#milkymist IRC log for Wednesday, 2012-03-07

wpwrakwe have other items from there, though. but no gao rong pins :)00:00
wolfspraulfinding quality suppliers in China (real Chinese, not Taiwanese) is harder than finding water in the desert, for sure00:00
wolfspraulwhich other items?00:00
wolfspraulwell as long as they are general parts and we don't make ourselves dependant on one vendor, it's ok00:01
wpwrakthat's the only .cn data sheet i have at the moment. then a few semi-obscure .tw00:02
wolfspraulproblem seems contained, good ;-)00:02
wolfspraulif we parameterize well and have multiple suppliers and just pick a chinese one because of price (assuming we think quality is ok), then that's fine00:03
wolfspraulbut in general you have to visit chinese companies00:03
wpwrakwe laso have a few farnell and one octopart "data sheet reference". haven't gotten to these yet. i'm sure they'll be entertaining :)00:04
wolfspraulthat's the reason so many chinese prefer to buy imported/foreign managed products, because finding quality chinese is so time consuming00:04
rohi really hope octoparts adds 'location' soon.00:04
rohas in 'i am in asia/us/europe' and list local sellers first00:04
wolfspraulthey exist of course, the problem is not 'china' but the fact that the market is essentially unregulated and chaotic, everybody tries whatever they want00:04
wolfspraulroh I am not following octopart closely but I do admire your passion for them :-)00:06
wolfspraulso if you tell me one day they are really good, I know I better listen00:06
wolfspraulfor now I think they are in a hopeless position00:06
rohwolfspraul: ;) i know how hard comparing sources on electronic parts can be. so i can understand what giant maddening problems they went against00:07
rohi think octoparts should add a rating system for sellers. could make finding quality sources easier.00:07
rohin the end its a tradeoff between quality and money, but thats ok as long as its transparent00:08
wolfspraulyes but nobody will want to be rated 'fairly' :-)00:08
wolfspraulI want to get away with my little cheat on quality00:08
rohno worries. we do qc much more paranoid than what i'd like to 'out'00:09
wolfspraulit's amazing if you think about it in low margin products, how big the difference is between skipping some tests or not00:09
rohbasically you only need to downvote sellers which sell broken/relabeled/fake parts00:10
wolfspraulanyway we all know the almost infinite details in this00:10
wolfspraulI like that boom starts with digikey00:10
wolfsprauland we later maybe add other 'quality' distributors00:10
wolfspraulone by one as we know them00:10
wolfsprauland then we parameterize as much as possible, making it easier to source parts from different manufacturers00:11
wolfspraulfinally I would try to reduce the number of manufacturers we source from at all00:11
wolfsprauland yes, the holy grail is quality and price00:11
wolfsprauland documentation00:12
wolfsprauland availability00:12
wolfsprauland ... :-)00:12
rohi think currently we just chose digikey to get good quality without thinking/testing ourselves, right?00:12
wolfsprauloh, I'm also very open minded to giving small new manufacturers a chance. that's something we can do that the large guys cannot risk.00:12
wolfspraulbut that needs to be a controlled thing then where we really trust that small/tiny manufacturer...00:13
wolfspraulroh: we are testing the entire assembled pcb, quite good nowadays00:13
rohtrue. but we have no incoming qc?00:13
wolfspraulcase by case, you can imagine everybody checks a bit00:14
rohjust asking.. i had weird screws in the bags etc.00:14
wolfspraulof course adam goes through, checks this or that00:14
wolfsprauleverything? nah, impossible00:14
rohmissing threads etc. i think one slipped through to adam even00:14
wolfspraulsystematic, following a plan? I doubt it, and it's not worth to formalize it to that level right now00:14
wolfsprauloh yes, I can imagine [threads]00:15
roheverything only works when a machine does it00:15
wolfspraulthen the SMT shop does checks before loading their machine00:15
wolfspraulbut we also need to do things practically because of our small runs00:15
wolfspraulI think the balance is right00:15
wolfsprauland as I think you know we suffered a little from what were probably fake TI schmitt-triggers00:16
wolfspraulI say 'probably' because you never really know and we cannot start an investigation or hire detectives :-)00:16
wolfspraulthere were also inexplicably high failure rates with the usb transceivers00:17
wolfspraulwell u know how it is...00:17
wolfspraulbut I am making 100% tested products, there has to be some work and risk in doing that00:17
wolfspraulit's fine00:17
wpwrakat low volume and high value per device, also some amount of rework is acceptable00:23
cladamw(J1/J2 datasheet) yes, i found that vendor changed their pin assignments number when I added/checked for insertion/removal, but the footprint in design files are correct to its real mechanical number. If i fix this, means that we also need to fix footprint library but will uncorresponding to previous rc3 version.01:28
cladamw(C2/C6) was kept to the same as previous codec lm4550b, since in rc3's design verification stage we agreed to the performance of audio noise after dropping wm9707 codec. If we'd like to follow wm9707's recommended parts which is also okay.01:31
cladamw(C6 polarized) This can be unpolarized if we follow wm9707.01:33
cladamw(R1) it's also came from previous codec.01:34
cladamwwpwrak, all those are messy and from that we tried to downgrade or upgrade with compatibility, so if we now ignore this, then we stick to all new codec recommendations. is okay to me, how about you ?01:37
cladamwalthough we already knew the new noise is less than older one.01:38
GitHub18[scripts] xiangfu pushed 1 new commit to master: http://git.io/NqUu2A01:43
GitHub18[scripts/master] update build with new softusb-input.h structure - Xiangfu Liu01:43
wpwrak(J1/J2) ah, good.01:46
wpwrakyes, i think we should update the design for the new codec. it's not a lot of things and those changes should be very safe.01:46
wpwrakand i think we're sure that we don't want to go back to the old codec ;-)01:47
wpwrakkeeping too much old stuff around just makes it hard to make changes in the future, because at some point, nobody will remember why certain things are there01:48
kristianpaulmigen and gtkterm ;D01:48
wpwrakcladamw: if you don't want to change J1/J2, then you should mention in the schematics that this is based on an older version of the vendor's design and that it's coordinated with the footprint.01:49
kristianpauli dont see clogb2 support in iverlog..01:54
kristianpaulneither verilator..01:54
cladamwwpwrak, okay that we can directly change/follow new codec. To (J1/J2) I'll mention it.01:55
cladamw(better illustration junctions) wpwrak I'll follow your pick on http://en.wikipedia.org/wiki/File:Schematic_junctions.svg01:57
wpwrakthanks ! :)01:58
cladamw(TVS on MIDI TX ?) don't know if add it. ;-)01:59
wpwrakif there's no explicit reason not to have it, then it should probably be added. i mean, i think it's the only place left01:59
wpwrakwhere we don't have TVS ;-)02:00
cladamw(common minimum voltage) on first schematic sheet which is good idea, so it's being to the 6.3V i think for the common.02:01
cladamwwpwrak, to use the same TVS as DVI ? or the audio in/out loop we'd have ?02:04
Action: kristianpaul considers again use clog2 instead of clogb202:05
kristianpaulnow where is that test bench....02:05
wpwraki think the ones for audio should be more than sufficient. the DVI ones seem quite sophisticated02:07
cladamwwpwrak, okay.02:26
cladamw(J1/J2) wpwrak, for a better safe, yes. after thinking. I'll edit J1/J2 schematic library to coordinate its assigned number(pin4 and pin 5 swapped).03:49
cladamwwpwrak, what do you mean ? >> "could we find a less creative placement of the "1%" at R164 and R165 ? "04:02
qi-botThe firmware (using branch) build was successful, checkout the VERSIONS for detail, see images here: http://fidelio.qi-hardware.com/~xiangfu/build-milkymist/milkymist-firmware-20120307-0337/04:17
qi-botThe firmware build was successful, see images here: http://fidelio.qi-hardware.com/~xiangfu/build-milkymist/milkymist-firmware-20120307-0640/07:19
qi-botThe firmware (using branch) build was successful, checkout the VERSIONS for detail, see images here: http://fidelio.qi-hardware.com/~xiangfu/build-milkymist/milkymist-firmware-20120307-0819/08:59
larscwell at least not all builds are a complete failure09:19
cladamw(R182) wpwrak, R182 which was still DNP from rc2, so I'll remove it from M1r4.09:55
cladamwwpwrak, good morning !09:56
cladamwwpwrak, what did it mean ? --> "nitpicking: could we find a less creative placement of the "1%"09:57
cladamw  at R164 and R165 ? :-)09:57
cladamw "09:57
wpwrak(R182) ah, another one. good :)09:57
wpwrak(R164) the location of the 1% is a bit strange. (before and slightly above the component reference)09:58
wpwrakbetter to put it either after or below the value09:59
cladamwoh...okay ;-)09:59
cladamwmy aesthetic standard is so unbelievable !10:03
wpwrakhmm. so you didn't change the junctions after all ?10:21
wpwrak(2a style vs. 3a)10:21
lekernelanyone feels like going to Rome? www.liveperformersmeeting.net10:34
lekernelparticipation deadline is march 1210:34
cladamwwpwrak, yup..since there are many of them, but it's connected with small circular dot so that i didn't change.10:52
wpwrakmmh. they're bad style. easy to get wrong.11:50
wpwrakwell, another item for the conversion-to-kicad style guide then11:51
qi-botThe firmware build was successful, see images here: http://fidelio.qi-hardware.com/~xiangfu/build-milkymist/milkymist-firmware-20120307-1123/12:02
cladamwwpwrak, hi the wiki R4 BOM, i put link firstly only, there's *.ods file uploaded. ;-) I'll do it later.13:02
wpwrakok :)13:18
wpwrakfor those wondering, it's  ssh-keygen -lf .ssh/whatever.pub18:33
wpwraksee also http://www.lysium.de/blog/index.php?/archives/186-How-to-get-ssh-server-fingerprint-information.html18:34
Action: lekernel is back from a nerdy tour at the local synchrotron18:56
wolfspra1lPadawan-: hi23:10
ThihiYeah, that's mine.23:25
ThihiApparently they sell articles as pdfs online too. It's actually a half page article in the magazine.23:26
wolfspra1lThihi: cool, thanks for the good work!23:33
ThihiHaven't seen the finished article myself yet.23:33
--- Thu Mar 8 201200:00

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