#milkymist IRC log for Wednesday, 2012-03-14

wpwrakwolfspraul: and there are analog domain differences in the transformer circuits that may or may not be real-life differences00:00
wpwrakif the circuit has subtle flaws, this could mean more bit errors, shorter maximum cable length, and so on. but i don't know if there's actually any issue. maybe all those things that look different aren't.00:01
wpwrakhmm. this kind of page scares me: http://thedmxwiki.com/definitions/dmx_terminator00:06
wpwrak"If the addition of a terminator causes problems in your chain, it is likely down to poor quality DMX Cable."00:07
wpwrakdoes that mean devices aren't expected to already have termination ? or are they trying to say multiple termination is okay ?00:07
lekernelsome devices have just the input connected directly to the output, and the RX circuit "tapped" on it (without termination)00:11
lekernelwhat we are doing is RX -> termination -> receiver (to 3.3V logic) -> transmitter (from 3.3V logic)00:12
lekernelthere's no problem with that00:12
wpwrakoh, i see. i saw a termination switch mentioned. i guess that's not something we'd want to consider ?00:15
wolfspraulreading about power switch for mmc worries me :-)01:00
wolfspraulat first glance the R4 bom cost went up about 30 USD from R3, and again at first glance I see a lot of quite expensive new switch/protection stuff01:00
wolfspraulthe will require some more detailed study and work later to analyze and optimize - not now01:00
wolfspraulmaybe after the whole thing is boomified would be a good time01:01
wpwrakyeah, we added quite a bit of stuff in the end01:11
wolfspraulwell. do you have any candidates that can be removed right away? :-)01:18
wolfspraulI think we leave everything as-is now, and wait until we have boom ready then we can do a really effective costing-down.01:19
wpwraki think it's more a question of removing functional groups than of tweaking components01:19
wolfspraulwhile our volumes are low we can afford a little over-engineered/crowded board01:20
wolfspraulbut maybe here and there less is more :-)01:20
wpwraki think it's best to start with many possibilities. then see what actually gets used.01:21
wolfspraulsure. I am also wondering whether there are unnecessary/overly conservative protection parts on the board now, electrical stuff that you wouldn't find in a normal consumer product01:21
wolfspraulbut like I said, I think we can leave everything as-is now, wait for boom, and then look at it effectively cost-wise01:22
wolfspraulI care less about being overly protective electrically than I care about expensive parts, too many different vendors, too many parts, parts with EOL problems, etc.01:24
wolfspraulalso large parts01:25
GitHub25[flickernoise] xiangfu pushed 1 new commit to master: http://git.io/ht4hzg01:38
GitHub25[flickernoise/master] Ctrl+H : display keyboard shortcut information - Xiangfu01:38
wpwraknice :)01:40
xiangfuwpwrak,  I even don't know those shortcuts before this patch. :-)01:41
wpwraki think only sebastien knows them. maybe :)01:42
wolfspraulxiangfu: can cursor left and right go backward and forward in the patch list?01:47
xiangfuwolfspraul, no.01:49
wolfspraulno I mean, of course they cannot do that right now. but can we add it? :-)01:50
xiangfuwolfspraul, Ctrl+H . again and again. it will display message one by one.01:50
wolfspraulcursor-left: previous patch. cursor-right: next patch01:50
xiangfuwolfspraul, oh. yes  of cause we can add it and it is easy :)01:50
wolfspraulhow about audio sensitivity?01:51
xiangfuthe audio sensitivity bug? not working on that yet.01:52
wolfspraul2 keys so we can increase and decrease audio volume01:52
xiangfuno such shortcut yet.01:52
wolfspraulI say just start simple and put it on cursor-up and cursor-down01:53
wpwrakxiangfu: btw, you may want to add "static" and "const" to the "help" array. also, it's not a good idea to use the same name for a global and a local symbol. if we enabled -Wshadow, you'd get a warning01:53
wpwrakif you only show one message at a time, you may wan to reduce the number of messages. e.g., combine increase/decrease in one. don't show Ctrl-H, because the user has found that one already01:55
wpwrakand you have F1 twice :)01:55
xiangfuF1. yes that is in different mode.01:56
xiangfusimple mode F1 means patch name01:56
xiangfuconfigured mdoe F1 means video-in01:57
wpwrakphew :)01:57
wolfspraulI'd say focus on simple mode01:57
wpwrakat least you don't have to use morse code to express your wishes ;-)01:57
wolfspraulno help in configured mode01:57
xiangfuI try to combine to one line. but the OSD display font is not very good on format. so I just give up and commit first :)01:58
wpwrakmaybe disentangle the hotkeys while we're at it ?01:58
xiangfuwpwrak, improve later. :)01:58
wolfspraulxiangfu: focus on simple mode01:58
wpwrakNB: if you take one of those little rf keyboards, about half the of those hotkeys may not even be available01:58
xiangfuwolfspraul, yes. agree. :)01:58
wolfspraulwpwrak: that's next01:59
xiangfuwolfspraul, we should re-enable all midi / osc etc patches on simple mode.01:59
wolfspraulthe shipment of 20 Riitek keyboards is on the way to Adam as we chat01:59
wolfspraulonce they arrive, I will remove the silicone keyboard for good01:59
wpwrakheh ;-)01:59
wolfspraulI will leave the remotes in the box until they are out of stock, but then not refill.01:59
wpwrakwhich model did you choose in the end ?01:59
wpwrakwill you keep the IR sensor ?02:00
wolfspraulI think we should keep it, yes.02:00
wolfspraulthere could be interesting applications later and if it's about BoM cost, there is A LOT of other stuff we can optimize.02:00
wolfspraulthis one http://www.riitek.com/product_Info.asp?id=5602:00
wpwrakokay. let's revisit the quesiton when the time comes for M1rc5 or M2 ;-)02:01
wpwrakand you switch from acrylic walls to something else. then realize you now need an IR-transparent window. etc. :)02:01
wolfspraulI wouldn't even care making it dysfunctional without removing the ics.02:02
wolfspraulnobody uses it anyway02:02
wolfspraulI will strictly just go step by step to improve actual product usage.02:02
wolfspraulin users hands, not theoretically02:02
wpwrakthen you may as well remove the chips :) well, in > M1r402:02
wolfsprauland we have quite a few guys now who want to use their m1 more, but they are facing this or that real-life problem. I work on fixing them.02:03
wolfspraulbtw, the two ADI codecs for the r4 design-verification run are with Adam now, vga and video-in (adv7181c)02:07
wolfspraulI just got fed up with ADI official channels and we bought in shenzhen grey, and 2 days later they are at Adam's02:07
wpwrakgreat !02:07
wpwrakheh :)02:07
wpwraklet's hope they are what they claim they are :)02:07
wolfspraulsure, will be02:07
wolfspraulI replied to all questions from ADI, but even though my mails got friendlier and friendlier, zero replies02:08
wolfspraulthe last one was the friendliest with me telling them that the problem had gone away thanks to shenzhen, and thanking them for their generous help02:08
wolfspraulI did get some "out of the office until xx-xx" auto-replies though02:09
wolfspraulbut of course I understand their problem, basically they cannot afford to write even a 1-line email to any customer odering less than.. what... 10k chips? 100k? :-)02:09
wolfspraulanyway, problem solved02:09
wolfspraulin the long run I am hoping we go all digital for video-in and out and can remove those codecs from our pcb02:10
wpwrakhmm yes. a good objective for 2022 :)02:11
xiangfuwpwrak, we have a lot of 'shadows a global declaration'02:12
wolfspraulnah, if we can pickup some speed it can be much faster. but one by one, sure. the product must stay working.02:12
wpwrakyes :)02:12
GitHub3[flickernoise] xiangfu pushed 1 new commit to master: http://git.io/rhheXA02:13
GitHub3[flickernoise/master] disable mouse click exit rendering mode - Xiangfu02:13
wpwrakwolfspraul: i don't think the global obsolescence process isn't seriously affected by our speed :) of course, you can just cut off analog-using users ...02:13
kristianpaulfinally! :)  (commit)02:13
wolfspraulwpwrak: oh, not at all [cut off]. let's see how the codec situation evolves.02:14
cladamw(C126) wpwrak thanks, Nice catch ! It can be 470nF which bom has it, see my replied. We didn't review carefully after changed from TSOP4838 to TSOP34838.02:37
kristianpaulwpwrak: btw where is that thread you made to catch user input from kbd?..02:39
cladamw(Varistors) all their symbols I changed them being as likely diac. so can't be faced as coil or bead or choke.02:39
kristianpaulah debug navre i guess02:40
cladamwAh~ from werner's great reviews, there're more mistakes discovered while I had made before. :( No excuse though.03:15
cladamwit seems that was quite hard to find mistakes in details in 1 ~ 2 persons. but one Werner beats our manual works.03:17
cladamwi hope the sooner boom system can really avoid all of these.03:18
cladamwwpwrak, don't know how to say thanks to you.03:18
qi-botThe firmware build was successful, see images here: http://fidelio.qi-hardware.com/~xiangfu/build-milkymist/milkymist-firmware-20120314-0248/03:28
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-20120314-0428/05:07
qi-botThe firmware build was successful, see images here: http://fidelio.qi-hardware.com/~xiangfu/build-milkymist/milkymist-firmware-20120314-0610/06:50
wolfspraulmaybe the bot should only announce failed builds?06:51
wolfspraulif we commit often and each commit triggers a build, then failure is far more important to be reported than success (which should be the norm)06:53
xiangfuthe build host is doing like:06:53
xiangfu1. check new commit compile milkymist-firmware --> check new commit compile a branch milkymist firmware --> check new commit compile ben Nanonote image.06:54
xiangfuthe step_3 is usually needs ~40 hours.06:54
xiangfusince I am working on the nanonote. so I disable build nanonote today. then it's will only build milkymist stuff. each build about ~2 hours.06:55
xiangfuwolfspraul, in fact I think those build success information is ok. even one message every 2 hours I can accept that :)06:58
xiangfuas lease I know build host is working :)06:58
wolfspraulok :-)07:06
wpwrakkristianpaul: https://github.com/milkymist/flickernoise/blob/master/src/shellext.c#L27607:55
wpwrakcladamw: ah, there were more issues ? checking the mails ... :)08:06
cladamwwpwrak, hi no, i'm just still digest your mails, i only replied that what i knew or i can do. ;-)08:07
wpwrakcladamw: btw, the "killer tool" for the reviews is "dsv". boom doesn't really enter the picture here. boom is more a "spend less time doing boring stuff like looking up 530 Ohm resistors" tool08:07
cladamw(TBD on my site) like 150uF with low ESR check, ...08:09
cladamw"dsv" okay, since your discoveries that I knew how they went through before. U3/C126/C120 were the problems we fixed but didn't review carefully. :(08:12
cladamwsince you may continue to review other sch pages, I'll reply asap. then later i generate a whole pdf after go through all reviews. Much thanks for your reviews. But i am curious why still no else can join. hehe :(08:15
wpwrakyeah, participation is a bit disappointing08:21
wpwrakjoerg has made a few small comments. maybe there will be more08:21
cladamwwpwrak, i am editing a page which will house to know idea: http://en.qi-hardware.com/wiki/Milkymist_One_Layout_Criteria/zh_tw08:23
cladamwbut one *ps file I may need your help to update. :-)08:24
wolfspraulthere is no/very little culture of collaboration in hardware08:24
wolfspraulwe try to jumpstart it but well, either we do something wrong or it's hard or just never will take off08:25
cladamwhttp://downloads.qi-hardware.com/people/werner/tmp/xbrd.ps  could you update this file according to 10*2 female dimensions ? No need now, but later when you're available.08:26
wolfspraulI think eventually it will08:26
cladamwoah..yeah..much thanks joerg too.08:26
wpwraki guess we're either lacking critical mass or didn't announce the review properly. maybe people didn't understand that this is a "free for all", not just the usual suspect slugging each other in public08:27
cladamwwpwrak, i just saw your replied on C120. since from M1rc1 the U3 we always used SO5032 with 10nF.08:31
wpwrakcladamw: updated: http://downloads.qi-hardware.com/people/werner/tmp/xbrd-top.pdf08:32
wpwrak(note the name change. in the future, we'll also gave other views, hence the addition of "-top")08:32
cladamwwow...so fast ! thanks.08:33
wpwrakalso note that this isn't a specification (yet), just a draft for possible dimenstions. layout will determine what actually works and i'll update the specification from that08:33
wpwrak(fast) the joy of parametric cad ;-)08:33
cladamwso I should remove " In practice..." as well as no need those mentions, better ?08:36
wpwraksince we're using SO5032, why not change C120 to 100 nF ?08:40
cladamwsince no experiments, and from rc1 to rc3, they works well, no ? of course we can change C120 to 100nF in M1r4( only 8pcs will be produced )08:46
cladamwso if the results from those 8 pcs after m1r4, then we keep 100nF, or think that 10nF, 100nF won't influence much at all. sorry i don't know now. :-)08:47
wpwrakyou can always try and rework an existing board ;-)08:49
wpwrakbut i don't think we really need to test this. going from 10 nF to 100 nF seems to be a safe change, particularly since it's what the data sheet suggests anyway08:49
wpwrakthe problem with "it works" is that we don't know where in the parameter space we are, also regarding component production tolerances08:51
cladamwmmm..okay, so I'll change C120 to 100nF for SO5032, thanks for explains. :)08:54
cladamwalso change U3's name to SO5032.08:55
wpwrakkewl, thanks !08:56
GitHub109[migen] sbourdeauducq pushed 1 new commit to master: https://github.com/milkymist/migen/commit/5c0cc6292c440e19dab5594477452769ca37dd7111:26
GitHub109[migen/master] fhdl: export log2_int - Sebastien Bourdeauducq11:26
wpwrakhmm ... no comment on R54 today ? :)13:04
lekernelwpwrak: you can play with the value, but it does have impact on the signal levels13:16
lekerneland the opto can pull 60mA/dissipate 100mW so it should be ok, even if it's not power efficient13:17
lekernelmax power you can get from that resistor is 23mW13:17
kristianpaulwpwrak: thanks !13:46
GitHub48[migen] sbourdeauducq pushed 1 new commit to master: https://github.com/milkymist/migen/commit/1665f293a619b6b989031fb5b293df9509f5760015:26
GitHub48[migen/master] bus/asmibus/hub: require finalization before get_slots - Sebastien Bourdeauducq15:26
wpwraklekernel (60 mA) ah, right. i got the wrong side.15:26
wpwraklekernel: still seems very low, though. did you experience problems with larger values ?15:27
lekernelcan't exactly remember... but yes, there has been some tweaking to get a proper signal15:40
lekernelthere's probably a large margin with the current value though15:40
lekernelbut values in the kohm range do cause problems15:42
wpwrakit should be input leakage plus anything we lose due to parasitic capacitance, no ?15:42
lekernelthere are problems with rise/fall times too15:43
lekerneloptoisolators are slow devices15:43
wpwrakwith a cut-off frequency of 1 MHz (> 20x the bit rate) and a parasitic capacitance of 50 pF, i get .... 3.3 kOhm15:43
wpwrakslow edges at the source should actually help, as far as moving parasitic capacitance is concerned15:45
wpwrakor at least not make things worse15:45
wpwrakdo you remember what value you had trouble with >15:45
lekernelwith the first optoisolator the problem was extreme15:46
wpwrakif it was 100 kOhm, that would make perfect sense. perhaps even 10 kOhm.15:46
lekernel(we changed it with that model later)15:46
wpwrakah, i see15:46
lekernelyou could pass only up to a few kHz, all the rest was filtered out - unless you put a ridiculously low resistor, which would damage the opto15:46
lekernelthis one has better characteristics15:48
lekernelyou can experiment with it if you wish, but as far as I'm concerned I believe in the old saying - if it ain't broken, don't fix it :)15:48
lekernelthe opto is operating within specs, gives a good signal, and uses a relatively small amount of power. what more can we wish for? :)15:50
wpwrakhmm, almost 3% of the entire power budget of the M1 ;-) (not counting USB, i.e., assuming 5 W = 1 A)15:52
wpwrakit's good that we don't have dozens of MIDI interfaces. else, M1 would feature a cooling tower like you find on modern CPUs ;-)15:53
GitHub78[milkymist-ng] sbourdeauducq pushed 1 new commit to master: http://git.io/ByEWgA17:32
GitHub78[milkymist-ng/master] asmicon: skeleton - Sebastien Bourdeauducq17:32
lekernellarsc: you're in Hamburg right?23:35
larscno longer23:37
larscmunich now23:37
--- Thu Mar 15 201200:00

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