#milkymist IRC log for Friday, 2012-03-23

wpwrakhmm .. does anyone happen to run a 2010 ubuntu ? if yes, i would be very interested in a conversion of  http://downloads.qi-hardware.com/hardware/milkymist_one/sch/20120305/Milkymist%20One%20R4%20-%2020120305.pdf00:12
wpwrakto postscript00:12
wpwrak(trying to figure out why i can't print m1 schematics anymore)00:13
wolfspraulwhy 2010 ubuntu?00:13
wolfspraulwhat command/util do you want to use?00:13
wpwraki'm running 2011.10 now and nothing i try works. i used to be able to print the schematics just find before i had to reinstall/upgrade.00:14
wpwrakjust xpdf, print to file00:14
wolfspraulok wait I try00:15
wpwraki tried: a) various versions of the m1 schematics, including rc3. b) various printer configurations in cups. c) remove cups, cat to /dev/usblp0 directly. d) convert on an old gentoo system (that didn't work either, but i don't know if that would have been the same in the past). e) power-cycles printer. f) factory-reset printer.00:16
wpwrakso now the only reasonable problems spots remaining would be the conversion or the USB communication itself00:17
wolfspraulhttp://downloads.qi-hardware.com/people/wolfgang/tmp/Milkymist%20One%20R4%20-%2020120305.ps00:18
wpwrakwell, or the printer itself regressed somehow, just coincidentally at the time i upgraded the pc.00:18
wpwrakthanks !00:19
wolfspraulwell, not sure whether that helps you00:19
wolfspraulI don't understand the problem you have00:19
wolfspraulI opened with xpdf and printed to file, as you said :-)00:19
wpwrakhah !00:19
wolfspraulthen I checked with ghostscript and it looks like schematics00:19
wolfspraulthen I uploaded00:19
wolfspraulnow you tell me :-) this helps you?00:19
wpwrakthat file works !00:20
wolfspraulgreat00:20
wolfspraulthat's a good way to start a day00:20
wpwrakso it's the PDF to PD conversion that's messed up00:20
wpwraks/PD\>/PS/00:20
wolfsprauladvance warning: Adam pinged me in PM yesterday about some layout options and prices, but I think we should do this all out in the open00:20
wolfsprauldemystify layout00:20
wpwrakwhat's the general picture ?00:21
wolfspraulif people in this channel don't object, we will have the chat here, it will be rather lengthy though, hope it's ok00:21
wpwrakaha. complex then :)00:21
wolfspraulotherwise if people feel it's off-topic, then into the more general #qi-hardware...00:21
wolfspraulno no00:21
wolfspraulit's just about prices, working hours, risks, etc. etc.00:21
wolfspraultypical open hardware stuff00:21
wolfspraulor 'hardware' in general00:21
wolfspraulI'll wait for Adam to show up soonish, otherwise he has to re-read so much00:22
wolfspraullayout is a weak spot in open hardware, well. one of many :-)00:22
wolfspraultogether with sourcing, mechanical, production testing, analog circuits, ... :-)00:23
wpwraklife is difficult :)00:23
wolfspraulI hope we can soon open up that black box as well, but first schematics & bom...00:24
wolfspraulthat's enough trouble00:24
wolfspraulhave we started the github repo for r4 kicad files?00:24
wolfspraulyes00:25
wolfspraul:-)00:25
wolfspraulboard-m100:25
wolfspraulthat must be it00:25
wpwrakyup00:26
wolfspraulok I will setup schhist to that url then00:28
wolfspraulshould work I think00:28
wolfspraul(I'll do it only after there are some .sch files)00:28
wpwrakunder m1r4/ ?00:33
wolfspraulyou want the revision number in a folder name?00:33
wpwrakpoor man's branching :)00:34
wolfspraulwe already have m 'one'00:34
wpwrakthat way, one can access older revisions without having to dig into history00:34
wpwrakyes, the m1 is redundant. just r4 ?00:34
wolfspraulok00:34
wolfspraulup to you00:34
wpwrakr4 it is then00:35
wpwrakkicad project m1.* ?00:37
wolfspraulsorry don't understand00:39
wolfspraulyou mean filename?00:39
wpwrakyes. m1.prj, etc. ?00:40
wpwrakerr, m1.pro00:40
wolfspraulsure00:41
wolfspraulcladamw: good morning! :-)00:41
cladamwgood morning00:42
wolfspraulcan you resend what you posted to me last night about the layout prices?00:42
wolfspraulwe should have the discussion in the channel so demystify layout a bit!00:43
wolfspraulit's open hardware :-)00:43
cladamwhow about I send to list ?00:43
GitHub172[board-m1] wpwrak pushed 1 new commit to master: https://github.com/milkymist/board-m1/commit/b4d4a477395427f9d9367bb9f8af09a0328ae49900:43
GitHub172[board-m1/master] r4/: added basic m1 schematics files - Werner Almesberger00:43
wolfspraulI think here is better00:43
wolfspraulbecause:00:43
wpwrakkewl. with notification :)00:43
wolfspraul1) there will be back and forth and I want werner to be closely in sync, get his advice/opinion00:43
wolfspraul2) the channel is logged, searchable, etc. can't be much better.00:44
wolfspraulalso I have not yet seen interesting/new feedback on the milkymist list, we could try the qi list but people are a little asleep there right now as well :-)00:44
wpwrakbtw, i think we should create our own symbol library. the default one suffers from a lack of uniformity. and there aren't actually a lot of things there we'll want to reuse anyway.00:44
wolfspraulbottom line: let's just power forward here in the chat00:44
wolfspraulcladamw: can you repost what you sent me last night?00:45
wolfspraulthat's a good starting point00:45
cladamw(6-layers if new routes for M1r4) NTD 96000 = ntd 32 * 3000(pcs of via)00:45
cladamw(8-layers if new routes for M1r4) NTD 105000 = ntd 35 * 3000(pcs of via)00:46
wpwrakso that's .... USD 320000:46
wolfspraul1 USD = ca. 30 NTD, yes00:46
wolfspraulhe has some more data first00:46
wpwrakso, same price00:46
wolfspraulwait00:46
wolfspraulthat's for a new from-scratch design, how they calculate00:47
wpwrakwhat does "pcs of via" mean ?00:47
wolfspraulthey count the vias :-)00:47
wolfspraulestimate rather00:47
wolfspraulremember this is a quote00:47
wpwrakinteresting metric :)00:47
wolfspraulwait a little, adam has some more numbers first00:47
cladamwhere the 'via' are different prices based on through via (6-layers), blind/buried via00:48
wolfspraulI think the final invoice will typically be based on working hours, but if complications occur etc. of course in the end the customer has to pay 'sufficiently' anyway. similar to let's say you hire a plumber...00:48
wolfspraulcladamw: also post the other numbers about working hour fee, cost for modifying existing layout00:49
cladamw3000pcs is roughly guessing not the exactly final quantities of vias00:49
cladamw(modification price on M1rc3) 15 hours * ntd 500/ per hour = ntd 750000:50
wolfspraul(that was the work from rc2 to rc3, I think)00:50
wpwrakso for them 6/8 layer is basically the same. i fact, they may need fewer vias with 8 than with 6 -> it's probably exactly the same00:50
wpwrak(in their estimate)00:50
wolfspraulmaybe that's why they quote the 8-layer vias a bit more expensive? :-) (just kidding)00:50
cladamw(modification price now on M1r4) ?? hours * ntd 800 /per hour = ???00:50
wpwrakwolfspraul: that's what i meant :)00:51
wpwrakhehe :)00:51
wolfspraulhourly rate increased from 500 ntd/hr (ca. 16 USD) to 800 NTD/hr (ca. 27 USD / hr)00:51
wolfspraulso...00:51
wpwrakinflation :)00:51
wolfspraulthose are all quotes00:51
cladamwarising from ntd 500 to 800/per hour is their rate now. not about types of vias.00:52
wpwrakhow difficult was  M1rc2 -> rc3  ?00:52
wolfspraulbut we have found this layout house to be fair, delivering good quality, charging for reasonable amounts of work, 'going the extra mile' when there are some hickups, etc.00:52
wolfspraulin addition to those numbers, we also need to think about risk00:52
wpwraksure. the hourly rate is very reasonable. even if they add margin hours :)00:52
wolfsprauland long-term reusability (which is low)00:52
cladamwno big deals of difficult from M1rc2 -> rc3.00:52
wolfspraulcladamw: do you remember how much we paid in the end for the first layout they did?00:53
wpwrak(no big deal) that's what i thought. so the delta may end up being expensive this time.00:53
wolfspraulI thought it was around 2000 USD00:53
cladamwwolfspraul, price for rc1 ?00:53
wolfspraulyes00:53
wolfspraulwhat I find unfortunate is that the layout is still this blob, black box00:54
cladamwmoment00:54
wolfspraulbut we have to do one by one, and first focus on schematics & bom now, get that into kicad00:54
wolfsprauland up the quality, documentation, reusability, etc.00:54
wpwrakyes, schematics first. can't do layout without them :)00:54
wolfspraulmy feeling is that there are actually not *that many* secrets in layout, it's just a matter of bringing them out one by one, documenting/annotating00:55
wpwrakbom isn't all that critical for the layout, since you'll usually have an idea of what package sizes you'll want.00:55
wolfspraulalso of course kicad's weakness in auto-routing00:55
wpwrakto put it politely :)00:55
wolfsprauland kicad's weakness in footprints00:55
wolfspraulyes but the alternative is to throw thousands of USD at this black box without any improvement for the future00:55
wpwrakwith fpad, footprints are fun to make :)00:55
wpwrakfpEd even00:56
wolfspraulsure, but one by one00:56
wolfspraullet's do a superb job on schematics & bom, it will take us weeks00:56
wpwrakoh, sure00:56
cladamw(layout fee of M1rc1) NTD5200000:56
wolfspraulmy feeling is with those numbers from adam, we should first ask the layout folks to spend a few hours trying to get our R4 changes into the existing 6-layer layout00:56
wolfspraultry that first00:57
wolfsprauland abort if things get out of control, let's say after X hours... how much? 20 hours or so? even more. I would try a bit.00:57
wpwrakcladamw: so M1r1 was 1/5 of the price for M1r4 ?00:57
wolfspraulone is quoted, one is actual00:57
wpwrakcladamw: that's because they already "know" the board ?00:57
wolfspraulno no00:57
wpwrakwhat was the quote for M1r1 ?00:57
cladamw52000/105000 roughly 1/2 ratio00:58
cladamw(M1rc1 = NTD52000)00:58
wpwrakoh, i missed a zero :)00:58
wolfspraulso I think we first ask them to modify the existing 6-layer layout00:59
wolfsprauleven though maybe if this were r1, we would try 8-layer right away00:59
wolfspraulcladamw: btw, let's also compare PCB costs00:59
wpwrakso the delta m1rc2->rc3 was only USD 25000:59
wolfspraulwe had 21 USD for 95pc of rc3 6-layer00:59
wolfspraulyes00:59
cladamwtoday I need to ask two vendor (Newheart & PLAB) for 8-layers01:00
wpwrakdoing things from scratch would be USD ~300001:00
wolfspraulhow much would 95pc of rc3 have been it if would have been an 8-layer board?01:00
wolfspraulwpwrak: that's a quote! :-)01:00
wpwrakcladamw: that's PCB makers ? or also layout people ?01:00
wolfspraulthey just estimate the number of vias to be 3000 and then multiply with 35 USD as a standard value for 8-layer01:00
wolfspraulsorry 35 NTD01:00
cladamw(Newheart) is only PCB maker01:00
wpwrakwolfspraul: (quote) sure. but are you saying quotes are meaningless ?01:00
wolfspraulpcb maker: newheart; layout: plab01:00
wolfspraulno01:00
wolfspraulbut the quote itself is done very fast, just via estimate01:01
cladamw(PLAB) it's a vertical ODM/OEM vendor who can make layout, PCB, SMT assembly works.01:01
wolfspraulwe have found their quotes to be conservative, but a good estimate01:01
wpwrakso it's fair to compare actual prices with quotes. qed :)01:01
wolfspraulyes, definitely01:01
cladamw(quotes only) yeah not exactly final price.01:01
wpwrakit's not like there was a hidden 10x factor in there01:02
wolfspraulno, the quotes definitely have value01:02
wolfspraulthey also represent their experience01:02
wolfspraulit's like Joerg told me years ago smt places do rough estimates of board soldering work/difficulty simply by the number of pins :-)01:02
wolfspraulwhy not? it's a good way to start... :-)01:03
wpwrakokay, then the delta first approach sounds reasonable. i think it'll fly out of the window the moment they see DVI, though :)01:03
wolfsprauland I guess layout quotes go by estimated vias...01:03
wolfspraulwpwrak: are you sure about that?01:03
wolfspraulif we are *sure*, then we can save the wasted effort01:03
wpwrakthey could still do a "reuse modules" approach, though. not sure how well this works.01:03
wolfspraulI would think/hope they try to reuse as much as possible even if they do a 'new' 8-layer pcb01:04
wolfspraulcladamw: right?01:04
wpwrakwell, is there an empty layer under the FPGA, with nice ground around it ? :)01:04
wolfspraulI mean there are things in the files like footprints etc. they are all fine01:04
cladamwwolfspraul, "how much would 95pc of rc3 have been it if would have been an 8-layer board?" this question may not make sense, since different kinda vias work in pcb maker. so i would only ask pcb maker to know a new rough 8-layer with new design with same board size.01:04
wolfspraulsure01:04
wolfspraulthat's what I meant01:04
cladamwoh, okay01:04
wolfspraulcladamw: if plab does a 'new' layout, how much of the current layout can they reuse?01:05
wolfspraulwe are starting with at least something that we know is pretty good01:05
cladamwso far i don't know if there have 'reuse module' for them. you can imagine they charge fee not realted to how they work on routing.01:06
wpwrakfor dvi, we need to bring .. 4 (?) pairs of high-speed differential lines across the FPGA. the FPGA should be a busy area. ergo ...01:06
cladamwbut we can imagine how transfer from 6-layers into 8-layers with 'reuse module' ?01:06
wolfspraulcladamw: can you export high-res PNGs of all rc3 layers?01:06
wolfspraulcladamw: well I am not worried about ripoff. I think we have found them to be fair in charging. they will charge us as much work as they did, which is fine.01:07
cladamwwell...only when i sit beside them, otherwise i don't think they tell me in honest. :-)01:07
wolfspraulsure01:07
wolfspraulI am not worried about ripoff, we know these guys and I trust them.01:07
wolfspraulwell, I do. maybe you don't? :-)01:07
cladamwyou meant PNGs for all layer ?01:07
wolfspraulyes01:07
wolfspraulhigh-res, but not too high. but so that we can make this discussion more 'real'01:07
wpwrak(pngs) we actually have the gerbers. gerbv can handle them01:08
cladamwi remembered Werner's script tool can do this very well.01:08
wolfspraul2000x2000 or so should show all features, I think01:08
wolfspraulah01:08
wolfspraulI'm behind01:08
cladamwbut from gerbv yes also can do this.01:08
cladamwwolfspraul, what's your point ? you wanted to print each layer of rc3 ?01:09
wpwrakmy script only does the top and the bottom. the outside of the board01:09
cladamwwpwrak, yours can't do each layer ?01:09
wpwrakthe the inner layers, gerbv is more useful anyway, because you can switch among layers. so you can follow the projection of a part01:09
cladamwwpwrak, i like yours top/bot very much also inlcude transparent function. Nice !01:10
wolfspraulgerbv can export all layers into pngs?01:10
cladamwwpwrak, oah..sure01:10
wpwrakcladamw: it combines the various surface layers into a pseudo-realistic rendering. there's nothing like that inside the board. sure, i could make it do that too, but that's not very useful01:10
wolfspraulok I will check it01:10
cladamwwpwrak, we can just use gerbv though, no problem.01:10
wolfspraulcladamw: so, we need: estimate on 8-layer pcb cost vs. current 6-layer one01:11
wpwrakwolfspraul: http://downloads.qi-hardware.com/people/werner/m1/tmp/front.png01:11
wpwrakwolfspraul: http://downloads.qi-hardware.com/people/werner/m1/tmp/back.png01:11
wpwrak(for the outsides)01:11
wpwrakfor the insides, the gerbers + gerbv, wherever they are :)01:11
wolfspraulcladamw: do you agree with werner that trying to add our changes into the 6-layer layout is most likely not going to work anyway?01:11
wolfspraulI think you should sit down with plab for a little and then maybe make a decision on that01:12
cladamwi now prefer to work in 8-layers01:12
wolfspraulok01:12
wolfspraulbut let's get pcb cost estimate first01:12
cladamwyes01:12
wolfsprauland I hope the actual plab cost is less than 105,000 NTD01:12
wolfspraulhope dies last :-)01:12
wpwrakso .. layer 1 is top. all the pads. that one's full by definition. layer 2 is ground also full.layer 3 brings out the middle ring of FPGA signals. layer 4 is ground again.01:13
wolfspraulafter we have succeeded in bringing sch+bom into kicad, we should try the same with layout too :-)01:14
wpwrakcorrection. layers 4/5 are ground/power. not sure which.01:14
wolfspraulwhat I mind more than the cash to plab is that we learn so little01:14
wpwraklayer 6 is the bottom has more busy fpga signals.01:14
wolfspraulwe could probably break down the entire m1 layout to not more than 50 rules and notes to follow, and all is safe - it becomes an easy drawing job01:14
wolfspraulbut since we don't really know a single one of these < 50 rules and notes, we deal with a black box...01:15
wolfspraulthat's my feeling only01:15
cladamw(rc3 layers) 1-Top, 2-gnd1, 3-inner, 4-gnd2, 5-power, 6-bot01:16
wolfspraulcladamw: ok - so do you have a clear plan now?01:17
wpwrakwolfspraul: (50 rules) objection. there are people present who have done working layouts before ;-)01:17
wolfspraulyou mean you know some/all of them?01:17
wpwraknot all. but some, certainly01:17
wolfspraulI just tried to ballpark the unknown, because I do think in the end it comes down to roughly that much01:18
wolfspraullayout is:01:18
wolfspraul1) lack of autorouter: tedious distraction01:18
wolfspraul2) even with autorouter: lots of manual work. without autorouter: crazy amount of manual work01:18
wpwrakyes and no. the best would be a semi-automatic router. a full autorouter tends to be useless.01:19
wolfspraul3) a short but important list of knowledge/criteria you *have to* follow. 'short' meaning 10 < x < 10001:19
cladamwi actually have seen ivorin with tones of layout rules fro her ***r company. Which is more than 50 rules.01:19
wolfspraulthat's what I meant, there is always manual intervention01:19
cladamws/fro/from01:19
wolfspraulcladamw: yes01:19
wpwrakcladamw: more like 50 pages ? ;-)01:19
wolfspraulthose are the combined layout rules of the head layout manager of benq, managing 10 layout teams01:19
wpwrakphew :)01:20
wolfspraulbut for m1, I think/guess about 50 important rules, and we know what matters *on our board*01:20
cladamwa very thicker notbooks inside that company to make nb.01:20
wolfspraulI could be wrong and it's 10001:20
wolfspraulbut my point is: we need to dab away at those rules one by one *eventually*01:20
wolfspraul:-)01:20
wolfsprauland werner may already know 5 or 1001:20
wolfspraulor 2001:20
wolfspraulwho knows :-)01:20
wolfsprauland that process is not currently starting in any meaningful way.01:21
wolfspraulwhich is ok while we have to focus on sch+bom01:21
wolfspraulso we keep paying plab to lookup the notes :-)01:21
wolfspraulcladamw: maybe you can extract a few from them? :-)01:21
cladamwhehe... we could always to learn or maybe i get them from somewhere in the end. :-)01:21
wolfspraulyes but let's not try to get a thick compendium01:21
wpwrak(6 vs. 8 layer) i wouldn't dismiss the possibility that one could make things fit with 6 layers, by rearranging what's already there. don't know if that would be successful in the end, though.01:21
wolfspraulthat's more like stealing trade secrets. we can learn one by one for our board, and document that.01:22
wolfspraulcladamw: maybe if you go to plab, you use the chance to learn the 5 most important rules for our board :-)01:22
wolfspraullayout rules i mean01:22
cladamwwpwrak, that's why i like to prefer 8-layers direct but pains would be the sunk money if go.01:23
wpwrak(plab) we also pay them for their individual experience. if you don't know how to do layouts, you'll be very slow and make lots of silly mistakes. things you know aren't right but that you don't see the moment you do them. i usually print my stuff and then start kicking out the stupid routes.01:23
cladamwwolfspraul, oh...will try to get more details this time.01:23
cladamwwpwrak, the points/problems are now that:01:25
wolfspraulcladamw: don't worry about cash cost. we have to balance many risks and future opportunities.01:25
wolfspraulwe have to do the right thing now01:25
wolfspraulmy #1 concern is to reduce risk and make functioning r4 asap, and push towards high-performance and opening up knowledge01:26
cladamw1. we don't know the real common rules on different functionalities excepts from each datesheet says.01:26
wolfspraulyes01:26
wolfspraulat least we know that - good start! :-)01:26
wpwrakcladamw: will the people at plab you'll work with remember earlier m1 revisions ? or will that be new people who don't know the board ?01:26
cladamw2. like power trays, digital in different frequecy, do we have common rules for ours ?01:27
cladamw3. as 2) differetial pairs, common modes routes, etc....01:28
wpwrakcladamw: you mean power trace widths ?01:28
cladamwyes.01:28
wpwrakwell, we know some currents ... :)01:28
cladamwi've seen ivorins's very clear rules: like :01:28
cladamw1, how bigger ring dia of via = how big current it goes through01:29
cladamw2. how/what kinda of corner trace should be as like when in different frequecy. etc01:30
wpwrakwe know some basics. e.g., to match high-frequency traces with the ground return01:30
wolfspraulneed to run out to buy a few cables, back in 10 min01:30
wpwrakthe corners are trickier :)01:30
wolfspraulkeep talking, it's good. we open up the can of worms :-)01:30
cladamwman... wpwrak so i would one day i can get them and comtribute into qi wiki server.01:31
wolfspraullet's start one by one01:31
wolfspraulnot push out to 'the big day'01:31
wolfspraulthat will not come01:31
wolfsprauleven when we mention things here we start to pull them out of the unknown one by one01:31
cladamwthen we know how we to against that house /or contributor in our qi layout community work together01:31
wolfspraulcladamw: did you see werner's question about whether the people at plab will be the same as before and remember m1?01:31
wolfspraulcladamw: like I said, let's start to document one by one.01:32
wolfspraullet's not try to document all at once - that will fail01:32
cladamwotherwise we still hard to manage qi projects's layout.01:32
wolfspraulwe just start, learn 'a few' of the most important rules01:32
wolfsprauland then it will grow over time01:32
cladamwwolfspraul, sure sure...i just commented only ...hehe ;-)01:32
wpwrakcladamw: do you have "The Circuit Designer's Companion" by Tim Williams ?01:32
cladamwit's a book or url ?01:33
wolfspraulbbiab01:33
wpwrakcladamw: that book covers a lot of the basics. alas, it doesn't go into HF.01:33
wpwraka book01:33
cladamwoah..nice01:33
wpwrakthey should have it at that famous taipei bookstore :)01:33
wpwrakwish i remembered the name. i think their logo has a dragon. but of course, a lot of chinese things have dragons :)01:34
cladamwwpwrak, " did you see werner's question about whether the people at plab will be the same as before and remember m1?" >>> the guy is the same, no worries, i know 'HER'. ;-)01:35
cladamwwpwrak, http://www.amazon.com/Circuit-Designers-Companion-Second-Engineers/dp/075066370701:36
wpwrakthat's the one01:36
cladamwhttp://www.wisewarthog.com/electronics/williams-the-circuit-designers-companion.html01:37
wpwrakoh, and there appears to be a 3rd edition. checking ...01:37
cladamwi go for reply Sebastien's fpga comments. will back...need to update sch & email again. ;-)01:41
cladamwwpwrak, haven't you explained R229 1M Ohm to him already ?01:41
cladamwif not I'll reply in list. ;-)01:42
wpwrakhm. the 3rd ed. has a whole page for PLDs and FPGAs, with room to spare. how generous ;-)01:43
wpwrakbut it's better to get for the updated version. the table of contents is nearly identical.01:45
wpwrak(R229) i'm actually not sure what it does :-)01:46
cladamwhttp://en.qi-hardware.com/mmlogs/milkymist_2012-02-02.log.html#t10:3101:47
wpwrakthat was with one side of the wire on the gate and the other between your fingers ?01:49
wpwraki.e., prior to connecting the other end to something interesting ?01:49
cladamwwpwrak, yes, first, i used my finger for it, but then I connected R229 to gnd.01:52
cladamwsecond, we you suggested me it should be good for a pullup. then I changed it from 10K to 1M ohm for pullup.01:52
cladamwdo you think he thinks that R229 no need it since fpga inside have pullup for it(gate pin) ?01:54
cladamwwpwrak, two reasons: 1- a pullup (R229) can also make D1 ON when in manufacturing mode[default]. 2 - a protection to prevent the gate from floating. 3- if we actually don't know to need it, since there are clamping diodes in the FPGA anyway. then we can DNP it for safe just in case.02:00
cladamwwpwrak, 1) is actually for diagnosis to me in factory 2) is to make protection02:02
wpwrakthe fpga doesn't have clamps. or at least not full clamps.02:08
wpwrakokay, 1) may make sense02:09
cladamwno clamping diodes ?02:16
cladamwhmm... okay02:17
wpwraknope. at least not on the high side. maybe to ground, but i'm not sure02:17
wolfspraulcladamw: I think about the layout more02:27
wolfspraulif we really go for an 8-layer layout, I think you should request that you can stay there and take notes while they do the work02:27
wolfspraulit is definitely worth your time02:27
wolfspraulif we can use this opportunity to really extract more routing knowledge, then I would feel much better about the cost02:28
wolfsprauljust a thought02:28
wpwraki would hate to have someone peek over my shoulder when doing layouts ...02:28
wolfspraulso hopefully it's not just an initial meeting with some instructions, then black hole, then 1 week later get the results02:28
wolfspraulwpwrak: oh actually they mostly work in pairs and have frequent review/checks during the work, so I don't think that's a problem02:28
wolfspraulplus adam would just be there with a little paper notebook and takes notes02:28
wolfspraulit's an idea only02:29
wpwrak(work in pairs) phew. scary :)02:29
wolfspraulhey, don't we know this from coding somewhere? :-)02:29
wolfspraulI don't want to be the manager who has to force you into that methodology :-)02:29
wolfspraulwhat was it again? something with X...02:29
wolfspraulextreme programming, right02:29
wpwrakyou forgot the "foolishly attempt to" ;-)02:30
wpwrakyup, XP. weird concept. not all of XP, but that bit.02:30
wolfspraulyou should work for aviation software, say the engine control02:30
wolfspraulthat is unbelievable, I had a friend doing this02:30
wolfspraulyou basically do paperwork 99% of the time, literally02:30
wolfspraulninety nine percent02:31
wolfsprauland I think 95% of that you would not even agree with :-)02:31
wpwraksounds like massive fun :)02:31
wolfspraulcladamw: so I just wanted to encourage you to ask them / request that. especially if you know the person already, maybe they are open-minded. you take notes, it would be very good for us... they need to understand that we want to learn, it's not about micro-managing them.02:32
wolfspraulbut I'm relaxed, whatever is possible is possible.02:33
cladamwwolfspraul, i "will" ask/request/try rather than i "would/could" since money comes from you. ;-) Agreed your "possible is possible". But i can't gurantee on this.02:43
cladamwit's like a scenario that:02:43
wpwrakcladamw: (R229) it's not the "firmware" that activates the pull-up. it's the fpga's hardware. so you only need R229 for this if there is a problem with the fpga02:44
cladamw1. we won't to use them after this, if they scare me this time then they still agreed me to work in pairs.02:44
cladamw2. no else next time for us possibly future. I foolish them or they hate me and no more deals in the end. :-)02:46
wolfspraulno no02:46
wolfspraulcladamw: you know I don't think like this02:46
cladamw3. 1 or 2 depends on my judgement/feelings on what they feel me after I ask/request them. ;-)02:46
wolfspraulwe will *always* try to consult/hire experienced people, if that creates value for us and if we can afford it02:46
wolfspraulbut of course I understand your point, the concern they might have about their 'secrets'02:47
wolfspraulbut I actually don't think that they think like that :-) so let's see02:47
cladamwwolfspraul, yes, i know since we're doing the things that things should be openmind at least and get nice/great feedback we pay.02:47
wolfspraulthe more they tell us about their work, the more we will value them02:47
wolfspraulyes02:47
cladamwwpwrak, ha..okay  correct me on NOT firmware. :-)02:48
cladamwit's default in fpga's hw itself in initial stage.02:49
cladamwwolfspraul, I'll always to ask them. no problem. like i said before. this time, I'll note more as possible. but aslo i don't guratee now to say they will. let's try.02:52
cladamwjust don't know if get a work "in pairs" goal.02:53
wolfspraulcladamw: sounds perfect02:55
cladamw_wpwrak, what's your latest version of KiCad ? Build: (2010-08-11 BZR 2448)-unstable ?03:59
wpwrakbzr 3378 plus patches :)04:03
wolfspraulcladamw_: I suggest you keep using what you use until you run into an actual problem04:03
wpwrakhehe :)04:04
cladamw_why ? Shouldn't we sync same tool together ?04:05
wolfspraulcladamw_: yes and no. I am trying to minimize resource waste.04:12
wolfspraulI think the chances that you will run into a problem are small.04:12
wolfspraulthere weren't that many changes and our use of kicad especially with that is planned right now is relatively superficial.04:12
wolfspraulupdating prematurely is just not worth it04:12
wolfspraulstart using what you have, and when you run into anything that doesn't work I think we will realize that quickly and can then still update04:13
wolfspraulotherwise we could all update all the time - too much time waste04:13
cladamw_okay04:15
wolfspraul seems wpwrak does not disagree04:26
wpwrakwe'll see :) there's always the risk of small version differences04:31
wolfspraulyes but I need to give Adam confidence. we have every day only once, and I believe we will waste less time first continuing with his 2448 version.04:32
wolfspraulI am *sure* that that will waste less time than proactively updating.04:32
wolfspraulbecause not many things have changed, and our planned use over the next weeks is relatively superficial.04:32
wolfspraulotherwise we will all be in an update-race04:33
wolfspraulthere are commits almost every day...04:33
wpwrakbut we're not tracking their head. we're based on a pinned version.04:33
wpwrakor are you updating the patches every day ? :)04:34
wolfspraul3378 was just the latest at that day when I last updated the cmdline patches04:34
wpwrakyup. and we haven't moved from that since. so it's not like we're updating daily, even if kicad mainline does04:35
wolfspraulwell I answered adam's question :-) if you have another answer then he has to pick04:36
wolfspraulI think first continue with 2448, and update once running into a problem, which I am sure we will easily identify given the type of work we do04:36
wpwrakfurther down the road, there will be a patch that links eeschema to boom. i don't know whether compatibility issues or that patch will come first. we'll see, i guess :)04:39
wpwrakin gta02-core we had the problem that people editing with old versions of kicad lost information when loading the file and then saved that. so some things mysteriously changed (until we noticed)04:40
wpwrakthose were truly ancient versions, though04:42
wolfspraulyes, when those things come we all uplevel in sync04:50
wolfspraulbut that is my point - no need *now* (today) for Adam to uplevel04:50
wpwrakno, not today ;-)04:53
cladamw(8-layers quote) from Newheart, NTD 97,293 @ 95pcs. Plab hasn't replied yet.08:58
wpwrakcladamw: (newheart) and what would be their price for 6 layers ?12:04
wolfspraulhmm12:06
wolfspraulthat looks like the 8-layer pcb would be 30 USD compared to ca. 20 USD for 6-layer - not good12:06
cladamw(newheart for rc3) ntd 57,57812:06
wolfspraulthe (large) size of our pcb bites us12:06
wolfspraulcladamw: maybe we still should try to stay at 6-layer first :-)12:07
wolfspraulkeep it all simple and cheap12:07
wpwrakpricy indeed. didn't expect such a big jump.12:07
wolfspraulthe difference would be 13 USD12:07
wolfspraulwell it's large, I think that matters12:07
wolfspraul(our pcb)12:07
wpwrakdo you think it's already "too large" for 8 layer ? that would sound strange to me12:08
wolfspraulno just guessing12:10
wolfspraulfact is, an 8-layer m1 pcb would be 50% more expensive (13 USD) than a 6-layer pcb12:10
wolfspraulat volume of about 10012:10
wolfspraulit may also be that more layers has higher one-time costs, i.e. a bit steeper curve to come down with increasing volume12:11
wolfspraulin other words - the difference will be less than 50% at 500 pc, even less at 1k pc, etc.12:11
wolfspraulor different manufacturers may have different optimizations of their process12:12
wolfsprauladam is getting a 2nd quote12:12
cladamwnewheart is a kinda of prototype sample pcb maker. let's see another next week.12:13
wpwrakhmm, at 4pcb the increment is also +63%12:21
wpwrakmaybe it's also a yield issue12:21
wolfspraulwe should not recklessly increase the minimum specs for a milkymist board, that's why when I see those complications I think we should give 6-layer a thorough try first12:22
wolfspraulit all goes back to layout... :-)12:22
wpwrak10 layer = +50% from 8 layer. 12 layer = i see exponential growth :)12:23
wpwrakso let's find out how good they are ;-)12:23
GitHub110[migen] sbourdeauducq pushed 1 new commit to master: https://github.com/milkymist/migen/commit/081b658e2d660abf40daf89b616658b00ad096e515:38
GitHub110[migen/master] Update copyright notices - Sebastien Bourdeauducq15:38
lekernelhttp://mako.cc/talks/20110402-when_fs_isnt_better/wfsib-SLIDES.pdf19:19
wpwrakouch ! neo1973 as a bad example. that hurt.20:40
GitHub110[llvm-lm32] jpbonn pushed 2 new commits to master: http://git.io/eZoREg22:56
GitHub110[llvm-lm32/master] Added CTTZ_ZERO_UNDEF and CTLZ_ZERO_UNDEF lowering. - JP Bonn22:56
GitHub110[llvm-lm32/master] Merge branch 'master' of github.com:milkymist/llvm-lm32 - JP Bonn22:56
--- Sat Mar 24 201200:00

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