2012-04-02 00:28 Martix: good morning :-) you still there? 2012-04-02 00:28 wolfspraul: yep 2012-04-02 00:41 cladamw has joined #qi-hardware 2012-04-02 00:44 kristianpaul: you probably saw that rtlsdr has made some more rounds, on slashdot, reddit, etc 2012-04-02 00:44 Martix: oh, hi, didn't see :-) 2012-04-02 00:44 wolfspraul: yup 2012-04-02 00:45 and people have trouble finding a good source :-) 2012-04-02 00:45 ;-) 2012-04-02 00:45 since I bought from the factory, I am sure I can get the right thing, and they are definitely not 'out of stock' 2012-04-02 00:45 and won't be, no matter how many sales come in 2012-04-02 00:45 oh wow 2012-04-02 00:45 are you interesting in re-sell? ;) 2012-04-02 00:45 well of course. the factories produce as much and as long as there are orders. 2012-04-02 00:45 oh no, definitely not 2012-04-02 00:46 it's already there, just some layers of mostly language confusion 2012-04-02 00:46 some reseller will pop up 2012-04-02 00:46 sure, just confirming 2012-04-02 00:46 plus in no time, an even better option may appear 2012-04-02 00:46 hopefully people will really buy and start diving into sdr land, to learn about the details that really matter 2012-04-02 00:47 I focus on milkymist, I know we are on a good path there. no time for rtlsdr right now. 2012-04-02 00:47 i bet for it,afaik for that price lets hope is not a just a sold-out season 2012-04-02 00:47 no 2012-04-02 00:47 btw the reference design comes from Elonics, not Realtek 2012-04-02 00:47 for the whole package? 2012-04-02 00:48 all under NDA, industry-typical... Elonics is a UK company 2012-04-02 00:48 oh sure 2012-04-02 00:48 sure, the real work is typically just done in one place :-) 2012-04-02 00:48 so Elonics makes the reference design, one Chinese company buys it or gets it for free under NDA, then lots of other Chinese companies get access to the files in one way or another 2012-04-02 00:48 diving into sdr land, yup that should be the path 2012-04-02 00:49 the only people who don't have the reference design are law-abiding foss citizens :-) 2012-04-02 00:49 rejon has joined #qi-hardware 2012-04-02 00:49 :-| ... 2012-04-02 00:49 I don't have it either and will not try to hunt it down, or sign the NDA. just fyi. 2012-04-02 00:49 and neither embark in a reversing process, again too much time and milkymist/qi focus 2012-04-02 00:50 i'm aware of people claiming already signing it.. 2012-04-02 00:50 oh no, reverse no sense now 2012-04-02 00:50 there are not many secrets in the reference design anyway 2012-04-02 00:50 what matters is stack, i think osmocom have some sdr primitives already, 2012-04-02 00:50 just a lot of tweaking and tuning, which is the hard part and important if you want to make your own product 2012-04-02 00:51 thats when osmo-sdr came to life :) 2012-04-02 00:51 but seeing that the Elonics tuner is so closed, well, one day hopefully we can find more openly documented chips 2012-04-02 00:52 perhaps, indeed 2012-04-02 00:53 or at least more documented driver, wich is the value i see from osmo-sdr too 2012-04-02 00:54 s/wich/that 2012-04-02 00:54 kristianpaul meant: "or at least more documented driver, that is the value i see from osmo-sdr too" 2012-04-02 00:54 Martix: I saw the CuBox on your conference program, is it interesting? 2012-04-02 00:56 wolfspraul: it's alternative to Raspberry Pi and Beagle Bone, but speaker is still waiting when CuBox and Raspberry Pi arrive, but it should be week before conference 2012-04-02 01:01 what's good about it? 2012-04-02 01:03 kristianpaul: I thought this was a nice 3min *intro* video into the rtl2832/e4000 dongle http://www.youtube.com/watch?v=FUQd9HOVTk8 2012-04-02 01:03 nothing much in it, but it shows the first steps and throws in some terminology that may make people want to go further 2012-04-02 01:04 ah yes i saw it 2012-04-02 01:04 i hope more people get in to gnuradio too, 2012-04-02 01:05 hah 2012-04-02 01:05 to be honest i never get it a try nothing, as i dont owned a usrp or receiver hw 2012-04-02 01:05 DocScrutinizer: what? :) 2012-04-02 01:05 just hah 2012-04-02 01:05 xD 2012-04-02 01:05 probably I was happy to finally know why I'm not asleep 3 hours before getting up 2012-04-02 01:06 :-D 2012-04-02 01:06 re 2012-04-02 01:07 also I thought maybe the chip isn't all that closed, it just lacks a 1200pp book about signal processing 2012-04-02 01:07 when do you sleep? :) 2012-04-02 01:08 book, good point ! 2012-04-02 01:10 or maybe both :) 2012-04-02 01:11 but if the driver can manipulate both internal PLL and LNA, what else is missing? 2012-04-02 01:11 besides the signal processing book :) 2012-04-02 01:16 I think a proper sdr is more than a PLL and a LNA 2012-04-02 01:17 you'll at very least need a ring modulator and a highspeed D/A A/D as well 2012-04-02 01:18 DocScrutinizer: the thing is only a receiver 2012-04-02 01:18 ooh 2012-04-02 01:18 then you don't need the DA 2012-04-02 01:19 http://web.archive.org/web/20100527005609/http://www.elonics.com/assets/PB_E4000_10_2009.pdf 2012-04-02 01:20 it would be difficult to make this information any less detailed ;-) 2012-04-02 01:21 :) 2012-04-02 01:33 well, the ring modulator is called "mixers" there 2012-04-02 01:51 [commit] Adam Wang: removed 74AUP1G08GW,125 from diodes_inc.lib (master) http://qi-hw.com/p/kicad-libs/75930c9 2012-04-02 01:51 [commit] Adam Wang: Merge branch 'master' of projects.qi-hardware.com:kicad-libs (master) http://qi-hw.com/p/kicad-libs/c5dea10 2012-04-02 01:53 wej has joined #qi-hardware 2012-04-02 02:22 rejon has joined #qi-hardware 2012-04-02 02:23 cladamw: i have some doubts about the scalability of the approach you've chosen for some of the schematics symbols 2012-04-02 02:25 cladamw: there are those that refer to a unique component, e.g., the audio codec. there, it makes sense to have one specific symbol for that part 2012-04-02 02:26 cladamw: but then there are those that are more or less generic. e.g., diodes. they may come in a number of different packages, but the number of pin variations is still quite small 2012-04-02 02:27 cladamw: so there are a lot of parts that can all use the same symbol. and the symbol should have a generic name, not just a vendor part number 2012-04-02 02:28 currently I don't have common idea on using 'same' specific symbol for parts's usual goal. for the one of audio codec is the beginning one of this time first doing. which is worse idea of my latest using a vendor chip name to include them. 2012-04-02 02:29 cladamw: now, i know that some people use a different approach and have basically one symbol per vendor part, with all the fields already set, and so on. but i'm very sceptical about the scalability of such an approach, since you'd have gazillions of duplicate symbols 2012-04-02 02:30 since there's few columns of 'properties' can fill up real p/n and description. 2012-04-02 02:30 the audio codec is okay. it's probably the only part with that pinout anyway 2012-04-02 02:30 ha....i finally still using this idea is : 2012-04-02 02:31 1. when user try to edit schematic, which information he should look for ? actually the p/n is rather than a specific symbols. 2012-04-02 02:32 the design should start with generic concepts. e.g., a p-FET instead of a specific diodes inc. part 2012-04-02 02:33 2. if we spend much time to find a 'specific' common smybol to corresponding to all same like AND /XOR gate then he doesn't need to find libraries/database to find 'good' AND gate symbol for his goal. 2012-04-02 02:33 stratman2 has joined #qi-hardware 2012-04-02 02:34 3. your specific unique 'symbol' is good /convience to passive parts or non -active parts., for ic chips, you are hard to let people know which on is which one. 2012-04-02 02:34 a universal AND gate wouldn't be possible with how kicad works, because each symbol needs to define all pins of the component. would be nice, though ;) 2012-04-02 02:35 4. for audio chip , i'm thinking even using wolfson.lib instead the p/n.lib 2012-04-02 02:35 oh, sure. for complex chips, MCUs, codecs, FPGAs, and so on, you definitely want a specific part 2012-04-02 02:35 for examples now. 2012-04-02 02:36 but there are more generic ones, e.g., diodes, diode arrays, transitors (BJT, FET, etc.), transistor arrays, gates, groups of gates, etc., where you have many packages with equivalent topology and also several vendors 2012-04-02 02:37 you collected many '74x1g***'.lib which user needs to scrolling and find what it looks like while placing parts or find it in Library Editor to see it. 2012-04-02 02:37 and in many cases also different parameters. e.g., a 741G00 will look pretty much the same in HC, HCT, LV, ... 2012-04-02 02:38 but if we gave them a real p/n, yes, it will create a lot of database(symbols), but it's easy to get part in schematic editot. ;-) 2012-04-02 02:38 and also a diode often doesn't look different in the schematics when it's 10 mA or 10 A :-) 2012-04-02 02:38 yes, i know, the real questions are we don't need to collect so much HC, HCT, LC, LVC now. :-) 2012-04-02 02:38 yes, once you've found it among the millions of parts, then it's easy ;-) 2012-04-02 02:39 so you already collected much of them. :-) 2012-04-02 02:39 Has anybody got mac os/x to talk to Ben? USB port where Ben is connected to, is not being recognized on my PC when I load a CD boot of linux. The Mac sees the Ben, so I guess next step is to get CD boot of MAC os/x (don't want to install linux on the mac). Other thing to try is just bite the bullet and get another PC and install linux, but want to avoid that - I am trying to update Ben with latest version of linux.. 2012-04-02 02:39 sure. but we should pick a structure that will not require a full reset when the library grows 2012-04-02 02:39 when people edited parts, they may think if the part is COMMON, like HC, HCT, even etc...yes they are. :-) 2012-04-02 02:41 but for reset IC ? usb power switch ? do you think that all chip vendors they will share their own pins assignments to a "Unique" one, the answer is probably not. ;-) 2012-04-02 02:41 for now, I'm thinking to if adding spartan-6 into xilinx.lib? 2012-04-02 02:42 for things like 74xx1Gxx, we could use a semi-generic structure. here's an illustration: http://downloads.qi-hardware.com/people/werner/tmp/out.pdf 2012-04-02 02:42 i don't have good idea now. since KiCad has included a same name library called "xilinx.lib" and there's one spartan-6 'XC6SLX25T-BG484' existed there already. 2012-04-02 02:43 if you open Logic > Single >, then you have a number of 74x1g07_* parts. they differ by the number of pins. 2012-04-02 02:43 yeah....yes, I agreed on those very few older generation chip TTL, CMOS, etc for collecting ic symbols. 2012-04-02 02:44 i agreed your ideas on that, but you still can't deal with those usb power switches, reset ics, etc.... 2012-04-02 02:45 i think reset ICs are semi-generic. the question is if you have a specific symbol for them :) 2012-04-02 02:45 usb power switches may be unique enough that it doens't make sense to try to generalize them 2012-04-02 02:45 for the audio codec, I think that I'll go back to create a wolfson.lib for WM9707... 2012-04-02 02:46 (xilinx) i would not touch the kicad libs. we should get rid of them. they're full of inconsistencies anyway 2012-04-02 02:46 (inconsistent font sizes and so on) 2012-04-02 02:48 KiCad does a real inconsistent symbols for spartan-6, it must not contributed by their community, since the best one is to follow current M1's 'bank' catagories to divide parts. 2012-04-02 02:48 i'd rather have one library per symbol (more or less. sometimes that doesn't make sense, but it's usually possible) 2012-04-02 02:49 (1 lib per sym) the reason is that it's easier to track revisions that way. and you have a lower risk of commit conflicts. 2012-04-02 02:49 yeah...so I just first created a name xc6slx45-2fgg484c.lib locally, not using a same name as 'xilinx.lib' 2012-04-02 02:51 (1 lib per sym) violates ( specific symbol per old TTL/CMOS. etc.) 2012-04-02 02:52 does it ? how ? 2012-04-02 02:52 for our spartan-6 p/n, now it violates (symbols collected per vendor name) 2012-04-02 02:52 phew ~ 2012-04-02 02:53 look in kicad-libs/components/, there's 74x1g07_4.lib, 74x1g07_5.lib, ... :) 2012-04-02 02:54 on lib per symbol. and three different symbols for the 74x1G07 family of gates (one for packages with 4 pins, one for 5, etc.) 2012-04-02 02:54 yes, you are with the same way to get roma to build a lot of symbols, a quite way same as mine to collect all p/n for one vendors. 2012-04-02 02:55 i wish kicad would let us treat them as a generalized 74x1G07, but that may be a bit difficult. and it may be confusing 2012-04-02 02:55 ah, but each of these applies to many vendor parts 2012-04-02 02:56 well. well...i won't touch your existing 74x***.... you can imagine that how we need to build own components.lib ? 2012-04-02 02:58 if I tried to find 'unique' symbol name into components for those parts in M1, then how much time i need to spend ? 2012-04-02 02:58 i only need a new 74x1G07 when i find one that has a different pinout. e.g., in a package with a different number of pins or a different assignment of functions to pins. but further package details, the family, vendor part naming schemes, etc., don't affect the symbol 2012-04-02 02:59 0.00001% of the time you'd spend later duplicating things indefinitely ? ;) 2012-04-02 03:01 we should have thought long term vs. short term way to go for it. I would say that specific / unique symbol is good but not suitable for every chip. :-) 2012-04-02 03:01 agreed 2012-04-02 03:01 it's mainly the simple things that have alternatives 2012-04-02 03:02 for example, when i met connector, i skipped and ignored to create new one. 2012-04-02 03:02 so i also agreed yours. ;-) 2012-04-02 03:02 extreme example: resistors. you wouldn't want to have one symbol for each vendor part number. or would you ? :) 2012-04-02 03:02 sure. 2012-04-02 03:03 that's the same 'Relativity' both you and me created. :-) 2012-04-02 03:05 also for example, how big compatibilities that audio codecs can meet and drop to place with the same symbol or footprint together ? it's very few. so i created a real p/n for it. 2012-04-02 03:06 things like reset chips are a bit borderline. there are lots of reset chips that all look very similar. but they may have different names for the same function, depending on the vendor 2012-04-02 03:06 but in the else fields, a SOT-23-3, it could be diode, led, reset ic, etc...they could be the same footprint but not same symbol name in each pin. 2012-04-02 03:06 so for reset chips, i think either way is fine 2012-04-02 03:07 yes, agreed with audio codecs. complex chips can't have a "generic" symbol 2012-04-02 03:07 diode, led, transistor, etc., all have different circuit symbols 2012-04-02 03:07 so when we build footprint, the 'specific' or 'generic' will be much the same approach to this way. 2012-04-02 03:07 let's not confuse circuit symbol and package 2012-04-02 03:08 no, not at all. think of the resistors again :) 2012-04-02 03:08 yeah,,,so when i created these days , i was also confused myself a bit. but in the end i still created the same symbols as vendor chip's drawing. ;-) 2012-04-02 03:09 rejon has joined #qi-hardware 2012-04-02 03:09 yeah... i think that we agreed on "not confuse" circuit symbol. 2012-04-02 03:10 i will always think a bit when creating a new symbol. Also find if KiCad has it already or not. ;-) 2012-04-02 03:12 i think a good test is if you think what part you'd use for this function. if you first have to go to digi-key to find the most common one, then the symbol should probably be generic. e.g., DVI connectors may have very little variation 2012-04-02 03:15 similar for USB. for USB < 3.0, the only thing that varies is the number of ground pads. so we should have one symbol for each ground configuration, and that's it. (if you look at the current library, it isn't structured like this yet for USB, because i didn't understand this pattern until recently) 2012-04-02 03:15 btw, do you know that why i didn't use your '74X1G08_6.lib' for M1's '74AUP1G08GW,125' ? since liked i said last week, i think that the best top view for chip pin name is to keep the same pin arrangement. ;-) 2012-04-02 03:18 well, we can have two variants. sometimes, you want to put more emphasis on the function, sometimes more on the physical appearance 2012-04-02 03:18 i'd rather to try make a symbol now to keep visually same orientation/top viewing from now on. then when user debugging, don't need to open datasheet to find it. Just look it at schematic. :-) 2012-04-02 03:18 you lose when it comes to big chips :) 2012-04-02 03:18 agreed. :-) 2012-04-02 03:18 SoCs, FPGAs, ... 2012-04-02 03:18 agreed too. 2012-04-02 03:19 (view) you could of course open the layout :) 2012-04-02 03:19 so when created SoCs etc, then we build them same as datasheet said. :) 2012-04-02 03:20 for MCUs, i usually take the physical view. mainly because i assign functions according to where i expect things to go in the layout 2012-04-02 03:20 when build M1 symbols nowadays, i dislike them now. maybe the same feeling from you for mine. :-O 2012-04-02 03:22 i haven't looked at the symbols yet :) so far, i'm only commenting on the general structure of the library 2012-04-02 03:22 M1's U2[ksz8001l] and U21[adv7181c] is good (same pin direction). :-) 2012-04-02 03:22 sure. the general structure i agreed. :-) 2012-04-02 03:24 M1's all others symbols[ bad and not easily to read] when debugging... as you know, for SoC, FPGA, the best to to seperate their functional pins as possible. :-) 2012-04-02 03:25 this may be more a tool problem than a symbol design issue 2012-04-02 03:25 mmm... if there's one, it would be good. 2012-04-02 03:26 so i think when chips's total # is under 48 pins, or 64pins, the best rules for creating symbol is: 2012-04-02 03:27 1. to keep pins assignments as same as chip's outline layout top viewing. 2012-04-02 03:28 2. find a exiting already generic/specific frequent symbols, people know already, like TTL, COMS, etc... 2012-04-02 03:29 3. but 2) at best to follow 1) as possible. some how in current electronic symbol drawing violate rules we expect above. ;-) 2012-04-02 03:30 i think it depends on what you want to show. also don't forget that symbols can be mirrored in the schematics. then it can't match the footprint even though it follows the physical layout. 2012-04-02 03:30 sure 2012-04-02 03:34 btw, in kicad, you can launch eeschema and pcbnew on different screens. when you click on a pin in the schematics, the cursor in pcbnew will jump to it 2012-04-02 03:34 and vice versa 2012-04-02 03:35 yeah. :) 2012-04-02 03:57 pabs3 has joined #qi-hardware 2012-04-02 04:00 pabs3 has joined #qi-hardware 2012-04-02 04:17 The build was successful: http://fidelio.qi-hardware.com/~xiangfu/build-nanonote/openwrt-xburst.full_system-20120331-2358 2012-04-02 04:19 rejon has joined #qi-hardware 2012-04-02 05:00 on component libs: it might be useful to just "steal" from standard libs but actually create a generic symbol for each part type in BOM and put all in one QI.lib 2012-04-02 05:01 this way you can have all the parameters of your component in the lib and use it as well for spice and whatnot 2012-04-02 05:03 and you have one pretty small lib to ship to user, instead of a plethora of vendor specific libs, plus your own tweaked versions of those when the original doesn't meet yout requirements 2012-04-02 05:12 creating BOM was a straightforward task then 2012-04-02 05:12 basically your lib with all the symbols *is* your BOM 2012-04-02 05:13 you could argue if you want to follow this approach even for plain vanilla 0402 resistors 2012-04-02 05:14 but then OTOH why not 2012-04-02 05:15 in the end it's just a copy from another (standard) lib to your QI.lib for each component you include to your schematic 2012-04-02 05:16 or a pick/edit/paste action, to add the particular component properties 2012-04-02 05:17 nothing terrible in doing that for 33R, 47R, nnnR. You have to do it *somewhere* anyway 2012-04-02 05:18 you could use 2 libs: an archive lib QI_we_had.lib plus a QI_we_recently_have.lib 2012-04-02 05:19 big advantage: all your components are of a uniform style. This usually isn't the case when you use vendor libs 2012-04-02 05:22 for the idea to match fotprints - it's a generally odd idea. You rather have symbols especially for huge components where you place all your pins in a way you get nice legible schematics. There's simply no use in somehow showing pinout in schematics 2012-04-02 05:24 e.g. when your MICP and MICN inputs of audio chip create a need to cross their lines in your schematic, then you simply swap them for that particular codec symbol so your lines don't cross anymore 2012-04-02 05:24 cladamw has joined #qi-hardware 2012-04-02 05:30 so you usually would have a number of single named pins as entities for all chips with >16 pins, and you place those into arbitrary boxes in any way you like. For things like 7400 4 NAND gates, you'd have 4 entities of one gate for the component and you place each gate wherever you like 2012-04-02 05:31 (single pin entities) sometimes it makes sense to group pins even then, e.g. for buses 2012-04-02 05:32 as you definitely don't want "pinout" in schem like "d0 d3 d1 d2" 2012-04-02 05:32 same might go for interfaces like SPI, I2C etc 2012-04-02 05:34 if you could have "macro" components consisting of several layers of sub-components down to the individual pin, and you may group and ungroup at arbitrary levels of that hierarchy, that would be really nice 2012-04-02 05:35 even nicer though, if you can have alternative hierarchies (trees) on the same pins (as leaves) for the same component, and switch between them 2012-04-02 05:36 Martix has joined #qi-hardware 2012-04-02 05:37 e.g some pins may have alternative primary function Addr17..32, secondary camera-IF 2012-04-02 05:38 you'd like to group these pins in two alternative predefined ways, so "user" can switch between the version of subtree he wants to use for that particular set of pins 2012-04-02 05:40 for simpler components like 7400 4*NAND you frequently find orphaned "spare" gates somewhere on a separate sheet or in a dark corner of schematic, when the EE used this approach. I always thought this is a very nice and clear way to document things 2012-04-02 05:42 and one more argument to do it this way: you definitely have best data to run design rule checks against your design 2012-04-02 05:43 no outputs of unused gates accidentally tied to GND or VDD, just because they were "forgotten" in schematics. 2012-04-02 05:46 ooh yes, and of course each 7400 also has a VDD and a GND pin entity. You usually place those next to the gates if all 4 gates are used on same sheet. You place them on a separate power sheet when there's no better place for them 2012-04-02 05:57 DocScrutinizer: kicad lets you have multi-part symbols. it doesn't let you rearrange their pins, though. 2012-04-02 05:57 morning 2012-04-02 05:58 hmm, also multi-multipart? 2012-04-02 05:58 (vendor libs) you mean the default kicad lib ? there are problems with that one. so it's better if we replace it. it also has only very few parts, so it's not an insane amount of work 2012-04-02 05:59 no, just one level. you're free to duplicate, though :) 2012-04-02 05:59 won't fly for my idea of implicit alternatives :-/ 2012-04-02 06:00 and for rearranging pins/sub-parts - that's pretty bad restriction 2012-04-02 06:01 I don't get what you mean by vendor lib = kicad default lib 2012-04-02 06:01 (one sym per part) i've heard of that approach. i think it's insane :) how long until you have all sorts of inconsistencies because something was changed without updating ancestors/children ? 2012-04-02 06:01 i was asking what you call a "vendor lib" 2012-04-02 06:02 xilinx 2012-04-02 06:02 which children? 2012-04-02 06:03 you mean a library where we group parts specific to a single vendor ? why would that cause inconsistencies ? 2012-04-02 06:04 if you're going to change all your resistors from 0402 to 0603, then of course you have quite some components symbols to change. You had to do exactly same amount of editing in BOM anyway, if you don't do it in CAD compnent lib 2012-04-02 06:04 children = items derived from the parent item. e.g., BOM entry from sym. "derivative sym" from "master sym". etc. 2012-04-02 06:05 hell, I mean the VENDOR libs that come from vendor. I don't suggest we create a xilinx lib for the one chip we use from them 2012-04-02 06:05 wtf is a sym component? 2012-04-02 06:05 how would it change? 2012-04-02 06:05 are you sure xilinx provides kicad symbols ? :) 2012-04-02 06:05 NO 2012-04-02 06:06 damn, I'm short in time, sorry. No spare time for fun talk 2012-04-02 06:06 hurry to work ! :) 2012-04-02 06:06 there's no such thing like master sym and child sym. At least I don't see any 2012-04-02 06:07 we won't go change all R from US to EU symbol 2012-04-02 06:07 well, that would be one such change :) 2012-04-02 06:07 if you wanna do this, you need a clever editor 2012-04-02 06:08 read: sed 2012-04-02 06:08 or even awk 2012-04-02 06:08 but it can be something smaller. like adjusting a font size. or just a small correction in a symbol. or a global style adaptation. 2012-04-02 06:08 so instead of having to edit a few dozen symbols you get to edit millions. not nice. 2012-04-02 06:08 if global styles are embedded in particular symbol definitions, then you already have a huge problem 2012-04-02 06:08 that's precisely the lack of scalability that makes EE such a mess :) 2012-04-02 06:09 "global style" can mean the grid you use. or how long you make pins. 2012-04-02 06:09 please don't exaggerate in such an extreme way 2012-04-02 06:10 well, look at it :) 2012-04-02 06:10 we're not building big blue here, end even that one has no separate schematics for each of the NNNNN nodes 2012-04-02 06:10 so no way we get beyond 1000 or 2000 components *total*, many of those are same 100R 2012-04-02 06:10 the goal is a shared library for everything in the qi-hw universe (and beyond, if anyone wants to reuse it) 2012-04-02 06:11 so what 2012-04-02 06:11 exactly what I suggested 2012-04-02 06:11 QI_what_we_currently_got.lib 2012-04-02 06:11 if that changes, you have to change all sorts of lists anyway 2012-04-02 06:12 so why not change the lib instead, and autogenerate the lists 2012-04-02 06:12 it's not only what we currently have but also what we may want in the future 2012-04-02 06:12 sigh 2012-04-02 06:13 naw, one symbol per part stands the workflow on its head 2012-04-02 06:13 what you have is typically a set of characteristics 2012-04-02 06:13 per kind!! 2012-04-02 06:13 sometimes, that's just one specific part (e.g., a specific SoC) 2012-04-02 06:14 maybe your workflow. evidently there are other workflows, in "professional" EE 2012-04-02 06:14 sometimes, it's a gazillion of parts that all have the same symbol. e.g., resistors 2012-04-02 06:14 i know that workflow you're describing exists 2012-04-02 06:14 where your design rule checks would warn you that this component isn't sourced yet, etc 2012-04-02 06:15 but it's not the only one 2012-04-02 06:15 now it's even gazillions WTF°! 2012-04-02 06:15 well, N(different types of resistors) :) 2012-04-02 06:15 I still can count all the component part entities of M1 in half an hour on schem 2012-04-02 06:16 probably a few 100k at digi-key alone. maybe a few ten k in our database 2012-04-02 06:16 let alone the number of different ones 2012-04-02 06:16 BS 2012-04-02 06:16 perhaps we should continue this when you don't have to run to work :) 2012-04-02 06:18 you need a 95.73kR. You check what you're using for R now, find the part at vendor in the series, and clone a symbol for it in your lib. Case B): no such part in vendor series, you find another vendor, andother R, you create a new symbol, where you borrow from whatveer you feel fit 2012-04-02 06:19 the overhead compared to editing your stock lists, your BOM, your whatnot, is marginal 2012-04-02 06:19 why clone ? you use the common R symbol and fill in the parameters you want. done. no need to update the global (!) component library (or create a project-local clone, etc.) 2012-04-02 06:20 the BOM is auto-generated from the parameters 2012-04-02 06:20 that's what boom does :) 2012-04-02 06:20 the global(!) has an exclamation mark only because you like to put it there. No other sound reason 2012-04-02 06:20 we can capitalize it ;-) 2012-04-02 06:21 meh, you don't like my concept, because your tools do it differently, and the tools define what can and should be done in which way. Fine with me 2012-04-02 06:22 OK, enough gfun talk eating my time 2012-04-02 06:31 a last quick remark: neither OM nor any of the "spinoffs" has ever wrapped their heads around two basic things: design rule checks, and JTAG boundary scan 2012-04-02 06:31 I suggest you try to find out why big companies use both methods 2012-04-02 06:32 jluis|work has joined #qi-hardware 2012-04-02 06:32 not sure what you mean by "Design Rule Check". the EDA tools we use have checking functions and we use them (of course) 2012-04-02 06:33 you're right with JTAG, though. i did some boundary scans in ben-wpan, but with a non-JTAG approach (no JTAG available) 2012-04-02 06:36 M1 has semi-automated functional test. but they're not at such a low level 2012-04-02 06:37 and wolfgang doesn't even want automated current monitoring during tests :) 2012-04-02 06:37 I want to make small incremental improvements that help a user *now* 2012-04-02 06:38 as opposed to being in lab-mode the next 100 years or longer :-) 2012-04-02 06:38 wpwrak: where can i find your collection of footprints again ? 2012-04-02 06:38 i finsihed my bom and am now figuring out which footprints i need to draw 2012-04-02 06:39 particularly i am looking for a bga84 ddr2 footprint 2012-04-02 06:39 i found the link to your kicad patches 2012-04-02 06:39 but i am failing to find the footprint store 2012-04-02 06:40 blogic: footprints are here: http://projects.qi-hardware.com/index.php/p/kicad-libs/source/tree/master/modules/ 2012-04-02 06:40 gracia 2012-04-02 06:40 blogic: no large BGAs or DDR there, though 2012-04-02 06:41 ok 2012-04-02 06:41 i tried to figure out how fped works last night 2012-04-02 06:41 did not advance that much though ;) 2012-04-02 06:41 will spend more cycles on it today 2012-04-02 06:42 blogic: (more cycles) good :) people generally find it easy to learn. surprisingly easy, actually 2012-04-02 06:43 actually i was wondering last night if it somehow does not work properly on my machine 2012-04-02 06:43 have you tried it inside awesome WM ? 2012-04-02 06:43 i had the impression that the WM was causing issues, as in mouse click seamed to be getting lost 2012-04-02 06:43 will investigate more today 2012-04-02 06:44 wolfspraul: hopefully < 100 years :) but one prerequisite for introducing such things rapidly would be a bit of a budget for buying instruments. else, we'd have to support the whole zoo that's the installed base 2012-04-02 06:44 (awesome) first time i even hear of that :) 2012-04-02 06:45 LunaVorax has joined #qi-hardware 2012-04-02 06:46 there shouldn't really be any problem with the window manager. fped keeps things _very_ simple. 2012-04-02 06:49 o i need to do anything magic to place a pad ? 2012-04-02 06:49 i am only able to add vectors somehow 2012-04-02 06:50 anything else i click on will always default back to the "select and move" tool 2012-04-02 06:50 you need two vectors to form a square or rectangle. then you inscribe the pad. 2012-04-02 06:50 e.g., 2012-04-02 06:51 1) vector (1mm, 1mm) 2012-04-02 06:51 2) vector (-1mm, -1mm) 2012-04-02 06:52 3) select pad, then go to the endpoint of the first vector, click and drag to the endpoint of the second vector. release and the pad is there 2012-04-02 06:52 click on the pad to select. then you can edit the name or change the type 2012-04-02 06:52 * blogic tries 2012-04-02 06:53 ahhhh 2012-04-02 06:53 (type) normal = normal pad; bare = without solder paste; trace = covered by solder mask; etc. 2012-04-02 06:54 kk 2012-04-02 06:55 now you can change the vectors and the pad will adapt 2012-04-02 06:55 and paste is with paste i assume 2012-04-02 06:55 awesome 2012-04-02 06:55 it was too simple ... i was looking for the magic ;) 2012-04-02 06:56 paste is just solder paste. you can use it to construct things. e.g., a pad that's only partially covered by solder paste (like the center of QFNs) 2012-04-02 06:56 hehe :) 2012-04-02 06:56 nikescar has joined #qi-hardware 2012-04-02 06:56 yes 2012-04-02 06:56 wpwrak: i need a central pad on a lqfp128 2012-04-02 06:57 wpwrak: DRC I basically mean sth like spice integrated into CAD 2012-04-02 06:57 blogic: you may find these QFNs useful: http://projects.qi-hardware.com/index.php/p/kicad-libs/source/tree/master/modules/qfn.fpd 2012-04-02 06:57 yes, saw them 2012-04-02 06:58 blogic: and here is a description of how they're constructed: http://projects.qi-hardware.com/index.php/p/kicad-libs/source/tree/master/modules/INFO 2012-04-02 06:58 wpwrak: to do this level of design check you need better data of a component than only "33 Ohm" 2012-04-02 06:58 DocScrutinizer: ah yes, that's something we're looking into just now 2012-04-02 07:00 (better data) especially if using wire-wound resistors ;-) 2012-04-02 07:04 indeed, which brings you from your unified R component concept to a "one definition per component kind" concept 2012-04-02 07:05 I'm all for inheriting properties in a parent-child design, if it's versatile enough 2012-04-02 07:05 but in the end you want a definition entity for a class of identical components in your design 2012-04-02 07:06 and identical (again) doesn't mean "it's a resistor" 2012-04-02 07:06 you mean "equivalent" parts ? 2012-04-02 07:06 it means "resistor wirewound 33R 1/4W" 2012-04-02 07:07 (plus a few other properties, for "spice") 2012-04-02 07:08 that's basically what the parameters do. they restrict what components can be used. if any additional characteristics are known, they're then picked from the parts database 2012-04-02 07:09 so you have all that information (that is, if it's generally available. not sure if inductance and such are readily available for, say, chip resistors), but it comes after the schematics 2012-04-02 07:15 well, I can continue discussion for next 30min, just with off times in between 2012-04-02 07:20 and i should get a bit of sleep :) 2012-04-02 07:29 ok, when your global lib has all the components like nn R n mW 2012-04-02 07:31 I just wonder how you're ding said change US-EU symbol *then* 2012-04-02 07:31 doing* 2012-04-02 07:33 the database has things like this: PANASONIC ERJ-6ENF16R9V FP=0805 P=1/8W R=16.9R T=R TC=100ppm/K TOL=1% V=150V 2012-04-02 07:33 fine 2012-04-02 07:34 and a link to a common symbol, or symbol directly attached? 2012-04-02 07:34 (T=R is the key that identifies this as a resistor. i didn't distinguish resistor constructions yet. should eventually, though) 2012-04-02 07:35 no no, this has nothing to do with symbols 2012-04-02 07:35 :-/ 2012-04-02 07:35 this operates off the BOM generated by the schematics entry program 2012-04-02 07:36 this BOM has component reference, footprint, and any parameters the user cares to specify 2012-04-02 07:36 you'd want to unify BOM and schem 2012-04-02 07:36 then boom looks for matching parts. of these, it picks the cheapest 2012-04-02 07:37 I could come up with a dozen instances where footprint determines schematic symbol 2012-04-02 07:37 (cheapest generally means "common". if you need something fancy, then you'd have to specify how it differs from what boom dug out) 2012-04-02 07:37 oh, sure 2012-04-02 07:39 ok, afk now, prolly til brunch break 2012-04-02 07:39 but it's also often an 1:N mapping (often with corner cases making it N:M) 2012-04-02 07:39 yep 2012-04-02 07:40 frohes schaffen ! :) 2012-04-02 07:40 (i don't think this has a good translation) 2012-04-02 07:45 dvdk has joined #qi-hardware 2012-04-02 07:59 Martix has joined #qi-hardware 2012-04-02 08:14 jekhor has joined #qi-hardware 2012-04-02 08:33 Martix has joined #qi-hardware 2012-04-02 08:52 jekhor has joined #qi-hardware 2012-04-02 09:15 wolfspraul has joined #qi-hardware 2012-04-02 09:16 Ewan has joined #qi-hardware 2012-04-02 09:23 panda|x201 has joined #qi-hardware 2012-04-02 09:42 Aylax has joined #qi-hardware 2012-04-02 10:17 MabYwen has joined #qi-hardware 2012-04-02 10:30 antoniodariuh_ has joined #qi-hardware 2012-04-02 10:50 GNUtoo has joined #qi-hardware 2012-04-02 11:22 DocScrutinizer has joined #qi-hardware 2012-04-02 11:25 wolfspra1l has joined #qi-hardware 2012-04-02 11:33 stratman2 has joined #qi-hardware 2012-04-02 11:41 Martix has joined #qi-hardware 2012-04-02 11:58 danke 2012-04-02 11:59 cladamw has joined #qi-hardware 2012-04-02 12:14 antoniodariuh_ has joined #qi-hardware 2012-04-02 13:23 wpwrak, have you edited several parts in per parts in KiCad ? i.e. using "Parts are locked" to be checked. :-) 2012-04-02 13:23 i only did once in avt2 for jz4720. 2012-04-02 13:24 you mean a multi-part component, like the FPGA ? 2012-04-02 13:25 wpwrak, yes! :-) Do you know how to create different graphic rectangle to the component body in different sub-part ? 2012-04-02 13:26 you can basically draw completely different things for each sub-part 2012-04-02 13:26 we used this extensively in gta02-core. lemme find it ... 2012-04-02 13:27 when i checked that icon "Parts are locked", f.g. I edited 'part A' with A size rectangle, but the graphic will still be placed in part-[B:F] which is not the right for the 'check' purpose. :( 2012-04-02 13:28 i'm not sure what "parts are locked" does 2012-04-02 13:28 and I imported KiCad's original one multi-part components. But they are different size rectangle. 2012-04-02 13:29 with "parts are locked" being checked, means the changes in current sub-part will not influence to other sub-part sheet. 2012-04-02 13:30 what you're looking for may be the "Edit pins per part or body" button, top button row, the last on the right 2012-04-02 13:31 yes, right, that's one. :-) 2012-04-02 13:31 hmm. svn,openmoko.org seems to be down. no examples from gta02-core then :-( 2012-04-02 13:31 if it's locked, any changes on current sub-part won't change else sub-part. 2012-04-02 13:32 if you select 'edit per part", yes 2012-04-02 13:32 wpwrak: hardware issues, hopefully back soon 2012-04-02 13:32 wow... :( that's bad news for gta02-core. 2012-04-02 13:32 cladamw: almost all the good stuff has been salvaged into qi-hw by now anyway :) 2012-04-02 13:33 hopefully i can get find msg in KiCad-users email threads. :) 2012-04-02 13:34 wpwrak, qi-hw is qi existed here and there! Good! :-) 2012-04-02 13:36 rektide has joined #qi-hardware 2012-04-02 13:37 argh. next time it's up we really should copy anything from gta02-core that still has value 2012-04-02 13:41 wpwrak, i used the same rectangle size for sub-parts in avt2. this means I needed to arrange a suitable expected size for them. But for creating huge SoCs or FPGAs, it may hard to. It must be some settings steps or editing steps in Library Editor. :( 2012-04-02 13:47 wpwrak: svn.om.org supposed to be up after move to new iron 2012-04-02 13:48 agni was down, and Roh didn't bother to reboot some of the vservers after hetzner swapped the NIC (or whatever). He told me it's scheduled for after the 'big move' 2012-04-02 13:49 Aylax has joined #qi-hardware 2012-04-02 13:55 wpwrak: you could ping roh about it 2012-04-02 13:57 roh: ^^^ 2012-04-02 13:58 wolfspra1l: well, i have a local copy of the svn ;) 2012-04-02 13:58 wpwrak, nice ... not sure if it's the s/w version problem. :) see http://tech.groups.yahoo.com/group/kicad-users/message/11442 2012-04-02 13:58 cladamw: you don 2012-04-02 13:59 t have to make them all the same size. they can have pretty much any shape 2012-04-02 13:59 bbl 2012-04-02 14:00 pabs3, DocScrutinizer: (svn.openmoko.org) so it's a self-healing problem then. good ;) 2012-04-02 14:01 cladamw: ah yes, be VERY careful if not in "edit per part" mode ;-) 2012-04-02 14:01 wpwrak, oah...sure...i tried to make different shape firstly but failed. :( Since even changes are in sub-part will all copy to else sub-parts no matter i check it or not. phew ~ 2012-04-02 14:01