we would have a hole in the case, unless we fix that as well (and I have a few cases in stock so it's not ideal)10:54
wolfspraulthere is this british company, forgot the name10:55
lekernelI checked them, it seems they make QFN chips only10:55
lekerneli'll try to find a TQFP one, will be much easier to test10:55
wolfspraulyes, wolfson10:57
wolfspraullekernel: do you think just changing the codec will fix the noise?10:58
wolfspraulor is this a shot in the dark?10:58
kristianpaulshoot shoot ! ;-)10:58
lekernelthere's a good chance i will10:58
lekernelit will10:58
lekerneland it's an easy test10:58
kristianpaulthats true10:58
wolfspraulwhy is there a good chance?10:59
wolfspraulthe chip we have now is so bad? how come it is selling at all then?10:59
wolfspraulmakes no sense to me10:59
wolfspraulI would rather think we are doing something on the board that is very unfriendly to a line-out signal11:00
wolfspraulsomething electrically outside the codec11:00
wolfspraulthat's a long shot too, but I just have no way to comprehend WHY changing the codec might help11:00
wolfspraulbut who knows, I'm always in favor of trying...11:01
wolfspraulI would love to fix line-out, we need a product that works really well in all details. and the noise is massive.11:02
wolfspraulbasically we have no working line-out on m111:02
wolfspraulwhich I can live with if needed, and will explain it like that11:03
wolfspraulit's a video machine :-) who wants line-out?11:03
wolfspraullekernel: if you have a few suggestions for codecs, Adam can get them and try as well, so we can speed up11:03
lekernelwolfspraul: you can look at the schematics and PCB design around the codec, but I did that already and found nothing unusual11:06
lekernelthat being said, bugs are often found by someone who did not draw the schematics :)11:06
lekernelwolfspraul: http://en.wikipedia.org/wiki/AC%2797#Codec_chips11:07
kristianpaulwolfspraul: btw did you finally test vide-in on your mm1?11:09
wolfspraulno time yet. every day I have work for 100 hours, pretty bad.11:12
wolfspraulif you say it works for you now, I will postpone my test for now.11:12
wolfspraulkristianpaul: with your video you want to say "it does work", right?11:12
wolfspraullekernel: I definitely cannot see anything in the schematics, unfortunately we pissed off Joerg who is quite good on audio issues.11:13
lekernelhm, is "we" means only me, or was there other problems as well?11:14
wolfspraulno need to point fingers, I will try to get him back. It's a joint failure, but that includes him as well. What can I do, we try to collaborate...11:21
wolfspraullekernel: for line-out noise, have you tried to see where the noise originates, i.e. whether it comes out of the codec already? is there anything in between the codec and the jack?11:33
lekerneljust passive components11:33
lekernelwolfspraul: cc me if you write to Joerg. since I contributed to pissing him off... :/11:37
lekernelwolfspraul: http://www.wolfsonmicro.com/products/codecs/WM9707/11:47
lekernelfound one eventually11:47
lekerneli'll check compatibility a bit more and order samples11:49
lekernelhmm... it definitely seems to be a product made for drop-in replacement of competitor's codecs :)11:57
Fallenoumaybe one day they will copyright pinouts to avoid concurrency with drop-ip replacement :p11:59
lekernelthe whole point of the ac97 standard is to allow that replacement12:00
lekernelpc market...12:00
lekernelit's the same with sdram for example12:01
lekernelthough the sdram jedec standard is usually more closely respected than the ac97 one :)12:01
lekernelac97 is a bit messy12:01
rohit is?12:05
rohlooked quite simple and sane (for some industry standard)12:05
rohisnt it just a loooong shift register with a data and a command chain and named registers in command mode?12:06
lekernelthe control signals are usually OK12:08
lekernelbut sometimes there are subtleties on things like the power supply voltages, pinout, and required external components12:08
lekernelthat wolfson codec supports 3.3 to 5V on all supply pins and requires very little external parts12:08
rohpinout is also defined by ac97 and afaik even upgrade compatible12:08
rohwolfson is quite cool.12:09
rohalso nice people with a clue about linux12:09
lekernelyup. but some manufacturers like to hook up proprietary extensions to NC pins12:09
rohlekernel: check different versions of ac97. some add features and revise pinouts12:09
rohe.g. there is a nc where newer versions have the spdif out12:09
lekernelyeah, but sometimes they use them for things like connecting capacitors for some onchip audio improvement circuit to work12:10
lekernelI don't think that's in the standard... is it?12:10
rohdunno. but ac97 isnt something loosely speced or so12:11
lekernelroh: ok, the wolfson codec requires decoupling of pin 32, but not the current one12:17
lekernelyou see, there are always problems12:17
lekerneland an additional cap on pin 3312:19
lekernelit's always small details, but pesky ones12:19
lekernelanyway, I think we can manage them, and I'll order samples now12:20
rohwhich ac97version is the board for?12:20
lekernelthe current chip is 2.112:21
rohuh. i see.. thats quite... well.. outdated..12:21
lekerneland the wolfson codec too12:21
lekernelah, also that wolfson chip can have SPDIF12:22
rohit is.. i see.. then they enforced physical compatibility in later versions. i mostly used the 2.3 spec for reading up12:22
rohspdif is speced in 2.312:22
roheh 2.212:22
lekernelah, funny12:22
rohhttp://en.wikipedia.org/wiki/AC'97 has a nice list of what feats came when12:22
lekernelanyway, I think we'll route the required pins for spdif to the internal audio header, as a totally unsupported easter egg12:23
rohthat would be cool12:23
rohits only one pin i think12:23
rohthe last one or so12:23
lekernelthere's also a "spdif enable" pin that needs to be pulled high12:24
rohthats not in the spec i read.. must be vendor specific12:24
lekerneli'll connect that one to the header too12:24
lekernelyeah, told you12:24
lekernelthere are tons of vendor specific extensions to ac9712:24
rohlekernel: maybe update to 2.3 standard12:24
rohi think its more specific and should make you have less work12:24
rohin using different chips atleast12:24
lekernelso far it's just 2 more capacitors12:25
lekernelare 2.3 chips compatible with 2.1 ones?12:25
rohi think so. mostly12:25
rohthe idea is the same, i think some pins changed which were nc before (and vendors did their proprietary extensions on)12:25
lekernelany chip you could recommend?12:25
rohuh.. not from my head. wolfson is a good start. ask them for a recommendation12:26
rohbut if i read the stuff right, you should have even sw compat. with the lastest revision.12:29
rohi think they only did new ones to root out all vendor special cases12:29
rohthe whole idea was to make it unneccessary to design a new board just to use another vendors codec12:29
rohfor the successor hda there isnt even a sw driver for different chips anymore. they need to be fully sw compatible12:30
lekernelyeah... but there wasn't hda on fpga boards when I designed MM SoC12:31
rohor? eh. no.. sorry.. its only the hw/sw interface.. still got a codec driver12:31
rohno problem. ac97 seems still to be used quite widely12:32
rohanalog devices also makes nice sounding chips (quality)12:32
rohand somewhere on the low end there is always realtek12:32
lekernelit seems AD is phasing out their ac97 codecs :(12:32
rohsomebody bought them?12:33
rohthere is still AKM for the quality segment12:34
lekernelI don't know... but all the AD ac97 references I searched for were "discontinued" with no replacement information12:34
wolfspraulthat's interesting, would like to hear the real story from an AD guy, will keep in mind...12:49
rohmaybe we need to find out what all the long-term users use12:50
rohthere is embedded hw with a 10 year gurantee for lines of devices12:50
Action: Fallenou just ran into compiler problem for pic3213:49
FallenouI am using a packed structure, so some of the fields are unaligned13:49
Fallenouam using a char * to walk through the struct13:50
Fallenouand then in the middle of it i switch to long int *13:50
Fallenoubut unfortunately the switch is not on an aligned address13:50
Fallenouthe compiler does not handle that case13:50
lekernelload the switch value to a temporary variable then13:51
lekernelor use an AVR :)13:51
Fallenoui am sending values through spi13:51
larscis the compiler supposed to handle such cases?13:51
Fallenoularsc: I guess since it does not produce warning or error13:51
Fallenouand since it generates code that generates an exception13:52
FallenouI am sending a struct through spi, sending first the length%4 values, in "byte width" burst, and then all the remaining part with 4-bytes-size burst13:52
FallenouI will try the other way13:53
Fallenoufirst sending everything in 4-bytes-size burst, and the remaining 3 bytes afterward :)13:53
Fallenouworks \o/13:55
kristianpaulwolfspraul: right15:14
wolfspraulright? (lost context)15:15
wolfspraulah ok "it does work"15:20
kristianpaulyeah, video-in context15:23
wolfspraulkristianpaul: why can I not tell any 'video-in' effect in your video?15:27
wolfspraulis there a way to display a static image in some part of the screen?15:27
wolfspraulI think for a casual viewer of that video, they may not even understand/see where the video-in part is :-)15:28
kristianpaulwolfspraul: he,sure there is a control panel for video-in15:29
wolfspraulas you can tell I never tried, so far15:29
wolfspraulI have the same cable problem as you now, need to get an adapter first15:29
kristianpauli got a cable15:30
kristianpaulwas easier15:30
wolfspraulkristianpaul: how much did you pay for the cable? how long is it?15:50
kristianpaulwolfspraul: i dint paid, i just borrowed, long... like 1m i think15:51
wolfspraulok got it15:51
kristianpaullekernel: the other video inputs can be enabled? _if_, is it a software/hardware(hdl) work needed to make it work?15:55
lekernelit's not other video inputs, it's for component video, and I don't know - never tested15:56
lekernelthe adv7181 datasheet should give you some hints15:56
lekernelmaybe they could also be used as other video sources to select from15:57
lekernelif you manage to get that to work, documentation or code would be much appreciated :)15:57
kristianpaulok, yeah, i was about to read the datasheet, just doing quick survey before :-)15:57
kristianpaullekernel: btw for the timing constrait check, is the sys_clk the only signal to care about i guess?16:02
lekernelno, it's not16:02
lekernelyou should check all the signals in the clock report16:03
kristianpaulok, yeah is logic procecure16:06
wolfspraulkristianpaul: maybe you can find out whether it is possible to connect 3 different composite cameras :-)16:12
wolfspraul(not that I think it's important now, but good to know if it's theoretically possible or not)16:13
Action: kristianpaul like the theoretically part16:15
kristianpaullekernel: For your reply seems this guy asumed out noise as manufacture hardware problem ;-)19:42
mwallelekernel: polling a IP bit should work without enabling the corresponding bit in IM, shouldnt it?22:47
lekernelmwalle: yes, it should work23:02
lekerneland I did that actually and it worked23:02
mwalleyeah it works, either i've used an outdated rom file or i finally trigger that RAM8 not initialized bug23:04
--- Fri Mar 18 201100:00

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