#milkymist IRC log for Friday, 2011-12-23

cladamwxiangfu, this command "echo 3 | sudo tee /sys/bus/usb/drivers/usb/usb2/../companion", do we need to concern it alway or forget it ?03:59
xiangfuforget it.04:02
cladamwxiangfu, so if i am right, this was to let usb speed to be a full-speed one instead high speed when we were not sure the root cause suffered from nor corruption. so we can now definitely use 1.8 usb cable @ high speed without any reflash problems. Can we?04:09
xiangfucladamw, yes. use 1.8 usb cable @ high speed will be ok.04:10
cladamwwpwrak, one thing I forgot to ask. When you were doing reliable testing on nor corruption, were you always using usb high speed mode to access ?04:11
wolfspraulxiangfu (and Adam): the issue was between full-speed and low-speed, not between full and high08:23
wolfspraul(ah, wrong :-))08:24
wolfspraulI got confused. this was the jtag cable, and indeed between full and high...08:24
xiangfuyes. and now it's not an issue anymore. since Werner have make sure that the RC4 fix the NOR problem.08:32
GitHub68[autotest-m1] xiangfu pushed 1 new commit to master: http://git.io/nd8L6Q09:03
GitHub68[autotest-m1/master] MIDI: clean RX before start test - Xiangfu Liu09:03
wolfspraulcladamw: hi there ;-)09:53
wolfspraulalmost finished the week - how is everything going?09:53
wolfspraulI saw you made lots of edits in the wiki, I'm waiting for that to settle down before I start studying the news :-)09:54
cladamwxiangfu, i just saw back log. Although Werner did fix nor corruption issue. But still not clarify my question about full-speed and high-speed. :-)09:54
wolfspraulyes I also think the two are unrelated, but we can start using the (normal) high-speed and see how it goes09:55
wolfspraulotherwise we drag all sorts of fear with us, without knowing what the actual problem is09:55
wolfspraulwe have crc checks etc nowadays, I trust that09:55
cladamwxiangfu wolfspraul so if Werner tell us that he was _always_ used high-speed mode to access/do experiments then i think that we can definitely say _yes_ we solved this. ?09:56
wolfspraulno 'solved'. we started using full-speed simply to rule out a high-speed problem.09:56
cladamwI don't know actually. :-) so I hope someone can really tell me. :-)09:56
wolfspraulyes, i tell you. forget the full-speed thing now.09:56
wolfsprauljust use (normal) high-speed with the jtag-serial cable09:56
cladamwwolfspraul, aha. okay. because i just thought Werner's experiments is extremely reliable, so I asked what speed he used. :-)09:58
cladamwwolfspraul, i just took back 0x74 from minbo09:58
wolfspraulbut the issues are completely independent09:58
wolfspraulthe reason we went from high-speed to full-speed on the jtag-serial reflashing was because we ran into this 1 board once where we thought high-speed caused a problem09:59
wolfspraulthat is unrelated to nor fixes09:59
wolfspraulbut we can also go back to high-speed now, and see what happens09:59
wolfspraulwe have crc checks, we have a more stable nor circuit09:59
wolfspraulwe cannot forever continue with the full-speed jtag-serial hack simply because "we once had a problem"09:59
wpwrakcladamw: naw, jtag speed was a dead end. i always used the default, which should be high-speed10:00
cladamwwpwrak, ha... great ! That's that i'd like to know. thanks !10:01
wolfspraulyes, good :-)10:01
wolfspraulso... when can we make the final rc3 report?10:01
cladamwwolfspraul, i did tried to clean up 'failure classification' wiki this morning.10:01
wolfspraulthe one and only mega report that will answer all questions once and for all? :-)10:02
wolfspraulcladamw: very good!10:02
cladamwso I think next week I can list them.10:02
wolfspraulI think we should soon come to a final rc3 report10:02
cladamwsince we still have three remaining boards. :-)10:02
wolfspraulok, good. then work on those of course.10:02
wpwrakwhat should be in report #42 ?10:02
wolfspraulrc3 production results10:02
wolfspraulwhich types of problems we ran into and what size they had10:03
wolfspraulafter that report I don't want to spend any more time on rc3, then it's 100% focus on rc410:03
wpwrakcladamw: "42" the answer it took millions of years to compute. (from the hitchhiker's guide to the galaxy :)10:03
cladamwwpwrak, what's your meaning of "#42"?10:03
wolfspraulso we (mostly Adam) should finish all rc3 work before that10:03
wolfspraulit sounds like we can be there in a week or two maybe?10:03
wpwrakafter that report, we'll probably have run out of rc3 problems anyway ;-)10:04
wolfspraulI am just patiently waiting, I know adam has the right priorities and will try to finish this asap.10:04
cladamwoah...man! All such stuffs were totally caused by me though ! surely. ;-)10:04
wpwrakwell, hardware problems. still have plenty left to do on the sw side10:04
wolfspraulyes but then onto rc410:04
wolfsprauldvi-i, and so on10:04
wolfspraulmore usb! :-)10:04
wolfspraulwhat not10:04
wpwrakyeah. M1 now wants usb ;-)10:05
cladamwwolfspraul, yes. after send out report, then go for else you listed above.10:05
wolfspraulok but let's give Adam the remaining time he needs to properly finish rc3 first10:05
wolfspraulcladamw: yes sure, I think we are in agreement. first completely finish rc3.10:05
wolfspraulfull report, yield, list problems, etc.10:06
wolfspraulwe have excellent data in the wiki, just needs to be cleaned up and summarized10:06
wolfsprauland the remaining boards too...10:06
wpwraksure. us westerners are slowed down by xmas anyway :)10:06
cladamw0x74 was removed fpga chip, so today I brought back.10:06
wolfspraulI have no comment now, there were lots of edits and I need to study all this a bit, the various boards and findings etc.10:06
wpwrakcladamw: why was it removed ? was there any visible/measurable defect on the chip ?10:06
cladamwand the others are 0x67 & 0x6d10:07
wpwrak(replacing large bgas to me has a bit of a ring of "kamikaze" to me :)10:07
wpwrakhmm, one "to me" will suffice. still need to get my morning caffeine.10:08
cladamwwpwrak, at that time we ( I and minbo) thought a short might be under fpga, so we removed it out. Before that I still have had 5 set of 'short' rc3.10:08
wpwrakcladamw: ah, so probably just another flux victim. good. these are easier to understand than fpga failures ;)10:09
wolfspraulcladamw: ok, so you are fully on track with rc3?10:10
wolfspraulyou know what to do and how to finish stuff etc? do you need any help or anything from me?10:10
cladamwso we picked one short (0x74) firstly. then decided if there's no short condition on the bottom side of chip and also measured without still 'short' then we illiminute fpga reason.10:10
cladamwthen we found even removed fpga, so still short. so I decided to stop remove fpga from else 'short' board.10:10
wolfspraullet's set a realistic goal for rc4:10:11
wolfspraul100% yield10:11
wolfspraul1 day testing for Adam, done10:11
cladamwthen just found all else short boards shorted by unqualified usb transceiver. :(10:11
wolfspraulyes another sourcing problem10:11
wolfspraulI want to source rc4 with boom and much more over digikey10:12
wpwrakno more shopping trips to chinese backyards ;-)10:12
cladamwwolfspraul, next Tuesday I'll go for ultrasonic cleaner visit, then I know how's going on the other two boards. :-)10:13
wpwrakwere those fake chips actually bought in taiwan ? (as the place of the merchant-sharism interface) or did you find them in china ?10:14
cladamwwpwrak, do you think what else functions of boom we can still improve in the future? Since i know you're not on this way now. :-)10:14
wpwraki'm very curious about how that ultrasound thing goes. i'm also still puzzled about minbo not having it. i wonder how they clean their boards after smt. i can't imagine they just leave all of them "dirty"10:14
cladamwin china10:14
wpwrak(china) mail order ? wolfgang sounded as if he knew those people personally. seems strange to me that he'd go shopping, then mail it to taiwan, though10:15
wolfspraulsure we know them personally, several visits also with Adam etc.10:16
wpwrak(boom) hmm, lots of things to do there :-( starting with the core logic10:16
cladamwwpwrak, well... i  would say that since i issued rework firstly then i also reworked then puzzled my works too. :-)10:16
wolfspraulwe will learn and improve10:16
cladamwcore logic?10:16
wpwrakcladamw: okay, with your rework, it's more understandable. you don't have the infrastructure for proper cleaning. so there's always a risk that you miss something.10:17
wolfspraulI am 100% certain though that if we source 'everything' (that's not the plan anyway) from digikey, we will run into problems there as well10:17
wpwrakwolfspraul: (learn and improve) sure. i'm just curious how the whole story connects. it's a nice story to tell to people :)10:18
wolfspraulcladamw: let me just confirm again: you are clear about the remaining rc3 tasks? do you need any help?10:18
wolfspraulplease let me know when you think you are slowly coming to an end...10:18
wpwrakcladamw: boom is currently written in Perl, weak syntax checking, slow operations.10:18
cladamwthis rc3 run, we made stupid way ourself, and without good verification then reworks more ... etc.10:19
wolfspraulyes but we are not supply bound :-)10:19
wpwrakaren't you case-limited now ? :)10:19
wolfspraul(cladamw - what I mean with "supply bound" is that even though the production process is slow and with painful realizations, we are not sold out)10:20
cladamwthis time with wpwrak's participation, we survived though. Or can't imagine. :-)10:20
wpwrakbtw, if you're low on cases, that may be an opportunity to set some boards aside for developer discounts. rc4 will get a new case anyway10:20
wolfspraulonly one new sideboard I think10:20
wolfspraulI'm not worried about developer discounts10:21
wpwrakcladamw: naw, you found most of the problems yourself :)10:21
wolfspraulwhat m1 needs is a real killer attractive marketing story, a story overall10:21
wolfspraulthat's my opinion and hasn't changed and we all work on that10:21
wolfspraulit's very unattractive right now, still10:21
wolfspraulif any serious developer comes along we will find the discount needed to satisfy that developer10:22
wpwraki think the "VJ instrument" story carries nicely. just needs better software. plus, the "hardware" overhaul sebastien is working on should also help quite a bit10:22
wolfspraulbut the lines in front of our store are not exactly very long ;-)10:22
wolfspraulcladamw: are you clear about rc3? need help?10:22
wolfspraulI just wanted to confirm since it's a friday and going into the weekend etc.10:23
wolfspraulwant to make sure all is good on your side :-)10:23
cladamwphew~ i should take a picture with all packed M1 retailed box which setting at my desk like a large building. :(10:23
wolfspraulyes, but where are the lines in front of your building?10:24
cladamwoah..yes I'll try to fix this 0x74 firstly.10:24
wolfspraulyou can shoot a photo down from your window showing the empty street :-)10:24
wpwraki think there are three groups now: 1) developers who find the platform sexy but are scared by the price tag. 2) VJs who would find it cool but who don't see a workflow yet and who probably haven't heard of it yet anyway. 3) people with fringe needs who'll jump on any chance to get someone to solve their problem.10:24
cladamwwolfspraul, joking...10:24
wpwrakfor 1), we can just make discounts. solved.10:24
wpwrakmost of the marketing successes so far seem to be in the are of 3), which worries me, because there lies the risk of lack of focus.10:25
wpwrakfor 2), i think we need to go to places where VJs hang out. fora, clubs, etc.10:25
wpwrakand we need to be able to show M1 the instrument they will use. not as a technology demo.10:26
wpwraks/M1/M1 as/10:26
cladamwwolfspraul, "are you clear about rc3? need help?" --> clear and no help needed now, thanks !10:26
wolfspraulah ok, good10:26
wolfspraulthen I just wish you a good weekend now, relax a little, and be happy about your Milkymist work!10:26
wpwrakwolfspraul: (empty street) step 1: stock. step 2: have room in the street for the crazy crowd of customers. step 3: launch product that makes crazy customer crowd camp in the street for a week pre-launch. so we're already 67% done ;-)10:27
cladamwoah ~ thanks ! and also wish (westerners and easterners even all ) happy xmas anyway.10:28
wpwrakhehe, thanks ! still lots of things to do there. today, it's infrastructure checks and tests ... (mainly the missile delivery systems)10:30
wpwrak(missiles) in argentina, there are fireworks on the 24th10:30
cladamwwpwrak, one quick question. :-) What's the most size of passive parts in ATUSB & ATBEN? 0402? 0603?10:42
cladamwxiangfu_, thanks ! I saw you updated the new test.bin. ;-)10:58
rohhm. about workflow...11:04
rohthe mm can do 'texture mapping', right?11:04
rohso we could do 'mapping' also? means clipping and doing transformations on geometry to 3dimensional stuff being a 'useable surface'11:05
rohso its 'just' sw which is missing ;)11:08
rohstuff like this http://dai.ly/g4bYa711:09
xiangfu_cladamw, that fixed the midi issue. I already updated the image under: http://milkymist.org/updates/2011-11-29/for-rc3/11:10
cladamwwpwrak, just git cloned ben-wpen, it's 0402. good. :-)11:30
cladamwxiangfu_, okay. I'll try it. :-)11:31
cladamwIn last time I tried to find what part caused 'short' so that removed all passive parts(49pcs)3.3V rail decoupling cap. and beads one by one. now needs to on by one soldering them back and measure. phew~13:14
wpwraksounds like fun :)13:27
cladamw( Xmas gift ) ==>  http://en.qi-hardware.com/wiki/File:M1rc3_packed_1.JPG13:31
xiangfu_cladamw, cool.13:38
kristianpaulnice gifts ;)13:40
kristianpaulcladamw: how long takes you packaga that 28 units?13:41
cladamwkristianpaul, i can't tell you how long i did in 28 units, but if said that one package only, the steps will be:13:43
cladamw1. prepared all accessaries firstly then followed http://en.qi-hardware.com/wiki/Milkymist_One_accessories#packing_instructions without assembled M1rc3 and case. ==> this may in 20 minutes and done13:46
cladamw but I definitely not did like that. so I needed to accumulate like said 5 ~ 10 sets rc3 board with full functions tested, then prepared folding retail box and else etc. :-)13:47
cladamw2. assembling m1 board with case, one needs 35 ~ 45 minutes since removed "small bits and pieces" of case films like this: http://en.qi-hardware.com/wiki/File:Rc3_bits_smallpieces_top_case_after_tearing_film.png13:50
cladamwso before next rc4, i have to find a good way to solve this. :-)13:50
cladamweach bits and small pieces needed to clean well. :(13:52
cladamw3. after assembled case with board, then rendering test again for make sure13:53
wpwrakhmm, one hour per box. about two weeks for rc3. tough :)13:53
kristianpaulno problem, adamn just wondering whare are the plans for packing M1's for next rc4, considering all this steps13:53
wpwrakyou need an assistant :)13:53
cladamw4. put it into folded-well retailer  box includes all stickers/lables. :-)13:54
cladamwwell...this is a start, then we know the details to go for plan in next. :-)13:55
cladamwway in time-multiplexed. ;-O13:58
kristianpaulbut not your hands ;)13:58
cladamwhope next time. :)13:59
kristianpaulgood thinking :)13:59
lekernelhi proppy15:40
lekernelwelcome back :)15:40
lekernelwhat's up?15:40
proppyhey, long time no see15:40
proppyI figured out I would idle there too :)15:40
proppydid you go to ccc 2011 ?15:41
lekernelthe camp? yes. but I'm not going to the congress this year.15:41
proppylekernel: congrats for the redesign of milkymist website btw15:41
lekernelwell it's just a wordpress theme :)15:41
proppylekernel: so you got a r0cket badge ?15:42
lekernelyou want it?15:42
proppyI have a friend here who he looking forward giving his badge away15:42
proppyI was wondering if you played a bit with it15:42
lekernelno, I didn't15:42
proppynp, milkymist is way more interesting to play with anyway :)15:43
wolfspra1lamazing, eh?17:51
wolfspra1lincompetence married to persistance17:51
wolfspra1l"The current thinking is that by removing the ability for this software to be upgraded, we can effectively treat the wireless networking hardware as a circuit, and importantly, not a malicious circuit that can be upgraded."17:52
wolfspra1lno wonder the FSF has lost any and all credibility and momentum to lead17:52
wolfspra1lmay they be happy in their non-free minds... :-)17:53
kristianpaulnon-free minds, heheh19:11
Artyomkristianpaul: hello19:26
kristianpaulHi !19:28
Artyomkristianpaul: I finally could update s3e port to the final version of MM SoC. I switched off all modultes that I don't need (like video in/out, midi, usb and so on). And I could test debugging with GDB. Everything works fine. So now I will try to port my program19:33
kristianpauloh, nice !19:34
kristianpaulbtw in you program, do you check for new  data? i guess all start by looking and accumt int, the wait tic occur and then check new data?19:37
kristianpaulthe clear some flags, load values to registers again, print data and wait for accum int..19:38
Artyomyes, you are right. I use polling of accum_int now. (I also played with reading status from memory but finally decided to use accum_int polling as it was in my old program)19:40
kristianpauland the gold code per satellite taken from namuru datasheet right19:42
kristianpaulbtw had you tested tracking for all of then?19:43
Artyomyes - initial value for shift register is taken from namuru datasheet.19:43
ArtyomRight now I have again problems with acquisition ;) I don't know the reason for now. May be I forgot something to change in namuru-port (Previously I used 80 MHz clock for correlator and now only 48 MHz).19:45
kristianpaulyeah, better to make correlator frecuency a constant inside the correlator code so the ecuation get update easilly19:47
kristianpauli noticed i have to calcultate again some values too..19:48
kristianpaulBtw then a TIC occuured means the a measure is going to start or had finished?19:54
kristianpaulacyually i think i just should care about accumt int itself plus new data not TIC19:55
kristianpaulas TIC seems to be in sync with accum int correct?19:55
Artyomhm... I think TIC is just a local time-signal.... It is used to synchronously lock chanles pseudoranges... Anaway it's the task for distant future...19:56
kristianpaulcause i was confusing with accumt int in some way :)19:56
Artyomyes, TIC and accum_int are synchronous because the are generated from the same clock.19:57
ArtyomWhat has confused you with accum int?19:57
kristianpaulacually more with TIC19:57
kristianpaulbut as you said, dont care for  it now20:00
Artyomkristianpaul: bye ;)20:18
kristianpaul""Use automake, always." Nobody wants to deal with the details of writing makefiles. Automake is ugly, but the ugliness can be ignored; like democracy, it is messy but it is still the best system we have. "22:42
--- Sat Dec 24 201100:00

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