2011-01-19 00:19 adamw_: heya ! question: for 4x4 mm qfn, is it okay if the center pad is just one square, with one square of solder paste, or should the solder paste be split into smaller portions ? 2011-01-19 00:21 wpwrak, this is good question! 2011-01-19 00:22 i knew you'd like it ;-)) 2011-01-19 00:22 either smaller portion or bigger than solder pad, it really depends on datasheet said firstly 2011-01-19 00:23 but datasheet always say recommended layout footprint ONLY. 2011-01-19 00:23 naw, data sheet says nothing ;-) 2011-01-19 00:23 they are not care more about 'solder paste' or 'solder pad' 2011-01-19 00:23 NXP do :) but they're the only ones, it seems 2011-01-19 00:24 but like in avt2 , the cpu die 2011-01-19 00:25 some chip manufactuer will very carefully on telling you the size of center 'solder pad', and define if connected to GND plate. 2011-01-19 00:25 so I could only tell you a normal spec. from pcb maker 2011-01-19 00:26 i'm talking about the AT86RF231 and the C8051F326. for the f326, the land pattern is given. central pad on the package is 3.15 x 3.15 mm, in the land pattern 3.25 x 3.25 mm (typ/median values) 2011-01-19 00:26 you could just split into at 5 mil nominal between pad and paste layer. 2011-01-19 00:26 for the rf231, i only have the package definition: 3.30 x 3.30 mm 2011-01-19 00:26 then you won't worry your design one day pcb maker won't produce it. 2011-01-19 00:27 so .. 125 um on each side ? 2011-01-19 00:29 just remember the clearance to be at least 5 mils : land pattern > 5 mil > central pad > 5 mil > solder paste. 2011-01-19 00:30 ah, okay. thanks ! 2011-01-19 00:30 so clearance please leave 0.127mm = 0.0254 * 5 mm 2011-01-19 00:31 hope this help you. 2011-01-19 00:31 i suppose, nobody worries about such things for other components. 0402 and such ? i.e., there, solder paste = pad size ? 2011-01-19 00:32 no ...it really depends on what kinds of product. some times always solder pad != pad size...more common!!! 2011-01-19 00:33 but if for your design, you could just use Milkymist foortprint for 0402, they are more common. 2011-01-19 00:34 that was I asked from pcb maker and smt vendors. 2011-01-19 00:35 oh, really ? hmm. ought to be a global setting then. 2011-01-19 00:35 i have had very care about the pad and solder paste on bga pcb design on bluetooth earphone before 2011-01-19 00:36 no, a said globe setting, we should always follow IPC specification, 2011-01-19 00:36 ah .. what does ipc say about 0402 ? :) 2011-01-19 00:36 wait...check for you. 2011-01-19 00:39 before you get digging into more precise suggestion on 0402 layout, please check http://en.qi-hardware.com/wiki/Milkymist_One_RC2_Layout_History 2011-01-19 00:40 scrolling down to see last 0402 picture I made stencil size. 2011-01-19 00:40 there's a link for http://www.xs4all.nl/~tersted/PDF_files/Plexus/tombstoning.pdf 2011-01-19 00:40 you should read this first. 2011-01-19 00:41 then you will know why people are caring about 0402 mounting. 2011-01-19 00:42 ah, at the very end :) 2011-01-19 00:42 then you could determine what the histories for good design on 0603/ 0402, 0805, or others. 2011-01-19 00:43 we've never produced over 3k products...you will hard to know the impact of bad 0402 footprint design. 2011-01-19 00:44 but I have met 1kk before for producing product....so they are important for MP. 2011-01-19 00:45 so I could only say to say with a 5 mil clearance, your board can be produced in all over the world. 2011-01-19 00:45 not just follow IPC ..:) 2011-01-19 00:48 for 0402, my pads are 0.6 x 0.45 mm. (0402-M from kicad-libs/modules/stdpass.fpd) 2011-01-19 00:49 would you also there make a 5 mil reduction ? 2011-01-19 00:49 the solder paste would then be about 0.35 x 0.2 mm. seems very small. 2011-01-19 00:50 0.6 x 0.45 mm is a central pad? or solder past size? 2011-01-19 00:51 that's the solder pad for 0402 2011-01-19 00:52 yes , 0.35 x 0.2 is too small 2011-01-19 00:53 please use http://en.qi-hardware.com/wiki/File:0603_0402_stencil_aperture_dimension.pngÂ Â for 0402 solder pad design. 2011-01-19 00:53 the recommended pad design I took from VIA for mother board industry. 2011-01-19 00:54 that's much larger than my pads ;) 2011-01-19 00:54 larger than your 0.6 x 0.45 mm 2011-01-19 00:54 my pad sizes come from the land pattern recommendations by various chip makers 2011-01-19 00:54 haha... 2011-01-19 00:55 hmm, i should have the analysis somewhere ... 2011-01-19 00:55 so what would you prefer? to let your board all over the world can use or few only? 2011-01-19 00:55 yes...analysis is good 2011-01-19 00:56 http://www.google.com/url?sa=t&source=web&cd=1&ved=0CBYQFjAA&url=http%3A%2F%2Fwww.ipc.org%2F4.0_Knowledge%2F4.1_Standards%2Fsmcwp001.pdf&ei=qno2TfTvFZHSuwOp8-GcBA&usg=AFQjCNFnNiZfoT5U1k2mKHb8coa4ZWgrFg&sig2=Ajhx23eW24Ijca3adbAN9Q 2011-01-19 00:56 check out this IPC for solder paste. 2011-01-19 00:56 you will get official answer. 2011-01-19 01:08 http://lists.openmoko.org/pipermail/gta02-core/2009-September/000563.html 2011-01-19 01:09 "My 0402" is a fairly tight version that's difficult to solder manually. since then, i've introduced a version that's 0.1 mm longer and is easily soldered manually, 0402-M 2011-01-19 01:10 aha....improvement from experienced life...quite good. 2011-01-19 01:10 ;-) 2011-01-19 01:11 here layout house have three catagories for 0402 footprint design.. 2011-01-19 01:12 i should ask them for sharing here....their experienced design was really confirmed by their much high quantities in customer end. 2011-01-19 01:12 some times to read more data like IPC is not bad. 2011-01-19 01:12 well, from manual soldering with an iron. reflow is a bit easier. i should build myself a proper reflow oven once. the toaster i use for this purpose has a hard time melting the solder but is really good at, well, toasting the FR4 (and oxidizing the traces in the process). not exactly how i want things ... 2011-01-19 01:12 but too many to learn. 2011-01-19 01:14 hmm..I have to run outside now. 2011-01-19 02:53 zrafa: gaah ! wireshark doesn't run. when loading its decoders ("dissectors"), it fails around "cisco" with Unhandled exception (group=1, code=6) 2011-01-19 04:19 wpwrak: did you try with a bit of swap on SD? 2011-01-19 05:42 Jay7, cool. stock OE or something else? 2011-01-19 05:42 stock OE 2011-01-19 05:43 but as I've said, I've not build any images for nn yet :) 2011-01-19 05:43 I'm building for zauruses mostly 2011-01-19 05:43 but DISTRO="jlime" and MACHINE="ben-nanonote" should be sufficient :) 2011-01-19 06:49 zrafa: (swap) nope, but there was also nothing in dmesg 2011-01-19 06:51 hmm, also jlime disables overcommit_memory. bad habits everywhere .... 2011-01-19 06:52 indeed, that helps. now at ipv6 ... 2011-01-19 06:53 ... ppp ... 2011-01-19 06:56 getting veeeerryyyy slooooowwww ... i think the ben really needs more memory :) 2011-01-19 07:15 and, eventually, the OOM killer got it. ah well, it's not that i really needed it ... 2011-01-19 07:15 ppp on the nanonote is too heavy? 2011-01-19 07:17 xMff: no, ppp probably isn't, but wireshark is :) 2011-01-19 07:18 ah... heh 2011-01-19 07:19 added pcap support to one of my wpan tools and wanted to show it off, but ... sneef 2011-01-19 07:22 maybe something like ettercap would be better suited? 2011-01-19 07:26 ah, nice. doesn't support IEEE 802.15.4, though. 2011-01-19 07:26 wow, not even tcpdump has it (at least not on my ubuntu host). how peculiar. 2011-01-19 07:35 [commit] Xiangfu Liu: bc, add -l, use the predefined math routines by default http://qi-hw.com/p/openwrt-packages/216706c 2011-01-19 07:40 jlime's tcpdump is better. heh :) gets very confused by my junk frames, though 2011-01-19 07:46 hmm, if a hypothetical ben successor had DDR SDRAM support (the Jz4720), then it could have 128 MB (32 Mb x 32) at about the same price as the current 32 MB 2011-01-19 07:46 s/Jz4720/Jz4720 doesn't/ 2011-01-19 07:49 the memory would also be much faster (4 x the burst rate. of course, setup eats some of this.) 2011-01-19 07:49 mm, 128 Mb, luxury 2011-01-19 07:50 kyak: get yourself a M1 ;) 2011-01-19 07:50 lekernel: i'll have to google that first :) 2011-01-19 07:51 what's that? 2011-01-19 07:51 http://www.milkymist.org/mmone.html 2011-01-19 07:51 ah, ok 2011-01-19 07:51 yeah, i could get it, but it seems like a very specific thingy! 2011-01-19 07:51 kyak: 128 MB seems to be around the "knee" in the price/capacity curve. anything smaller doesn't lower the price much, anything bigger grows proportionally in price or worse 2011-01-19 07:52 lekernel: so M1 is the official acronym, not MM1 ? 2011-01-19 07:52 wpwrak: are you talking about embedded systems? because desktop "knee" is perhaps around 2Gb 2011-01-19 07:52 kyak: in fact, it's the least specific thingy you could imagine ;-)) 2011-01-19 07:52 dunno, I use both interchangeably. M1 is faster to type and easier to pronounce 2011-01-19 07:53 kyak: embedded, yes. well, i didn't plot the graph. it just feels like that when i compare a few price points. 2011-01-19 07:54 wpwrak: specific thingy.. i'm not sure, because i can actually buy a desktop PC for that much money MM1 costs 2011-01-19 07:54 it's around 850 Euro, if i'm not mistaken 2011-01-19 07:54 you are mistaken 2011-01-19 07:55 oh! 2011-01-19 07:56 kyak: btw, be careful with GB vs. Gb. the capacity of memory chips is usually specified in mega- or giga-bits, Mb or Gb. so you get 1 Gb (32 M x 32), which means that it has 1 gigabits with the bus interface organized as 32 M words of 32 bits each. 2011-01-19 07:56 its 380.0Â¬ 2011-01-19 07:56 i'm sorry :) 2011-01-19 07:57 still, i bought 1.5 years ago Shuttle K45SE and it costs less than that 2011-01-19 07:57 and it's muuuuch faster 2011-01-19 07:57 that's why, MM1 is very specific 2011-01-19 07:58 it's much faster? 2011-01-19 07:58 uhm 2011-01-19 07:58 lekernel: ... compiling linux kernels ;-) 2011-01-19 07:58 the fpga on MM1, when properly programmed, beats any CPU core, even the high end ones 2011-01-19 07:59 it is, 2Gb DDR, E6500Â Â @ 2.93GHz Pentium etc 2011-01-19 07:59 lekernel: any core at any task ? 2011-01-19 07:59 lekernel: yes, fpga, so very specific :) 2011-01-19 07:59 kyak: mind your capitalization ! :) 2011-01-19 08:00 oh, well for texture mapping for example, it's certainly faster than that CPU core 2011-01-19 08:00 wpwrak: so tired of hittin ctrl+a with my left hand all day, lazy to hit Shift :) 2011-01-19 08:00 even the current design 2011-01-19 08:01 i would like to have some many-purpose device 2011-01-19 08:02 I wonder how you can argue that FPGAs are "very specific" :) 2011-01-19 08:02 and i think that MM1 is very limited-purpose.. like you said, it "can be programmed" 2011-01-19 08:03 but can it be programmed to do all tasks faster than regular PC? 2011-01-19 08:03 are you seriously comparing a FPGA with a general purpose cpu? if you really want to do that, you have to implement a similar cpu on the fpga and compare that to the cpu 2011-01-19 08:04 lekernel: btw, in terms of marketing, i have some doubts about the attribute "interactive". as i understand things, that's basically the key feature as far as a VJ is concerned. but i wonder if "interactive" is strong enough. after all, if it was just some macbook application where you have can manipulate the colors a little while the loop is running, that would be "interactive", too. 2011-01-19 08:04 going the "implement a similar CPU" way for this purpose is stupid and will definitely yield poor performance results 2011-01-19 08:04 of course it will, but comparin g an fpga with a general purpose cpu is at least as stupid as that 2011-01-19 08:05 using partial reconfiguration and novel hardware synthesis approaches is very likely to be a lot more interesting and will work if you do it right 2011-01-19 08:05 but it's a hard thing to do, and there are intermediate approaches like manually implementing dedicated acceleration cores, like milkymist does 2011-01-19 08:06 wejp: you may want to consider that many fathers think their kid is the best and most beautiful :-) 2011-01-19 08:06 wejp: in fact, you can compare them from the users' point of view. "What can i do with that, and how fast?" 2011-01-19 08:06 wpwrak: yeah, right ;) 2011-01-19 08:06 it's an interesting research subject, really 2011-01-19 08:07 i'm not saying FPGAs aren't worth using them, but they are totally different form a cpu, so comparing them isn't all that useful 2011-01-19 08:07 lekernel: (partial reconfiguration) oh yes. that's what i'm looking for 2011-01-19 08:07 oh, on the contrary it _is_ useful :) 2011-01-19 08:07 oh yeah? 2011-01-19 08:07 but you need a way more complex metric than GHz 2011-01-19 08:08 yeah, a silicon compiler that runs Python or Ruby extremely fast using PR on FPGA would be a wonderful thing 2011-01-19 08:08 i'd settle for C :) 2011-01-19 08:08 there is no such metric. you can compare sepecific tasks and how an implementation on a fpga compares to one on a specific cpu, but that's not comparing fpga and cpu but rather comparing different implementations of a specific problem 2011-01-19 08:09 wejp: its the question if one wants a synthetic or an real world comparison 2011-01-19 08:09 synthetic ones are just generating not much saying marketing numbers. 2011-01-19 08:10 i consider synthetic comparisons to be rather useless 2011-01-19 08:12 wpwrak: C would be a horrible language for this purpose 2011-01-19 08:13 [commit] Xiangfu Liu: update the qi-hardware sources mirror address http://qi-hw.com/p/openwrt-xburst/bf27ef7 2011-01-19 08:13 [commit] Xiangfu Liu: bash-files: update the package url http://qi-hw.com/p/openwrt-xburst/e3999a9 2011-01-19 08:13 [commit] Xiangfu Liu: add 127.0.0.1 BenNanoNote to hosts file http://qi-hw.com/p/openwrt-xburst/14981ac 2011-01-19 08:13 [commit] David KÃ¼hling: fix DNS timeout problem when applications resolve the FQDN http://qi-hw.com/p/openwrt-xburst/5299335 2011-01-19 08:13 [commit] David KÃ¼hling: undo hosts file fix.Â Â Xiangfu was quicker and did it better :) http://qi-hw.com/p/openwrt-xburst/a7c8d1f 2011-01-19 08:13 [commit] David KÃ¼hling: Adapt config.full_system to new emacs package name.Â Â Add emacs japanese http://qi-hw.com/p/openwrt-xburst/5337fe8 2011-01-19 08:13 [commit] Xiangfu Liu: update package to known-good revisions, thanks David Kuehling http://qi-hw.com/p/openwrt-xburst/c6eda58 2011-01-19 08:13 [commit] Xiangfu Liu: config.all_packages: remove mp3, mp4, rootfs, u-boot http://qi-hw.com/p/openwrt-xburst/698f0ec 2011-01-19 08:13 [commit] David KÃ¼hling: Add gfortran compiler support to the toolchain http://qi-hw.com/p/openwrt-xburst/57daa05 2011-01-19 08:13 [commit] Xiangfu Liu: configu.full_system: remove IB and Toolchain http://qi-hw.com/p/openwrt-xburst/1ea482b 2011-01-19 08:13 [commit] Xiangfu Liu: using 'strip' instead of 'sstrip' http://qi-hw.com/p/openwrt-xburst/672539f 2011-01-19 08:13 [commit] Xiangfu Liu: config.full_system, add math stuff and some new pakcages http://qi-hw.com/p/openwrt-xburst/0a3ed89 2011-01-19 08:13 [commit] kyak: put qi openwrt-packages git on top in feed.conf http://qi-hw.com/p/openwrt-xburst/c137e6a 2011-01-19 08:13 [commit] David KÃ¼hling: use ccache also for C++ compilation http://qi-hw.com/p/openwrt-xburst/1ce0cb7 2011-01-19 08:13 [commit] Xiangfu Liu: enable the toolchain option for Octave and libgfortran http://qi-hw.com/p/openwrt-xburst/536fd95 2011-01-19 08:13 [commit] Xiangfu Liu: config.full_system-add-more-php5-modules.patch http://qi-hw.com/p/openwrt-xburst/6e2a0c1 2011-01-19 08:13 [commit] Xiangfu Liu: config.full_system: add package sqlite2-cli http://qi-hw.com/p/openwrt-xburst/7872756 2011-01-19 08:13 [commit] Xiangfu Liu: feeds.conf add @revision to feeds.conf http://qi-hw.com/p/openwrt-xburst/da282ed 2011-01-19 08:13 [commit] Xiangfu Liu: INSTALL_FGORTRAN needs those two options http://qi-hw.com/p/openwrt-xburst/52c8c73 2011-01-19 08:13 [commit] Xiangfu Liu: config.full_system: add recently new packages http://qi-hw.com/p/openwrt-xburst/07b9ae1 2011-01-19 08:13 lekernel: how's that ? 2011-01-19 08:13 with all its pointers, side effects and ad hoc code everywhere e.g. for memory allocation, string handling, error reporting, etc. 2011-01-19 08:13 oops, some merge I think :-) 2011-01-19 08:13 lekernel: or is this just your way of saying that you don't like C ? :) 2011-01-19 08:13 and it's not even productive to write such code, compared to high level languages 2011-01-19 08:14 C is fine when you need fast code on a RISC CPU 2011-01-19 08:14 period 2011-01-19 08:14 seems that xiangfu's boss has returned ;-) 2011-01-19 08:15 rebase 'master' on upstream 'backfire', last upstream backfire commit date: [Mon Jan 17 02:15:27 2011] 2011-01-19 08:17 lekernel: i'd convert C to a more regular form, i.e., SSA, and then extract patterns from that. the fact that you have C on top doesn't really matter at that point, but you get "lean" operations without gazillions of dependencies and side-effects 2011-01-19 08:19 lekernel: i wouldn't be surprised if there was also research on alternative but similarly regular formats than SSA. after all, SSA is pretty old already and the work on obtaining efficient algorithms for the basic transforms seems to be more or less done. (haven't been following developments in the field for the last years, though) 2011-01-19 08:20 SSA has nothing to do with problems posed by unrestricted usage of pointers and ad hoc code for every basic operation 2011-01-19 08:22 C has the traditional CPU architecture built into its fundamental design 2011-01-19 08:23 and you want to move away from this traditional architecture 2011-01-19 08:25 xiangfu: please help me resolve this "Pull is not possible because you have unmerged files." after git pull -\ 2011-01-19 08:27 kyak: do you have un-commit modify ? 2011-01-19 08:28 xiangfu: there is one, and several untracked files 2011-01-19 08:31 kyak: I think it's will goto an temp branch. what is the output of "git branch" 2011-01-19 08:31 on master 2011-01-19 08:32 i think i understood you.. git branch temp; git checkout master? 2011-01-19 08:32 kyak: no. 2011-01-19 08:33 kyak:Â Â when do rebase the git will goto a temp branch first. so I guess if the git pull, will also goto a temp branch first. 2011-01-19 08:34 kyak: can you wait me some minutes. I found there some file should not commit, let me delete them first. 2011-01-19 08:34 not sure.. it could do it in one step 2011-01-19 08:34 sure 2011-01-19 08:35 i think i have problems mostly because i'm working on master branch.. maybe it's better to branch always, as explained in git tutorials 2011-01-19 08:36 then i can always update master, and then move my updates there 2011-01-19 08:37 when you do git pull , you can run "git stash" backup you un-commit first. 2011-01-19 08:37 git: you'll never know when it kicks you :) 2011-01-19 08:37 then after pull, run "git stash pop" 2011-01-19 08:37 yeah, i do it sometimes, when i have conflicts 2011-01-19 08:39 but in this case, dozens of "added" files appeared for soem reason 2011-01-19 09:01 kyak: too bad, there are some temp files commit. I need sometime cleanup it. 2011-01-19 09:03 kyak: for m1, I think we should not get stuck in fpga vs. asic thinking 2011-01-19 09:03 I hope we are able to establish milkymist for the long run, and build greats products with it 2011-01-19 09:03 then we have to dramatically lower the barrier of entry to use fpgas in free software - we leave the path to asic open whenever that makes sense anyway 2011-01-19 09:04 what I like about m1 is that it is a product, so we can skip all the theory talk and focus right away on making something that works, same as with the Ben 2011-01-19 09:05 it is not the fpga's fault that software tools to utilize its power have been developing so slowly, in free software it's even worse 2011-01-19 09:07 the tools are so bad, I think I can compare it to a normal let's say Intel CPU, and then all apps are run in Python. No C compiler all, let's say the C compiler for some reason only works for the kernel and the Python interpreter. 2011-01-19 09:07 then people run all their apps in Python, weather forecase, databases, everything. 2011-01-19 09:08 that would certainly grind down even the latest and greatest stuff, and it's what I think is being done to FPGAs currently 2011-01-19 09:09 I agree we need to get the price of m1 down, but first we need to make it a full product/package, then start to bring the price down. It will take some time... 2011-01-19 09:20 [commit] Xiangfu Liu: update the qi-hardware sources mirror address http://qi-hw.com/p/openwrt-xburst/efd9373 2011-01-19 09:20 [commit] Xiangfu Liu: bash-files: update the package url http://qi-hw.com/p/openwrt-xburst/b4d7537 2011-01-19 09:20 [commit] Xiangfu Liu: add 127.0.0.1 BenNanoNote to hosts file http://qi-hw.com/p/openwrt-xburst/aaaef18 2011-01-19 09:20 [commit] David KÃ¼hling: fix DNS timeout problem when applications resolve the FQDN http://qi-hw.com/p/openwrt-xburst/2a870e5 2011-01-19 09:20 [commit] David KÃ¼hling: undo hosts file fix.Â Â Xiangfu was quicker and did it better :) http://qi-hw.com/p/openwrt-xburst/38a5421 2011-01-19 09:20 [commit] David KÃ¼hling: Adapt config.full_system to new emacs package name.Â Â Add emacs japanese http://qi-hw.com/p/openwrt-xburst/d0e0372 2011-01-19 09:20 [commit] Xiangfu Liu: update package to known-good revisions, thanks David Kuehling http://qi-hw.com/p/openwrt-xburst/ef77ca6 2011-01-19 09:20 [commit] Xiangfu Liu: config.all_packages: remove mp3, mp4, rootfs, u-boot http://qi-hw.com/p/openwrt-xburst/5bbdc66 2011-01-19 09:20 [commit] David KÃ¼hling: Add gfortran compiler support to the toolchain http://qi-hw.com/p/openwrt-xburst/fae2ee8 2011-01-19 09:20 [commit] Xiangfu Liu: configu.full_system: remove IB and Toolchain http://qi-hw.com/p/openwrt-xburst/94ddef2 2011-01-19 09:20 [commit] Xiangfu Liu: using 'strip' instead of 'sstrip' http://qi-hw.com/p/openwrt-xburst/708d901 2011-01-19 09:20 [commit] Xiangfu Liu: config.full_system, add math stuff and some new pakcages http://qi-hw.com/p/openwrt-xburst/b0ccf2b 2011-01-19 09:20 [commit] kyak: put qi openwrt-packages git on top in feed.conf http://qi-hw.com/p/openwrt-xburst/60dec5b 2011-01-19 09:20 [commit] David KÃ¼hling: use ccache also for C++ compilation http://qi-hw.com/p/openwrt-xburst/a63d39d 2011-01-19 09:20 [commit] Xiangfu Liu: enable the toolchain option for Octave and libgfortran http://qi-hw.com/p/openwrt-xburst/8e334fb 2011-01-19 09:20 [commit] Xiangfu Liu: config.full_system-add-more-php5-modules.patch http://qi-hw.com/p/openwrt-xburst/752f820 2011-01-19 09:20 [commit] Xiangfu Liu: config.full_system: add package sqlite2-cli http://qi-hw.com/p/openwrt-xburst/ebf59b7 2011-01-19 09:20 [commit] Xiangfu Liu: feeds.conf add @revision to feeds.conf http://qi-hw.com/p/openwrt-xburst/b458f29 2011-01-19 09:20 [commit] Xiangfu Liu: INSTALL_FGORTRAN needs those two options http://qi-hw.com/p/openwrt-xburst/351a2b5 2011-01-19 09:20 [commit] Xiangfu Liu: config.full_system: add recently new packages http://qi-hw.com/p/openwrt-xburst/3adb62d 2011-01-19 09:33 wolfspraul: i would not wonder if the mm1 also sells well as a develboard 2011-01-19 09:34 wolfspraul: i compared some prices and actually all develboards seem to be more expensive (and have less interfaces) when they use the same, similar or bigger fpga. the only ones (fpga develboards) priced below 350us$ are the ones with spartan3 or really small spartan6 or similar sized chips 2011-01-19 09:38 wolfspraul: ok, let's think about MM1 as "FPGA computer". What i can do with it? It is positioned as a VJ station, and i don't really understand how i make use of it... 2011-01-19 09:41 hard to answer for me because I haven't use it much yet myself 2011-01-19 09:41 I like to work late at night, and I like to listen to music while working. 2011-01-19 09:41 so that's a starting point :-) 2011-01-19 09:41 once I have my own m1, I will definitely let it project something - what I don't know yet 2011-01-19 09:42 i think roh is right, it will be used as a dev board mainly 2011-01-19 09:42 it shouldn't be too distractive because I'm still working... 2011-01-19 09:42 I believe in software as the enabler. 2011-01-19 09:42 software has to get easier and easier - once that's the case people will drive it into more use cases 2011-01-19 09:42 as a 'dev board', well let's see I don't know 2011-01-19 09:43 I doubt I can understand the motivation of someone buying a 'dev board' per se. Need to know what comes next. 2011-01-19 09:44 I think a lot of people who buy a dev board buy it out of an almost automatic habit. 2011-01-19 09:44 for example I've heard this several times: 2011-01-19 09:45 company decides to use chip X for a new project, company buys one (or multiple, depending on team members) 'official' devboards for that chip 2011-01-19 09:45 they totally don't care about the actual features of the devboard, it's just a habit, a way of running a project 2011-01-19 09:45 in that case there is no chance to replace that devboard with an m1 2011-01-19 09:46 it gives them a baseline. basically completes the documentation :) 2011-01-19 09:46 Hi everyone ! 2011-01-19 09:46 yeah 2011-01-19 09:46 I have a technical question 2011-01-19 09:46 wpwrak: and often not used at all in the end. 2011-01-19 09:46 Does an emulator of the CPU used in the Ben Nanonote do exist ? 2011-01-19 09:46 also many projects get canceled, but by that time the devboards were already sold... 2011-01-19 09:46 wolfspraul: yeah. it's kinda like insurance :) 2011-01-19 09:48 LunaVorax: I think qemu works to a degree, but I don't know details maybe others do... 2011-01-19 09:48 Ok wolfspraul 2011-01-19 09:48 It's MIPS architecture right ? 2011-01-19 09:48 kyak: it is too early to come up with really convincing use cases. 2011-01-19 09:48 the VJ is one 2011-01-19 09:49 I am _sure_ we will come up with others, but it depends on us, how we get software to run on it, software as the enabler of use cases... 2011-01-19 09:49 then those use cases need to be marketed too 2011-01-19 09:49 kyak: have you seen this? http://visikord.com/ 2011-01-19 09:49 it's something that m1 probably could do relatively easily 2011-01-19 09:50 a nice usecase I think 2011-01-19 09:50 not seen 2011-01-19 09:50 bundle m1 with camera and projector, and you have a nice dance augmentation system... 2011-01-19 09:50 sounds cool.. only i 'm not a dancing type :) 2011-01-19 09:51 this is also guess-land, but I guess it would be easy to implement real-time video & audio compression, say into Ogg Theora, and then stream it over Ethernet 2011-01-19 09:52 maybe you can do this with other computers too, but you do need some 'computer' to do it, why not an M1 box... 2011-01-19 09:52 dmx and midi offer interesting and unique use case opportunities too, need to see which are the low hanging fruits there in terms of getting software ready 2011-01-19 09:53 roh had the idea to hack the video-in signals to what? a 3-channel scope? 2011-01-19 09:53 forgot 2011-01-19 09:54 kyak: hi seems now, everything is ok. remove duplicate merge commit. 2011-01-19 09:54 kyak: do you lose anything? 2011-01-19 09:57 wolfspraul: do you know if m1 would have the same customs tratamiento than nn?.. I mean, if I would buy one I would like to know if it would be the same from a custom point of view (I know that I should know better than you perhaps, after all, I live here :P .. but maybe you already know some m1 features customs would like to use for disturbing buyers) 2011-01-19 09:58 wolfspraul: tratamiento=treatment :P 2011-01-19 09:58 zrafa: the power supply 2011-01-19 09:58 m1 coult be more complicated depending on where 2011-01-19 09:59 e.g. the us seems to think: 'powerfull fpga == weapons grade technology' ;) 2011-01-19 09:59 wpwrak: ah :( .. but we could ask to remove that from package before shipping right? 2011-01-19 10:00 zrafa: that is, unless wolfgang got the http://imagenes.acambiode.com/empresas/5/5/1/3/55135020082167565066496756554557/productos/Etiqueta Marca S.gif 2011-01-19 10:01 zrafa: yup. that's how i solve this sort of issue. did it twice already successfully. (not sure if they would have caught the power supply if it had been there, though.) 2011-01-19 10:01 roh: (weapons technology).. ah.. that will not happen here I guess (Arg). They will think "what is fpga?" 2011-01-19 10:01 roh: well if they look in google, maybe find the us customs thinking "it is weapons grade technology" :P 2011-01-19 10:03 zrafa: i am in germany and had customs hassle with getting a beagleboard from the us 2011-01-19 10:04 i don't think weapons are a problem at the moment. right now. drugs are the thing they have to watch out for. at least until the latest scandal (about a ton of cocaine airlifted from argentina to spain) has calmed down a little :) 2011-01-19 10:04 in the end i gave digikey the finger and used different hw 2011-01-19 10:04 kyak: if you don't lose anything, we can do 1. git fetch -aÂ Â 2. git reset --hard origin/master, for force local master sync with server 2011-01-19 10:04 if i can avoid it i dont buy anything from the us. its just pain. the rest of the planet is easy compared to that. 2011-01-19 10:05 roh: ah, I see. 2011-01-19 10:05 wpwrak: the tunesian ex-first lady supposedly airlifted about 1.5tons in gold out of the county while fleeing ;) 2011-01-19 10:06 wpwrak: yeah.. and all for that tiny cocaine package for personal use 2011-01-19 10:07 hrhr 2011-01-19 10:07 zrafa: we can ship it without the power adapter 2011-01-19 10:07 then we can ship the power adapter separately, see whether it passes, or you get the power adapter locally 2011-01-19 10:07 you are still in Argentina, right? 2011-01-19 10:07 roh: yeah. private cargo flights seem to be popular these days ;-) 2011-01-19 10:08 wolfspraul: yep, Argentina 2011-01-19 10:08 wolfspraul: and the power adaptar should be easy to get here I guess 2011-01-19 10:08 sometimes i wonder if there should be a wiki documenting such oddnesses 2011-01-19 10:08 wpwrak: I have never seen that logo anywhere (watching for it now), but it's mostly a printing problem :-) 2011-01-19 10:08 Wasn't there a Nanonote2 wishlist or something, I though I saw one some weekd ago 2011-01-19 10:08 like 'open logistics and sales'-wiki 2011-01-19 10:10 wolfspraul: :-) some of the big names have it. ibm, asus, fujitsu, etc. 2011-01-19 10:12 LunaVorax: i wonder if the "nanonote2" (officially "ya") is really still on the roadmap. it's been a long time since i heard about any plans for that. 2011-01-19 10:13 wpwrak: i think thats a more difficult question. 2011-01-19 10:13 Too bad if it has disapeared, the nanonote lacks some important features (like video out and usb host) 2011-01-19 10:13 there is the want. the could... the experiences and then the opportunies 2011-01-19 10:14 the last being a combination of money and avail materials (chips which one can buy not only talk about (etc)) 2011-01-19 10:15 LunaVorax: video out? whatfor? 2011-01-19 10:15 well vga out to plug on a bigger screen :o 2011-01-19 10:15 m-) 2011-01-19 10:15 I was working for a mod to turn the nanonote into the board of an arcade cabinet 2011-01-19 10:15 will not happen. waste of money (and mechanically unfeasible) 2011-01-19 10:15 But the lack of video out stopped me 2011-01-19 10:15 Like, real quick 2011-01-19 10:16 roh: yeah, basically no growth in development resources, and i suppose neither in finances 2011-01-19 10:16 i built a mame-o-mat (arcade machine mechanics with a pc inside) once.. running linux and mame and hundreds of games 2011-01-19 10:17 Yeah roh my idea is a bit more complex 2011-01-19 10:17 Maybe not the right place to talk about it 2011-01-19 10:17 real badass-hack. running the mame on fbdev, having someÂ Â menu infront and shell hacks around that. using a second keyboard pcb connected to the mechanical buttons 2011-01-19 10:18 But if I had the knowledge for it I would have took the nanonote's schematics and modifieds them just to add/remove the features I wanted 2011-01-19 10:18 LunaVorax: now here's your chance to learn :) 2011-01-19 10:19 the point is: the nanonote is a low end multimedia computer (low end because its <1ghz cpu, no hw-accel gpu etc) 2011-01-19 10:19 I think it's starting way too high for a beginning wpwrak :P 2011-01-19 10:19 but thats ok. anything high end would suck the batteries dry asap 2011-01-19 10:19 roh, not a problem 2011-01-19 10:19 LunaVorax: for video out, there are chips that can do basic vga out. e.g., the propeller. you would be extremely limited in what you can display, but if you connected it to the ben, you could, say, have a text screen with relative ease. 2011-01-19 10:20 wpwrak, I was thinking about graphics actually 2011-01-19 10:20 I guess that's a different story then 2011-01-19 10:20 LunaVorax: make small steps. build the ben-blinkenlights board. then add an mcu that blinks on its own. then make it talk to the ben. and so on. 2011-01-19 10:21 I should make tinyer steps wpwrak since I understood 50% of your sentence 2011-01-19 10:21 :D 2011-01-19 10:21 But I'm much into learning programming for now 2011-01-19 10:21 And messing around with my arduino stuffs 2011-01-19 10:22 More advanced electronics may (or not) happend later 2011-01-19 10:22 LunaVorax: ben-blinkenlights is a rather simple board: http://www.almesberger.net/misc/ben/blinken/blinken-100831.jpg 2011-01-19 10:23 Oh right, 2011-01-19 10:23 Oh ! 2011-01-19 10:23 LunaVorax: ah, arduino. that's already waaaay more advanced ! 2011-01-19 10:23 So that was that strange picture on the wiki 2011-01-19 10:23 ;-)) 2011-01-19 10:23 arduino is advanced ? 2011-01-19 10:24 LunaVorax: you can do a lot with it. of course, it has its limits, but you're not starting from zero 2011-01-19 10:24 Yeah 2011-01-19 10:26 wpwrak, what is the purpose of this ben-blinkenlights anyway ? To command them using the Ben Nanonote ? 2011-01-19 10:26 I think I anwered my own question 2011-01-19 10:26 answered* 2011-01-19 10:27 yes. the real purpose was to put the 8:10 card slot to use. proof of concept that we can indeed drive something there. 2011-01-19 10:28 this project then spawned a few more advanced designs: a passive cable/board combo to program the c8051f32x family of microcontrollers and then the atben board of ben-wpan 2011-01-19 10:30 Hum ok, pretty unexpected for me 2011-01-19 10:30 illustrates the process of small steps: first do something trivial to make sure the basics (signals, software, mechanical) are okay. then build more ambitions things on it. well, ben-wpan is admittedly a fairly huge leap, because this also adds the whole complexity of RF. 2011-01-19 10:31 wpwrak: btw: atmel solves the 8051-misery finally 2011-01-19 10:32 atmega8U2 32U2 16U2 series is also in the 2.0-2.30US$ region like the c8051f32x 2011-01-19 10:32 and it also got real usb, dfu etc 2011-01-19 10:33 roh: usb sans crystal ? 2011-01-19 10:34 dunno. i think yes. it has an internal 8mhz rc which it also can use to generate the 48mhz via usbpll 2011-01-19 10:35 my point is mostly not details, but the part that the reactions i get from embedded hackers when it comes to 8bit cpus and 8051 (from my pov understandable) are different but all in the region of *run 2011-01-19 10:35 * 2011-01-19 10:36 roh: http://www.atmel.com/dyn/resources/prod_documents/doc7799.pdf section 20.2, page 185 says you still need a crystal 2011-01-19 10:36 *eeek* *yikes* .. stuff like that. avr most of them like.. pic most of them also despise now 2011-01-19 10:36 roh: well, sdcc isn't the nicest compiler, that's true. but 8051 isn't *that* horrible. 2011-01-19 10:37 true. still its a deadend nobody wants to even be near. (the arch) 2011-01-19 10:37 cisc. ugly to read. complicated to do simple stuff 2011-01-19 10:38 yeah, everybody uses it. the obvious hallmark of a deadend ;-))) 2011-01-19 10:38 bad clockcycles vs getting stuff done ratio 2011-01-19 10:38 nice single-bit i/o operations 2011-01-19 10:38 so what. every serious mcu has that 2011-01-19 10:38 even the z80 clones 2011-01-19 10:38 peripherals suck a bit, though. but i'm not sure how much of this is the 8051's fault. 2011-01-19 10:39 single-bit i/o ? no, not really 2011-01-19 10:39 still pain nobody wants to see again. (peripherals from that time) 2011-01-19 10:40 most had theses mcs51 learning computer experiences in school. 2011-01-19 10:41 well, the 8051 arch doesn't bother me so much. the sdcc compiler bugs are a bit of a nuisance, though. but also that got better. 2011-01-19 10:41 single bit iolike sbi cbi? or what do you mean? 2011-01-19 10:41 i don't know the assembler. but that sounds about right, yes 2011-01-19 10:41 set/clear bit 2011-01-19 10:41 wpwrak: i think you can understand why people avoid such cpus after having worked with avr and similar post 1980 archs 2011-01-19 10:42 btw: sbi and cbi are actually used by gcc in avr in some situations. means the gcc already does quite some cool optimisations 2011-01-19 10:44 i've worked with z80, 6502, 8086, ia32, pic, m8c, avr, arm, had some peeks at sparc, mips, vax, and propeller, and i don't find the 8051 overly repulsive 2011-01-19 10:44 maybe some people are just sissies ;-) 2011-01-19 10:44 about the crystal.. whats the issue there? 2011-01-19 10:44 space, cost, yield 2011-01-19 10:44 even for a serial one needs one to have not shitty baudrates 2011-01-19 10:45 the usb host gives you an accurate clock. use it :) 2011-01-19 10:45 only as long as its plugged. its usb. it doesnt mean permanent connection 2011-01-19 10:46 often enough, it does. but yes, if you unplug, the scenario changes. 2011-01-19 10:46 if your device gets bus powered only thats correct. still you need some special purpose pll for that inside the chip 2011-01-19 10:47 sure. but the problem has been solved. that is, unless you insist on avr. then you still have to wait until atmel figure it out, too ;) 2011-01-19 10:47 i dunno how that works in detail.. even every usb hub i know has a crystal. and every serial2usb also 2011-01-19 10:48 (ftdi have also caught up at long last. of course, their is the evil kingdom of secrecy, with the predicable result on the quality of free software support.) 2011-01-19 10:48 wpwrak: my point mostly is: with 8051 you are riding a dead horse and nobody wants to hop on and help out/send patches develop extensions, reuse code etc. thats what opensource is all about. (community of developers) 2011-01-19 10:48 with avr you are sitting in a melting pot of people who are motivated to do stuff. or abuse your hw with different fw to stuff we never thought about 2011-01-19 10:49