#qi-hardware IRC log for Saturday, 2012-12-29

qi-bot[commit] Werner Almesberger: libubb/swuart.c: don't open/close UBB (master) http://qi-hw.com/p/ben-blinkenlights/661420900:01
qi-bot[commit] Werner Almesberger: swuart-chat/chat.c: open/close UBB explicitly (tracking API change) (master) http://qi-hw.com/p/ben-blinkenlights/ace55ff00:01
qi-bot[commit] Werner Almesberger: lpc111x-isp/: Ben-based in-system programmer (ISP) for NXP LPC111x chips (WIP) (master) http://qi-hw.com/p/ben-blinkenlights/eea048400:01
qi-bot[commit] Xiangfu: cgminer: enable all IP allowed to access the API (master) http://qi-hw.com/p/openwrt-packages/fddf50a03:27
wpwrakpity that wolfgang isn't at EHSM. might have been an opportunity to pimp what qi-hw is doing.10:34
Fallenoua qi-hw stand at ehsm would have been great :)10:39
wpwraki was more thinking of trying to see if, say, CERN could be interested in sponsoring/contracting/... some of out activities10:47
wpwrakso that's less a thing you do at a booth but a thing you discuss over lunch or dinner10:48
wpwrakon the hardware side of qihw, we have a bunch of things, but all of the recent stuff is fairly scattered. so the project isn't quite in a quite booth-worthy state10:50
wpwrakhmm, i'm getting good at that stuff. lately, all the little boards i make just work, maybe except for some minor soldering problems.11:27
wpwraknot so long ago, it took me a number of spins just to kill the severe design flaws11:28
wpwrak(right now: nxp lpc111x mcu and a DC-DC converter. admittedly none of this rocket science, but still)11:28
wpwrakthree more major subsystems in this design, though: memory card, a fancy switch, and an acceleration sensor. card holder and sensor are new, the switch i've used before.11:30
kyakpractice makes perfect, they say11:34
larscYou may want to think about a change in career if you get worse at something the more you do it ;)11:37
wpwraknaw, that means that you've reached nirvana11:38
qi-bot[commit] Werner Almesberger: lpc111x-isp/lpc111x.c: restructure code (WIP) (master) http://qi-hw.com/p/ben-blinkenlights/548757811:50
qi-bot[commit] Werner Almesberger: libubb/swuart.c: use mlockall to prevent page faults with interrupts off (master) http://qi-hw.com/p/ben-blinkenlights/e4f8b9711:50
qi-bot[commit] Werner Almesberger: lpc111x-isp/lpc111x.c: option -v to control verbosity level (master) http://qi-hw.com/p/ben-blinkenlights/c4d64fa11:50
qi-bot[commit] Werner Almesberger: lpc111x-isp/lpc111x.c: add printf-style dialog() variants (master) http://qi-hw.com/p/ben-blinkenlights/0553fed11:50
qi-bot[commit] Werner Almesberger: lpc111x-isp/lpc111x.c: flesh out device identification (master) http://qi-hw.com/p/ben-blinkenlights/90d39d311:50
larscwhitequark: you are famous :)16:44
kristianpaulwpwrak: time to get into ceramics? :-)16:48
wpwrakkristianpaul: hmm, ceramics ?17:02
kristianpaulor was other tech you were interested?17:04
kristianpaulanyway have fun as always :-)17:05
kristianpauli really like those tiny boards and software tools you made for then 17:05
wpwrak(other tech) ah yes, quantum-chemistry :) still not quite sure how to get into that. it seems the more advanced the tech, the messier the lab17:09
wpwrak(tiny boards) thanks :) using anything specific ?17:10
kristianpaulnope using but i refer a lot to it to friends that get in to electronics using free software17:13
kristianpauldamnm there is a way to rever a git stash drop ?17:14
virickristianpaul: does it appear in the reflog?17:16
kristianpaulviric: hi17:17
kristianpaulviric: how i do see that?17:17
kristianpaulgit log ?17:18
viricgit reflog17:18
virichello kristianpaul 17:18
kristianpaulnope nothing17:18
virickristianpaul: the bottom of 'man git-stash' includes a section about recovering drops17:19
kristianpaulahh yeah17:22
Action: hozer pings wolfspra1l18:44
hozeris qi-hardware an actual company, or more the idea of community-driven open source hardware?18:45
hozerI would like to hear some details on how you got fpgatools working on the lx9 part (and figure out how I can get it to work on an lx25 part)18:47
kristianpaullx25 is BIG18:52
hozeryes ;)18:52
hozerwell, I have some xc3s200's, but those are a different architecture18:53
hozerI (in theory) have a bitstream of the Leon-sparc core compiled for an LX2518:53
hozermy goal is to be able to compile some usefull CPU design to a bitstream using fpgatools18:54
wpwrakhozer: if you want to help, perhaps you could get/make an lx9 board, familiarize yourself with the existing work, then try to explore the lx25. the structures should be very similar but it's reasonably certain that it'll take a lot of time to map them.18:55
wpwrakand qi-hw is idea/community, no company18:55
kristianpaulqi :-) yeah18:57
kristianpaulhozer: thats quite big 18:57
kristianpaula leon-sparc wow18:57
kristianpauleven in verilog is BIG18:57
kristianpaulfor me at least18:58
kristianpaulindeed wpwrak, hozer could try that lx9 and the blinking led 18:58
hozerwhere can I click 'order now' for an lx9 board with a LED to blink ;)18:58
kristianpaulhum well wait and ask wolfspra1l :-)18:59
kristianpaulhozer: are you good with floorplanning?19:00
hozerwpwrak: I'm getting ideas in my head now to write some code to load random bitstreams and see what happens, and automate it19:00
hozerkristianpaul: give me some code and I could figure it out. The problem I have with fpgas is the awful proprietary tools19:00
kristianpaulthey are not that awful at all i think19:01
kristianpaulsee xilinxs there are lots of docs19:01
hozerthere's also 19GB of crap polluting my disk just so I could compile for a single fpga19:02
kristianpauland tools for lots of specific things like modifying a bitstream for example19:02
hozerand I'm pretty sure the mapper/floorplanner/whatever are all single-threaded19:02
hozerkristianpaul: some of my venom for Xilinx comes from using their tools in 1998, when they were windoze-only19:03
hozerat which point I decided FPGAs will never take off until there is a full-open source toolchain like GCC to go from a high-level languate all the way to bitstream19:04
hozerhere's a chip layout question for you: Why does everything have to be a 90 degree angles19:06
hozerWhat prevents someone from layout out an FPGA with hexagonal cells19:06
hozererr, what prevents someone from designing an FPGA layout where the LUTs/elements/etc are in a hexagonal honeycomb packing?19:07
wpwrakmaybe because the old algorithms would break and nobody knows how to make new ones ;-)19:08
kristianpaulhow old are those IC primitives?19:09
hozerkristianpaul: you mean all the right angle stuff?19:10
kristianpauldont know i'm very aware of IC 19:11
kristianpauland CAD related to it19:11
hozerkristianpaul: apt-get install alliance19:12
wpwrakhozer: (order now) xiangfu has the design for an lx9 board, which i think is the same wolfgang is using. not sure how DIY-friendly his current work is, though. he's been experimenting with BGAs ...19:12
hozeryou can try laying out some transistors19:13
kristianpauli'm not*19:13
kristianpaulwpwrak: you dont like BGA dont you? :-)19:13
kristianpaulwell is not so DIY friendly PCB making yes19:14
hozerI would rather pay someone else to burn BGAs and send me a board and a blinky-led bitstream19:14
kyakexcuse me, are you talking about a DIY FPGA board that is based on lx9 and that can blink leds as a proof that it actually works?19:16
wpwrakkristianpaul: i dislike the cost of industrial SMT. and i like a reasonable yield. so yes, no fan of BGAs :)19:16
wpwrakkyak: that would be the idea19:17
wpwraksort of a starter kit19:17
wpwrakof course, fpgatools for lx45 would be nice. we could use them on milkymist :)19:18
kyakwpwrak: and it's different from starter kits by xlinix because it's going to be open source?19:18
wpwrakthat is, unless there are bugs that burn up the chip. we still don't know if that can happen in real life.19:18
whitequarklarsc: how am I?19:18
wpwrakkyak: existing starter kits may be an option, yes. depends on whether the comm protocols are suitable, whether their price point is okay, and so on.19:19
hozerwpwrak: you use lx45 for milkymist?19:19
wpwrakwhitequark: you should start with questions like "where am i ?" and "what year is this ?"19:19
hozerhow many MGT's are there on the chip you use19:20
whitequarkwpwrak: was replying to "20:44 < larsc> whitequark: you are famous :)19:20
wpwrakhozer: M1 has an XC6SLX45-2FGG484C19:20
whitequarkbesides, asking about the year could soon be quite relevant.19:20
wpwrakwhitequark: yeah, if the party picks up speed early :)19:20
kyakwpwrak: and what is it going to be at the end of the day? i mean, it's a starter kit for something?19:20
wpwrakkyak: access to (previously) secret knowledge :)19:21
hozerwpwrak: any chance I could talk you into doing a board revision that some multi gigabit transcievers brought out to an Infiniband CX4 connector?19:21
wpwrakhozer: not really :) maybe you could talk sebastien into such things, though19:22
hozerkyak: It's a starter-kit for FPGA development with fpgatools19:24
wpwrakwhitequark: may take a while until lars is back. he should now be in the process of getting drunk with the other ehsm folks19:24
whitequarkwpwrak: yeah, quite sad actually I didn't make it to ESHM19:24
whitequarkfirst I was busy getting my intl' passport back, then some stupid bureaucraticc crap tried required me to stay here19:25
kyakhozer: i see19:25
whitequarkand actually I didn't need that19:25
hozerregarding blowing up chips with bitstreams: I'm quite sure you can configure an LX45 with gigabit serdes to make fire19:25
kyakis fpgatools a reverse-engineering attempt or it is documented and it is some kind of redesign of xilinx proprietary toolchain?19:26
larscwhitequark: your blogpost about the rpi was mentioned in one of the talks19:26
hozerkyak: from-scratch fpga toolchain19:27
kyakyeah, i'm looking at it. But still not clear19:27
hozerit looks like cleanroom reverse-engineering19:28
wpwrakkyak: it's reverse-engineering of the bitstream semantics, and then some minimalist toolchain that goes on top19:28
wpwrakkyak: alas, not much documentation19:28
kyakwpwrak: i see.. probably stupid question, but is it legal?19:29
wpwrakkyak: i don't see why it wouldn't be19:29
hozerthat probably depends on if you've accepted the Xilinx tool bs end-user agreement19:29
hozerif you have never used (or at least removed all copies of) the tools of the beast, the beast has no power over you :P19:30
hozerthis means that all I can do is throw money at the problem and contract someone else to do the reverse engineering :P19:30
kyakok.. from the other hand side, why is xilinx's toolchain closed? They sell hardware, not the software. Or is it "just because"?19:31
larsckyak: they sell the software19:31
larscthe tools are key19:31
kyakoh, they do.. ok19:31
hozerthe tools hide the hardware bugs19:31
kyakheh, that's nice :)19:32
larsckyak: there is a free version of the tools, but it has a limited set of features19:32
kyakso it means everyone who plans to base their design on xilinx FPGAs, must also buy their software.. very smart19:32
hozercan the webpack (free version) compile the milkymist SOC?19:32
hozereven then, I think the software is still a loss-leader for the hardware19:33
kyaklarsc: is it the same with altera?19:33
hozerthe only business reason that makes sense to me is get designed sucked into 'your' tools so they always choose your fpga19:33
hozerdesigners get sucked into , I mean19:33
kyakit's pretty strange to me anyway, because in MCU world it seems a little bit different.. For example, TI officially supports gcc and Makefiles for some of their DSPs19:34
wpwrakkyak: in the fgpa world, there's a big conspiracy to keep things secret19:35
wpwrakkyak: actually, no everyone is in the cabal. e.g., the cypress psoc 3/5 also have some clpd-like blocks in there, and that one is open.19:36
kyakwpwrak: so this is the secret knowledge you refered to :)19:36
wpwrakaye :)19:36
wpwraknow, the next step would be for them to encrypt the bitstream19:37
wpwrakthat would attract all the crypto hackers, who'd see this as a highly welcome challenge19:38
whitequarkisn't it that Xilinx's bitstream is currently fully documented in the datasheet?19:38
wpwrakand so on. we'll see how far that arms race will go. of course, the sillier the lock-in protection gets, the more attractive it will be to leave the cabal19:38
kyakwpwrak: it sounds like you are almost waiting for that :)19:39
wpwrakwhitequark: at some level, yes (at least i think so)19:39
whitequarkhozer: bitstream encryption is not the same as lock-in19:39
whitequarkyou do encryption to avoid reverse-engineering by your competitors, no?19:39
hozerwpwrak: cypress... hrrm.. they have a fab a 15 minute train ride away from me....19:40
whitequarkit's not like xilinx encrypts something trying to prevent you from developing your own translator19:40
wpwrakkyak: seeing someone repeat other's mistakes reaffirms my faith in humanity :) but yes, xilinx could do the smart thing and just accept the new situation, or even embrace it. depends a bit on how open-minded they are.19:42
hozerum, is there an open-source DFSG-compliant tool that I can program the Cypress PSOC's with??19:43
kyakwpwrak: they and altera own the market, so they probably don't care, since everybode is already in cabal19:43
wpwrakhozer: the psoc 5 have some arm core19:44
wpwrakhozer: the psoc 3 are 8051, so sdcc is your friend19:44
hozersdcc is nice, I'm more interested in the stuff to program the analog gadgets 19:45
wpwrakno major witchcraft required :)19:45
wpwrakwell, there are registers your code can set ... :)19:45
wpwraklast time i checked, the documentation was very detailed, if a bit overwhelming19:46
hozerfrankly, I'd rather layout my own analog chip than screw around writing the code needed to set the registers the right way19:47
wpwraktheir application notes are probably all based on their IDE, so they're of limited use. they may still explain the underlying principles, though.19:47
hozerso unless Cypress is going to make http://www.cypress.com/?rID=41083 open source, they just lost my business19:47
hozeralthough I still might approach the fab and see if they'll make my chip isntead19:47
wpwrakright. i also prefer setting up my own cattle farm instead of going to the restaurant around the corner and order a steak :)19:48
wpwrakjust ignore that psoc designer19:48
hozerI can send you some soymilk if you want19:48
hozerwpwrak: I'd rather have someone else raise the cattle, but I'll buy from a small local producer and not McDonalds19:49
wpwrakbah. i'm right in the middle of the planet's GMO soy empire ;-)19:49
hozeror brazil19:49
wpwrakwell, logistically the middle19:49
wpwrakbuenos aires, argentina19:49
hozerhow is the crop going, anyway...19:50
hozerso... total subject change... is anyone talking about open-source drones for precision farming applications down there?19:50
hozerit's hilarious that you bring up the farming analogy... One of my long-term goals is to be able to fabricate silicon on my farm19:51
wpwrakit got a little wet this spring. we'll see what that does. last season was very good. and the country needed that, because they completely screwed up the fuel industry. so basically all the record earnings from soy exports went right into fuel imports.19:51
hozerbecause right now I can fix up 20 year old farm equipment for 5% of the cost of a new John Deere and it does 95% of what the new thing does19:52
hozerand in 20 years I'll need to replace random CAN controllers that nobody makes anymore19:52
whitequarkreminds me of how wpwrak replied when I posted a link to an electronic caliper19:53
wpwrakhozer: (drones) hellekin may have heard of such things19:54
hellekinhozer: I never heard of such an application yet, but I guess it's a good question for open source ecology.19:58
hellekinthat application would change from https://twitter.com/dronestream19:58
wpwraknice. reverse-feeding dc-dc converter draws only some 50 uA. so no problem with setting the board up for programming (via the ben, of course)20:02
hellekinhozer: I asked on #hackerspaces (where brimstone came from) and [thor] there knows about someone using GIS to monitor their farm. Looks like you could go there and have a chat :)20:05
wpwrakhmm. seems that this year isn't destined to end well :-(20:24
hellekinwpwrak: why?20:25
wpwrakfirst my dc-dc converter developed a short under a heavy load20:25
wpwrakthen, when trying to remove the regulator chip, the air pump of my soldering station stopped working20:25
wpwraklet's see if it just need some cooling down. though it never did that before ...20:26
wpwrak(heavy load) well, still only about 1/5 of the rated maximum output of 1.6 A. but with 15 V input (the chip is good for up to 17 V)20:28
wpwrakthe 10 Ohm load resistor heated up nicely, though. so it was good while it worked.20:28
wpwrakand of course, a saturday at 17:30 before three "sundays" is about the perfect time for an equipment failure. @#$%^20:38
kyakmaybe it's Fate itself telling you to do something else during these three days20:48
hozerwhat are you working on20:48
wpwrakkyak: oh, the 1st is already booked for something quite different. hangover removal :)20:49
wpwrakhozer: the controller for a LED-based text (or whatever) display device.20:50
hozerwhat display are you using20:50
wpwrakbasically it'll be the big brother of this one: http://www.almesberger.net/misc/ant/edd1.jpg20:51
wpwrakjust a line of LEDs20:51
hozerooooh fun20:52
hozerI want to put one of those on a wind turbine20:52
wpwrakthis device has 16 pixels, 2 LEDs for each. so 32 LEDs20:52
wpwrakthe new one has 64 LEDs, individually switchable20:52
wpwraknaw, just red.20:52
wpwrakthis is an earlier version of the 64 LED variant: http://downloads.qi-hardware.com/people/werner/tornado/proto1122.jpg20:53
wpwrakfunction test: http://downloads.qi-hardware.com/people/werner/tornado/proto1122.ogv20:54
hellekinLOL... So practical :) <     wpwrak> kyak: oh, the 1st is already booked for something quite different. hangover removal :)20:55
wpwrakthat controller used an atmega168 which is a bit underpowered for the task. and it didn't have a voltage regulator, so the uSD card would freak out with all the voltage jumps20:55
hellekinwpwrak: my girlfriend likes it :)20:56
hellekinnot the hangover, but the prototype :)20:56
wpwrakhehe ;-)20:56
wpwrakthe old one (32 LEDs) is for shaking the stick. the new one is for spinning it at the end of a string. less work, better picture :)20:57
wpwraki also de-fancies the design a little. the "antorcha" ("torch") has a ben-wpan built in, so images and firmware can be updated over the air. drawback: it needs a ben+atben to do anything useful.21:01
hozerwhat is a ben-wpan21:01
wpwrakthe new one has just a uSD slot. so it can carry a lot of images without needing an external device. plus, it's easier to find something that can talk uSD than anything that can talk IEEE 802.15.421:02
hozerheheh, why not wifi? ;)21:03
wpwrakhozer: an IEEE 802.15.4 board for the ben nanonote: http://downloads.qi-hardware.com/people/werner/wpan/web/21:03
wpwrakhozer: got any small wifi solution that doesn't suck NDA-wise ? :)21:03
hozerum, probably the one we use an GNUradio on an fpga for? ;)21:04
wpwrakwell, or suck in all other regards, like that microchip critter that only does 802.11 (no a, b, g, ...)21:04
wpwrakhozer: that uSD slot costs me cents :)21:05
whitequarkwpwrak: wait. is there even such a thing as "just" 802.11?21:06
wpwrakwhitequark: oh yes :)21:06
whitequarkI always thought it was just a name for a family of standards21:06
whitequark>IEEE 802.11 is a set of standards for implementing wireless local area network (WLAN) computer communication in the 2.4, 3.6 and 5 GHz frequency bands.21:07
whitequarkwait 3.6?!21:07
whitequarksince when21:07
hozerhrrrm... 3.6... sounds like a good backhaul channel for open-mesh21:07
wpwrak" IEEE 802.11y-2008 extended operation of 802.11a to the licensed 3.7 GHz band. Increased power limits allow a range up to 5,000 m. As of 2009, it is only being licensed in the United States by the FCC."21:09
wpwrakah, seems that microchip recently made some new modules with more modern characteristics21:11
whitequarkwpwrak: why would you want to use anything by a company which produces PIC and ENC28J60?21:15
wpwrakwhitequark: in this case: finding sufficiently open small wlan is difficult21:16
whitequarkwhat about atheros?21:17
wpwraklast time i checked they didn't have anything suitable. all their embedded-friendly stuff was quite closed.21:18
wpwrakit's not only a question of linux drivers but also one of access to chip/modules, design information, and so on21:18
wpwrake.g., there is a linux driver for the ar6k (the one in the openmoko gta02) but you can't get the chips or hw design information. modules are difficult to source and you don't get firmware updates.21:20
whitequarkfor at least one chip they published firmware sources21:21
whitequarkbut updates are a different thing, yeah.21:21
wpwrakcompilable and installable firmware sources would solve the update problem :)21:22
whitequarkit is even for a 11n chipset21:23
wpwrakthings get more relaxed when you move to larger chips, e.g., for laptops. but they're usually too big for small embedded systems.21:23
whitequarkit depends. connectivity line STM32's have USB OTG. I agree this is far from ideal, through.21:24
wpwrakyou also need to go up to 5 V if doing USB. so you're forced to make a rather complex system before you can actually use that stuff.21:25
wpwrakand then you're still limited to modules, which increases your sourcing risk.21:26
wpwrakand of course, you need lots of space for all that21:26
whitequarkif you want less space, maybe it isn't a good idea to use 802.11 after all?21:28
wpwrakoh, 802.11 would be fine. it's just that you can't get the chips unless you're a big player.21:29
viricdamn it... does 3.7.1 have a different specification for mtd partitions in cmdline?22:00
virica 3.7 vs 3.6 change I mean22:00

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