#milkymist IRC log for Wednesday, 2011-07-20

aw0x30: keeps rendering well through over night01:54
awadded one "Rendering" column to show the result: http://en.qi-hardware.com/wiki/Milkymist_One_run_3_schedule#Test_Results03:14
kristianpauloh, ISE 13.2 lets check out..03:33
aw0x30: U7's input signal from fpga is very good, it's ouput pin 4 is not 5V pulse, then I re-soldering U7 pins again then still no output well. so I decided to replace a new one later.04:28
wpwrak_aw: but pin 5 (of U7) is a clean 5V ?06:11
awwpwrak_, yeah...much clean 5V06:11
wpwrak_aw: sounds nasty then.06:12
awR56 220 ohm also good06:12
wpwrak_aw: have you measured U7.4 to ground resistance (when unpowered) ?06:12
awdon't know, just soldered new one. ;-) let's test. :)06:13
awhmm...good idea. No.06:13
awgood..now PASS. ;-)06:16
wolfspraulaw: you record the U7 findings and exchange in the test results spreadsheet, right?06:20
wolfspraulwe did pretty well on the rc2 run, with the 'notes' column remembering all actions, so we should do that again06:21
awwolfspraul, recording now. ;-) also will add another test logs. ;-)06:21
wolfspraulgreat06:21
wolfspraulthen on to 0x31 :-)06:21
wolfspraulso we have the first full-function board with 0x30 ?06:21
wolfspraulprobably 0x2F had too many reworks and we consider it internal/Adam use only06:22
wolfspraulcorrect?06:22
awbtw, we need to tell xiangfu that when we test only like MIDI from menu selection, after i press "h", test tool should tell me a "h" selected. ;-)06:22
wolfspraulmaybe you can also take a picture of 0x30 on both sides to document the wire...06:22
awyes, let's let 0x2F to delievered by Adam firstly..06:23
wolfspraulbut 0x30 looks perfect now, right?06:23
awyes, take pictures later06:23
awyes06:23
wolfspraulmaybe you can leave 0x30 rendering for 24h or so06:24
wolfsprauljust plug in the power, boot, let it sit somewhere06:24
wolfspraul"24h rendering test" is a good benchmark, a 'full day'06:25
wolfspraulman we could invent so many more tests. temperature cycles for example.06:25
wolfspraulor drop tests (with case assembled of course)06:26
wolfspraulor or or06:26
wolfspraultemperature cycles would be quite interesting actually, I think. but very time consuming so maybe later...06:26
awhttp://downloads.qi-hardware.com/hardware/milkymist_one/production/rc3/test_results/30-results06:27
awha...so far haven't tested a 24-hour on 0x3006:28
wolfspraulyes, but I propose you do that06:28
wolfsprauljust let it render, put a small note on it marking the start time, so we know it's really 24h and don't forget the times06:28
awlet's me do this on just plugging it on power and put it aside another place.06:28
wolfspraulyes, and mark the time, just put a note on it06:28
awnow my table is messy though. ;-)06:28
wolfspraulotherwise we forget, especially with more boards later, it will get chaotic06:28
awsure06:29
awhttp://downloads.qi-hardware.com/hardware/milkymist_one/production/rc3/test_results/30-results06:29
wolfspraulyes saw it06:29
wolfspraulaw: I am not proposing this for rc3, but what do you think about the value of testing real temperature cycles?06:29
wolfspraullike 20 times up to 60 degree, down to 0 or less, back up, etc.06:29
awno no, i added a newer test on the bottom06:29
wolfspraulI think most OEMs are doing temperature cycle testing, do they?06:30
awyes, why not! they indeedly do that.06:30
wolfspraulwe have never done it, and we don't have programmable cases either to heat and cool in cycles06:30
wolfspraulyes I know :-) but we never did06:30
wolfspraulso I don't even know what interesting or uninteresting results one could find06:30
wolfspraulbut since everybody is doing that, there must be something valuable there...06:31
wolfspraulanyway06:31
wolfspraulnot in rc3!!!06:31
wolfspraultoo time consuming06:31
awbut I can only to use a heat air to rise temperature to 60 degree, then blow cool air down...cycling06:31
wolfsprauland we don't have the equipment either for controlled heating and cooling06:31
wolfspraulno no06:31
wolfspraul:-)06:31
wolfspraulI said: not for rc306:31
awyes,06:31
wolfspraulI'm just saying that's one big test area that Sharism Ltd. has never actually done right06:31
awor go find a labortory to do test after thic rc3.06:32
aws/thic/this06:32
wolfspraulI think a small case that can do programmed heating/cooling shouldn't be too expensive06:32
wolfspraulsomeone can build one with a few Arduinos :-)06:32
wolfspraul(I am joking Adam, please don't start working on this...)06:32
wolfspraulwe keep that for later when we are Apple sized06:33
awoah~yeah...i saw Make magzine had have someone did such things before with a small box alike.06:33
wolfspraulthere you go06:33
wolfspraul:-)06:33
awwell...time to record and check 0x3106:33
wolfspraulI am wondering what kind of value one can get from temperature cycle testing.06:33
wolfspraulthat would be the first thing I would want to understand. need to talk to someone at a large OEM.06:34
wolfspraulyou can get a lot of data, OK. but why is the data valuable? What kind of decisions do you make based on that data?06:34
awsure sure temperature test chamber can verify design06:34
wolfspraulyou change certain components?06:34
wolfspraulyou can measure/predict the life expectancy of a board, do the temperature cycles imitate aging? etc. etc.06:34
wpwrak_wolfspraul: one thing you'd find are parts with incorrect tolerances, be it by design or by manufacturing06:35
wolfspraulprobably you could also throw humidity tests in there...06:35
awand also capture which part is not on best balanced value, like wpwrak_  did very well on checking C238's value for lower bound or upper bound validation.06:35
wpwrak_wolfspraul: not sure what temperature changes will find. could be just temperature gradients (e.g., no all parts heat/cool at the same speed), could be thermal stress. you'd need a lot of cycles for the latter, though.06:36
awand with them can know also parts tolerance if took well consideration while designing... ;-)06:36
wolfspraulI just know that all OEMs do temperature cycle testing quite consistently, and we don't.06:36
awyeah..."thermal stress" to figure/verify design06:36
wpwrak_aw: heh, you did the low/high tests. i merely suggested values :)06:36
wolfspraulbut it's very time consuming, that's for sure06:37
wolfsprauland needs a lot of boards06:37
wolfsprauland it will generate a lot of data that then needs to be analyzed, not discarded06:37
wolfspraulwe cannot do it now06:37
wpwrak_wolfspraul: i think it would be good to have tests at the ends of the allowed/expected temperature range as soon as possible. see if the board still works on a hot day. see if it still works if you bring it in from the cold.06:37
awwpwrak_, well we are not close and/or big company, but with your tools or design's parameters all taken into considerations then design goes better. ;-)06:38
wpwrak_wolfspraul: M1 should be subjected to rough use, so such things matter much more than to, say, a desktop PC that sits in a nice 20-30 C all day long06:38
wolfspraulI think it matters for all consumer electronics.06:38
wpwrak_wolfspraul: it does. but more for some than for others :)06:39
wolfspraulbut I do not have the resources now to go into that, maybe we can do some very simple tests like you said with 'hot day', but most likely the results will be inconclusive then.06:39
wolfspraulI think the only way to tackle it is to make the cycles reproducible, with a controlled heating to a controlled temperature, several boards (at least) for comparison, etc.06:39
wolfspraulotherwise you will look at noise06:40
wpwrak_wolfspraul: M1 is aimed at people who move around, encounter unknown and non-ideal working conditions, may have time pressure, have very embarrassing failure scenarios, etc.06:40
wolfspraulso they have to buy 2!06:41
wpwrak_that helps if one flat out fails. but what if there's a design problem that breaks it in general < 10 C ?06:41
wpwrak_and you're VJ'ing a christmas party in moscow ? arrived late, because of too many drunk drivers in the streets06:42
wpwrak_and yes, you can't do much more than simple tests for now. still, you should do them :)06:43
wolfspraulthere will be no systematic temperature cycle testing for the rc3 run06:43
wolfspraulbut I have an eye on it, for sure. first would like to understand the value, i.e. how many good feedback to expect. and the setup and execution needs to be in a controlled way.06:43
wolfspraulnot random, then we will only look at noise06:44
wpwrak_for cycle testing, you'd first have to find out what to look for, yes06:44
wolfspraulI do have extensive experience in how _NOT_ to do temperature cycle testing :-)06:44
wolfspraulso I know there is something to discover there, but need to be careful the first time around to really extract value06:44
wpwrak_but you should be able to, say, put a M1 in the fridge and see how it likes that. or take a hair dryer to it06:44
wpwrak_ah, what was the negative experience ?06:45
wolfspraulyes, we can try. but I'm telling you the results may be inconclusive.06:45
wolfspraulbecause we don't really know the temperature some part of the board were cooled/heated to or not06:45
wolfspraulin the end we will ignore the results, whatever they are06:45
wolfspraulnot good06:45
wolfspraulso the fridge/hair dryer idea, don't know. fridge maybe better than hair dryer.06:46
wpwrak_(ignore results) there's that risk :)06:46
wpwrak_fridge is easier and more controlled :)06:46
wolfspraulit has to be reproducible, and you most likely want to subject several boards at once to the same test06:46
wolfspraulthis being so time consuming, you easily get lost in noise, or too much data points06:46
wpwrak_for getting a first feeling, i'd say one board would be enough. you'd only want to go into more extensive tests is there are problems06:47
awbtw, last time we had have good repeated MIDI test image to test. so i just used that to see signals, that's good tool. so while signals always sending then I can keep on eye on circuit.06:47
wolfspraultoo easy to discover noise06:47
wolfspraulso first we need an approach that we believe masks the noise06:47
awhope later DMX or other accessory can have such great tool to debug. ;-)06:47
wolfspraulthen do the test06:47
wpwrak_e.g., if you say the minimum operating temperature should be 10 C, test it at 0 C. that'll take care of most variations.06:47
wpwrak_(noise) only if your tolerance margins are too small06:48
wolfspraulat our former employer (that's where my only limited experience comes from), temperature cycle testing was a joke06:48
wpwrak_the expected outcome should be 0 problems, with any board that passes QC06:48
wolfspraulpeople knew they had to do it, but nobody knew what to do with the results :-)06:48
wolfspraulso we need to avoid that06:49
wolfspraulanyway, it won't happen for rc306:49
wolfspraulif we can sell the rc3 run and go for rc4, maybe we can do some more investments along those lines06:49
wpwrak_well, those phones hardly worked even if you didn't stress them at all ;-)06:49
wpwrak_remember #1024 ? (-:C06:49
wolfspraulit would be cool for sure, also to establish a quality standard, certain standardized cycles06:49
wolfspraulbut those cycles must relate to actual user problems, not be some random number06:50
wolfspraulit must correlate with actual life expectancy of the boards...06:50
wolfspraulthat's the tricky part, I think testing and collecting results is relatively easy.06:50
wolfspraulso we need someone with experience to tell us which cycles to test in order to verify the design, the life expectancy, how to mimick aging, etc.06:51
wolfsprauljust my thinking...06:51
wolfspraulthe famous 80/20 rule surely applies to that subject as well06:52
wpwrak_again, cycle testing != probing the operating range. cycle testing is much more advanced. but if you get complaints that M1 crashes in the middle of a show on a hot night or doesn't boot, etc., your VJ customers will not like to hear that you never tested it beyond 28 C06:53
larsc'put device into fridge for 30 minutes before operation'06:56
wpwrak_"operate between 20 and 25 C. maximum humidity 50%."06:57
wpwrak_reminds me of those labels on clothes. something made of cotton, "hand-wash cold. do not dry."06:58
wpwrak_in the old days, those labels told you what the fabric and the color were designed to survive. nowadays, it's what the bloody lawyers think will cover the company's ass.06:59
wpwrak_i find those most interesting that specify a maximum temperature that's well below normal body temperature. "warning: actually wearing will ruin it" ? or maybe it's their line for zombies and other corpses.07:05
larscdo you wear your shirts on the inside?07:15
awhttp://en.qi-hardware.com/wiki/File:M1_rc3_hw_fix2_bottom.JPG07:22
awhttp://en.qi-hardware.com/wiki/File:M1_rc3_hw_fix2_top.JPG07:22
wolfspraulaw: can you find a wire with black insulation?07:24
wolfspraulwould probably look a little better...07:24
wolfspraulalso I'm not sure about the tape, maybe some drops of silicone or so are better along the way? or not?07:25
awoah~ yeah...you're right. ;-)07:25
wolfspraulbut small drops, only to fixate the wire07:25
awneed to buy black insulation one. ;-)07:26
wolfspraulyes07:26
wolfspraullooks better07:26
wolfsprauland how about small drops of silicone instead of tape?07:26
awyes, that picture haven't used silicon, seems that I'ven't assembled them into case. so not glue them yet. ;-)07:27
awwill do glue after I confirm m1 board fit with case well then. ;-)07:28
awafter 0x31, i assemble them.07:28
aw0x30 now let it rendering to 24 hour test07:30
wpwrak_larsc: on the inside .. like on my body, below a jacket (in winter) ? yeah :) besides, ambient temperature around here can easily exceed 30 C (in summer, occasionally even in winter - global warming is fun :)07:42
larscwpwrak_: i meant on the iside of your body ;)07:48
lekernelaw, so you had a defective midi optoisolator?08:15
wolfspraulno, I think it was a schmitt-trigger08:17
wolfspraulanyway, 0x30 is in perfect condition now and doing a 24h rendering test08:18
lekernelok08:18
lekerneland what about the 0.7A board?08:18
wolfspraulhe's working on that now (that's 0x31)08:18
wolfspraulalso I think he had an idea how to improve the DMX test, I didn't fully understand it but improved test is always good. hopefully xiangfu can help on that one.08:19
wolfspraulxiangfu: did you see adam's comment about the midi & dmx test above?08:20
wolfsprauladam wrote "btw, we need to tell xiangfu that when we test only like MIDI from menu selection, after i press "h", test tool should tell me a "h" selected."08:21
wolfsprauland also "last time we had have good repeated MIDI test image to test. so i just used that to see signals, that's good tool. so while signals always sending  then I can keep on eye on circuit.08:22
wolfspraul"08:22
wolfspraulso maybe the midi and dmx tests can continue to go in circles, until a button is pressed? then we could just have one test image and it can be used to look at signals as well, if needed.08:22
xiangfuok. I will look into ti08:24
xiangfuit08:24
wolfspraulbasically you run the test, if it fails you print FAIL and finish. if it succeeds, you print "PASSED (retrying until button press)" and continue to run the test again08:25
wolfspraulthat way the same test image can be used to measure signals if needed, and if not needed it's just one more button press08:26
wolfspraulI don't understand what Adam meant with the "press 'h'" comment, maybe it's clear to you?08:26
wolfspraulthat retrying thing should be done for midi & dmx for now08:26
xiangfunot sure. I can talk with Adam08:27
wolfspraulok, but the retrying until button press is clear?08:27
wolfspraulonly print it once, so it doesn't clutter the log, but keep retrying indefinitely until button press08:28
wolfspraulsorry I mean 'key press' (on the keyboard)08:28
xiangfuI will working on those 1 hours later. a little bit tied08:28
wolfspraulsure, it's not urgent I think. just an idea for improvement.08:28
wolfsprauloh I think I understand Adam's comment with the 'h selected' now. Maybe he means that when he presses 'h', there is no record of that in the log (=on screen). So it just needs to add a printf to document that 'h' was in fact pressed. That's my understanding.08:33
awwolfspraul, yes, so that i know what item I selected. ;-)08:34
awprint that item again like "h ->" or somethings08:35
aw1. foreign matter on fpga's top body, that one I not sure if it's like burned inside when high currents was happened on my first time power on.09:19
GitHub168[extras-m1] yizhangsh pushed 1 new commit to master: https://github.com/milkymist/extras-m1/commit/f7dc40599b6d262f49f05a1623c47e1bfe883d0909:22
GitHub168[extras-m1/master] added shadow to text 'Live' - Yi Zhang09:22
aw2. now try to use Eliminates the law thus I'd like to disconnect L7 to cut off 3V3 (which is the most power sources in m1 ), if now high current occurs, might be some where short about 3.3V circuits (including fpga's supplies)09:22
aw3. if no high current occurs, then narrow potential failure area though. ;-)09:23
aw for 1) if this to be true, cant see bga solderabilities condition under fpga unless a X-ray took. ;-)09:28
awhttp://downloads.qi-hardware.com/hardware/milkymist_one/production/rc3/test_results/0x31_fpga_body_top.png09:32
wolfspraulaw: you think that 'foreign matter' comes from inside the fpga?09:34
awnot sure now, i just took picture09:35
wolfsprauldoesn't look like it. let's see what the L7 test shows...09:35
awNo09:35
awjust spot somethings now. ;-)09:35
wpwrak_(foreign matter) could be a scratch09:36
wolfspraulaw: I don't remember we had this type of problem in rc2, let's hope it's not something bigger in the whole rc3 run...09:36
wolfspraulbut one by one, now they are all soldered already anyway09:36
wolfspraultwo boards were good09:37
awjust measured C98 (3V3)  to ground resistance (when unpowered) then is 30 ohm, which is not correct to 0x30( 1M ohm )09:37
awsurely I disconnected L7, and now I can very surely impendance 3V3 to ground (unpowered )is very BAD. :(09:38
awnow keep using  Eliminates the law. if someone has new idea, let me know. ;-)09:39
wpwrak_30 ohm would be just 100 mA :)09:39
wpwrak_is this a two-wire or four-wire measurement ?09:39
awyeah...so any steps on power on again needed to be solid judgement first09:40
awtwo-wire09:41
wpwrak_one test would be to just power the board and probe for the component that gets hot. of course, if you're not sure that the board is already nearly dead, that's no a nice test09:41
wpwrak_a thermal imager would help. wish they were affordable ...09:42
wpwrak_(2 wire) okay, so it's anything between 0-30 Ohm :)09:42
awbut 0x30 has at least 1 M ohm . ;-)09:43
awso I don't want to power L7 again... could be immediately damage some where . ;-)09:43
awunpowered , i can just measure quickly somewhere. ;-)09:44
wpwrak_1 M is nice. but i wonder how reliable that is. but yes, the difference 1 M vs. 30 Ohm (or whatever) is striking09:44
awyeah...1M vs. 30 ohm already..damn09:44
lekernel_what happens when you power that board with L7 removed? no more overcurrent I'd guess?09:49
lekernel_this just looks like some short circuit or bad chip ...09:50
awi guess too. but before power on again....i am looking more..hope spot some else09:50
lekernel_if all else fails, you can use the brutal technique: apply power without current limitation, and see where the smoke comes off09:50
awha~see directly some chip burned. ;-) also a good idea though. ;-)09:50
lekernel_there's a risk to burn PCB traces, though09:50
awno... ;-) I don't like smoke air though. ;-)09:51
lekernel_so i'd keep this technique as 'last resort' ;)09:51
awha09:51
wolfspraulnormally would be x-ray now, if we cannot compartmentalize more with L7 on09:52
lekernel_or you can use an intermediate - use current limitation and power on the 3.3V rail directly... it must get hot somewhere09:53
wpwrak_do all chips have the right orientation ? :)09:54
wolfspraultoo late for xray today, hmm. can't we remove some other parts of the circuit after l7?09:54
wolfspraulthe 'check where it gets hot or burns' technique is a little unusual I'd say :-)09:54
wolfspraulalso, Adam may get more boards soon. it's a huge difference whether this is an isolated case on 2-3 boards across the run, or affects 50% of the run (for example)09:55
wolfspraulso if he looks at the next 5 boards and they are all ok, he would normally postpone looking at 0x31 now09:55
wolfspraulbut if we get another 1-2 like that in the next 5 -> big problem :-)09:55
wolfspraulso... aw - when do you get more boards? does minbo have xray? can we remove some other parts after l7?09:56
wolfspraulthose are my ideas09:56
wpwrak_(check where it's hot - unusual) naw, that's actually about the first one you learn ;-)09:56
wolfspraulno factory works like that09:57
wolfspraulxray09:57
wpwrak_no, hopefully not the factory ;-))09:57
awfrom now on, every board before I tested. I'd like to quickly measure (unpowered) impedances on TP1 ~409:59
wolfspraulgood idea10:00
wolfspraulhow to continue on 0x31 now?10:00
aw0x31 , even I get good condition without L7, this cant help me restore 0x31 now.10:01
awso let's not do any more on this 0x31 now10:01
lekernel_no, but you know there are no further problems on the power supply10:01
awwhen after finish all batches rc3, then back to see this board...10:02
wolfspraulwell, that's assuming it's an isolated case10:02
wolfspraula bit risky after seeing 3 boards10:02
awlekernel_, sure , i agreed. but I have other things to do not just this board . ;-)10:02
wolfspraulaw: when do you get more boards?10:02
wpwrak_wolfspraul: do you suspect it's a through-hole problem ? if not, knowing now or later wouldn't make much of a difference10:03
awthey said to me yesterday at fast on this Friday. at last on Saturday.10:03
awwpwrak_, hup?10:03
wolfspraulno idea, can be lots of things10:03
wolfspraulit's just an economic problem, not technical. is this 1 board of 80, or 40 of 80? we don't know10:04
wolfspraulthat's the #1 thing to know now10:04
wpwrak_actually, can a board without through-hole do anything useful ? (such as blink some leds)10:04
wolfspraulif it's 1 of 80, we may never investigate further10:04
wolfspraulever10:04
wolfspraul:-)10:04
awokay...i can still quickly to know if other circuit caused that, when disconnected L7, not big deal. ;-)10:04
wolfspraulwe use the board as a spare part lab supply10:04
wpwrak_aw: if it was a through-hole problem, then identifying the problem may provide useful feedback for the fab (because they haven't done through-hole yet). if it's something else, the damage is already done.10:05
wolfspraultoo late for feedback10:05
lekernel_there's almost nothing through-hole on those boards10:05
lekernel_it's merely the connectors10:05
awactually i'd like to jump to assemble case with m1 to settle how/where i must to glue.10:05
wolfspraulthat's what the test of the first board provides10:06
awthough hole problems?10:06
wpwrak_lekernel_: so would it make sense to do a power-up test (with current measurement) on boards after SMT, before TH ? (during production)10:06
lekernel_no10:06
wolfspraulnah come on, it's a short somewhere. this is normal. biggest problem is that adam cannot quickly look at some more boards, then we wouldn't even discuss this.10:07
awwell...rc1 & rc2 went through a real through hole reflow process before. will it suddently happened strange thing on rc3?10:07
wolfspraulso we don't know how rare/common the 0x31 problem is across the run10:07
wolfspraulit's a bit unsettling to me, with us only seeing 3 :-)10:07
wolfspraulbut well10:08
wolfspraulno risk no fun10:08
awi bed something if need to damage this board. probably my soldering skill will be the number one to kill it. ;-)10:08
lekernel_btw, does the AOI work this time?10:08
lekernel_I remember seeing an horror picture with capacitors on run2 that the AOI didn't spot10:09
awthe remaining boards I didn't bring back. I still request them to do AOI. since this time they said the qty is enough to collect hard data to train machine.10:09
wpwrak_(aoi) cool !10:10
awi could only say these firstly three boards haven't been scanned by AOI and soldered through hole parts by myself. :(10:10
wolfspraullekernel_: AOI is also a matter of how much time is invested to make it find things10:10
wolfspraulyou can always let boards run through the AOI machine, but how useful that actually is is another matter10:11
awnow, without L7: 0.38A only as we guessed, not big deal...since I compared to 0x2f on TP1~Tp4. ;-)10:14
wolfspraulaw: which parts are powered by L7 ?10:16
wolfspraulor behind L710:16
awfor other 3V3 circuits, it will spend more times to eliminate. ;-)10:16
awphel....fpga, ethernet,flash, usb transciver, vga encoder, ...10:18
lekernel_do all chips have the right orientation?10:18
wolfspraulok10:19
wolfspraulmaybe put this board aside and see what the next 3-5 look like10:19
wolfspraulor xray10:19
wolfspraulprobably faster and better than soldering around10:19
awyes,10:19
wolfspraulaw: does minbo have an xray machine?10:19
awyou can imagine those boards are pass through smt machine..so they should be all on the same directions excepts the fisrt panel they starts to mount. ;-)10:20
awwolfspraul, no10:20
awlast time the jtag rc2 that smt vendor also don't have xray machine.10:21
awso I'd like ti investigate this 0x31 after we finish all batch...then if more boards like this, I bring them to do x-ray scan.10:22
aws/ti/to10:22
wolfspraulyes, sounds good10:22
wolfspraulyou are right, even if we want to do xray, we pool together with other boards first, that's much more economical10:23
awi think i now to try assemble a case with 0x2F. ;-)10:23
wolfspraulyes10:23
lekernel_wow there are so many M1 production and other pictures on the qi hardware wiki10:29
lekernel_we should print big panels and posters with all of them ;)10:29
wpwrak_pin them on the chinese wall ;-)10:37
aw_the original version of case's 3 buttons  is composed of three parts, and not easy to stick together, so currently button is going to be composed of chemistry process?13:16
aw_last time I assembled case skipped those three buttons. ;-)13:16
aw_now need to feel how those buttons assembled together though.13:17
wolfspraulaw_: the buttons are made of three parts, but you will get them assembled13:26
wolfspraulif you only have unassembled buttons right now, you can try with some glue to get them together :-)13:26
wolfspraulthere are three parts, from inside to outside like this: 1) a little bigger thin part 2) smaller thin part 3) smaller thick part, same color as case13:27
aw_yes, i am zoom in this http://en.qi-hardware.com/w/images/c/cf/M1_vga_buttons.jpg13:28
wolfspraulyou can try any glue you have, but it's really terrible to put them together and make them fit through the hole, let alone keep it all clean13:28
wolfspraulfirst you have to peel off the white film from the 3 pieces13:29
aw_can i use Instantaneous rubber to glue those three parts together?13:29
wolfspraulthen you need to glue 2 times, between 1/2 and 2/313:29
aw_yes, peel off already13:30
wolfspraulyou can try any glue13:30
wolfspraulit will be a mess anyway ;-)13:30
wolfsprauljust try to get them together in _any way_ that works, no hesitation. it's very hard to do it well.13:31
wolfspraulyou may also need to remove some glue leftover with a knife, because the tolerances of getting the button through the hole are small, and the button can easily get stuck13:32
wolfsprauljust experiment, and no worries13:32
aw_yes,, i notice that condition.13:32
lekernelcyanoacrylate glue applied in small quantity at the right spot isn't that much of a mess13:48
lekernelbut if you do it wrong, it does become quite a mess indeed13:49
lekernele.g. if you put too much glue, some of it leaks from the sides and blocks or glues other things13:49
aw_umm..good that at least have 1.5mm space between front case and board, so no problems on shape up an ellipse on a 30awg wire.13:51
wolfspraulcorrect13:53
wolfspraulthe wire will be safe13:53
aw_yes..i put a little & thin then slippery few glue on inner of circumference. ;-)13:53
aw_now just wait few times to let them stuck well.13:54
aw_btw, tomorrow I need bring two case to find someone help me drill two holes: one is microphone , the other is the upper case for usb cable. ;-)13:55
wolfspraulhmm13:59
wolfspraulwhy do you need a mic hole?13:59
wolfspraulfor the jtag cable, if you do holes, I suggest you make a hole into the backpanel, near the power plug14:00
wolfsprauland use a regular usb cable14:00
wolfspraulthe upward pointing usb cable is meant to be used with the top acrylic completely removed. of course you can also make a hole into that one if you like :-)14:00
aw_since need to drill a jtab usb hole, so accompany a mic hole. ;-)14:02
aw_lekernel, yes, i used 3M cyanoacrylate glue. :-)14:02
GitHub131[extras-m1] yizhangsh pushed 1 new commit to master: https://github.com/milkymist/extras-m1/commit/d5d176e39a172ce18e7888d43533165ac3f5508f14:10
GitHub131[extras-m1/master] removed shadow for better printing result - Yi Zhang14:10
aw_my some screws are the same to http://en.qi-hardware.com/w/images/c/cf/M1_vga_buttons.jpg14:10
aw_so next version from Roh will be the same as that high resolution picture?14:11
wolfspraulyes14:12
wolfspraulthat same color, buttons assembled14:12
aw_got it.14:12
aw_I'll upload my steps of assembled pictures though. ;-)14:13
GitHub192[milkymist] sbourdeauducq pushed 1 new commit to master: https://github.com/milkymist/milkymist/commit/e0c8d1d415933a3ed6e39b2b23209b1878aee5ed14:29
GitHub192[milkymist/master] TMU prefetch (WIP): split tag memory results between fragment FIFO and fetch unit - Sebastien Bourdeauducq14:29
aw_http://en.qi-hardware.com/wiki/Special:NewFiles14:35
aw_not finished yet. remaining screws to tighten midi & dmx rca connectors. :)14:37
aw_good, two rc3 being rendering whole night. ;-)14:41
wolfspraulstrange, which color is that case?14:41
aw_0x2f with case. :)14:42
wolfspraulit looks like some dark blue?14:42
wolfspraulI thought you have mostly purple cases?14:42
aw_yes14:42
wolfspraulalright then14:42
aw_don't know. i got ten with the same dark blue. ;-)14:42
wolfsprauloh14:43
wolfspraulI lost overview over the colors14:43
wolfspraulassembly looks correct14:43
wolfsprauldid you put the rubber feet on?14:43
aw_yes, i have feet14:43
aw_not used only14:44
wolfspraulput them on so it's fully assembled. also you will scratch your table surface otherwise.14:44
wolfspraulbecause of the screws14:44
wpwrak_lekernel: the little red box on milkymist.org is cute ;-)15:06
aw_wondering how degree inside of case...let's see tomorrow if it will rendering also good. ;-)15:08
GitHub182[milkymist] sbourdeauducq pushed 1 new commit to master: https://github.com/milkymist/milkymist/commit/de40a8628c6dd2127bd4b5edf3ce5765c52f66eb15:12
GitHub182[milkymist/master] TMU: buffer: arbitrary depth - Sebastien Bourdeauducq15:12
kristianpaulbe aware jtag pod could increse temperature a bit..15:16
kristianpaulbut just, wondering same here :)15:16
GitHub66[autotest-m1] xiangfu pushed 1 new commit to master: https://github.com/milkymist/autotest-m1/commit/0381f55ac8021c203336f0b6283889e65191415315:31
GitHub66[autotest-m1/master] output more info, format a little - Xiangfu Liu15:31
GitHub93[autotest-m1] xiangfu pushed 1 new commit to master: https://github.com/milkymist/autotest-m1/commit/7822c25f7737eba2e3a8b50dcb1509264d2176e315:55
GitHub93[autotest-m1/master] MIDI: forever test until 'e' pressed - Xiangfu Liu15:55
lekernelnew lm32 version http://www.latticesemi.com/dynamic/index.cfm?fuseaction=view_documents&document_type=65&sloc=01-01-08-11-48&source=sidebar16:04
lekernel(I have not checked what is new... diff it if you want)16:04
wolfsprauloh nice, I didn't think lm32 was even actively maintained anymore16:13
wolfspraulI should contact them and try to sell them an m1 :-)16:14
lekernelI don't know if they changed anything16:15
lekernelin the March version, there was nothing new in the processor core itself16:15
wolfspraulsure, but I'm serious. Once I have units I will fire an email in their direction, to marketing or so, telling them to buy it to see a good use of their core.16:15
lekernelalso, it seems they have used the open source license for LM8 as well16:15
wolfspraulyes, that's another reason to contact them16:15
kristianpaulLM8 have GCC support as well?16:15
kristianpauli never undertood that..16:16
wolfspraulonce I have established contact with someone, I can bringup the issue of the unfortunate licensing/legal maze, see whether they want to improve it16:16
wolfspraulbut I need a product so I have something to talk to them about16:16
mwallekristianpaul: iirc there was a press announcement about gcc support for the lm816:42
mwallelekernel: where do you have the information about the lm8 license from?16:47
wolfspraulthey emailed him I think :-)16:48
wolfspraulLattice's way to execute and communicate licensing decisions is messed up16:49
wolfsprauljust look at the lm32, same thing. first you have to register to download a free as in beer software, then generate source files from it, then some of those source files fall under an open source clause in a difficult to read license16:50
wolfsprauland the headers of those source files are even unfortunately written with "confidential and proprietary" standing right next to the reference that will eventually explain that it is actually open source16:50
wolfspraulargh16:50
wolfspraulI wish they cut put one source tarball for download, without registration, and the licensing headers and information in that tarball would be crystal clear16:51
wolfspraulbut I doubt it will happen, at least not soon16:52
GitHub119[milkymist] sbourdeauducq pushed 3 new commits to master: https://github.com/milkymist/milkymist/compare/de40a86...40a265616:58
GitHub119[milkymist/master] TMU prefetch: use miss mask - Sebastien Bourdeauducq16:58
GitHub119[milkymist/master] TMU prefetch: remove useless buffers - Sebastien Bourdeauducq16:58
GitHub119[milkymist/master] TMU prefetch: serialize simultaneous texel cache misses - Sebastien Bourdeauducq16:58
mwallenew copyright notice on every file: http://paste.debian.net/123531/17:10
wolfspraulmwalle: oh wow!!! I was so wrong17:14
wolfspraulthanks a lot for the heads up, that's good17:14
wolfspraulwpwrak_: did you notice? http://paste.debian.net/123531/17:15
wolfspraulmaybe someone will read all the fine print in the new arrangement, but it definitely looks better than before17:16
mwalle+// Version 3.817:30
mwalle+// 1. Feature: Support for dynamically switching EBA to DEBA via a GPIO.17:30
mwalle+// 2. Bug: EA now reports instruction that caused the data abort, rather than17:30
mwalle+//    next instruction.17:30
mwalle+//17:30
mwallefor lm32_cpu17:30
wpwrak_wolfspraul: so that's on the files that had a nasty copyright notice before ?18:12
kristianpaulyeah, better than beffor indeed18:14
wpwrak_kewl. so maybe the rants were useful :-)18:15
kristianpaulalso, the link sebastien posted, saids [R] wich is not true, cause as soon you click on it download start, no registration required18:16
kristianpauls/saids/point a18:16
kristianpaul"LatticeMico System 1.3 on Linux" right?18:17
wolfspraulkristianpaul: I think you may have registered in the past.18:19
kristianpaulyes,18:19
wolfspraulwpwrak_: I don't know exactly which files it applies to etc and won't dig in there now. I want for another guy like that crazy dude, then I do it ;-)18:22
wolfspraulthe actual intention and meaning was and is clear since 2006, for anybody who acts in good faith18:22
kristianpaul(good faith) yeah :)18:23
wpwrak_wolfspraul: yeah, but it's good to have it in writing18:23
mwallewpwrak_: yes all files, i'll post a patch to the ML later18:24
wolfspraulsometimes it's really enough. it's as if we ask people to pray 5 times daily in front of the temple of rms, just to be sure they are really serious.18:24
wolfspraulI'm more interested to find someone at Lattice for a marketing cooperation. That's be harder though since they have this idea that the open core helps them promote their proprietary peripherals :-)18:25
wolfspraulthe Milkymist strategy doesn't quite fit into that plan...18:25
wpwrak_wolfspraul: (enough) the problem is that you're putting yourself in the position of (supposedly) authoritatively interpreting and representing their license. this is simply not a place you should be, for all sorts of reasons.18:27
wpwrak_(marketing) well, once you get their attention, word may spread :)18:29
kristianpaulwhat's that er1?18:30
kristianpaulah, nv jtag related..18:30
kristianpaulok, there is lm8-elf-gcc now where is the source...18:33
Action: kristianpaul dowloading Archivo:diamond_1_3-base-92-i386-linux.rpm18:37
mwalle* 2010-09-32 lekernel Removed debug ROM (made external), trace module << 09-32? :)19:25
mumptaikristianpaul, ?19:26
kristianpaulmumptai: thats was to achieve 90Mhz and 1024x769 experimental rendering.. i remenber20:04
mumptaiqi-bot said you were looking for me?20:04
kristianpaulme? ahhh, sorry not20:05
kristianpaulmistake from my part20:05
mwallethe lm8 core has the same license header like the lm3220:42
mwallelekernel: one of my patches was bounced (>40kb)22:05
lekernelmwalle, ?22:13
lekernelah, yes22:14
lekerneljust checking email now22:14
GitHub138[milkymist] sbourdeauducq pushed 4 new commits to master: https://github.com/milkymist/milkymist/compare/40a2656...31d8a8622:18
GitHub138[milkymist/master] lm32: sync to upstream v3.6 sources - Michael Walle22:18
GitHub138[milkymist/master] lm32: new license headers for lm32 source files - Michael Walle22:18
GitHub138[milkymist/master] lm32: update changelog - Michael Walle22:18
lekernelmwalle, thanks for the patches!22:18
mwallelekernel: i just checked, that the core still synthesize, but i think it should still work ;)22:19
lekernelgit head is the 'experimental' branch anyway :)22:20
mwallegn822:25
lekernelgn822:25
lekernelbtw, they have removed the assignments delays in the new version?22:26
lekernelWesley mentioned that the previous one added delays everywhere22:26
lekernel(which is bad)22:26
lekernele.g. they would do register <= #5 value; everywhere instead of register <= value;22:27
wpwrak_maybe they fixed some bugs in their FPGAs/synthesis ;-)22:28
mwallelekernel: no they didnt :(22:31
wpwrak_wimps :)22:33
kristianpaulyou may use the CLOCK_DEDICATED_ROUTE constraint..23:12
kristianpaulwhat?23:12
kristianpaulfor time saving i'll back to 13.2....23:13
kristianpaul12.1*23:13
kristianpaul13.1*23:13
kristianpaulno way, same issue, but i just remove a wire from a reset signal and got all this..23:28
--- Thu Jul 21 201100:00

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