#qi-hardware IRC log for Sunday, 2014-02-09

whitequarkokay I noticed I forgot to drill it00:01
wpwrakor just move to japan and let the earthquakes do the stirring ;)00:15
whitequarkdamn! no free lunch for me00:26
whitequarkthe laser printer has a systematic error00:26
whitequarkactually, hm, no, I don't think it's printer, I've explicitly verified that printer has less than .05mm deviation on 10cm square00:27
whitequarkinkscape? I bet it's inkscape.00:27
whitequarkhah, an interesting physical effect: etchant is so concentrated that if I leave a drop of water, it doesn't dry to leave some white markings01:20
whitequarkrather it dries to become a sphere with a hard crystallized shell and soft "sludge" inside01:20
whitequarkooo you'll like this03:00
whitequarkwpwrak: DocScrutinizer05: hole electroplating http://www.instructables.com/id/Inexpensive-method-of-industrial-level-quality-PCB/03:00
whitequarkhowever it is not your typical ink-based process03:00
whitequarkrather you dunk the board in an activation solution and then heat it depositing an initial layer of copper crystals03:00
whitequarkI'm actually rather inclined to try this03:01
whitequarkprobably more out of curiosity rather than practicality, but who knows03:01
whitequarkthe required calcium hypophosphite is actually easy to obtain here, $7/kg and no restrictions03:20
whitequarkthere are companies which readily ship such stuff to individuals03:20
wpwrak... for they know it'll kill them before they can do anything seriously dangerous with it ;-)03:50
DocScrutinizer05whitequark: (thru hole .ru patent) nice, but what I'm missing are details regarding electroplating04:58
DocScrutinizer05http://talk.maemo.org/showthread.php?p=1411750#post141175007:11
wpwrakvictory at last ;-)08:37
wpwrak"Quantity 1000" ;-)08:38
DocScrutinizer05yoh, sufficient since we probably can get more when we really would need to09:12
wpwrakyeah. and your first batch is safe now.09:13
DocScrutinizer05let's hope it's not ;-P09:13
DocScrutinizer05I still hope for a first batch of 1500 ;-)09:13
DocScrutinizer05since most likely there never will be a second batch 09:14
wpwrakwell, depends on how good your PR will be :)09:14
wpwrakwhy ? if you manage the first, another run will be considerably easier09:14
DocScrutinizer05not from the orders perspective09:15
DocScrutinizer05everybody interested will order for first batch. I don't see a massive run starting after first batch got built09:16
DocScrutinizer05press coverage will happen on prototypes09:16
DocScrutinizer05or PV the latest09:16
DocScrutinizer05maybe the later announcement of "Neo900 privacy care package" might start a 2nd run09:19
DocScrutinizer05VPN and VoIP09:19
DocScrutinizer05"1666 'free' voice minutes with a 100MB/month dataplan"09:20
DocScrutinizer05"100% encrypted phonecalls between Neo900 and to all SIP phones supporting ZRTP"09:21
DocScrutinizer05"Neo900 group recommends twinklephone for KDE"09:22
nicksydney_anyone has tried to buy any IC from this website http://www.semiconductorstore.com/ ? seems like their price is very reasonable10:25
nicksydney_wow! another open source hardware store site http://www.inmojo.com/10:37
DocScrutinizer05http://www.semiconductorstore.com/ seems down for me10:41
DocScrutinizer05500 - Internal server error.10:41
DocScrutinizer05There is a problem with the resource you are looking for, and it cannot be displayed.10:41
nicksydneyDocScrutinizer05: weird...just tried 10mins ago10:44
nicksydneyseems like not credible looking at the current site now :)10:44
whitequarkoooo I really like this resist12:30
whitequarkI printed out a calibration pattern, it's quite clear even at 30s@3W exposure12:31
whitequarkand optimal time seems to be 120s12:31
Action: paul_boddie was just reading the Neo900 shipping to Russia arguments12:31
paul_boddieSheesh! Is it "hearsay" when stuff like that is actually reported on the BBC?12:32
paul_boddiehttp://www.bbc.co.uk/news/world-europe-2587886912:32
nicksydneyDocScrutinizer05: the website http://www.semiconductorstore.com/  is up :)12:32
whitequarkpaul_boddie: yes it is a huge problem here12:32
whitequarktaxless customs limit was initially planned to be lowered to $150 per *month*12:33
whitequarkthough it seems to have been changed to $150/package like in .de12:33
whitequarkthat being said, dhl and fedex didn't really like to work with private individuals before that12:33
paul_boddieThey started arguing a bit about tax-free limits here in Norway again.12:34
paul_boddieThe right-of-centre parties want the limit to go up from 200 NOK to 400 or even 1000 (per shipment).12:34
paul_boddieBut some business organisation wants it to go down to 0.12:35
whitequarkeh?!12:35
whitequarkwhat is the point12:35
paul_boddieThe point of the limit? It's the threshold above which you pay tax on the value of the goods.12:36
paul_boddieMoving it up or down either "punishes" or "rewards" local business, supposedly.12:37
paul_boddieI wouldn't have a problem paying VAT on everything shipped from abroad provided that...12:37
whitequarkI see12:38
paul_boddie(1) I could deduct foreign VAT, and (2) the postal service didn't rip everyone off with their "service" of importing the goods.12:38
whitequarkI wouldn't really mind paying VAT too, even with its ridiculous 30% rate, I can afford it12:38
whitequarkbut the problem is that interaction with customs is absurdly painful12:39
paul_boddieI just don't want to pay double VAT. Oh, and VAT on the post office's "service" charge.12:39
wpwrak(russian restrictions) if the article is as accurate as the one about argentina then i would take it with a few grains of salt ;) http://www.bbc.co.uk/news/world-latin-america-2583620812:43
wpwrak(the title is correct, but what follows is half-truths at best)12:43
whitequarkwpwrak: I'll be able to speak about it concretely after I ship the next parcel in a few days..12:44
wpwrakyou must have made their major importers watchlist by now ;-)12:45
paul_boddiewpwrak: Well, who is truly able to keep on top of what the Argentine government is doing?12:45
paul_boddieIt's Whitequark Industries now!12:45
wpwrakpaul_boddie: not even themselves ;-)12:45
whitequarkwpwrak: that is quite likely actually12:45
whitequarksince I truthfully declared what is classified as "industrial equipment"12:46
whitequarkit is both flattering and inconvenient at the same time :p12:47
paul_boddiewpwrak: I think people just started to give up trying to follow them. The Economist refused to print their statistics and more or less called them liars.12:48
wpwrakhah. what would Marx say about that ? industrial equipment in private hands ? never ! comrade whitequark, you have been a very bad boy. we'll have to confiscate all your toys.12:48
paul_boddieNo, he's surely seizing the means of production from the borgeois ruling classes.12:49
wpwrakpaul_boddie: well, that is certainly true. the official inflation figures are in the 10% p.a. range. more credible sources have 25-30%12:49
wpwrakpaul_boddie: ah, the new strategy - just buy up all the means of production from the capitalist class enemy, so they'll starve. clever.12:50
paul_boddieWell, if the capitalists are Anglo-Saxon, they'll probably sell you the means of production and rent it back just to get some quick cash.13:01
Action: whitequark doesn't get it13:03
paul_boddieI think it was another wpwrak joke about life in the former USSR.13:06
whitequarkno, the anglo-saxon one13:06
paul_boddieBut the selling off and renting back thing is classic modern capitalism.13:07
paul_boddieBusiness owners thinking that they have too much money "tied up" in assets and want some more to spend, so they sell the assets off but still need them to actually make money.13:08
wpwrakyeah, lots of fun stuff there. also in the form of cross-border leasing. usually pretty ruinous.13:09
paul_boddieYou get this in "economically liberal" countries where they privatise essential services or industries: there's a "lottery winner" moment of being given lots of cash, but then you have to pay for everything.13:10
paul_boddieWhere "you" is the state and, of course, the individual (via taxes or fees for things that used to be included in taxes).13:11
whitequarkah13:11
paul_boddieJust be sure not to fall into this trap: https://www.youtube.com/watch?v=DcC_iEHiFZs13:14
wpwrakwell, the taxpayer pays either way. but usually prices go up, service quality goes down, and maintenance gets neglected when you take that route. there are countless examples for this.13:15
Action: whitequark . o O ( what if I add H2O2 to (NH4)2S2O8? )13:38
wpwrak"what if we dropped it from higher up ?" http://what-if.xkcd.com/57/13:46
whitequarkexactly13:48
apeletelarsc: documentation done: http://apelete.seketeli.net/writing-an-musb-glue-layer.html13:49
apeletecomments welcome :)13:49
larscYou are the musb expert, not me ;)13:50
apeleteyou're kidding me :)13:54
larsclooks good, very nicely done13:56
whitequarkDocScrutinizer05: ooo you've been right all along and I was wrong13:58
whitequarketching doesn't noticeably cool the solution and it was just cold from before13:59
whitequarkI'm now etching at 30° and it's almost done in 20 minutes13:59
apeletelarsc: looking at how to convert this in dockbook format to submit it to upstream kernel documentation14:00
apelete(actually written in rst format: http://git.seketeli.net/cgit/~apelete/about-me-hacking-and-tao.git/tree/content/writing_musb_glue_layer.rst)14:00
apeletelarsc: looks good enough to you ?14:00
larscyes14:02
apeleteokay thanks, will fix a few things (forgot to add acknowledgements chapter to the table of contents) and convert it14:04
pcercueiS2is it safe to lock a spinlock in the interrupt handler?14:08
whitequarkI don't think so14:08
larscyes14:09
larscbut you need to make sure to use spinlock_irqsave when you lock the same spinlock outside a irq handler14:10
larscor spinlock_irq14:12
larscif you know that the spinlock will always be called with interrupts enabled you can use spinlock_irq14:12
larscif you don't know use spinlock_irqsave14:12
larscpcercueiS2: https://www.kernel.org/doc/Documentation/spinlocks.txt14:20
CYB3RpcercueiS2: hi! could you please share your jz-boot configuration?14:27
pcercueiS2ehrm14:27
pcercueiS2I don't have any specific configuration14:28
CYB3RI mean that config files14:29
pcercueiS2I don't use anything else than the default config14:30
CYB3Rso you just 'nprogram' your image to nand?14:31
pcercueiS2yes14:34
pcercueiS2nerase one block first14:34
pcercueiS2then nprogram14:34
CYB3RDo you have only 2 binary images? Bootloader and ubifs?14:36
pcercueiS2well I never flashed an UBI image using jz-boot, when I did it it was through a running system14:38
pcercueiS2but I believe you can do that, yes14:39
pcercueiS2beware, ubiboot will run a kernel stored on a UBI volume, but it does not understand ubifs14:40
CYB3RpcercueiS2: I can't understand it. How could it load a kernel from ubi volume without ubifs?14:41
pcercueiS2well, UBI is not a filesystem, it's a FTL14:42
pcercueiS2ubifs is a filesystem designed to run on top of UBI14:42
CYB3RWhat does FTL stands for?14:43
pcercueiS2basically, UBI provides you with volumes (sub-partitions, sort of)14:43
pcercueiS2where you can store raw data, or ordered data with a filesystem14:44
pcercueiS2flash translation layer14:44
pcercueiS2it's mostly a software layer that provide wear levelling and handling of bad blocks14:45
CYB3RIs it kind of abstraction on NAND flash?14:48
pcercueiS2yes14:49
pcercueiS2without it, it would die in no time14:49
pcercueiS2SD cards have such a layer too, but it's internal to the card and not exposed to the SD card reader14:50
CYB3RDo UBI and MTD have similar functions?14:50
pcercueiS2MTD gives you low-level access to the physical blocks of the NAND14:53
pcercueiS2UBI spreads its logical blocks on the NAND device; two consecutive logical blocks in UBI may not be physically consecutive14:54
pcercueiS2you could store data directly on the NAND via the MTD nodes, but you'd trash your NAND in no time14:56
pcercueiS2that's why UBI exists14:56
CYB3ROkay, it's some memory controller logic to manage bad blocks and all this things. This is a good thing. How do I start using UBI?14:57
pcercueiS2I think ubiboot expects the UBI to exist at a specific address and size on the NAND14:59
pcercueiS2basically create a file of the correct size, create an UBI layer on it, create a volume named "kernel", then write there your vmlinuz.bin file15:00
pcercueiS2and then flash the image at the correct address using jz-boot15:01
CYB3RCan I write a file to volume with no filesystem?15:01
pcercueiS2yes, and that's what ubiboot expects15:02
pcercueiS2for detailed steps you'll have to search the web, I don't remember the commands15:02
CYB3RI don't understand this part about writing file to volume with no filesystem15:02
pcercueiS2...15:03
CYB3RDo you use dd with an offset for this?15:03
CYB3RI'll try searching tools for ubi15:03
CYB3RBtw, is SDRAM initialisation code in UBIBoot taken from ingenic's nand-boot?15:08
DocScrutinizer05whitequark: 30min sounds exactly OK19:15
DocScrutinizer05((<CYB3R> Can I write a file to volume with no filesystem?)) hardly since then you got no bad block management at all19:40
DocScrutinizer05since there's no bad block handling on hw level and in mtd drivers, you need something in your "filesystem" drivers to do that. dd and similar tools don't have any idea about bad blocks19:42
DocScrutinizer05ubifs does19:43
DocScrutinizer05usually the bootloader checks first 4 blocks to find a vmlinuz image header. vmlinuz afaik is a "self extracting archive" and should manage (and skip) bad blocks while decompressing. During flashing the tool that is flashing your kernel will need a method to handle and "skip" bad blocks19:45
DocScrutinizer05wpwrak probably knows much better than I do19:45
pcercueihey larsc 19:57
pcercueiat which hour should I show up tomorrow?19:57
DocScrutinizer05ooh, sorry for the nonsense about vmlinuz, when you use UBIboot it is obviously the UBI FTL that handles the bad blocks, and thus you need a UBI aware tool to flash vmlinuz to NAND20:09
DocScrutinizer05I don't know what exactly uBoot is using for FTL, and I never looked into UBIboot before20:10
--- Mon Feb 10 201400:00

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