#qi-hardware IRC log for Saturday, 2012-02-25

wolfspraulgood morning :-)01:18
abushcrafterforggood morning.... arr is 2 am02:02
whitequarkDocScrutinizer51: APE?10:34
wolfspraulwpwrak: I'm wondering how susceptible to RF interference atben/atusb is compared to a RF keyboard11:36
wolfspraulso for example, you take your rii rf keyboard, and you have a lot of 2.4 ghz wifi noise around - do you experience problems with the keyboard? how about atben/atusb in comparison? any insights?11:37
wolfspraulI think I will start to watch this a bit more, and test in noisy environments...11:38
wpwrakwolfspraul: (interferences) like wlan, it is affected by interferences12:44
wpwrakthe rf keyboard will have the same issue. it's just a little harder to notice.12:45
wolfspraulok, that's quite vague :-)12:46
wpwrakalso don't forget that received energy drops with distance squared. as long as sender and receiver are close to each other, interference rarely matters. it's when distances grow when the problems start12:46
wpwrakoh yes, RF is all about being vague ;-))12:47
wolfspraulyou have been using both the RF keyboard and atben/atusb quite a lot, and you have at least 1 wifi network as well, maybe a few more from neighbors12:47
wpwrakthere are countless factors that affect this. also frequency separation, channel width (of the "good" signal but also of the interferers), signal path characteristcs, etc.12:47
wolfspraulwhat RF technology is used by the RF keyboard?12:47
wolfspraulyes sure, but I am wondering how atben/atusb does compared to say the 'typical' RF keyboard12:48
wolfspraullet's say you are in a very wifi-noisy environment12:48
wolfspraulmaybe the answer right now is simple: we don't know12:48
wpwraki think it's some proprieatry 2.4 GHz stuff. presumably somewhat similar to 802.15.4. perhaps with less efficient modulation.12:48
wpwrakatben/atusb should get a little further than the average rf keyboard12:49
wpwrake.g., mine is good only up to 3-4 m straight line of sight. atben/atusb can do 7 m and more around corners12:50
wpwrakatben and atusb operate at comparably low power. only 2 mW. in the 2.4 GHz band, you can usually go up to 100 mW. so other IEEE 802.15.4 devices (those with extra amplifiers) may be able to go almost ten times as far.12:52
DocScrutinizerwhitequark: Application Processor Environment13:02
DocScrutinizerbasically the "linux cpu"13:03
whitequarkDocScrutinizer: nope, I only looked at the file which was downloaded into the modem at powerup13:04
whitequark(it doesn't have its own flash)13:04
DocScrutinizeryep, I know that design13:07
DocScrutinizerthe very first file transferred to modem is mcore, basically the OS13:09
DocScrutinizerthe domain I work for13:09
DocScrutinizerthen this file gets checksummed and started, and it loads next file which usually is NS (network signalling, the LTE/UMTS stack)13:10
DocScrutinizerthough after mcore got loaded, the load sequence of other load-modules is not strictly defined, they may change depending on packaging13:12
DocScrutinizerwhitequark: could you pastebin the output of strings?13:19
DocScrutinizerand maybe also a od|head -n 10013:20
whitequarkDocScrutinizer: sure, I'll be back in a hour13:20
DocScrutinizerno access to all that any sooner than monday ;-D13:21
DocScrutinizertoday is WEEKEND, and my RF-thermomenter tells me it's 14°C outside \o/13:22
whitequarkDocScrutinizer: there are 9683 strings longer than 32 bytes13:26
whitequarkpastebin chokes on that.13:26
DocScrutinizerpastebin is nasty13:27
DocScrutinizertry another service13:27
whitequarkbtw, http://pastie.org/345517013:27
whitequarkalso, that BP is xgold626, if that'll tell you something13:28
whitequarkbaseband processor?13:31
whitequarkand the second B stands for?..13:32
DocScrutinizerDBB and ABB, for digital and analog, what we call companion chip at OMAP13:32
whitequarkah. in the pre-Android motorolas, they were AP and BP (Application Processor and Baseband Processor). looks like the naming is different in different companies.13:33
DocScrutinizerI'm sure there's no ISO standard for that ;-)13:33
whitequarkok. the whole firmware is at http://rghost.ru/download/private/36701695/c1cb9e90fba71927ec3b0f0c9dc4c6a7/d616cef15099b5365d15f1347bc5fdc5b59be524/fw.tar.gz13:33
DocScrutinizerAP(E) still used13:33
whitequark(and the phone is Samsung Galaxy S II)13:34
DocScrutinizer"Save as..." (help me out please ;-D)13:34
DocScrutinizeraaah :-D13:35
whitequarkthe particular thing I like about it is that AP only shares an UART, a PCM pipe and a shared memory area13:35
whitequarkwhich is distinct from the main memory13:35
whitequark*shares with BP13:35
whitequark(I have schematics in a level 3 service manual, too)13:35
DocScrutinizerI think there are several different design options. You can have so called bridgeless design where BB and APE share same memory13:38
DocScrutinizerI always wondered how NS guys want to make sure no APE process spits into their broth13:40
DocScrutinizerwith this bridgeless design13:40
whitequarkmemory protection?13:41
DocScrutinizerwell, the mmu on AP can do a good job not to expose this memeory to userland13:41
DocScrutinizerstill I think it's a kinky design, like UMA13:42
DocScrutinizerwas that UMA?13:43
DocScrutinizerwhere you have to decide how much of your RAM will get eaten by gfx-card?13:43
DocScrutinizerwell, usually we talk to APE (or rather APE to 'us' [modem]) via a mipi HSI interface13:45
DocScrutinizerusing CAIF as logical layer / protocol layer13:45
DocScrutinizeryesterday I investigated reset details in M7400 and discovered a porn_core (SIC!)13:47
DocScrutinizermade me LOL13:47
Action: pabs3 wonders if Google would accept Qi in the GSoC13:59
DocScrutinizerwhitequark: the firmware.bin seems to have not a single string that looks familiar to me14:48
DocScrutinizerare you sure this device is using a STE modem?14:49
DocScrutinizerwhitequark: you say you have schematics, so what's the BB chip used there?14:53
DocScrutinizerI honestly feel like wasting my time14:54
DocScrutinizerwhitequark: google doesn'T yield a single hit on search terms of both your device and "my" modem chipset14:57
DocScrutinizerhttp://www.google.de/search?q=thor+m5730 however has a hit on "Samsung Galaxy S 4G Teardown - Page 2 - iFixit"15:00
DocScrutinizeralas http://www.ifixit.com/Teardown/Samsung-Galaxy-S-II-Teardown/5861/1 dosn't break down to chip name level, but neither on http://guide-images.ifixit.net/igi/GVUGCVdx2BMSHVe3.huge nor on http://guide-images.ifixit.net/igi/dko2oBWWbgsSvhKq.huge I can spot a STE DB57x0 chip15:18
DocScrutinizerit seems the S2 is using a Samsung original "Exynos 4210" cgipset that probably also has wireless15:34
DocScrutinizerat least I couldn't find any mentioning of the wireless radio chipset used in S215:36
DocScrutinizernow CYA15:36
DocScrutinizerout for some fun in the big bluebox15:36
qi-bot[commit] Xiangfu Liu: cgminer: support icarus by upstream, fix the multi-icarus support (master) http://qi-hw.com/p/openwrt-packages/01bb56a17:02
ArtyomKristianpaul: hello! :)17:36
kristianpaulI noticed you are playing with interrupts ;)17:42
kristianpaul"uart-data-transmition occupies a lot of cpu-time" hmm17:44
kristianpaul"Delay, phase and frequency discriminators must be optimized! Otherwise tracking of more then 2"17:44
kristianpaulchannels will be impossible.17:44
kristianpaulhow do you know it occupies lot of cpu time?17:46
kristianpaulbtw i havent checked your namuru verilog port17:47
kristianpaulbut i'm aware there are some posible related to a not very good mix of blocking and non-blocking asigments17:47
kristianpaulthat could explain at first un-stabillity on the correlator17:48
kristianpaulbtw what ISE version are you using?17:49
ArtyomI use 12.4 (if I remember correctly)17:55
kristianpaul13.4 is last and recomended vesion for milkymist soc :)17:57
ArtyomIt was very easy to check how long interrupt are processed with accum_int output and socillosccope17:57
kristianpaulnot very constant graph?17:58
kristianpaullast time i did was very funny to see that on scope17:58
ArtyomWhen uart transfer occures (approximately once per 1 second) then interrupt-function lasts more then 1 ms. That is not good17:59
ArtyomThe same picture when 3 channels are working in parallel17:59
kristianpaulWhat keep busy uart? too many printf/put i guess18:00
ArtyomBut there are a lot of possibilities to optimize the code. So it shouldn't be a problem18:00
kristianpaulyes it is damn slow.. i remenber :/18:00
kristianpaulnot use too much printf ? ;)18:00
Artyomyes, I try to use it as few as possible. But I cannot refuse from it because it is the only way to output data from program to external world18:02
ArtyomMay be I should look for a faster uart-core. Or write my own18:02
kristianpaulmay be buffer first and transfer later, 128Mb is plenty for some samplint?18:02
kristianpaulwrite own.. could be18:02
kristianpauli know this uart is no desing for heavy transfers18:03
ArtyomBut I want to check how tracking is working. I want to compare results with other receivers18:03
ArtyomSo data must be output in real-time mode18:03
kristianpaulwhat other receivers you mean? i mean.. ergh what other receivers allow debug tracking?18:04
kristianpaulWhat is your soc speed?18:04
kristianpaulwich gcc toolchain version are you using for lm32?18:04
kristianpaulfor those transfer the best will be ethernet using udp(tftp) my opinion18:05
kristianpaulbut yeah, is clear m1 soc was not designed for such tasks..18:05
ArtyomI mean I want to compare Doppler for one of the satellites generated by my program with Doppler calculated with some hardware receiver18:06
kristianpaulah, benchmark Lab :-)18:06
Artyomgcc.... I think I downloaded some scripts for MM1. And they downloaded gcc, gdb and so on.18:07
kristianpaulah good18:07
ArtyomWell, yes. Ethernet can be a good solution. But not for me ;)18:08
kristianpaulGet an M1 !! ;-)18:08
kristianpaulof course this dont work out of the box.. still code to write for those features..18:09
kristianpaultftp works just as client last time i cheked..18:09
ArtyomI played with it too. Downloaded software in MM SoC18:10
kristianpaullunch, back in a some minutes18:11
Action: kristianpaul back18:18
ArtyomI use 48 MHz system clock (16*3)18:48
ArtyomKristianpaul: It's very interesting for me if you could run namuru-core and my program on your MM1. May be you will be able to run your system with higher frequency18:53
kristianpaulWell, i could do another port yes18:56
kristianpaulTo be sincere i dont like the baseband implementation method18:56
kristianpaulbu yes i could, plus solder the maxim receiver..18:57
kristianpaulI think clear register must be separate18:57
ArtyomWhat do you mean by "baseband method"? I think you can use your front-end18:58
kristianpaulyes sure i can use fronted18:59
kristianpaulArtyom: i meant basically that there are separate address for writing and reading same register18:59
kristianpaulanyway i could run it yes19:01
kristianpaulArtyom: i want to be able for example to read back a value when writing prn code for example19:01
kristianpaulthats is currently no posible19:02
kristianpauland implement it yes, but i will need two register for that..19:02
Action: kristianpaul stubborn19:03
ArtyomYes, prn-reading is not realized in namuru...19:04
kristianpaulhe indeed ;)19:04
ArtyomBut what are the reasons for such implementation? Just curious...19:10
kristianpaulI want to debug when is that register cleared :)19:10
kristianpaulI just debug19:10
ArtyomWhy not to use simulator for this task?19:17
kristianpauli havent written the test bench, but good point19:21
Artyomoh, seems I should publish mine...19:22
kristianpauli prefer write test in software for now :)19:24
Artyomto test cores in hardware? ;)19:25
kristianpaulwhy? because i dont know what xilinx sintheize at the end and posible timing bugs i dont know off..19:26
kristianpauli dont trust xilinx tools19:26
ArtyomYeah, I understand you very well. I also tried to start hardware testing as soon as possible. 19:28
ArtyomBut after spending a huuuugggeee amount of time on finding simple errors in verilog/vhdl I prefer to simulate as much as possible...19:30
kristianpaulI use gplcver fot simple errors19:31
kristianpaulbut yes i got the point19:31
ArtyomI used icarus-verilog. But I want to test verilator. As simulating in icarus verilog is extermely slow for namuru-testing19:32
kristianpauloh indeed19:34
whitequarkwolfspraul: was the ethernet cable included with m1 specifically designed to be exceptionally hard to remove?19:51
whitequarkbecause I can't do that without two screwdrivers.19:51
kristianpaulhmm no19:51
whitequarkwhich isn't very convenient or a well-known requirement for ethernet cables.19:51
whitequarkkristianpaul: it has some really stupid kind of rubber latch19:52
kristianpaulyou need remove top of the case to plug usb cable19:52
whitequarkthat's easy19:52
kristianpauland use jtag/serial etc19:52
kristianpaulbesides that..19:52
qi-bot[commit] Werner Almesberger: M1 build process update (master) http://qi-hw.com/p/wernermisc/fe7a33423:37
--- Sun Feb 26 201200:00

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