#qi-hardware IRC log for Thursday, 2012-03-01

DocScrutinizerI should've answered: please change serial number of that defective device, to rule out something in driver fails on exactly that particular serial#00:02
DocScrutinizer"dear microsoft, we bought 500 workstations with win7. All are identical hw and same sw and config. one of those machines freezes screen after POST, while bios config works flawlessly. Please fix that bug in your OS!"00:04
wpwrakbut yes, serial numbers sometimes do matter. particularly if they're in some sort of ID PROM chip :)00:07
wpwrak(sun used to have that. and we had one - at EPFL - that may have been flaky. of course, the heavy EMI from the experiments in the lab in the basement, right underneath the machine, didn't help)00:08
DocScrutinizerno, you should not connect the cut pin to GND. You should leave it NC, so user can connect it to GND or arbitrary other signal he likes, and replace the post with a new uncut one00:11
wpwrakhmm. you mean to give more routing options ?00:12
DocScrutinizerno use in connecting GND to a not existing PIN when we already got 3 more GND lines00:14
wpwrakhmm. making sure it's GND would enhance the barrier role on J21. but yes, that's admittedly not a strong argument.00:17
pabs3zenlunatic: I guess donating your time to work on it is better00:25
kristianpaulroh: about that cool p2p pager project03:29
kristianpaul802.15.4 cheap tranceiver is ... in $$$ ?03:30
kristianpauli liked rfm12b-like modules they are less than 2usd last time i remenber03:30
kristianpauli asume 802.15.4 is like starting from 5usd?03:30
kristianpaulbtw you also mentioned cdc-ethernet03:31
kristianpauldoes it out of the box  (some how?)03:31
rohit was chips for something below 10E i think03:31
rohwhole device below 2003:31
kristianpaulwouldt be cool just plug a nanonote to computer and get instant connectivity..03:31
rohthe mcu was 3$ or so03:32
kristianpauli think we may share similar issues with cdc-ethernet :)03:32
rohsure. but its not for 'lots of ip' but rather 'send a sms through the jungle'03:32
kristianpaulreally cheap mcu03:32
kristianpaulis nice,03:32
rohyeah. i need to get here here ;)03:32
kristianpaulfor example i iamgine my nanonote, plug to computer and cdc-ethernet works of of the box then i get the message, beep !03:32
kristianpaulbeep and buzz ;)03:33
roh i asked her about how we could get it onto the nn, but we dont know yet03:33
kristianpaulall encrypted pager right on my device.. but network dont work that automatically03:33
kristianpaulyes, as you said it requires some routers infraestructure as well 03:33
rohthe nn doesnt do host-usb which would make it easier. but i think the stm32 should have enough pins so we can interface to spi too03:34
kristianpaulwell !! invite her, no more to said, please share links blogs, etcc :)03:34
kristianpaulwhat is spi for in that case?03:34
rohto connect it to a ubb and the nn03:34
wolfspraulkristianpaul: very nice blog post from Artyom03:35
kristianpaulyou also metioned 6Km, is not that too much power for unlicesed bands?03:35
kristianpaulwolfspraul: oh, indeed !03:35
kristianpaulyeah :)03:36
kristianpaularm down, milkymist up ;)03:36
kristianpaullekernel will be happy ;-D03:36
wolfspraulwell, let's be as inclusive as possible03:37
wolfsprauldismissing arm is not smart and good imho03:37
kristianpaulinclusive, what you mean?03:37
kristianpauloh i see03:37
wolfspraularm has been doing fabulous work the last 20 years and that brings us 2 USD chips nowadays that can do unbelievable things today03:37
kristianpaulyes not, but may be for some aplications move to a fpga is much interesting03:37
kristianpauland will start to see this more often03:37
wolfsprauloh sure, I defend milkymist to every last bit03:37
wolfspraulit's the next thing03:37
wolfspraulbut dismissing arm is not right03:38
wolfspraulI need to be careful about that as well, it's all too easy.03:38
kristianpauloh no, i was kinding of course :)03:38
wolfspraulmilkymist today is a tough nut to crack03:38
wolfspraulit's expensive, buggy, features at first sight look more like from a tech museum03:38
kristianpaulbut yes..03:39
wolfspraulguess how it felt to work at ARM in the mid-nineties, when Intel was coming out with one market victory after another...03:39
wolfspraulso milkymist can win if it also creates a new culture of collaboration03:40
wolfspraulnot just some new tech, where the existing tech is already there, written off, cheaper, better, etc.03:40
wolfspraulthat's how I look at it, at least03:40
kristianpaulcollaboration,yes, more software suopport ports etc.03:40
wolfspraulbest documentation, best test instrumentation, friendliest to newcomers, friendliest to free sw stacks running on it, friendliest to research, and so on03:41
Action: kristianpaul try to be at least friendlier03:41
wolfsprauloh I think we are partially there already03:42
wolfsprauljust try to email ARM :-)03:42
kristianpaullol, no03:42
kristianpauldocumentation is good, could be better but is there.. just a bit hard to find..03:42
kristianpauland in french sometimes if you want a getting started guide ;)03:42
wolfspraulmy list was not meant to describe today, but the goal of how it can truly succeed03:42
wolfsprauldefinitely not by dismissing ARM03:43
wolfspraulthat's just ridiculous actually, if you look at milkymist today with a calm eye03:43
kristianpaulcalm eye, and the result of a big list of missing features yes of course nobody can defeat that03:45
wolfspraulsure, more people need to join and contribute back03:45
wolfspraulbut that is increasingly happening03:45
wolfspraulto the extent that such things happen at all right now, in collaborative IC desig03:45
kristianpaulyeah, i got a mail yday froma guy with DE0 wanting to port milkymist soc to it :)03:46
kristianpaulwanting or wip, i dont remenber..03:46
wolfspraulinvite him here (or #milkymist)03:46
kristianpauloh yes..03:46
kristianpaulwaiting he ping me :)03:46
wolfspraulI think there's something wrong with *all* mailing lists on qi-hardware.com ;-)04:07
wolfspraulthat shall be investigated now!04:07
wolfspraulhmm. there could be a flood now, urgh.04:12
DocScrutinizerthanks for all the thanks :-)06:46
wpwrakthe mailing list problem seems to have been very selective. seems that all of my mails made it without trouble.07:18
wpwrakso maybe the milkymist list wasn't affected ?07:21
wolfspraulsure, milkymist is not on the qi server07:27
wpwrakah, right. that explains it07:29
qi-bot[commit] Xiangfu: PR2028: Milkymist USB: forward MIDI messages. fixed 2012-02-29 (master) http://qi-hw.com/p/wernermisc/9fda79908:30
zumbiviric: doncs ja veus... per aci estem10:52
wpwraksigh. i'll never understand how that mailing list archive works. in the month of february, the mails archived in september (!) shrank by some 3-4%13:03
wpwrakor, and march 2011 grew by 1 kB :)13:04
wpwrakdamn all those time travelers ! can't they be a little more careful about leaving traces all over the place ?13:04
viricI see you trust your brain a lot13:14
viriclet me recommend you drugs.13:14
wolfspraulwpwrak: those are pipermail bugs13:15
wolfspraulit's buggy and the speed of bug fixing is measured in decades13:16
wolfspraulthat's just how it is13:16
kyaki think wpwrak is about to write his own archiver :)13:16
wolfspraulsometimes I do understand sebastien's GNU anger13:16
Action: pabs3 shudders at his mailman & pipermail memories13:17
wpwrakviric: yes, it's probably a bad idea to do statistics when sober :)13:20
viricwpwrak: the speed of bug fixing of a piper? 13:20
viricpipers live slow13:20
wpwrakdon't follow the piper ;-)13:26
qi-bot[commit] Werner Almesberger: ircstats/ML: updated to include Feb 2012 traffic (master) http://qi-hw.com/p/wernermisc/456786813:33
qi-bot[commit] Werner Almesberger: Merge branch 'master' of projects.qi-hardware.com:wernermisc (master) http://qi-hw.com/p/wernermisc/e99d4af13:33
Action: DocScrutinizer51 yawns a bit14:01
DocScrutinizer51boooooring work14:02
DocScrutinizer5110% real work getting done is much. 90 are overhead for tools and orga14:03
DocScrutinizer51while the 20 on top of 100 are for ranting ;-D14:04
DocScrutinizer51figure: no decent reset for peripherals of SoC14:05
wpwrakpower-cycling ? ;-)14:16
DocScrutinizer51ohmy, they hired me for the wrong division, might have been of better use in proofreading chip datasheets/specs14:36
Artyomwolfspraul: you judge too strong... There are number of reasons why I had to refuse from ARM now. But before describing them I would mention that I don't exclude possibility to use it again in developing my current project.18:41
ArtyomReasons to switch to MM SoC:18:42
Artyom1) Most commercial devices use one-chip solution. Couple of examples: Javad's Triumth chip (ASIC) is based on leon3+custom_cores. Namuru project uses single FPGA to run soft-cpu + altera's peripheral cores + famous_namuru_correlator. gnss-sensor also uses single FPGA to run custom gnss-core and leon3.18:45
ArtyomTechical resons:18:45
Artyom2) ARM that I used for first experiments didn't allow to run namuru-core in FPGA and ARM on the same clock frequency. This is not only limitation of demo-board that I use but also limitation of ARM's interfaces.18:47
Artyom3) With ARM system I could use very limited memory (less then 100 kb) comparing to tens of Mb in MM SoC. Ohnestly speaking this is a limitation of the demo-board that I use. There is potential possibility to use external SDRAM-memory, but I don't know how fast it is for my tasks.18:49
Artyom4) It is easier to debug design with one chip then with two chips. I use less programmators. Reliaility of one-chip solution is higher. Price lower. 18:52
Artyom5) For GPS/GNSS tasks FPGA is required in any way. There is no possibility to make correlator in ARM. (Ok, software solution is possible but that would be expensive solution which will require extra-difficult pcb design, soldering BGA-chips, solving signal-integrity issues and so on. That is why I don't consider software-correlator alternative ). So ability to refuse from ARM/DSP/any other...18:56
Artyom...controller seems attractive.18:56
ArtyomBut as i mentioned I don't exclude possibility to use my board with ARM. But this is because I use rather "weak" FPGA (spartan3e500).18:59
ArtyomAnd of coarse MM SoC also have weak sides:19:10
ArtyomUART is very slow for my tasks, lm32 core is synthesised without cache for spartan3e wth xilinx soft.19:18
whitequarkwpwrak: what do you think of this...19:29
whitequarkonce there was an STM32. it still works, except for one pin19:29
whitequarkit's 3.6V19:29
whitequarkand on that pin, a voltage of 15V was applied from a tantalum 10u cap19:29
whitequarkif I'd pour 120% H2SO4 on it and examine it under a microscope, would I see a crater where the protection diode should have been?19:30
kristianpaulArtyom, had you compare you port with this https://github.com/fallen/milkymist-avnet ?20:39
kristianpaulperhaps there is a workaroud..20:39
--- Fri Mar 2 201200:00

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