#milkymist IRC log for Tuesday, 2011-04-12

aw_lekernel, what's the mark on your WM9707SCFT/V chip? is it "WM9707S"?01:07
kristianpaulVj meeting was canceled and i lost 2 hrs finding the place today .. :(01:18
aw_kristianpaul, could you scope your 24.576MHz (Y1) easily? It should be easily use your scope via 'auto set' button. :-)01:37
wolfspraulaw_: why should he scope Y1? any bug we are hunting?01:51
wolfspraulkristianpaul: too bad :-) that happens, keep your spirits high01:51
aw_wolfspraul kristianpaul , no big bugs found/hunt, just want someone can help me know what's amplitude it should? since i replaced mine to wm9707, and no xtal pulse though.01:54
aw_no problems on must help me. I'd just want to someone's condition though.01:57
wolfspraulaw_: kristianpaul is traveling and does not have access to his scope for at least another day or two.01:59
aw_wolfspraul, um..okay.02:00
aw_lekernel, hi07:59
aw_lekernel, i followed your wm9707 replacement h/w patches, keep external capacitors except added 10uF on pin32, dnp 1M on crystal. this should work but mine is not.08:01
aw_lekernel, just want to confirm/compare your board now. :-)08:01
aw_my crystal doesn't oscillate. and check all pins with good powers and grounds.08:03
terpstra_does anyone know of a free gigabit etheret MAC that supports a streaming mode on wishbone?08:49
terpstra_(by streaming i mean !DMA)08:50
lekernelaw_: did you boot the board?08:58
lekernelwhen the board is unbooted the ac97 chip is in reset. not sure the oscillator stays enabled then.08:58
xiangfulekernel: <xiangfu> http://pastebin.com/whFmt36g09:19
xiangfu<xiangfu> a patch for change the video-in size. but I guess the line:58, 68, 73 in this patch is not good. please give me some tips on MTKLIB  :)09:19
xiangfuping09:25
aw_lekernel, i didn't boot board. I am pretty sure that Y1 doesn't need to stay under booted stage, I compared my other boards with unbooted. it oscillates.09:31
xiangfu_<xiangfu> lekernel: <xiangfu> http://pastebin.com/whFmt36g09:35
xiangfu_<xiangfu> <xiangfu> a patch for change the video-in size. but I guess the line:58, 68, 73 in this patch is not good. please give me some tips on MTKLIB  :)09:35
xiangfu_(seems something wrong with my network)09:36
aw_lekernel, did i maybe miss somethings or parts wrong? a hint would be good though.09:36
aw_lekernel, lm4550b can still oscillate under AC97_RESET is low while unbooted. not sure if i still missed what.09:45
aw_good, now i can see my ac97 signals but oscillating waveform is quite not stable so that my right line out randomly audio on randomly off, but left line out always on.09:55
lekernelaw_: it is not specified whether the oscillator should be on while the chip is on reset. national semiconductor and wolfson can do whatever they want there and possibly different things.10:05
aw_interesting..10:05
aw_lekernel, yup, i think so.10:06
aw_lekernel,  probably i need replace other value for C28/C29.10:07
lekernelyeah, maybe10:10
lekernelalso, maybe it is your probing which makes the oscillator unstable? is it still unstable with the probe disconnected?10:11
lekernelyou're using a x10 probe, right?10:11
aw_yes, i am using x10 probe, now it's very stable no matter I touch Y1's pad or not.10:13
aw_i did nothing though. now it's stable, but my LINE_R is good, LINE_L is sometime ON /OFF. i can see signals on C19's pad differently.10:15
aw_also randomly have noise appeared on both channel. not stable though.10:15
aw_keep monitoring.10:16
aw_but there's an fact that 'white noise' is less than lm4550b. this is definitely i confirmed here. this is good. :-)10:18
aw_compared to lm4550b's performance http://en.qi-hardware.com/wiki/File:M1_rc2_0x1e_LINE_OUT_R_scopes_no_patches.ogv10:19
aw_the white noise stays roughly 500mV on lm4450b.10:19
aw_wm9707 is lower than 200mV. thus SNR ratio actually works better than lm4550b. :-)10:21
lekernel200mV is still a lot10:22
lekernelI measured a lot less than that10:22
aw_actually lower than 100mV here. my work now is not good on soldering  i think. now it comes with Intermittencely noise.10:24
aw_does your wm9707 have 'pop' sound between 'muted' and 'unmuted'?10:26
lekernelno10:26
lekernelit doesn't make any noise or pop except one pop when you turn the power off10:26
aw_i see, so my work was just still have somewhere's wrong. or parts by my heated.10:27
aw_so you used 10uF on pin32, right?10:28
lekernelyes10:28
lekernel10uF ceramic10:28
aw_good, okay10:28
aw_me too.10:29
lekernelbtw how is the rest going?10:31
lekernelthis audio codec is the last thing you're testing?10:31
lekernelbefore rc3?10:31
aw_no, still have zener/fuse & video littlefuse waiting for final test.10:32
aw_once they are all done, we can go for rc3.10:33
wolfspraulroh: I'm a bit worried about the buttons and tolerances on the inside11:09
wolfspraulI've had several cases where the pressure onto the electrical switches is so high that they are stuck (always pressed down)11:09
wolfspraulthis happens particularly after putting the top cover on and tightening the screws11:09
wolfspraulnot sure what to do...11:09
wolfspraulmaybe we should move the switches into the pcb a few mil?11:10
wolfspraulor we change the design of the 3 button pieces?11:10
wolfspraulit's really only the thickness of that one slightly larger circle that is in between the switch and the acrylic side element11:10
wolfspraulI'm worried that the design is too unforgiving on the inside, so some percentage of cases (let's say 10-20%) will exhibit the 'buttons stuck' problem11:16
aw_wolfspraul, although i didn't know how these buttons's problem/situations. do you think that we can just use another 'long' tag push button then from the case surface we can easily push the long 'tag'?11:17
aw_wolfspraul, can you imagine my 'long tag' push button to solve this, suffer from 'button stuck'?11:18
aw_the tact switch with long length tag which goes though the thickness of case?11:19
wolfspraulhmm11:24
wolfspraulaw_: any change is risky, because what we have now is not bad, and tested well11:25
wolfspraulalso the current solution is nice because there is no glue between the electrical switch and the button11:25
aw_if M1 is i use only, i'll change all 3 tact switches to be like SIE's tact switch, it have long tag. although it's ugly from outside outlook view.11:28
aw_well.. forget about my idea. :-)11:28
lekernelroh: just wondering, how can we cut the current case design out of metal and how expensive would that be?11:51
kristianpaulaw_: hi, i'm traveling home today at night, so i hope tomorrow help you with you measurements14:28
aw_kristianpaul, hi thanks, no needs now, i can check my others though.14:30
kristianpaulsure?14:31
kristianpaulok14:31
rohlekernel: how? much more expensive laser or plasma-torch. also waterjet is an option15:52
rohbut it would be quite heavy15:53
rohand crazy expensive15:53
lekernelplasma torch is expensive?15:53
lekernelthey have plasma torches at tacheles... just not mounted on CNCs ;-)15:53
rohlekernel: well.. doesnt help15:54
rohits very imprecise15:54
rohfor stuff <1cm width you usually use a waterjet15:54
lekernelI know... but my point is it's not such an unattainable/expensive technology, is it?15:55
rohalso.... none of the methods works without a lot of work after (remove brill etc)15:55
rohlekernel: i dont know exact prices, but afaik material price as well as manufacturing cost AND working time would atleast double. so we end up with >150E a case minimum15:56
lekernelhow does apple cut their cases?15:56
rohwhich ones?15:56
lekernelmacbook pro for example15:56
lekernelinjection?15:56
rohsome the cnc-mill.. in china. in a process which only is feasible if you recyle ALL the ground up aluminium15:57
rohalso it only works if you have whole halls of cnc machines and do masses.15:57
rohits still expensive tho.15:58
rohi find that impressive.. dont get me wrong... apple are the first consumer-electronics-company who does such manufacturing methods. yet it doesnt work for smaller scales16:07
rohif you'd ask me what machine i want to get access to next: a machine to form plastic over molds. vacuum-deep-molding16:08
lekernelwhat about that? http://en.wikipedia.org/wiki/Electron_beam_machining16:13
rohEDM sounds fancy.. but afaik only works on metal16:14
rohbtw.. its the only machine which can do 'real rectangular corners'16:14
lekernelexcept for the x ray hazard it looks very nice16:14
rohinside-corners16:14
lekernelshouldn't even be super hard to DIY... probably easier than a laser cutter for metal16:15
rohEDM/EBM is also huge, expensive and does not work with all material16:15
lekerneland cheaper too16:15
lekernelyeah, but for cutting metal16:15
rohon the other side... metal corrodes.. so plastic gives a better finish for less money anyhow16:15
lekernelimo the price is only because certified vacuum components are expensive like hell16:15
rohfor cutting: use waterjet on metal.16:15
lekernelbut there are ways around that16:16
rohor depending on the thickness: NG-YAG or so.16:16
rohco2 laser is not good for metal at all16:16
lekernelimo the major problem I see with building such a machine is getting your ass x-rayed16:16
rohso.. there is no single manufacturing technique which is anywhere in the range of owning machines besides co2-lasers16:17
rohguess why we have one ;)16:17
lekernelif we can have a small one that can work with x-ray safe voltages (< 12kV or so), perfect16:17
rohedm isnt that mulipurpose how you think16:18
lekernelseem to me it's like your laser cutter but for metal :)16:19
lekernelwhich would already be very nice16:19
rohalso one needs a vaccuum chamber16:19
lekernelyeah sure16:19
rohlekernel: the cheapest way to cnc metal is either laser, waterjet or plasmajet16:19
rohin that order.16:19
lekernelin the hackspace in paris we have some turbopumps laying around already. such a project would put them to good use.16:20
rohwaterjet gives the nicest results, but laser and waterjet have a thickness-limit16:20
rohplasmajet can go through thicker stuff, but is less precise16:20
rohalso all three machines need 'space'16:20
rohmuch more space16:20
lekernelwe even got our hands on a ion pump (which works by trapping ionized gas molecules and spinning them in circles using magnetic+electric field... quite crazy stuff for ultra low vacuum)16:21
lekernelalso, what about photoresist+chemical attack?16:21
rohon metal?16:22
lekernelyes16:22
rohguess how pcbs are made16:22
lekernelyeah. but what about a thick piece of metal?16:23
rohbut you wont get great z-precision by 'continuing ecthing'16:23
lekernelyeah... they have this problems with semiconductors too. which they solved using special formulations of the etchants.16:23
rohi dont see any gain in there16:24
rohwaterjet is fine and comparedably easy to handle. just a bit expensive in room and money (but less than the other methods)16:24
carlobar hi, im trying to compile the kernel for milkymist SoC, and i get this eror:    arch/lm32/kernel/early_printk.c:43: error: implicit declaration of function out_be32, someone can help me?16:33
lekernelwhere did you get the source from?16:33
lekernelhttps://github.com/milkymist/linux-milkymist works for me16:33
carlobarim going to try that, thanks16:35
mwallelekernel: minimac2 tx irq is pulsed?17:13
lekernelyes17:13
lekernelbut not rx17:13
mwallekk :)17:14
Action: lekernel is making a bunch of flickernoise screenshots18:53
CIA-48flickernoise: Sebastien Bourdeauducq master * r21c418b / src/filedialog.c : File dialog: properly get file extension - http://bit.ly/gkOeD118:56
carlobarhi, i got the kernel on https://github.com/milkymist/linux-milkymist, but compiling i got the same error: arch/lm32/kernel/early_printk.c:43: error: implicit declaration of function out_be32. Maybe is something wrong with my linux configuration?20:16
mwallelekernel: should i remove minimac1 from qemu?20:24
mwallecarlobar: i guess there is a missing include (which declares out_be32)20:24
mwallecarlobar: mh out_be32 isnt provided anymore for lm32 arch20:29
carlobarhi mwalle, and what can i do?20:31
mwalletry writew() instead20:31
mwalleerm writel20:31
carlobarok, thank you20:33
larsccarlobar: #include <linux/io.h>20:35
mwallelekernel: and could you change the microudp timeout loop to use the timer, too? this way it'll work in qemu too20:35
mwallelarsc: out_be32 is ppc specific20:35
larscoh out_be3220:35
mwalledunno the exact equivalent20:35
larsciowrite32be20:35
larscwritel is litte-endian20:35
lekerneluhm, why did it work for me?20:36
mwallelekernel:  in qemu?20:36
lekernelmwalle: (remove minimac1 from qemu) yes, I think so20:36
larsclekernel: CONFIG_EARLY_PRINTK=n20:36
lekernelno, compiling the linux kernel20:36
lekernelah, maybe20:36
lekernelmwalle: (using timer) yes20:37
lekernelshould do that20:37
lekernelcounting cycles is just a quick hack20:37
lekernelminimac2 works neatly now... been using ftp quite intensively for the past 3 hours or so without bug20:38
lekernelthere is still a reset problem of the PHY, but if you've not run into it at boot time ethernet is fairly stable20:40
lekernelbtw it's clearly the PHY, pulsing its reset line when the problem occurs resolves it20:40
lekernel(without rebooting)20:41
lekerneland the symptom of it is just dysfunctional ethernet, no driver crash as before :-)20:42
mwallelekernel: btw the ac97 'problem' only affects lineout?20:45
lekernelseems so, but tbh I have not measured20:45
mwallei desoldered three ac97 codecs today.. maybe one of them is drop in compatible, ad1980, alc650, and stac922020:47
mwallemh the latter is a hd one20:47
carlobarwith CONFIG_EARLY_PRINTK=n the error was solved, but now i got other: kernel/posix-cpu-timers.c:287: internal compiler error: in lm32_block_move_inline, at config/lm32/lm32.c:785... this is because of wrong configuration in timers?20:49
mwallecarlobar: whats your compiler version?20:50
lekernelmwalle: I can send you a wm9707 if you want20:51
carlobarmwalle: gcc version 4.4.1 (GCC)20:51
mwalle carlobar lekernel: i think the latest working version was some 4.5.x right?20:55
larsc4.5.x workes find (somthimes)20:57
larscsometimes20:57
lekernelfor c yes20:59
lekernel4.5.1 works20:59
lekernelc++ seems broken in all 4.5.* and works in the 4.4 lattice release20:59
mwallelekernel: btw on the qemu ml there was some consens on first disable opengl by default, then make configure more intelligent in the way it links to the needed libs21:01
mwallethe disable by default patch wasnt merged, instead they are trying to figure out the real problem, but it seems everyone got stuck somewhere21:02
mwallewasnt merged yet..21:03
carlobari'll try to compile the kernel with the lattice toolchain.. i'll write about the results latter21:04
lekernelcarlobar: fwiw i'm using the 4.5.1 rtems toolchain21:08
lekernelhaven't tried 4.5.221:10
lekernel4.6 seems totally broken21:10
lekernelnew shots: http://www.milkymist.org/flickernoise.html21:56
rejonhot21:59
mwallelekernel: nice22:00
mwallegn822:35
lekernelgn822:54
wpwraklekernel: would be nice to see them in motion :)23:03
lekernelyeah, yeah, coming...23:03
lekernelunless something goes wrong, tomorrow there will be a party at tacheles and they would get filmed in action :-)23:04
--- Wed Apr 13 201100:00

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