#milkymist IRC log for Friday, 2012-03-02

wolfspraulcladamw: good morning02:05
wolfspraulSebastien mentioned something yesterday that I wanted to reiterate and explain before you go to the layout house (plab)02:06
cladamwgood morning02:06
wolfspraulhe said "when you have the first R4 boards, send me one right away so I can verify whether we can still clock sdram at 166mhz"02:06
wolfspraulwe are making many (good) changes and improvements in R402:07
wolfspraulbut we also need to be careful about regressions02:07
wolfsprauland memory performance is maybe one of the highest most important things on the board02:07
wolfspraulso when you go to layout, keep an eye on this and try to extract as much as possible knowledge from them on how we can keep sdram performance up02:07
wolfspraulwe rather sacrifice some of the other improvements than taking a hit on sdram performance02:08
wolfspraulof course ideally we want all :-) new features and still the highest sdram performance we can achieve02:08
wolfsprauldoes this make sense?02:08
cladamwmake sense, there's no else bigger changes than keeping sdram performance.02:09
cladamwso i'll always ask house and keep my eye to see the placement/route works02:10
wolfspraulyes, and try to get advice from them02:10
wolfspraulwe need to learn learn learn02:11
cladamwof sourse we make use of extra un-used pins which may/or already impact the performance in sdram, so far we don't know if it will. but IF house tell me the truth existed from thier experiences before, then I take their experiences rather than our expected changes, and feedback irc /or list to know.02:12
wolfspraulyes02:13
wolfspraulwe want both02:13
wolfspraulwe want our changes and the best possible sdram performance ;-)02:13
wolfspraulbut sdram performance is indeed very important02:13
wolfspraulthat's what Sebastien reminded me of and of course I think we all agree02:13
wolfsprauland yes, from the first R4 mini-run (maybe 8?), one board goes to Sebastien right away to verify just that...02:14
wolfspraulbut before that we have to keep an eye on it, I am hoping the layout house can give us good advice02:14
wpwrakdid we get any feedback from the layout house for M1rc3 ?02:16
wpwrak(it was the same, wasn't it ?)02:17
cladamwme too. ;-) I'll edit the most info in this page before sending design files to house: http://en.qi-hardware.com/wiki/Milkymist_One_Layout_Criteria02:17
cladamwwpwrak, what did you mean ? What expected info you point ?02:17
wpwrakcladamw: no expectation. i'm just curious if they said anything and if yes, what.02:18
cladamwi need to edit subtitle for this R4 criteria later02:18
wolfspraulI think they say a lot of things02:19
wolfspraulmaybe Adam can edit the wiki with some of what they say02:19
wpwrakgood. so they communicate :)02:19
wolfspraullayout is one of those areas where we still by and large stay away, so there is very little discussion about layout preferences, criteria, etc.02:20
wolfsprauloh of course02:20
wolfspraulit's a small company (haven't been there though)02:20
wolfspraulthen Adam also goes there to sit down with them and go through the layout, for hours02:20
wolfspraulcladamw: correct?02:20
wolfspraulif needed, multiple times02:20
wpwrakgood. there are often design details you don't see in the schematics.02:21
cladamwthe problem is may be my cooperated attitude does still not get used to 'record' everything details in whatever any words they told to me. So probably it's my problem not house for all previous histories.02:21
wolfspraulyes but relax02:22
wolfspraulit is extremely hard to take notes of this kind of verbal sit-down co-work situation02:22
wolfspraulthere's just too much back and forth02:22
wpwrakyeah. i just want to get a feeling for how that communication works. what we have to prepare, what to expect, etc.02:23
wolfspraulAdam starts with a wiki page, like the one he pointed to02:23
wolfspraulbut then it gets messy02:23
wolfspraul:-)02:23
wolfspraulalso when he's at their office, not sure whether he has the internet connection and time/chance to update the wiki right there02:24
wolfspraulat least so far that probably did not happen02:24
cladamwthe behaviours among like KiCad or close Cad work, the differences may cause such unexpected results, but I'll always try to get inquires from 'experiences' which most Chinese guys are still not very feel free to share though. But this time I'll try. ;-)02:24
wpwrakhe should go there, sit in an isolated room, and communicate with them only via IRC. then we could log the conversation ;-)02:24
wolfspraulthen he comes home at night, with dozens of details being discussed, and it's hard to update the wiki02:24
wolfspraulcladamw: you don't need to tell them anything about free or share02:24
wolfspraulmaybe you can update the wiki page live when you go there?02:25
wolfspraulthat might be easier, if it's possible technically02:25
cladamwwolfspraul, the most problem is that I'm not sitting always besides layout guys. when I went to their office, my work is that:02:26
wolfspraulhow big is that company?02:26
cladamw1. tell them my request on each part/circuit/footprint02:26
cladamw5 person to handle ODM/OEM job, but only one guy she can route, and when in meeting, I meet their PM/route guy. only three, included me.02:28
wolfspraulahh. so much we can learn from them :-) They must be doing tons and tons of layout, and know a lot of little helpful tricks :-)02:29
cladamw2. step by step goes through each item/list I concerning, and tell my work priority. meanwhile discuss with them.02:29
cladamwso I can make notes while discussing.02:29
cladamwwolfspraul, oha..yeah..sure they bet!02:30
cladamwproblem is on me to digs the treasure among them.02:31
cladamwand skills on 'record' & make 'notes' .;-)02:31
wolfspraulsure02:31
wolfspraulthat is very difficult, I know exactly because I tried a lot, sometimes successful, sometimes so-so, sometimes failed02:31
wolfspraulat some point you get carried away in the discussion, and all sorts of things are being discussed back and forth, and it becomes impossible to record from memory later02:32
wolfsprauland if you put a dictation device (recorder) there, you have 2 problems: 1) they may feel bad, and no free discussion will ever start 2) editing the recording is a tedious job that will require many hours of work02:32
wolfspraulso that also doesn't work02:33
wolfspraul:-)02:33
wolfsprauljust try to take some notes02:33
cladamwhttp://en.qi-hardware.com/wiki/Milkymist_One_Layout_Criteria#Criteria_for_M1R402:33
cladamwjust make a subtitle, if you think out some thing else, don't hesitate to add them in. I'll sum up surely after we finish M1R4 list review.02:34
cladamwwolfspraul, you are very right on analysis on how they/vendor feel the cooperation with us. always be the challenge.02:36
wolfsprauloh you cannot put a recorder there02:37
wolfspraulI wouldn't like that either02:37
cladamwif they are not busy, it's always okay. if not, things go worse, but if they busy, I used to not push them , and make another appoinement to meet them.02:37
cladamwyeah02:38
wolfspraula secret recording may work, it's not nice though (using your notebook mic or so)02:38
wolfspraulbut even then you have tons of work editing all that - don't do it02:38
wolfsprauljust take classical notes, then edit wiki02:38
wolfsprauland sure, we are a tiny customer02:38
wolfspraulso what02:38
wolfspraulit's like say you hire a lawyer, for some small legal trouble you have with your landlord (better not ;-))02:39
wolfspraulthe lawyer may have giant clients that give him work all the time02:39
wolfsprauland those clients will have preference, always02:39
wolfspraulthat's ok02:39
wolfspraulbut at the same time the lawyer will (should) do a proper job even for his small one-off clients02:39
wolfsprauland so will the layout house02:39
wolfsprauljust no stress on this02:39
wolfspraulif they have to push us back, reschedule an appointment, etc. all fine.02:40
wolfspraulif we would be a big customer, we would demand them to do that to their small customers as well, to give us priority :-)02:40
wolfspraulI would, I admit.02:40
cladamwoh...indeed a nice balance/reasonable thinkning.02:41
cladamwthat's the thing how it goes. well...just always to probe/check them out. ;-)02:42
wolfspraulyes. I practice this every day in China, as you can imagine.02:42
wolfspraulin life, someone is first, and someone is last.02:42
wolfspraulI like to drive my Chinese friends crazy by opting to be the last one, he he02:42
wolfspraullike when entering a bus, I go to the end of the line saying "today I want to be the last one"02:42
wolfspraulha ha (really, I do that)02:42
wolfspraulthat gives me such satisfaction. and they think I'm some alien from Mars, everybody fighting with all sorts of little tricks to go 1 spot ahead in the line...02:43
wolfspraulbut there are days when I do that too, and I do it well. and I will be the #1 :-)02:43
wpwrakthey probably assume you're secret police or such ;-)02:43
cladamwit's sometime pretty good idea , really, they even call me before and after Chinese lunar new year to know when I will go there/house to have a meetings.02:43
wolfspraulnah, it's just a culture play02:43
wolfspraulbut I really do it, it's relaxing02:43
wolfspraulsomeone has to be last, right?02:44
wolfspraulmight as well be me today! :-)02:44
wolfspraulI can't stand this stressful "I have to be #1" all the time, just takes the joy out of life...02:44
wolfsprauldefinitely OT in #milkymist, sorry :-)02:44
cladamwat that time I would say oah yes, nice that you called me about this, and this time the smt run we surely select your vertical system to make it. or etc...02:44
wolfspraulway to go layout house! :-)02:44
cladamwwpwrak, http://downloads.qi-hardware.com/people/adam/m1/tmp/m1r4/DVI-ISingle_20120302.pdf02:59
cladamwthe EN/FAULT I followed ds to use external 10K.03:00
wpwrakhmmm, how about moving the DVI connector and surrounding components to a separate sheet ? it gets a little crowded ;-)03:02
cladamwand changed orignal C188 1uF/0402 to 10uF/0805 as ds suggested.03:02
cladamwwpwrak, ha..just to finish sch then we(with house) to crazy route brainstorming. ;-)03:03
cladamwdon't forget we still can put some passive parts at pcb bottom side. ;-)03:03
cladamwwell... just need to make good placement before route. ;-)03:04
cladamwwpwrak, no extra sch sheet to add, please. ;-)03:05
wpwrakbah, bottom-mount the fpga. that would be fun ;-)03:05
wpwrakwhat's wrong with extra sheets ?03:05
cladamwno, i just printed this dvi-i sheet only.03:06
cladamwsecond, letmme generate all again to update. ;-)03:06
wpwrakah, interesting. 0805 is half the price of 0603. that's why you picked it. good.03:11
wpwrakthe U32 circuit looks fine03:12
kristianpaulergh, i'm lost if i want a bug free rtems what should i do?..03:16
kristianpaulxiangfu's scripts apply those un-merged patches? or..03:17
cladamwwpwrak, thanks check03:17
Action: kristianpaul need filesystem support for osgps porting.. soo anyway.. the rtems03:17
wpwraka bug-free rtems ? hmm. detach yourself from reality ? :)03:18
kristianpaulwpwrak: well, bug-know free ;-)03:18
wpwrakthen you want those patches, yes03:19
kristianpaulha ! and yes i had troubling with getting catch keystroke, is amazing that i remnber gave up and swich back to bios..03:19
kristianpaulnow i need an OS again :)03:19
kristianpaulat least you tell me there is a way to implement yaffs2 against a baremetall app :D03:20
kristianpaulwpwrak: ok, i'll find then ! :)03:20
xiangfukristianpaul, this one : compile-milkymist-firmware.sh have merged patches from wernermisc.git03:21
cladamw(latest R4 draft)http://downloads.qi-hardware.com/hardware/milkymist_one/sch/tmp/MILKYMISTONE.pdf03:21
xiangfukristianpaul, all you needs is update all milkymist git repo. then apply wernermisc.git patches for rtems.03:21
cladamwwpwrak, another one mechanical hole added in fgpa sheet.03:22
cladamwI'll start to sum up my histories about all changes then send to list for review if you don't see any big problems. ;-)03:23
wpwrakah :) maybe mention in the comment that it's a mechanical hole ?03:23
kristianpaulxiangfu: thanks °03:24
kristianpauls/°/!03:24
cladamwaha...mention to like : make strong fix for suffering from push extension board when plugging ?03:25
cladamwor show me the best mention you like. ;-)03:25
wpwraknaw, just say something like "mechanical mounting hole"03:25
wpwrakotherwise, readers will wonder what the heck this "connector" does :)03:25
cladamwhup? hehe..okay.03:26
cladamwoh ... yup yesterday you found it long :)03:26
wpwrakschematics aren't a good place for writing long stories :) what concerns me most are things that can't be easily identified. such as J27 and friends.03:28
cladamwtrue03:32
cladamwwpwrak, updated and changed J21 J22's p/n to also FCI p/n : 87606-310LF you selected.03:42
cladamwthat p/n is being the final but can be also alternative. as long as we find another same spec but good price.03:44
cladamwgood price sometimes doesn't show up available quantity in digi-key though.03:44
wpwrakyeah03:49
wpwrakwe should select 2-3 choices. also for mating parts. and the little plastic plug. then you could order a few of each and test that they really fit.03:51
cladamw(keying) http://downloads.qi-hardware.com/people/werner/m1/tmp/plastic-key.jpg03:52
cladamwwe should get few samples about this too to see if can fit as well.03:52
cladamwi had have male 40*2 headers here. so it's okay03:53
wpwrakwe should check that the digi-key parts fit. there's quite a bit of variation on, say, mating length03:54
cladamw40*2/2.54mm header is very  popular in normal electronic shop.03:54
wpwrakoh yes, that it is :)03:55
cladamwsure, but female connector on main board, the most parameter is the height said 8.5mm03:55
cladamwand check their suction hole's depth to let male plugin.03:56
wpwraknot quite :) the main parameter is the mating length. the overall height is influenced by it (and should be more or less have a fixed offset).03:56
wpwrakyup, the "hole depth" is the mating length03:56
cladamwthe FCI, Harwin and Molex have parts to as the same 'mating length', yeah...we can leave this task later when order parts.03:57
wpwrakyeah, it's more a documentation issue. for the "expansion connector spec"03:59
cladamwtime to review mistakes or stupid connections or underline secrets we didn't find among those changes we made. phew~03:59
wpwrakwe basically want to have a set of parts we know will work03:59
wpwrakso if someone runs into trouble, we can offer them a safe choice03:59
wpwrakwhether they take it or not, well, not really our responsibility :)04:00
cladamwyup, to promote a good application on expansion board, we should edit a specific wiki page or document to introduce this important feature.04:00
wpwrak(stupid secrets) hehe ;-)04:00
wpwraki like LaTeX :)04:00
cladamwWhat's that 'LaTex'?04:01
cladamwhttp://en.wikipedia.org/wiki/LaTeX ?04:02
wpwrakyup :)04:02
wpwraki used it for the midi manual. and i think sebastien used latex for the FN handboek as well04:03
cladamwalright, so do u have else opinions/comments on draft ?04:03
wpwrakit's the tool of choice if you want good-looking documents04:03
cladamwoh, for document tool.04:04
wpwrak(draft) is there anything specific you want me to look at ?04:04
wpwrakoh, and joerg suggested to NC the J21/J22 pins we use as a key. so the GND wouldn't actually connect in schematics or layout.04:05
cladamwno, if you don't have else to remind me. then I'll sum up and post to list for a whole review. and if I don't describe well somewhere then pls directly reply to replenish. ;-)04:06
wpwrakhis reasoning is that this gives some more layout choices for expansion boards. i'd agree with this.04:06
cladamw"On the expansion board or connector side, one would simply04:09
cladamwcut the respective pin. Joerg suggested to cut pin 4 of J21 and pin 1 of J22. With04:09
cladamwthis kind of keying, connectors could not be reversed, would04:09
cladamwbe harder to shift, and it would also be difficult to insert something for J21 at J22 and vice versa. We should probably still connect the "cut" pins to GND, just in case."04:09
wpwrakhmm, the SW2 comment may still be a tad cryptic. how about: "About the numbering: there used to be also SW1 and SW3 (and BTN1/3), but we found that one button is all we need, and removed them." ?04:10
cladamwwpwrak, maybe i misunderstood this content,04:10
cladamw(SW2) okay,04:10
wpwrakwhat joerg suggested contradicts my suggestion to ground them. but i think it's okay to follow his suggestion. so instead of grounding the pins that will get a plastic key (J21.4 and K22.1), NC them.04:11
cladamw(keying) I was thought/understood Joerg's 'cut' which to cut the pins on extension board not M1 side. am i right?04:12
wpwrakbonus: you can also see in the schematics where the key is04:12
wpwrakon M1, it would be a plug (that little plastic key). but yes, the trace would be cut (or just not exist), too.04:13
cladamwso let J21.4 and J22.1 be NC(was GNDs) in M1 side, is that correct ?04:14
wpwrakyup04:14
cladamwokay...i change them . ;-)04:15
cladamwwpwrak, updated. :)04:29
Artyomwolfspraul: Am I understand right that milkymist-group is mainly focused on developing MM SoC and MM1 schematic? pcb-layout is made by third party company?04:35
wolfspraula number of external companies/suppliers are involved04:37
wolfspraullet's go through04:37
wolfspraulfirst there is one that is culture-compatible, that's www.raumfahrtagentur.org04:37
wolfspraulraumfahrtagentur has been doing the mechanical design and case manufacturing04:37
wolfspraulall files are published freely, free tools were used (qcad), etc.04:38
wolfspraulthe electrical board design is still primarily held in Altium files, the Altium files of course are published04:38
wolfspraulfrom the Altium files we create the GERBER files that are used to make the PCBs, by newheart.com.tw04:39
wolfspraulthe GERBERs are also somewhere04:39
ArtyomAnd who have access to Altium? It's a commercial tool.04:39
wolfspraulfor the layout, we use the service of a layout house, PLAB www.plab.com.tw04:40
wolfspraulso we give them the Altium files that only contain the schematics, and they add the layout (basically), for a fee, and return the files to us04:40
wolfspraulyes, agreed. so we want to switch to KiCad :-)04:40
wolfspraulthen we use a SMT place MinBo in Taipei (no functioning website)04:41
wolfspraulassembly and testing at Adam's home factory ;-)04:41
ArtyomAnd how expensive is PLAB service for a boads like MM1? Does PLAB works with KiCAD? Do they make signal-integrity analysis?04:42
wolfspraulaccessories are sourced from various vendors, mostly in China (cables, camera, keyboard, remote from Taipei). All of those are documented in the wiki.04:42
wolfspraulgood question04:42
wolfspraulespecially since we have another round of using PLAB just next week or so :-)04:42
ArtyomAnd who is producing PCBs?04:42
wolfspraulthink of PLAB like a service-for-hire04:42
wolfsprauljust wrote: newheart.com.tw04:43
wolfspraulso PLAB is a layout house that is doing layout for OEM/ODM in Taipei04:43
wolfspraulbasically you hire them because of their experience, they are fast, etc.04:43
wolfspraulso no, definitely, if we would switch to KiCad we remove our ability to work with PLAB, unless just for giving them GERBERs and asking for feedback04:44
wolfspraulbecause PLAB has professional layout people and they are so crazy fast in the tools they know, that's why people work with them04:44
wolfspraulthey will typically support PADS Layout/Mentographics, Altium04:45
wolfsprauldefinitely no kicad or geda04:45
wolfspraulwe searched all of Taipei, no professional company even has heard of them04:45
wolfspraulsame in China, though China is huge and there may be some small places somewhere04:45
wolfspraulso if we move to kicad, we start with the schematics & bom04:45
wolfspraulthe layout stays in altium, for now04:46
ArtyomIn what format do they take electic schemes? Can they convert from KiCAD or any other free program to their tools?04:46
wolfspraulwe do all these steps one by one so it's practical. no dogma. we want to make functioning products.04:46
wolfspraulthey 'can', of course04:46
wolfspraulyou pay for their working hours04:46
wolfspraulit's like a lawyer04:46
wolfspraulyou can hire a lawyer to start all sorts of crazy lawsuits for you04:46
wolfspraulas long as you pay he will not care that much :-)04:46
wolfspraulthose are essentially paid-hands that are very good with a few tools, like pads/orcad/altium04:47
wolfspraulwhen you take out those tools, there is not much value from hiring them04:47
wolfspraulto do the layout for a board like m1 is not that expensive, for example if you give them altium files with reasonable quality schematics, and zero layout, maybe it will cost 2000 USD or so to create the entire layout04:48
wolfspraulif you just want to tweak/make changes to an existing layout, a few hundred USD should be enough04:48
wolfspraulso we do things in parallel a little04:49
wolfspraulon one hand we continue to use the Altium files, just next week or so again for R4 - hire PLAB to update the layout04:49
wolfspraul(and very importantly: get their feedback/advice)04:49
wolfspraulon the other hand start entering schematics and bom into kicad, connect to boom sourcing tool04:50
wolfspraulthen we see what works, and improve later04:50
wolfspraulArtyom: does this answer your queestion?04:50
ArtyomDo PLAB guaranty somehow the quality of their design? That it will work fine for high-speed signals?04:50
wolfspraulhe04:50
wolfspraulno04:50
wolfspraulyou pay them for their work04:51
wolfsprauldoes a gardener you hire for your garden guarantee you that your flowers won't die?04:51
wolfspraulno04:51
wolfspraulyou can change the gardener though04:51
wolfspraulbut of course, that is precisely the reason why you want to work with someone like PLAB04:51
wolfspraulbecause they do tons of layouts, all the time04:51
wolfspraulthey have seen it all :-)04:51
wolfsprauland they know what people are doing in new high-speed designs04:52
wolfspraulin fact they do it04:52
ArtyomYeah, this helped to understand the process :) Thank you for your explanation :)04:52
wolfspraulover 50% of new computer designs are originating at Taiwanese OEM/ODM companies04:52
wolfspraulso there is a really serious cluster of knowledge about those things in Taipei04:53
wolfspraullike *serious*04:53
ArtyomLast questions: As I understand there are people that live in Europe, China and other places that cooperate with all the companies that you have mentioned?04:53
wolfspraulsort of, yes04:53
wolfspraulit's all very simple though, think of a few friends04:53
wolfspraulI think we all operate from our homes :-)04:54
wolfspraulyes04:54
wolfsprauljust went through mentally :-)04:54
wolfspraul100% from-home 'corporation' :-)04:54
wolfspraulAdam is in Taipei04:54
wolfspraulah hold on04:55
wolfspraulraumfahrtagentur has a big studio! :-)04:55
wolfspraulthe mechanical design guys in Berlin04:55
ArtyomAnd one more question: What is the price for manufacturing MM1 PCB in small quantities (I mean production of new pcb not existing one)? As I understand it's a 6-layer pcb with minimum line width about 0.125mm?04:58
wolfspraulthe prices are all in the wiki05:01
wolfspraulone sec I find it05:01
wolfspraulactually I want to make the entire accounting public, except for money paid to individuals :-)05:02
wolfspraulthinking about setting up an openerp/tryton server05:02
wolfspraulhere is the data for the 3rd run http://en.qi-hardware.com/wiki/Milkymist_One_RC3_BOM05:05
wolfspraulwe made 95 PCBs at a cost of 21.02 USD per pcb05:05
wolfspraulso the total was 95*21.02 = 1996.90 USD05:05
wolfspraulthere are many parameters that influence PCB pricing05:06
wolfsprauldozens05:06
wolfspraulI remember the older numbers vaguely, for your reference05:06
wolfspraulin the first run of 6pc, we paid 900 USD for the PCBs05:06
wolfspraulso that was 150 USD / pc05:07
wolfspraulmaybe we made 10pc PCBs, but only 6 soldered05:07
wolfspraulin the second run of 40 pc, we paid about 40 USD / pcb05:07
wolfspraulthe total was about 1600 USD05:07
wolfsprauland in the third run the total was 2000 USD (for 95 pc)05:07
wolfspraulso you get the idea05:07
wolfspraul900 USD / 10pc, 1600 USD / 45pc, 2000 USD / 95pc05:08
wolfspraulthe price per pcb would continue to come down a lot if we order higher volumes05:08
wolfspraulyou see the curve and you can imagine, if you order 200pc, 500pc, 1000pc, price per pcb will drop easily below 10 usd, and further05:08
ArtyomAs I understand for small quantities of pcb the main part of the price is "production preparations".05:09
wolfspraulor even just company overhead05:10
wolfspraulthink of the lawyer again05:10
wolfspraulthe pcb shop may have 100 people to feed05:10
wolfspraulthat means accounting, management, etc.05:10
wolfspraulat some point they cannot accept an order/customer who will give them 10 USD revenue05:11
wolfspraulplease keep in mind that there are real people everywhere05:11
wolfspraulat newheart, I think whatever you make you will have a minimum of 500 USD or so05:11
wolfspraulif you want to make each pcp cheaper, just order more :-)05:12
wolfspraulprice will come down A LOT05:12
wolfspraulbut there is overhead for each customer/order, of course05:12
ArtyomI asked these questions because I want to compare them with prices that are offered on russian-market. As I understand for difficult designs (with traces width less then 0.15) some pcb-production companies make orders in china.05:13
wolfspraulI will share all my numbers with you05:13
ArtyomI mainly interested in experimental-pcb-designs. At least for now...05:13
wolfspraulthat's the whole point of our open manufacturing project :-)05:13
ArtyomSo only extra-small quatities...05:13
wolfspraulif you really want to make something, Adam can help you05:13
wolfspraulbut please only if you really really want to manufacture something05:13
wolfspraulbecause I am paying Adam's time ;-)05:14
wolfspraulit all comes down to working hours, really05:14
wolfspraulfor terminology, China != Taiwan05:14
Artyomok :) When I will have ready gerbers I will asl your advice ;)05:14
wolfspraulin general you can say this:05:14
wolfspraulvery welcome05:14
wolfspraulis your work/files open?05:14
Artyom*ask05:14
wolfspraul(just curious)05:15
wolfspraulso yes, I want to help, definitely05:15
wolfspraullet me give you some more guidance05:15
wolfspraulI don't know South Korea and Japan well, only Taiwan & China05:15
wolfspraulfirst thing to keep in mind05:15
wolfspraulSouth Korea and Japan are options that are surely intersting, but I just know very little05:15
ArtyomMy personal project are opened. There are also designs that I have to make for the place where I work. And in this case "it depends..."05:15
wolfspraulnow, Taiwan in general is about 2* as expensive as China05:15
wolfspraul'in general'05:16
wolfspraulTaiwan has a huge Japanese influence05:16
wolfspraulit was a Japanese colony for 50+ years05:16
wolfspraulTaiwan is tiny (25 million people, massive tech cluster in Taipei)05:16
wolfspraulChina is huge, 1.3 billion people, dozens of huge cities05:16
wolfspraulso...05:16
wolfspraulTaiwan in general is an excellent place for small runs05:16
wolfspraulthey have very educated people05:16
wolfspraullots of experience05:17
wolfspraullots of brain05:17
wolfspraulvery reasonable prices05:17
wolfspraulJapanese influence (quality), Chinese influence (efficiency)05:17
wolfspraulI think I can say that without upsetting Adam :-)05:17
wolfspraulI can highly recommend Taipei as a place for innovative small runs, pcb or smt or layout or mechanical work05:17
wolfspraulChina (mainland) is more difficult05:17
wolfspraulit's cheaper, but it's wild05:18
wolfspraulchaotic05:18
wolfspraulabsolutely no quality standards05:18
wolfspraullike *zero*05:18
wolfspraulyou cannot imagine05:18
wolfspraulI often meet Chinese designers who admire my pcbs and say they are searching in mainland china for years and cannot find 1 quality pcb maker...05:18
wolfspraulin the long run from all the chaos and competition in mainland, some quality companies may emerge05:19
wolfspraul'may'05:19
wolfspraulbut I have full access to both sides, China and Taiwan05:19
wolfsprauland I am happily picking Taiwan for pretty much everything I can05:19
ArtyomI read on one russian forum about china-pcb-makers. The main idea: many companies are very different... Some can produce rather good things.05:20
wolfsprauldoes this make sense?05:20
wolfspraulyes, sure05:20
wolfspraullike I said05:20
wolfspraulwild05:20
wolfspraulno standards05:20
wolfspraulvery very chaotic05:20
wolfspraulmost companies in China that are consistently making high quality products are under foreign (including Taiwanese) management05:21
ArtyomBut actually it's difficult for me to judge. I've never been in either place.05:21
wolfspraulI think I can say 90% :-)05:21
wolfspraulmy advice to you is simple: follow and use the same small Taipei shops I am using05:21
wolfsprauland Adam can help you, as long as the time investment is controlled05:22
wolfspraulChina is cheaper, theoretically. about half price.05:22
wolfspraulbut you are in the jungle05:22
wolfspraulmost quality shops in China are under foreign management, and that will drive the price up again05:22
Artyomok, when the time will come I will take into account all possibilities. And thanks again for your explanations. I don't know a lot of things :)05:22
wolfspraulsure no problem, you are very welcome05:23
wolfspraulespecially if your projects are open, I'd love to support them with the limited means that I have05:23
wolfspraulone last thing about China05:23
wolfspraulso I said 90% of the quality companies in China are foreign-managed05:23
wolfsprauland for the remaining 10%, most are huge05:23
wolfspraultotally not setup for small runs, prototypes, innovation05:23
wolfspraulyou cannot work with huawei, for example05:24
wolfspraulunless you want say 1 million PCBs, or better yet 10 million PCBs05:24
wolfspraulat that level they may do it for you in their factories, at their (high) quality standards05:24
wolfspraulbut what we are looking for is an intersection between cheap, low-volume, high-brain05:24
wolfspraulthat's tough, and generally pick Taipei and follow the shops I'm using05:25
wolfspraulsince I'm doing this for years and saw A LOT of shops :-)05:25
ArtyomThanks for the time you've spent for me. It's time to leave for me. bye.05:28
wolfspraulcu05:28
Fallenoumorning !08:54
Fallenou(new release) excellent news !08:54
cladamwphew~just fixed 46 warnings in R4. ;-)09:38
cladamwrc3 we left 18 warnings, R4 is 19. 16/19 is the same as rc3 which is okay. Working on rests.09:46
cladamw(removed L19) this makes compiler telling warning for having multiple names(GND & VIDEOIN_AGND), it's okay.09:49
cladamwone shows for Bidirectional Port and Unspecified Port objects, it's also acceptable to me. Since rc3 have one too. ;-) Well but at best to fix it as well. :)09:52
cladamwthe last is Floating object.09:54
cladamwcompiler I dislike you but also like you more. :)09:55
wpwrakcladamw: removing warnings is always good ;-)12:34
wpwrakcladamw: when programming, there are sometimes people who don't that. they simply don't ask the compiler for warnings and ignore those it generates anyway. when i run into such a program, the first thing i always do is turn on the usual warnings. then run a compilation and see what happens.12:36
wpwrakcladamw: i almost always find a bug or two this way ;-)12:36
wolfspraulI cannot wait to move all this into KiCad and get the quality up12:36
wpwrakyeah, in kciad, we can also add our own checking12:37
wolfspraulI finally have my m1 setup to be running continuously besides me, which is so inspiring12:39
wolfspraulnothing will stop me now from a) improving it little by little b) emailing poor targets from my long marketing leads list about how great it is :-)12:40
wolfspraulhey, you "500 usd screensaver" commenters - I'm ready!12:40
wpwrakhehe ;-)12:40
wpwrakyes, it's important to have one available all the time12:40
wpwraki should actually build some sort of rack so that i can have both units up and running12:41
wolfspraulthere's two things out of this already12:42
wolfspraulfirst, I will look around for the cheapest & largest flat TV I can find that supports VGA in12:42
wolfspraulthat way I can hang the TV somewhere and turn it into a M1 art box12:42
cladamwwpwrak, oah..yup. i need to check more and update my histories in wiki.12:42
wpwrakthat projector you gave to xiangu sounds kinda nice12:42
wolfsprauland second, I will buy the cheapest openwrt supported router I can find to turn it into a ethernet<->wifi gateway for m112:42
wpwrakheh :)12:43
wolfspraulthen I have a complete solution to turn any TV screen into something really different12:43
wpwraktry to find one that operates on 5 V. that way, you could share power12:43
wolfspraulone by one12:43
wolfspraullike I said, I feel inspired having this run besides me now12:43
wolfspraulfinally12:43
wolfspraulwithout fan, I cannot handle the noise of a fan12:43
wpwrak;-)12:43
wolfspraulnow TV, openwrt wifi gateway...12:44
wpwrakbtw, did you show midi to jon ?12:44
wolfspraulnot yet12:44
wolfspraulJon and I are both working 24/7, we meet all the time12:44
wpwrakphew. the looong road.12:44
wolfspraulno no12:44
wolfspraulI need to get it to work here myself first12:45
wolfspraul'almost' there :-)12:45
wpwraknapoleon made it to moscow and back in just a few months. i think i'm on that "kill the screen saver meme" mission already longer than that ...12:45
wolfsprauloh that comment is great12:45
wolfspraulit shows that the commenter is 100% stuck in his PC thinking12:46
wolfsprauleither he wants to get out there or not12:46
wolfspraulif not, keep commeting about 500 USD screensavers12:46
wolfspraulif yes, follow me on explaining the Milkymist journey12:46
wpwraknext phase: you need to make your own midi-using patch (or adapt an existing one)12:46
wpwrak(screen saver) i was more thinking of you and jon pushing M1 as an unattended video generator. as opposed to an instrument.12:47
wpwrakimaging going to an instrument shop looking for a violin. they try to sell you a model that has as its main feature (as far as they advertize it) a fully self-playing mode. what does that say about the instrument ? what does that say about those who buy it ? :)12:49
wpwrakmaybe it's actually a great violin. but you'll probably never find out ;-)12:50
wolfspraultoo hard to follow on a friday night12:52
wolfspraulI will re-read this thought tomorrow morning :-)12:52
wpwrak;-)12:52
wolfspraulI was just thinking about experimenting with a large prism in front of my laser projector12:52
wolfspraulcladamw: have a nice weekend!12:53
wolfspraulthis was another great milkymist week with lots of good things happening12:53
cladamwhave a nice weekend you guys too !12:54
wolfspraulI don't think any sales came in, but that's on my side and I need to just work more diligently on that too.12:54
wolfspraulwe shipped out the sunday order though12:54
wolfsprauloh, and another one from taipei in-progress12:54
wolfspraulwe get there :-)12:54
wpwrakbtw, a while ago, it seemed as if yi was going to work on the FN handbook. is this (still) on ?12:54
wolfspraulin theory yes12:54
wolfspraulin practice don't know, you know how hard it is to prioritize and focus12:55
wolfspraulor should I re-read some old schedule plans of yours? :-)12:55
wpwraknaw, i'm just curious who's taking care of documentation12:55
wolfspraulshe just mentioned to me today how frustrated she was that that idea still wasn't really started...12:56
wpwrakthe handbook is a good start but needs more work12:56
wolfspraulyes12:56
wolfspraulif I tell Yi that you remembered her plan she will be happy and motivated, I will surely do so12:56
wpwrakmaybe also a different format. there's relatively little on a page, making it a bit inconvenient to use online12:56
wpwrakalso, maybe the latex could be generated from some simpler markup. one that we could also use on the M1. that way, we'd get build-in help.12:57
wpwrak(motivated) very good ! :)12:58
wpwrakbtw, how can i find out whether a changes soc still meets timing ? will the build fail if it does ? will it issue a warning but let me proceed with a broken design ? or will it leave me in the dark completely ?13:43
wpwraks/changes/changed/13:44
cxadamsquick question, is there a convenient way to get a screenshot from an M1? wolfspraul?13:46
Fallenouwpwrak: at the end of synthesis you have a big table with timing constraints13:46
Fallenouwhen there is a (*) it means it's not met13:46
Fallenouand after the table it should write "All timings were met"13:46
Fallenousomething like this13:46
Fallenouit's definitely written13:46
cxadamsoh, I found it13:47
lekernelcxadams: http://milkymist.org/wiki/index.php?title=Flickernoise_user_manual#Capture_screenshot13:47
cxadamsthanks lekernel13:47
lekernelwpwrak: it will print a warning message13:47
wpwrakwould the "All timings [...]" be in system.srp ?13:48
wpwrakno "met" or "MET" in the file13:49
wpwrakwhat would the warning message look like ?13:49
Fallenoulet me check13:49
wpwrakah, i may have found it. system-routed.par has "All constraints were met."  is this the one ?13:52
Fallenouyes13:52
FallenouI just found it ;)13:52
Fallenouand you have a big big table before that13:53
Fallenouwith "PATH"13:53
Fallenouif you get a *, it's bad13:53
wpwrakkewl. thanks !13:54
wpwraknow let's see if it can get at the pc of that navre ...13:55
Fallenouyou're tring to setup a continuous integration server ?13:55
wpwraknaw, i'm just trying to find out why softusb goes completely mad13:57
wpwrake.g., i remove one debug message -> it doesn't even print the banner (NB: that debug message is never even triggered in that scenario)13:57
wpwrakor i replace one debug message with another one -> navre hangs the M1 (alas, i didn't jump on that one in time and have to figure out again how to reproduce it)13:59
wpwrakwe never read from MM_SOFTUSB_CONTROL, do we ?14:05
lekernelwpwrak: not afaik, why?14:32
lekernelit can be read though14:33
wpwraki'm abusing it now to read the PC :)14:33
wpwrakkinda convenient: mdump prints more-or-less successive PC values14:33
lekernelhmm, navre shouldn't crash the M1, unless it keeps asserting IRQ14:41
wpwraki see three ways how it could do this: 1) IRQ, 2) keep on producing data faster than the LM32 can read it, 3) set the data index to an invalid value (>= period)14:44
wpwrakah, and 4) send data that triggers some other bug in FN ;-)14:44
wpwrakbtw, i'm setting up some environment variables for logging into the M1. so far, I've used M1_HOST, and M1_USER, and M1_PW. do they sound good for a "semi-official" feature ? (i.e., insecure as hell, but we can expect the M1 to be properly firewalled in this case)14:48
wpwrakor should they be MILKYMIST_... ? (a bit of a mouthful in all-caps) or something else ?14:50
lekernelMM_ ?14:53
wpwrakhehe :) okay, MM_ then14:53
wpwrakand let's hope disney don't sue us :)14:53
lekernelwe already have MMDIR and M2DIR for -ng14:54
wpwrakah yes, then M1 might be confusing14:55
wpwrakpart 2: should here be default values ? mine are MM_HOST=m1, MM_USER=foo, MM_PW=bar14:56
wpwrakso .. shall there be defaults ? and if yes, what shall they be ?15:09
lekernelI don't think we need defaults here... everyone's has their network set up differently, and people would need to set login/pwd in the M1 GUI anyway15:11
wpwrakit's more for people who don't even want to bother with setting up the env variables. lazy bastards like me :)15:14
lekernel_I don't think we need defaults here... everyone's has their network set up differently, and people would need to set login/pwd in the M1 GUI anyway15:18
wpwrakit's more for people who don't even want to bother with setting up the env variables. lazy bastards like me :)15:18
GitHub56[milkymist] wpwrak pushed 4 new commits to master: https://github.com/milkymist/milkymist/compare/f0936ed...f1a088617:32
GitHub56[milkymist/master] softusb-input/Makefile: minor cleanup (whitespace and .PHONY) - Werner Almesberger17:32
GitHub56[milkymist/master] tools/asm/pfpu: renamed M1_{HOST,USER,PW} to MM_* and removed defaults - Werner Almesberger17:32
GitHub56[milkymist/master] softusb-input/Makefile: added target "upload" to upload softusb-input.bin to M1 - Werner Almesberger17:32
wpwrakhmm, should this be as scary as it looks ?18:18
wpwrakWARNING:Timing:3225 - Timing constraint OFFSET = OUT 5 ns AFTER COMP "vga_clk" ignored during timing analysis18:18
Action: Fallenou didn't care about it but not sure if it should scare you18:19
wpwrak(just stumbled upon it, while watching paint dry, err, synthesis complete)18:19
wpwrakit makes me think of the weird DAC issues we run into from time to time18:30
wpwrakand the warning looks as if things were done in the wrong order18:32
wpwrakwell, after going to migen, perhaps we can try to reduce that flood of warnings a little. having a generator (migen) between source and the xilinx tools will also help in cases where ugly work-arounds are needed to get rid of warnings18:34
GitHub67[milkymist] wpwrak created usb-debug (+2 new commits): https://github.com/milkymist/milkymist/compare/ff72330^...5cdd93f22:20
GitHub67[milkymist/usb-debug] softusb: added access to Navre's PC via CSR - Werner Almesberger22:20
GitHub67[milkymist/usb-debug] softusb: added the usual set of -W... flags and corrected code accordingly - Werner Almesberger22:20
wpwraklekernel: does this look suitable for master ? https://github.com/milkymist/milkymist/commit/ff723302cd0725486fbd0624ed860f08d8856d9322:22
wpwraklekernel: navre.ld says about data LENGTH = 32K. shouldn't that be 8K ?23:40
wpwrakphew23:55
--- Sat Mar 3 201200:00

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