#qi-hardware IRC log for Saturday, 2010-09-11

xiangfuHi all.00:03
xiangfuI am try to add this option to kernel command line: 'g_ether.host_addr=da:40:9a:d8:78:2a'00:03
aisacoi lo'ai xiangu00:03
xiangfubut I am not sure if the 'MAC Address' is free to use?00:04
xiangfuaisa: Hi. :)00:04
aisaxiangfu: you mean not already allocated to a vendor or device?00:05
xiangfuaisa: yes.00:06
aisaStart with this?:  http://www.iana.org/assignments/ethernet-numbers00:06
aisaVMWare assigns dynamic mac addresses, so there must be some kind of experimental or open range.00:06
aisaand there are most definitely blocks that aren't available,00:06
aisabut as a worst case, purchase a network card, record the MAC address,00:06
aisaand then never use that card ever again :-)00:07
aisaI may misunderstand what you're trying to do...00:07
wolfspraulI think 54:52: is reserved as a 'local' range or so00:12
xiangfuaisa: nanonote also use the dynamic mac address. but when we reflash the kernel. the mac address changed. then we must remove the old one in .ssh/know_hosts.00:13
xiangfuwe just need a static mac address for NanoNote00:13
aisahow about something like 3l:33:70:....?  :-D00:14
xiangfuaisa: if we can just pick one dynamic mac address for static?00:15
aisaI'm actually not clear how USB=>ethernet works,00:15
aisaso I may not be helpfully contributed to the conversation.00:16
wolfspraul54:52:00:'B':'N':'N' :-)00:16
wolfspraulcan't find the document that says 54:52 is reserved for local use...00:18
xiangfuwill offline for 5 ~ 10 mins. back later :)00:22
wolfspraulxiangfu: there's a bit that needs to be set, see http://en.wikipedia.org/wiki/MAC_address00:26
wolfspraulif the bit is set, it's 'locally administered'00:26
wolfspraulso forget the 54:52 stuff, don't know where I got this from. the locally administered bit is not even set there.00:28
wolfspraulanything 02:00:00:00:00:00 should work. you can set any of the zeroes to a radomly chosen 'Ben NanoNote' identifier00:29
xiangfuwolfspraul: rename to trunking_* done.00:40
aisaw00t!  I want to update my repos.00:40
aisawhat did we do last time?00:40
aisaoh right, editted .git/config00:41
aisaxiangfu: openwrt_trunk => ??? and openwrt_backfire => ???00:41
xiangfuoepnwrt_trunk --> tracking_trunk00:42
xiangfuopenwrt_backfire --> tracking_backfire00:43
xiangfuaisa: ^00:43
wolfspraulxiangfu: he, again :-) cool, I think now everybody is happy, we see...00:43
aisaI'm deleting the remote branches in my local repository:00:45
aisagit branch -rd origin/xburst00:45
aisaI've fixed the wiki too.00:47
xiangfuaisa: thanks.00:57
qi-bot[commit] Xiangfu Liu: this u-boot patch fix the c_size. it should 22bit. http://qi-hw.com/p/openwrt-xburst/67f552002:29
xiangfujust test one commit for make sure commit is works fine. now the email(update local to rename the branch) just sended. :)02:32
wolfspraulxiangfu: do you know why the sound modules are not statically compiled into the kernel?03:37
wolfspraulbtw - my minimal rootfs is actually booting now, my mistake with the 'unable to mount rootfs' was that I didn't do an nerase - forgot how sensitive ubifs is to random data behind the actual filesystem...03:38
xiangfuwolfspraul: no. larsc decide not compile into kernel.03:38
xiangfuwolfspraul: I also don't know why.03:39
wpwrakgaah. way too much stuff.03:43
wpwrakwolfspraul: data point: (from early last week) usp did a wifi chip design03:43
wpwrakwolfspraul: data point: everybody but red hat seems to be unhappy about glib, yet not quite ready for a real fork03:44
qi-bot[commit] Niels: don't perform action if a modifier is pressed http://qi-hw.com/p/nanomap/bf6b56603:45
qi-bot[commit] Niels: move handling of application parameters to main.cpp http://qi-hw.com/p/nanomap/4e98efe03:45
qi-bot[commit] Niels: add option to skip existing tiles while downloading; based on patch by Mikhail Peselnik http://qi-hw.com/p/nanomap/bfd639f03:45
wpwrakdata point (to no one in particular): despite my most accurate retelling of my experience making my usrp2 work, christoph didn't die laughing, although i think he was close :)03:46
wpwraknow i have to figure out how to get my noise floor for +/-5Mhz or more below -30 dB ...03:46
wpwrak(mac address) the company_id will come back for 802.15.4, too03:48
wolfspraulhmm, overloaded03:50
wolfspraulusp did a wifi chip design?03:50
wolfspraulwho is usp? and why do you mention this?03:50
wpwrakwolfspraul: so it seems. how to procees ?03:50
wpwrakwolfspraul: usp = university of sao paulo (info from maddog)03:51
wolfspraulwhy should christoph laugh about your usrp2 work?03:51
wpwrak(usrp2) because it's particularly kafkaesque :)03:51
wpwrakin fact, kafka himself may have drawn inspirations from it :) all the instructions lead you in the wrong way, etc.03:52
wolfspraulwhere is the url of that wifi chip design project, is it properly (freely) licensed, and is it actually working?03:52
wpwrak(wifi) i still have to figure out these things ...03:53
wolfspraulunfortunately academia has degenerated to 'who can make the most outlandish claims in a paper with a straight face' competition, so I won't hold my breadth...03:54
wpwrakwel, there's a broad range for "wifi".  see that 802.11 module by microchip :)03:55
wolfspraulanswerin my own questions - no url, license 'not yet decided' (it's cool stuff, talking to qualcomm, indian government, south african government, and others), working: sure :-)03:56
wpwrakso i don't know what they really did. but then, from what i've seen so far, they're serious about their work.03:56
wpwrakprobably an abandoned project by now (my guess)03:57
wpwrak(gov) usp tend to have their finger on the pulse of the brazilian government ;-)03:57
wpwrakso my main concern wouldn't be in the quality of their work (in the case of the wifi) but what may have been lost in the telling03:59
wpwrakanother data point: ESA apparently put some SPARC core on some satellites. that core should be Free and implementable. might be a plan B for milkymist.04:04
wpwrak(plan B) sometimes, it's good enough to have a plausible plan B to advance with plan A ;-)04:04
wolfspraulesa has this leon stuff04:06
wpwrakthat may be the one04:07
wolfspraulbut how does it help with Milkymist?04:07
wpwrakit's an lgpl core whose non-open roots may be older than 20 years04:08
wpwrakmilkymist may still encounter IP issues04:08
wpwrakthe probability of encountering IP issues decreases with the availability of alternatives ;-)04:09
wolfspraulwhich ones?04:09
wolfspraulsure, but I'm not paying the fear tax04:09
wpwrak(quite drastically so, in fact :)04:09
wpwrakit's the antidote for the fear tax :)04:09
wolfspraulsince I don't pay the fear tax, I don't need the antidote either04:10
wpwrakheh ;-) but it's good to have a remedy for the taxman, too :)04:10
wolfspraulif you think there is an actual 'IP issue' we should look at it and clean it up sooner rather than later. and if we are not aware of any, that's it at that moment.04:10
wpwraki think it's a conditional ip issue. if lattice feel they should create some problems, they could.04:11
wpwrakif the feel it's pointless, they won't.04:11
wpwrak(or, of course, if they feel they should support what we're doing)04:12
wolfspraulanybody can 'create problems' if they feel like04:12
wolfsprauldo you really see a problem somewhere or just spending too much time looking at nothing?04:13
wolfspraulor you just can't believe that a company 'open sources' something, like Android?04:13
wolfspraulGoogle holding back some patents, for tax day?04:13
wpwraksure. my point is that the existence of the free and proven to be useful SPARC dramatically increases our BATNA04:13
wolfspraulcome on, we can spend all our lives discussing this stuff...04:14
wolfspraulI've looked at SPARC for a long time.04:14
wolfspraulmet with people, companies04:14
wolfspraulI'm sure Sebastien would know more from the tech side.04:14
wolfspraulI feel much better with Mico3204:14
wolfspraulSparc is used in military and space, quite a lot.04:15
wpwrakin his paper, he didn't mention that sparc variant. he just commented on another, much more complex one.04:15
wolfspraulthose are customers who are paying 10,000 USD and more per chip.04:15
wolfsprauloh the LEON project is very well known04:15
wpwraki don't have a problem with mico32. what i'm saying is that the possibility of using a sparc could make mico32 a more stable choice.04:16
wolfspraulI forgot all the details that led me to eventually join Milkymist.04:17
wolfspraula quick look at Wikipedia says LEON has a dual-licensing model and a commercial license if you want to embed it in a 'proprietary product'04:18
wolfspraulI would think, already from the distance, that there are many more IP issues there. Are they OK with someone manufacturing LEON in ASIC under the LGPL?04:18
wpwraksounds fair so far04:18
wolfspraulI'm always very careful about dual-licensing models.04:19
wolfspraulLattice is not doing that, they have an open core (totally open from their perspective), and proprietary enhancements around it.04:19
wolfspraullike Android04:19
wolfspraulLattice believes that the quality of their proprietary peripherals will make someone pay them anyway, then open core is just to get started (for them).04:19
wolfspraulnow - if we take the open core, and add open peripherals around it - that's our choice.04:20
wolfspraulI will try to talk to Lattice still one day, just not super high priority right now.04:21
wolfspraulyou can ask Sebastien what he thinks about LEON, I actually never talked to him about it.04:21
wolfspraulI met with several companies last year that are making military ICs around Sparc. Interesting. Had no idea what kind of businesses exist :-)04:22
wolfspraulbut once one chip costs 10,000 USD the whole calculation kind of changes, in terms of 'moq', tape-out fees, etc.04:22
wpwrakhmm, can't find clear information on the leon's licensing situation04:22
wolfsprauleven 100 chips is already a million USD then, can get a lot of stuff done :-)04:23
wolfspraulbut I can't work with these companies to make chips for consumer electronics, ever04:23
wpwrakbut leon is based on a very old sparc design. so there should be no relevant patents.04:23
wolfspraulfor the instruction set, yes04:23
wolfspraulfrom an IP perspective, there is nothing, absolutely nothing, wrong with Mico3204:24
wolfspraulit's as clean as it can get, comparable like I said maybe to something like Google's Android04:24
wolfspraulLattice has strategic motivations to open up this core04:24
wpwrakalso the sparc architecture is quite old by now :) well, perhaps missing a few years. but not much04:24
wolfspraulnot altruistic04:24
wolfspraulbut the opening itself is done in a clean and professional way, and a company is standing behind it04:25
wpwrak(mico32) again, i'm not saying that there's anything wrong with mico32. i'm just saying hat there may be a plan B.04:25
wolfspraulif there would be a patent issue against Mico32, I would expect Lattice to defend it, after all they have (proprietary) business attached to it04:25
wolfspraulSebastien mentioned several recently, didn't see LEON though04:26
wolfspraulwe need to ask him about LEON04:26
wolfspraulLEON at least is definitely real04:26
wpwrakwhen talking to people who think all you have is plan A, and they control plan A, that may give them ideas. having a plan B may eliminate these ideas.04:26
wpwrak(sebastien and leon) yup04:27
wolfspraulyes, but: Lattice has said this is an open IP, they themselves stand behind it.04:27
wolfspraulthat's as if Google would suddenly come out with a press release saying "sorry guys, we have some Android patents and everybody has to pay now"04:27
wpwraki was thinking of this more as a data point of possible use for your meeting :)04:27
wolfspraulLattice's business reason is the proprietary IP around the open core04:28
wpwrakis it the ip or the chips ?04:31
wolfspraulI think IP, but I don't know Lattice's business very well.04:31
wolfspraulthey probably license whole IP packages to customers, with their own open core, plus proprietary peripherals/enhancements.04:31
wolfspraulif you look at it that way you also see why they need an open core: to spread their design, which otherwise would be seen as totally obscure/proprietary.04:32
wpwrakthat would be good for us04:32
wolfspraulyes but I am guessing, I need to talk to some real people first04:32
wpwrakwell, there can be many reasons for "open". sometimes, the term gets twisted. but it's good if you can obtain clarification.04:33
wolfspraulwpwrak: wrong. Lattice mainly sells FPGAs and CPLDs04:39
wolfspraulthey seem to have not been doing very well in the last 10 years, if Wikipedia is painting a realistic picture my feeling is they had a lot of changes of direction.04:40
wolfspraul770 people04:40
wpwrakthat's indeed why i'm a bit concerned.04:41
wolfspraulso who knows when and why they started this Mico32 effort, and whether the people in charge of the strategy back then are even still there :-) Publicly there has not been much new stuff been released around Mico32 since 2008, I think.04:41
wpwrakthere's a certain tendency among "closed" companies to "open" things when they are already broken beyond repair.04:42
wpwrakso this could be that04:43
wpwrakor it could be somthing else, of course L(04:43
wpwrakL( = :)04:43
wolfspraulyes. so let's leave it there until someone actually talks to someone from Lattice.04:44
wolfspraulthe world is still driven by humans04:44
wolfspraulit's amazing what you can find out sometimes by just talking to people04:44
wolfsprauland we should quiz Sebastien about LEON. good point!04:45
wpwrakyup. sounds good to me. i just wanted you to know about the leon alternative, because sebastien dismissed all of sparc based on a different design.04:45
wpwrak(humans and talking) indeed :)04:45
qi-bot[commit] kyak: NanoMap: "skip downloaded tiles" is upstream. http://qi-hw.com/p/openwrt-packages/74176f004:46
wpwrakfor the talks with lattice, i would encourage them to call the "mico32" the "lattice mico32". i mean, who even recalls they still exist ? :)04:49
qi-bot[commit] kyak: NanoMap updated to Commit bfd639fd99e6e77cb58b9649a7ddb1c03068ef2d http://qi-hw.com/p/openwrt-packages/0a3c8a204:54
lekernelwpwrak, I did mention ESA's SPARC in my paper (2.2, §2)05:25
lekernelit's well documented, clean and probably bug-free; but slow and bloated05:26
lekernelif you want to talk to Lattice, there's Johnathan on the Milkymist list who's a Lattice employee05:33
lekernelhe subscribed/posted by himself, I did not talk to them05:33
kyakxiangfu: hi07:20
xiangfukyak: hi07:20
kyakshould i now change from openwrt_backfire to tracking_backfire?07:20
kyakgit pull says:07:21
kyak * [new branch]      tracking_backfire -> origin/tracking_backfire07:21
kyak * [new branch]      tracking_trunk -> origin/tracking_trunk07:21
kyakYour configuration specifies to merge with the ref 'openwrt_backfire'07:21
kyakfrom the remote, but no such ref was fetched.07:21
xiangfukyak: yes. you better do that.07:22
kyakwhy is it called like this?07:23
xiangfukyak: tracking_backfire is better then openwrt_backfire.07:32
xiangfuit's more apropos :)07:32
kyakall right07:35
wolfspraulkyak: no rename is complete with at least another, immediately following rename07:48
kyakwolfspraul: so true :)07:50
bartbesI still can't build gforth..09:48
bartbesstil the same error too:09:51
bartbesvm.push_string("GetAllSetNames"); vm.push_cfunction(get_all_set_names); vm.set_table(-3);09:51
bartbesnot that09:51
bartbesmake[3]: *** No rule to make target `kernlb.fi', needed by `gforth'.  Stop.09:51
bartbes(preceded by asm_fs not being found)09:51
qi-bot[commit] Wolfgang Spraul: added config.miminal http://qi-hw.com/p/openwrt-xburst/e19e45610:18
qi-bot[commit] Xiangfu Liu: config.full_system: add at, file, gsm-utils, i2c-tools, jpeg-tools http://qi-hw.com/p/openwrt-xburst/ea5d4e511:18
wpwraklekernel: ah, so that included the LEON. i thought you were talking about some later SPARC, as the LEON is supposedly simple. so it isn't quite that nice :(11:34
wpwrakwolfspraul: (renames) someone should collect quotes for the weekly news ;-)11:35
lekernelwpwrak, i talked both about opensparc and grlib/leon/esa sparc (which are the same project)11:56
wpwraklekernel: ah, i see. didn't realize they were the same. pity. would be nice to have more viable choices around.11:57
lekernelwell LM32 is quite nice, even if widely disregarded11:59
lekerneleven clones of crappy processors like the 6502 seem more popular11:59
wpwrak(6502) hehe, that's true :)12:00
lekernelbut I guess this is the hardware hacker equivalent of the "dancing pigs" problem of the security hacker12:01
lekernelalso it seems (to me) that some FPGA begineers tend to prefer crappy stuff because they think properly designed stuff would be too complicated for them12:03
wpwrakheh :) well, 6502 is arguably well-established. some may even have childhood memories :)12:03
djbclarkkristianpaul: sorry what was the question?12:05
wpwrakheh, another "gpios via mmap" binding. now we have C, LUA, Gforth, Python in the works.12:07
wpwrakhmm, how to design a DIY anechoic chamber for RF testing ... wikipedia says it can't be done cheaply :-(12:20
wpwrakor maybe just wait till late at night, when most RF generated in immediate response to human actions disappears :)12:20
qi-bot[commit] Werner Almesberger: Updated to do list. http://qi-hw.com/p/ben-wpan/d14e55b14:27
qi-bot[commit] Werner Almesberger: atusd/tests/psd*: first transmit power spectral density measurement http://qi-hw.com/p/ben-wpan/4c8117514:27
wpwrakDocScrutinizer: Q: when picking small ceramic capacitors, there are the general-purpose type and special RF caps. the parameters that are quantified look pretty much the same. so is there really that much of a difference ?16:20
DocScrutinizerpossibly yes16:21
DocScrutinizergeneral purpose may have unwanted resonances and whatnot, at RF freq16:21
DocScrutinizerwpwrak: (anechoic) I guess wikipedia is right16:26
Action: DocScrutinizer goes googling "dancing pigs"16:28
wpwrakDocScrutinizer: (resonances) hmm. wish they'd specify such things in way suitable for parametric matching.16:31
DocScrutinizerwpwrak: the parasitary effects aren't specified or tested for on components that are 'not for RF usage'. So they simply can't specify such things precisely, asd they may vary largely even on one lot of components16:37
wpwrakoh, i see. so it's really not good to pick the non-RF ones.16:38
lekernelhow high is RF?17:30
wpwraklekernel: up to 2.48 GHz17:31
osokurohi all.17:51
lekernelah, pretty high :)17:51
lekernelI have the dream of DIY MMICs17:51
lekernelthe microwave band (and beyond) is really hard to get with, and a lot of circuits are non-reproducible because they use very specific/hard to find/super-expensive components17:52
osokuroAny idea how sputtering differs from PVD or CVD?17:56
lekernelsputtering is pvd (afaik)17:57
osokuroAh, okay.17:58
lekernelwith CVD instead of evaporating the final material directly, you use for example a gas that decomposes itself (via a chemical reaction, hence the name) into the final material on your target17:58
lekernelthat's how I understood it17:59
osokuroAh! You've just taught me something. Thank you.18:00
osokuro(because I did not know the difference)18:00
lekernelthe main problem with all that off-the-shelf vacuum equipment is it's expensive as hell18:01
lekernelit's convenient, but it's just horribly expensive18:01
lekerneleven simple things like http://www.getupanddiy.com/projects/thin-film-sputtering-machine/images/voltage%20feedthrough2.JPG18:01
lekernelare about 300 euros18:01
osokuroThat's a high-voltage anode or cathode?18:02
lekernelit's just a stupid feed through, ie isolated wire that goes into the vacuum18:02
lekerneland you are supposed to spend 300 euros on each18:02
osokuroSo the real key is someone figuring out a cheap way to grind/lap the mating surfaces.18:03
lekernelyeah, with little outgasing and vacuum tightness18:04
lekernelmaybe making vacuum chambers out of glass (like CRTs) would be cheaper18:04
osokuroand then asking "do we really need to use difficult-to-process stainless steel or nickel alloys.18:04
osokuroOoo. Now there's a thought.18:05
osokuroNice thick borosilicate or maybe even soda-lime glass.18:05
lekerneland for feed throughs you can melt the glass and insert the electrode (the actual technique for acheiving a tight and reliable joint is a bit more complex though)18:05
lekernelbut they are virtually free, they only cost some gas, electric wire and a bit of time18:06
lekernelnot 300€18:06
lekernelI found a French manual from the 1920s that explains how to do :)18:07
osokuroAhhh. And the technique is applicable to very thick walls?18:08
lekernelhmm... if you can melt it, yes :)18:09
lekernelblowing glass is something I should learn someday, but so much things, so little time18:10
osokuroSounds like it's a good fit with your goals. Might be a priority.18:10
lekernelso is finishing milkymist, writing fpga synthesis tools and what not :)18:12
osokuroSpeaking of which, is MM seen as a potential replacement for the XBurst in a future nanonote-like platform?18:13
lekernelever seen that?18:13
osokuroHah! That's fantastic!18:14
lekernelyeah, Wolfgang thought about it, and maybe it'll happen in the future18:14
lekernelbut it's not a priority for me18:14
lekernelthis whole PWL website is about DIY vacuum tubes and neons18:14
lekernelwhen I get a chance I should travel to Warsaw and meet up with this guy :)18:15
lekerneli'm doing 25% of the road tomorrow btw :p18:16
osokurowhich road?=18:16
lekernelmoving from Paris to Berlin, then Warsaw is about 1000km more in the same direction :)18:16
osokuroAhhh. Then that's a long weekend!18:17
osokuroWe'll be well served in building out future in looking back at our past.18:18
lekernelhmm... less than that actually, only 600km from Berlin... but i've heard Polish roads are crazy18:18
osokuroI knew an operator of antique radio sets who died a few years ago. Made his own bakelite circuit boards from scratch. Wish I'd asked him how.18:19
lekernelthere is like a dozen antique radios lying in my mother's attic... I used to repair them when I was in high school18:20
lekernelor use the tubes for other purposes, like making pirate radio transmitters18:20
lekernelthat was fun18:20
osokuroI believe it.18:21
lekernelI could get them easily and cheap at local second hand markets18:21
lekerneleven bought a box of more than 100 various tubes in new condition for 15€ one day18:22
lekernelfrom an old repair shop that closed18:22
osokuroGuess that's why I'm hanging around here.18:23
osokuroNearly everything we've built is useful. Many people today forget that.18:23
osokuro(can anyone point me to simple instructions for flashing Debian to the nanonote?)18:28
qi-bot[commit] Werner Almesberger: Removed DTC123JE transistor files. (Was used only in 20100903 design.) http://qi-hw.com/p/ben-wpan/876a5ca23:31
qi-bot[commit] Werner Almesberger: Applied clock voltage divider fix and corrected too closely spaced via. http://qi-hw.com/p/ben-wpan/f17613b23:31
qi-bot[commit] Werner Almesberger: Added Digi-Key catalog data for automated BOM processing. http://qi-hw.com/p/ben-wpan/6fc656f23:31
qi-bot[commit] Werner Almesberger: bom/: automatic BOM generation (work in progress) http://qi-hw.com/p/ben-wpan/1a5d9f523:31
qi-bot[commit] Werner Almesberger: Corrected small footprint errors. Made 0R "inductor" a resistor. http://qi-hw.com/p/ben-wpan/5d28e1b23:31
--- Sun Sep 12 201000:00

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