#qi-hardware IRC log for Wednesday, 2011-10-19

DocScrutinizer(sw for antenna simulation and R&D) rarely anything. Basicaly it's still not possible to *accurately* compute antennae01:32
DocScrutinizerthere are a few algos for well known models, lika yagi01:33
DocScrutinizerof course for parabolic it mostly should work as well01:33
DocScrutinizerbut calculate the gain in dB of a random wire bent to some shape of your liking? nope01:34
wpwrakoh, there are simulators. but they're kinda hostile ... (or non-open)01:44
DocScrutinizeryes, and not really accurate enough. AFAIK all industry tests with real models, as they yield different results than the simulators01:47
DocScrutinizerseems it's a variation of the 3+ planets problem. Really hard and expensive to calculate, and never exact01:50
DocScrutinizerspecial class of strange attractors01:51
wpwrakmaxwell's demon :)01:51
DocScrutinizerwasn't the maxwell daemon only to know, but not to extrapolate?01:52
wpwrakmaybe he has other skills :)01:53
DocScrutinizeroooh I first got it right, then mixed up maxwell with laplace01:54
DocScrutinizercan't see how maxwell's daemon is related01:55
DocScrutinizerbtw I reinvented maxwell's daemon in school when I was 14 years old :-D. My physics teacher wasn't completely convinced I never before heard of maxwell's daemon when I asked him about where the energy goes when just pouring warm water into cold water, and if it was possible to "unmix" them again01:57
wpwrakheh ;-) and at 15, he flunked you when you proved the existence of "dark matter", right ? :)02:01
wpwrakand then, at 16, they made you see a therapist when you explained that there are nearly weightless particles, that almost never interact with matter, that spontaneously change into other, similar, particles and back, and that travel at the speed of light or slightly faster02:04
DocScrutinizernah, none of that02:20
DocScrutinizerbtw I'm not convinced about existence of dark matter even today02:20
DocScrutinizerat least that stuff must be really strange, as it's obviously not forming planets or stars, despite allegedly obeying law of gravity02:22
wpwrakyeah, it's one of the more dubious concepts. has that strong smell of aether on it :)02:22
wpwrakmaybe dark stars :)02:22
qi-bot[commit] Xiangfu Liu: nanonote-files: update feeds.conf for next release (master) http://qi-hw.com/p/openwrt-packages/fe24c6b03:17
qi-bot[commit] Xiangfu Liu: makfa: add makfa.dump to files. not generate every build (master) http://qi-hw.com/p/openwrt-packages/bf7720107:43
kyakviric: hey, that sounds great. What kind of functionality for offrss do you think about?08:16
virickyak: something like getting a list of only-titles, for feeds in a group08:17
viricTo handle so much data as you need... there should be some code 'removing posts' there08:18
kyakthat would be usefull for some feeds, where only title matters08:18
virickyak: there are some people publishing almost only title and hyperlink08:19
kyakyes, there are :)08:19
virickyak: for your use case, you could use some 'rss2title URL | grep ...'  :)08:19
kyakthat's basically what i did before, curl URL | grep | sed & etc08:21
kyakand than i have the download magically in rtorrent08:21
kyakfrom my opinion, the best new feaure would be a "mark all as read" (for a group or for a feed) - but i don't know how hard is it to implement it08:22
virickyak: easy easy08:31
kyakviric: well, when you need testing, just let me know ::)08:38
viricok :)08:46
whitequarkwpwrak: remember that my LVDS thingy?14:37
whitequarkIT WORKS!!14:37
whitequarka console with a penguin is displayed at least :D14:37
wpwrakcongratulations ! :-) that was quite a battle14:40
whitequarkbut now I can make happier quite a few people14:59
whitequarkI already have a second order :D14:59
whitequark(a battle) IMO, a pretty regular debugging routine, isn't it?15:01
whitequarkI don't think that even atben or of course the ben itself were any easier15:01
wpwrakwell, they were different :)15:10
wpwrakthe basic stuff worked quickly. but getting the RF right was a little harder15:11
qi-botThe build was successfull, see images here: http://fidelio.qi-hardware.com/~xiangfu/compile-log/openwrt-xburst.full_system-10172011-2340/16:03
kyaklarsc: do you which of the USB Gadget Drivers would work for Ben? I tried a few so far, none of them works (except for USB_ETH)16:18
whitequarkwpwrak: http://rghost.ru/26276911/image.png :)16:20
larsckyak: i think i only tested USB_ETH so far16:21
whitequarkkyak: try the CDC-ACM16:23
whitequarkit may work16:23
kyakwhitequark: that i tried, it doesn't work16:23
whitequarkit works on 4750l, and it has the same usb gadget as 474016:23
kyakwhitequark: the composite driver or..?16:24
whitequarkkyak: the composite driver?16:25
kyaklarsc: i see.. i thought USB_CDC_COMPOSITE would work (CDC ECM for Linux and CDC ACM for Windows, which doesn't have ECM driver)16:26
kyakwhitequark: i'm trying now the USB_G_SERIAL16:26
wpwrakwhitequark: nice !16:28
whitequarkkyak: IIRC 4740 usb gadget does not have enough descriptors or whatever for ECM+ACM tow ork16:28
whitequark*to work16:28
whitequarkwpwrak: thanks16:28
larsckyak: are you getting any error messages in dmesg?16:43
kyakno, it's quite at all16:44
kyaklike i'm not plugging anything16:44
kyakdamn, kernel_menuconfig is so broken, i can't configure the kernel correctly..16:45
kyakwhitequark: yeah, USB_G_SERIAL works fine (at least on Linux, still doesn't work for Windows)17:13
kyakbut this config is not so interesting anyway..17:14
whitequarkkyak: as I've said, when I've looked at it, it turned out that the UDC does not have sufficient resources to run the combined configuration17:25
whitequarkunless you'll hack it to do so, of course17:25
kyakthat's pretty strange though because i want it to run one configration a time17:35
kyaknot both configurations concurrently17:35
whitequarkkyak: then, you should not use a combined one17:36
whitequarkjust compile both ECM and ACM  as modules and load the one which you want to use currently17:36
whitequarkit's proven to work17:36
whitequarkor ETH, that works with any gadget17:37
kyakwhitequark: yes, compiling both as modules makes sense..17:39
whitequarkit's the way it works17:40
kyakwhitequark: are you saying that ETH supports ACM, too?17:40
whitequark   17:46
whitequarkkyak: nope, it doesnt17:47
whitequarkthey are different gadgets17:47
whitequarksorry, crappy WiFi17:47
kyakwhitequark: ok, thnkas17:56
ArtyomHi Kristianpaul18:05
kristianpaulArtyom: hello!18:16
Artyomhow is your work with gps?18:17
kristianpaulsigned problem solved, but i have some problems when reading memory :/18:17
kristianpaulmy fault..18:17
kristianpaulI'm working on solving, actually you are using a DCM in your board isnt?18:18
kristianpaulThe problem is that namuru core clock is 8Mhz and the SoC clock is 80Mhz18:19
Artyomyes, I use DCM to generate 80MHz clock for FPGA from 16MHz clock from my front-end18:20
kristianpaulyes, well i'm working on solve that, i just found a solution but is vhdl so i need port it 18:21
kristianpaulso it should avoid cross clock domain issues using a fifo18:22
kristianpaulbtw i was checking your code, so i think can be easy ported later, also i'll re-use that mprintf function :)18:22
kristianpaulSo far, thats my status18:22
kristianpaulHow is yours?  what about your students projects?? all related with gnss-sdr? :-)18:24
Artyommprintf is not mine ;) I've found it in internet ;) 18:24
ArtyomI switch to some other tasks related with glonass. especcially with the new cdma signal. 18:26
kristianpaulah sure, yes :)18:26
Artyomand i also spend some time on planning future front-end. I want to make dual frequency glonass/gps front-end18:27
Artyomstudents are rather lazy and unpredictable ;) Actually I was the same... 18:29
kristianpaulWhat about processing? still want to use the usb for transfer data to pc or  want do soemthing in same board with arm+ fgpa¿18:30
Action: kristianpaul laaaazy18:30
ArtyomI want to make something universal: the device should be able to transmit data to PC through USB 2.0 and to process data autonomously18:32
kristianpaulthats fair18:33
ArtyomAnd I want to split analogue and digital parts. Each part should be made on a separate pcb18:34
kristianpaulthat will be great, also will allow re-use the anloag part in other projects ;-)18:35
ArtyomBut now I'm mainly busy with software processing... And I have to prepare couple of articles...18:37
kristianpaulArtyom: what you think about i run namuru at same clock of my soc, and uses dcm to to something inverse to decime with the coming I/Q signals??18:37
ArtyomWhat is the quartz-generator that you want to use?18:38
kristianpaulsorry i don understand18:39
ArtyomGPS devices usually use TCXO or even OCXO.18:39
kristianpaulah, you mean the clock for the SoC?18:39
ArtyomI mean what kind of clock do you want ot use?18:39
kristianpaulfor namuru i want to use the same clock as the main soc18:40
kristianpaulyes, my front end have own TCXO18:41
ArtyomSo as I understand you want to use mylkimist clock for your gps front-end?18:42
kristianpaulbut the problem is that data is sampled at 8mhz, how i can still read it if my system (and namuru)uns at 80Mhz18:42
kristianpaulmay be18:42
Artyomor you want to use gps front-end clock as the main clock?18:43
kristianpaulI wish i could, but the exp connector i have dont have pins for mail clock signals afaik18:44
kristianpaulthats why i'm dealing with two clocks and crossing its domains18:45
ArtyomSo your problem is that front-end and mylkimist use different clocks?18:45
ArtyomOnce again18:46
ArtyomCan you take front-end clock and use it in mylkimist?18:46
kristianpauli could try yes, hoping synthesis will pass trought 18:49
kristianpaulright now i'm suing that front-end clock in a slave core (namuru) of the mylkimist18:49
kristianpaulbut not in the whole soc18:50
Artyomyou can use DCM in FPGA to multiply front-end clock. Like it's done in my example18:50
kristianpaulyes i think could try at least18:51
Artyomsorry... how you do it now?18:52
kristianpauli can change some pin asgiments in the  .ucf files plus modify a bit the dcm milkymist already uses18:53
kristianpaulright now namuru is a slave core in the milkymist soc, it uses clock from front-end and there is a no very good clock domain crissing "bridge" that is giving me problems :/18:57
kristianpauli think thast was you asked for, right?18:57
ArtyomI think the simplest way is to use front-end clock for everything...18:59
kristianpaulSure, i do think same, actually in the end i will still need the milkymist cpu for the aquisition and tracking18:59
kristianpaulso that make sense...19:00
kristianpaulI could try that o the milkymist one board or implement a small soc in other board i have19:00
kristianpaulhow much ram do you need for acquistion and tracking algortym only?19:01
Artyomthat's a good question ;)19:03
ArtyomHow can I measure it?19:04
kristianpauli guess srtating by the size of your osgps19:05
Artyomi will check it tomorow19:06
kristianpaulmay wpwrak knows how stimate amount of ram a c program will need.19:07
Artyombtw is it difficult to implement in fpga a cpu+other? Never done it before...19:07
kristianpaulno at all19:08
kristianpaulwell you'll hit this clock domains issues, but always there is a workaroudn :)19:09
kristianpaulyou know milkymist one already, isnt? cpu is mico32 from alttice19:09
Artyomyes, i've read a little about it. btw I thoought that mico32 can only be used with lattice fpga's...19:10
kristianpaulof course is not:) i tought same then foudn milkymist project a year and half from now :)19:11
kristianpaulthere is people i  cern i read at #milkymist using mico32 in altera fpga's so thats good too19:12
Artyomand there are problems with licence and etc?19:12
Artyom*no problems19:12
kristianpaulzero, there was a minor missing clarification in mico32 license but that was fixed in last release19:12
kristianpaulas we stay away from mips and arm implementation in fpga, all is okay so far :)19:13
Artyomand what tools are used to compile programs? And how everything is working? ;) Is there any tutorial in the web?19:15
kristianpaulmico32 is supported by gcc19:15
kristianpaulthere is also a rtems port19:15
kristianpaulwich curerntly works as the base software in the milkymist one first market aplication wich is video synthesis19:16
kristianpaulHere there are some presentations19:16
kristianpaulhere a good tutorial http://lekernel.net/blog/2011/08/intervention-materiel-libre-et-milkymist-au-master-i2l/19:17
kristianpaulyou can skip french :)19:18
Artyomlast question: what is the minimum hw requirements to implement mico32 in FPGA?19:19
kristianpaulalso you are invented you join #milkymist at any time19:19
kristianpaulthis is nice slide too http://lekernel.net/presentations/masteri2l/conf/mm_i2l.pdf19:19
kristianpaulsorry this http://lekernel.net/presentations/masteri2l/tp/mmtp_i2l.pdf19:20
Artyomthanks for links :)19:21
kristianpaullet me find requirements in LE19:21
kristianpaulbtw here is a small port https://github.com/fallen/milkymist-avnet19:24
kristianpaulprety basic just lm32, uart sram and vga19:25
kristianpaulArtyom: http://paste.debian.net/138098/19:27
kristianpaulIs around 2500 LE for spartan6, i still looking to confirm19:28
kristianpaulso is pretty small i think19:29
Artyomyes, I should test it ;)19:30
kristianpaulabout if it works, well check some videos http://www.youtube.com/user/xiangfuliu19:31
kristianpaulbut, everything work for the VJ app, and still somw work to do for getting linux run better on it (missing drivers)19:32
kristianpauland of course best usb stack and support for it19:33
kristianpaulah, just in case http://en.wikipedia.org/wiki/Milkymist19:35
Artyomthe idea to use only one fpga without additional cpu is attractive. The only thing that terrifies me is debugging19:39
kristianpaulthere is a gdb support http://milkymist.org/wiki/index.php?title=Using_GDB_in-system_debugger19:40
kristianpauljtag is WIP, you may want ask mwalle at #milkymist for mode detail19:41
kristianpauli personally use serial port and bios for debuging :)19:41
kristianpauli still on the printf and led blink era :) 19:42
C-Keenwhich is not a bad approach19:42
C-KeenIn my experience debuggers help only find certain bugs quickly, mostly doing a post mortem analysis of some kind of memory dump19:43
kristianpaulalso there is qemu port, i'm not sure but i think is a valid way to debug software once you learn how customize the port :)19:46
whitequarkyou'll then need to write both software and port updates to be compatible with the real "hard"ware19:48
whitequarkcompared to only the software19:48
kristianpaulyou write hardware you can trust more ;)19:49
whitequarkkristianpaul: hm?19:51
kristianpaulwhitequark: in qemu 19:52
kristianpaulsimulate hardware behavior at register layer, etc.. i guess19:53
whitequarkkristianpaul: well, it's easier to do the same error in qemu & sw layer19:55
whitequarkjust because you write both19:56
whitequarkand then you'll think for a week why it does not work on the real thing19:56
Artyomand what is qemu? ;)20:02
kristianpaulIs a computer emulator like propietary equivalents, vmware virtual pc20:03
ArtyomKristianpaul thanks for all the information. Very useful for me :)20:04
Artyomit'm time to sleep for me... bye :)20:06
ignatius-I've been playing around with the accelerated port of MPlayer to the Ben. I can get the video, but no the audio.23:11
--- Thu Oct 20 201100:00

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