#milkymist IRC log for Monday, 2011-05-09

azonenbergkristianpaul: I set up the google code project "homecmos" and google group "homecmos"00:03
azonenbergLab blog is on the to-do list00:03
azonenbergRight now i have just been posting status updates to my facebook :P00:04
azonenbergBut that needs to change00:04
azonenbergprobably gonna set up another blogspot for it00:07
azonenbergI already have one, "siliconexposed"00:08
azonenbergfor my IC reversing work00:08
kristianpaulBtw, what are considered material related to semicoductors00:08
azonenbergYou mean semiconducting materials? Or materials used in fab / research00:08
kristianpaullast one00:08
azonenbergSilicon is the main actual semiconductor used, germanium and gallium arsenide are used in microwave stuff and LEDs00:09
azonenbergPhosphorus, boron, and arsenic are used for P and N type doping00:09
azonenbergAluminum for interconnect wiring, replaced by copper in newer high-speed chips00:09
azonenbergCopper ions tend to diffuse and contaminate PN junctions so the copper is usually isolated by some kind of barrier material00:10
azonenbergtitanium nitride and ruthenium are two i recall specifically00:10
azonenbergFor dielectric normally you use SiO200:10
azonenbergnewer processes use high-K dielectrics like tantalum pentoxide00:10
azonenbergWhich i was actually considering using in my process but for a totally different reason (hardmask for wet etching, the stuff is pretty impervious to KOH but is etched quickly in dilute HF)00:11
azonenbergLow-K dielectric between wiring layers of newer chips, dont know of any specific materials though00:11
azonenbergFor processing, various photoresists and developers00:11
azonenbergKOH, tetramethyl ammonium hydroxide (TMAH), EDP, and other water-amine complexes for anisotropic wet etching of silicon00:12
azonenbergHF for wet etching of SiO2 and some metals, other acids for other metals00:12
azonenbergLots of fluorine based gases for RIE00:12
azonenbergAnd chlorine, halogens in general00:12
azonenbergSF6, CCl4, ClF300:12
azonenbergArgon, as an inert gas for sputtering and purging of stuff00:13
azonenbergNitrogen, for forced-air drying of wafers, purging, and in some cases reactive sputtering to form nitrides and oxynitrides00:13
azonenbergthere are more but those are the ones that come to mind in particular00:13
kristianpaulSourcing those should no be hard i guess :-)00:14
azonenbergLol, many of them are hard00:14
azonenbergBut my processes will be using a much smaller subset of those materials00:14
azonenbergI'm actually going to start documenting a lot more stuff in the near future, probably going to start the blog tonight00:14
kristianpaulwpwrak: (386dx) wow now i realize the µm times :-)00:15
azonenberg4004 was on a 10um PMOS process iirc00:15
azonenbergCan you imagine making a 1:1 scale working replica? :D00:16
azonenbergThere will def be some process changes though, 4004 was not made to use high-K dielectric00:16
azonenbergIt probably used polysilicon gates, my process will be using metal00:16
kristianpaullets change that to a navre core IC ;)00:17
kristianpaulat least you mips core was tought to fit in your photolithography research :o00:17
azonenbergWell, the first step is MEMS as i can get away with much looser tolerances etc there00:17
azonenbergThen, single transistors00:17
azonenbergsimple CMOS logic cells00:18
azonenbergAnd if i can pull that off - a couple of standard cells plus interconnect wiring00:18
azonenbergThen its time to start thinking bigger00:18
azonenbergAnd smaller, as well00:18
kristianpauland smarter than intel ;-)00:19
azonenbergLol not just yet, i have a ways to go still00:19
kristianpaul"Why not change the world?" nice slogan, you catch it very well :-)00:22
wpwrakin what other industrial process but semiconductors is the substrate used ? i.e., what are the chances of sourcing it internationally00:22
azonenbergI'm only concerned about US vendors right now as thats where i'm located00:23
azonenbergbut there are a lot of international fabs00:23
azonenbergin fact, *most* fab is done in the far East - china, taiwan, etc00:23
wpwraki'm thinking of poor agrarian countries, like argentina :)00:23
azonenbergOh... you are not likely to find semiconductor grade silicon there very easily lol00:24
azonenbergyour best bet will be ebaying etc probably00:24
kristianpaulazonenberg: Is it only you, or it will be an interdiciplinary research?00:25
azonenbergYou mean, am i the only person on the project?00:25
azonenbergAs of now i'm the only really active person but i'm bouncing ideas off of a bunch of friends and colleagues00:26
wpwrakmaybe go to the beach, get a few bags of sand. then make some kitchen experiments. after all, the secret is in the impurities, right ? :-)00:26
azonenbergGood luck getting five nines (or probably six or eight, actually) purity silicon out of sand at home00:27
azonenbergI'm doing the fab, not the materials production00:27
kristianpaulwpwrak: are you swiching from ceramics to sand now? :-)00:29
wpwrakkristianpaul: what did i do with ceramics ?00:30
wpwraki kinda like the picture of a little x/y blue ray laser "printing" me some chips :) you know, kinda "make cpu". first runs llhdl, then prints the chip. voila. nicely integrated process.00:31
azonenbergWell, its a little more complex than that00:31
azonenbergThe bluray module won't have alignment capability00:31
azonenbergYou'll use it to make the masks00:31
kristianpaulwpwrak: dunno, i just tought had something in mind.. thats all :)00:31
azonenbergBut then a contact aligner will be necessary to actually fab the chip layer by layer00:32
wpwrakazonenberg: for alignment, if you can get a substrate that's smaller than the movement range of the x/y laser, you could just align with chip edges (photo sensor on the other side, chip blocks laser light or not)00:37
azonenbergToo imprecise00:37
azonenbergi'd need theta, magnfiication, etc00:37
azonenbergSubstrate dimensions will not be precisely controlled00:37
azonenbergI'll be fracturing along cleavage planes00:37
wpwrakmaybe you could scan the outline ?00:38
azonenbergI'll need optical alignment using purpose-built alignment markers on the chip surface00:38
azonenbergThe thing is, doing that with laser direct write will be tricky00:38
azonenbergDoing it by hand under a microscope is a lot easier00:38
azonenbergThe standard alignment mark is a cross on the bottom layer00:39
azonenbergthen a hollow cross on the next one00:39
azonenbergthat you fit over it00:39
azonenbergYou have one at each side of the die to detect slight theta shifts00:40
wpwrakbut they're mainly used for the same laser, right ?00:40
azonenbergMost production systems don't use lasers00:40
wpwrakif we assume direct exposure00:40
azonenbergThey use electron beam direct-write (similar to the laser process i'll be using) to create a photomask in chrome on copper00:40
azonenbergAnd then you run that through a contact or projection exposure system00:41
azonenbergthe key is that the mask-fab process is distinctly separate from the alignment and exposure process00:41
wpwraki'm thinking of the x/y BR laser. if you can scan the chip's outline, you could probably reproduce its orientation and position quite accurately00:41
azonenbergoutline of the die?00:42
azonenbergbear in mind that i'll be taking little fractured pieces of silicon and then patterning a square in the middle of them00:42
azonenbergAlso, scanning will require optical pickups, a camera, vision software00:42
wpwrakperfect. the more fractured, the more information :)00:42
azonenbergI'm not saying it's impossible00:42
azonenbergBut it will definitely not be easy and is certainly not a rev1 or even rev2 capability00:42
azonenbergReal production steppers do something similar00:43
azonenbergThey scan the alignment markings on the first mask level and line the new mask up to that00:43
azonenbergall done automatically with no human intervention00:43
wpwraknaw. you already have the BR laser. put a photo diode on the back. check if you have a beam or not. take lots of data points. filter/average/etc.00:43
azonenbergThe key is getting good resolution00:43
azonenbergwe're talking submicron here, potentially00:43
azonenbergyou cant just use a photodiode, you'd need a full camera00:44
wpwraki would expect that you can get good resolution statistically00:44
azonenbergYou could, but that means much more intensive DSP capabilities00:44
wpwrakpc time is cheap ;)00:44
azonenbergNot for a realtime process00:44
azonenbergIf you can't automate it faster than a human can align it?00:44
azonenbergNot worth the trouble00:44
wpwrakdo you care so much about time ? for a DIY ultra low-volume process, you can omit a lot of things the industry does00:45
azonenbergExactly, such as automated alignment00:45
azonenbergWhich is why i am using a manual alignment process instead00:45
wpwrakindustry is all about throughput and yield. none of this matter much to you. for you, everything is in the setup cost.00:45
azonenbergWhich is cheap to set up and requires minimal hardware00:46
wpwrakdoesn't the manual process need fancy extra equipment ?00:46
azonenbergDid you read my paper?00:46
azonenbergIt was pasted to the channel a few times00:46
azonenbergIt's just a converted microscope (that still functions as a regular microscope, in fact)00:46
wpwrakheh, i guess i should :)00:46
azonenbergprint the mask pattern on a laser printer at 200 microns per pixel (the limit of my cheap printer, 150 is doable sometimes but closely spaced lines will blur together)00:47
kristianpaulwpwrak: http://colossus.cs.rpi.edu/~azonenberg/papers/litho1.pdf00:47
azonenbergAnd use the microscope optics run backwards to project the pattern onto the die00:47
kristianpauljust in case ;)00:47
wpwrakkristianpaul: thanks !00:47
azonenbergThe first process i described there got me to 15um per pixel on the actual die00:47
azonenbergThe second one is still being refined, i've hit 5um per pixel but it didnt expose evenly across the field00:47
azonenbergi need a more intense UV source and a diffuser00:47
azonenbergwhich is going to be ordered next payday :P00:48
wpwrakazonenberg: okay. but that's the process that gives you a very limited field. i'm talking about the x/y BR laser. that shouldn't need a microscope, right ?00:48
azonenbergMy plan, actually, is to combine the two00:48
azonenbergUse the laser direct write process instead of the printer to make masks00:48
wpwraklaser through the microscope ?00:48
azonenberguse a low-power objective (say 4x) to project this mask slightly reduced onto the die00:48
azonenbergand align through the microscope00:48
wpwrakah, okay. i thought you'd go 1:1 with the laser00:49
azonenbergI considered it00:49
azonenbergBut this lets me get away with lesser tolerances on the laser00:49
azonenbergalso, direct write is slow00:49
azonenbergThis way, if i fix a bug on metal 300:49
azonenbergi can reuse the M1/M2 etc masks00:49
azonenbergand only redo the M3 mask00:49
rohmaking navre on silicon would be so cool00:50
azonenbergnavre is what? softcore CPU?00:50
rohcould enable a fully idea->hw toolchain for avr stuff00:50
rohnavre is a avr core in verilog.00:51
wpwrakhmm. i still kinda like the idea of a maskless process. fewer moving parts :)00:51
azonenbergwpwrak: Maskless is nice in theory but is awkward00:51
azonenbergMy thinking is, sputter a microscope slide in metal and direct write on that00:51
rohso one could do a 'no non-free parts' adruino ;)00:51
azonenbergthen use the mask in my current process00:51
azonenbergAlso, has anybody tried synthesizing navre in an ASIC tool?00:53
azonenbergrather than fpga00:53
azonenbergas in, getting actual gate counts etc00:53
wpwrak(asic tool) probably not. at least i didn't see lekernel mention anything like this01:26
rohi think one also needs some io units01:49
wpwrakazonenberg: hmm, seems that a BR reader would already have a strong enough diode. makes things nicely inexpensive :)01:49
rohnow its only a core with some sie for usb01:49
wpwrakroh: some gpios probably wouldn't be too hard to implement. even output-only would be sufficient for showing off ;-)01:50
wpwrakwell, you need some input mechanism for loading the code. spi perhaps.01:51
rohwpwrak: true. but then i think it would make more sense adding some units like serial etc.01:52
rohdoesnt even need to be a atmel style copy, could be different. do changes in avr-libc01:52
wpwrakroh: i'd just focus on plain brag value, i.e., core and whatever else is easiest to do :)01:53
kristianpaulargg, 000002:06
kristianpaulwhat i missed..02:06
kristianpaulWB Write: 00001000=12345678 acked in           1 clocks02:06
kristianpaulWB Read : 00001000=00000000 acked in           1 clocks02:06
CIA-48flickernoise: Xiangfu Liu master * r2b78537 / src/performance.c :04:02
CIA-48flickernoise:  add missing head file <osd.h>04:02
CIA-48flickernoise:  merge 'firstpatch' and 'simple_current' to 'current_patch'04:02
CIA-48flickernoise:  press 'F1' when rendering will display the patch name04:02
CIA-48flickernoise: Signed-off-by: Xiangfu Liu <xiangfu@sharism.cc> - http://bit.ly/iDLjQ804:02
azonenbergwpwrak: Agreed04:04
azonenbergMy first project will be a simple MEMS comb drive, once i get that working i'll make a transistor04:04
azonenbergthen gates04:04
azonenbergFirst actual CMOS project will be a ring oscillator04:05
azonenbergThat alone, in a 5um homebrew process, will be worth quite a bit for bragging rights lol04:05
wpwrakoh yes ;-)04:27
azonenbergEven if it is just a "hello world" circuit lol04:27
azonenbergBut there will be a lot of work needed before i'm even close to there...04:28
wpwrakyou can always do that 4 GHz hexacore another evening ;-)04:28
azonenbergLol, i dont plan on doing that any time soon :p04:29
wpwrakyeah, you'll want that 8 nm process for it. so that's a couple more years.04:30
wpwraki still find the idea of converting two BR drives into an xy plotter intriguing. maybe even some of the drive electronic could be reused.04:33
azonenbergonly one drive04:34
azonenbergjust get a precision linear stage04:34
azonenbergput the single axis perpendicular04:34
wpwrakmmh. one axis would be the sled. what drives the other axis ?04:35
wpwrakor would you spin the die, like a disc ?04:35
azonenbergThe sled would be sitting on top of a linear actuator of some sort04:37
azonenberglike used for an automated microscope focuser or similar04:37
wpwraki don't know these. but another BR or DVD sled ought to be a cheap linear actuator, no ?04:38
wpwrakpreferably the same BR, to keep the drive circuit simple04:38
wpwrakjust pull the laser. use it to scare pigeons or such04:39
rohhm. how much H and V move do you need on the laser?04:46
rohthe 'sled' isnt the only thing moving there. usually one moves the laser on a cheap and imprecise sled and has a magnetic lens holder04:46
rohmeans 2 moving coils around the lens. one for focus, one for fine-alignment04:47
azonenbergI was planning on having a relatively loose tolerance on the direct-write system04:47
azonenberggetting maybe 50 microns per pixel or so04:47
azonenbergand then further reducing in the projection system04:47
azonenbergI just want something that can go better than the 200um/pix of my cheap laser printer04:47
azonenbergand will make darker masks (chrome on glass)04:48
rohso if you need only small movements... use the H coil only... not the dc servo04:48
azonenbergand the masks will be up to maybe 1cm^204:48
azonenbergwhich is, in fact, about the travel range of the servo end to end04:48
rohhm. how about using commercial services for stuff like that? friends of mine give away digital images and get film back for projections on whole walls04:49
azonenbergSure, that's an option04:49
rohits really tiny and sharp.. and not that expensive04:49
azonenbergAnd thats definitely at least 50um resolution04:49
azonenbergBut is it chrome on glass, or on plastic?04:49
azonenbergIf plastic it's less stiff and flat04:50
azonenbergand you can get focusing etc issues04:50
azonenbergWhich is a problem with my current system, actually04:50
rohi think its plastic. but there should be options for glass since that stuff gets quite hot in the projectors04:50
azonenbergBut is it ink on glass?04:50
azonenbergOr metal04:50
azonenbergI dont think its metal04:51
rohdunno. will ask04:51
azonenbergThe whole idea of the direct write system was to get metal-on-glass masks04:51
azonenbergLaser printer or commercial print shop, same deal04:51
azonenbergit's still think plastic04:51
rohmaybe the machines used to copy to microfiche would be of some help04:52
wpwrakazonenberg: the sled should be able to travel a bit more than 3 cm (for a 12 cm disc). so you'd get at least 9 cm2. plenty of room for that hexacore ;-)04:52
azonenbergLol, no way am i making 3cm^2 dies04:52
azonenberg3cm x 3cm*04:52
azonenbergTrying to keep something that big planar and in focus? Not happening any time soon04:53
wpwraka bit of extra range could be useful for loading/unloading, calibration, etc.04:53
azonenbergi'm shooting for 1cm x 1cm as a more realistic size04:53
azonenbergHowever, that's per die04:53
azonenbergBeing able to make a stepper, and put multiple dies on a wafer in one session04:53
azonenbergWould be very cool for the long run04:53
wpwrakyeah, you probably want some optical feedback for beam size at some point in time. i wonder how the writers do it. when writing, what sort of feedback do they get ?04:54
wpwrakaye. DIY mass production :)04:54
wpwrakfor planarity, if you have a feedback system, you could just have focusing area between the dies, where you expose until the beam is right. then interpolate.04:59
azonenbergI'm going to need a street between dies anyway for dicing05:00
azonenbergWhat about direct write through the microscope05:00
wpwraki still wonder how the focus feedback works in BR drives, though. must be something simple.05:00
azonenbergMicroscope camera05:00
azonenbergRealtime feedback of spot size etc05:00
azonenbergAnd would permit alignment as well05:00
wpwrak(street) good point. may as well "pave" it :)05:00
azonenbergthen just get a motorized stage05:00
wpwrakmicroscopes tend to be expensive equipment05:01
azonenbergI have one already is the point05:01
azonenbergand its a $800 model05:01
azonenbergnot a $10K mitutoyo or zeiss05:01
wpwraksee. a BR drive is something like USD 100 :)05:01
azonenbergBut then you need all of the control electronics, optics, assemblies to hold it all together05:01
azonenbergAnd you'd need a scope for inspection anyway05:01
azonenbergand failure analysis05:02
wpwrakthe optics ought to be in it already05:02
azonenbergNobody is going to be doing 5um fab if they can't even see 5um features05:02
azonenbergA microscope is implied05:02
wpwrakassemblies: a base plate and a few spacers, i guess05:02
wpwrak(inspection) yes, that's true. still, it could be a lot simpler than one that you could use for exposure05:03
azonenbergNot really, this scope is about the bare minimum you'd want for work like this05:04
azonenbergyou cant use a biological scope for opaque samples easily05:04
azonenbergyou need through-lens illumination05:04
azonenbergand low-power 30-50x stereo ones arent enough power to see 5um features clearly05:04
azonenbergtrust me, i've tried :P05:04
wpwraki actually wonder how small you can get with a digital camera. not 5 um, that's for sure, but they should also have a pretty good resolution if you get close enough and manage to get some light source05:05
wpwrakas a poor man's microscope05:05
azonenbergPixel size is normally a few microns but you can't actually resolve that clearly05:05
azonenbergthey have bayer filters05:05
azonenbergso 10+um is your actual resolution05:05
azonenbergthats assuming you have a condensing lens to properly focus everything, ideal best case05:06
wpwraksomething 10-ish wouldn't be too bad05:06
azonenbergrealistically i think 50 is more reasonable05:06
wpwrakprobably something like 15-30 um. this image http://downloads.qi-hardware.com/people/werner/tmp/zoom-13_6mm.jpg05:17
wpwrakis about 13.6 mm wide05:17
azonenbergYes, its 13.6mm wide05:18
azonenbergso you say that's 15-30um per pixel05:18
azonenbergBut can you actually resolve two lins at that spacing?05:18
wpwrak1848 pixels05:18
azonenbergOr is it just empty resolution05:18
azonenbergpixels with no meaningful data05:18
wpwrakit's the nominal pixel width05:19
azonenbergI'm willing to be 3-5 pixels is your minimum actual resolution05:19
wpwrakso that's about 7.4 um per pixel. now, how many pixels do we need for a real-life edge ...05:19
wpwrakthat would be 20-40 um05:19
azonenbergRealistically, 10pix is what i see in most cheaper sensors05:21
azonenbergSo 20-8005:21
azonenbergAbout what i expected05:21
wpwrakdunno how much of the blur i have is sensor, focus, or motion. but yes, it's not quite good enough for your 5 um structures, let alone the 1-1.5 um of a 386dx05:23
azonenbergOn the other hand, my nice scope...05:24
wpwrak(focus) the camera doesn't autofocus properly at this short range, so i moved it until the image looked sharp in the LCD viewfinder. so i can probably squeeze out a few more um ;-)05:25
azonenbergthose lins are pretty small lol05:25
azonenbergiirc around 400nm pitch05:25
azonenbergthis is with contrast turned all the way up using oil immersion05:25
azonenbergthose marks are scratches on the sample05:26
azonenbergthin plastic diffraction grating05:26
wpwrakis the url correct ? the host doesn't resolve05:26
wpwrakhmm. whois has it.05:26
azonenbergHaving some DNS issues05:27
azonenbergis the correct IP05:27
wpwrakhmm yes, nice resolution :)05:28
azonenbergNot nearly as nice as this http://www.drawersteak.com/downloads/chippics/pic16f59/pic16f59_005.jpg05:29
azonenbergThat's on the zeiss supra 55 SEM at my school's cleanroom lol05:29
azonenbergmetal 3 of a 350nm 3-metal chip at 50,000x05:29
wpwrakand full of life ;-)05:30
azonenbergi was doing a failure analysis job for a consulting customer and decided to play for a few minuts afterwards on some samples i had lying around05:30
azonenbergNo, actually05:30
azonenbergthose arent bacteria05:30
azonenbergLook at the one at top left05:30
azonenbergit's cracked open05:30
azonenbergThose are glass beads from the packaging material05:30
wpwrakdunno what cracked bacteria look like :)05:30
wpwrakheh ;-)05:30
azonenbergLol, they dont show brittle fracture05:31
azonenbergAfter the 150C boiling nitric acid to decap05:31
azonenbergfollowed by rinses in acetone and isopropanol05:31
azonenbergIt's about as sterile as you can get lol05:31
azonenbergStored inside a closed container before being brought into the cleanroom05:31
wpwrakwell, it could still be dead bacteria :)05:31
azonenbergI'd expect them to be completely lysed by the HNO305:31
azonenbergthe other giveaway is the perfect spheres and large size range05:32
azonenbergI mean, you could be right05:32
azonenbergI didn't actually try IDing them05:32
azonenbergi was just trying to get practice in focusing the thing just right05:32
azonenbergidk if you've ever used an electron microscope but they're a little tricky to get good images out of05:32
wpwrakthe picture is pretty spectacular05:32
azonenberggetting the last little bit of focus is hard05:32
azonenbergyou have two axes of astigmatism, two of aperture alignment, and focus to adjust05:33
wpwrakso you'll have all you need to debug that 8 nm process for the hexacore ;-)05:33
azonenbergCleanroom time is not cheap05:33
azonenbergi cant afford to use it a lot05:33
azonenberg$188.50 an hour05:33
azonenbergWhat i do instead is, when i have to use it for work05:33
azonenbergSneak in at the end and shave 15 minutes off the bill when i send it to the customer05:33
wpwrakhehe :)05:34
azonenbergThat way i don't have to pay for gowning, pumping down vacuum, etc05:34
azonenbergOnly for the time i actually used on my own stuff05:34
azonenbergcombined shipping, you could say ;)05:34
wpwrakit's all about efficiency :)05:34
CIA-48mtk: Xiangfu Liu master * rd00cd70 / lib/keymap.c :07:39
CIA-48mtk: keymaps: add French_layout_05807:39
CIA-48mtk: Signed-off-by: Xiangfu Liu <xiangfu@sharism.cc> - http://bit.ly/lFAvOq07:39
lekernelxiangfu: why do you persist adding new features to this code you will remove?08:10
xiangfulekernel: first make it work then make it better :)08:10
xiangfuI wrote down all shotcuts here: http://www.milkymist.org/wiki/index.php?title=Flickernoise_user_manual08:11
lekernelonly a short press is needed to boot08:15
lekernelxiangfu: your other patch is ok08:26
xiangfu(short press) thanks, I always hold it until screen show up :(08:36
xiangfuforget to take a look serial output.08:36
lekernellong press makes the splash screen appear09:18
lekernelshort press doesn't09:18
awlekernel, you are right on that we've used faulty 1206 beads in rc2.11:03
awwhich came from the same vendor but two shipments for rc1 and rc2.11:04
awmeanwhile I fully made wrong measurements by ignored resistance from my extended dc jack cable.11:06
awafter replaced right beads for L10/L11, the drop is 3.3mV ~ 3.5mV.11:10
awwith calibrating my cable loss, the correct measured "5V" net voltage is 5.199V @ 5.25V input, 4.951V @ 5.0V input, 4.699V @4.75V input.11:12
awe.g. : when 4.75V in, the total divided dropped voltage = 0.0033 (VL11)+ 0.0035(VL10) + 4.699(V5v net) = 4.7436V, which is exact closely to 4.75V input.11:16
awcorrected formula = 0.0033 + 0.0035 + 0.0378 + 4.699 = 4.743611:19
awbe noted that Vfuse = 37.8mV @ 44.7°C11:20
aw_lekernel, hi i just wrote some msg log, you may read a bit while you were not here.11:22
aw_lekernel, pls check http://en.qi-hardware.com/wiki/Protection_of_Reversed_Polarity_on_DC_plug-in#Sch._D-111:22
aw_lekernel, just see the bottom table, forget all others.11:23
lekernelwhat is 'gnd loop cable loss' ?11:24
aw_also be noted that my room temperature here is 33.3°C11:24
lekernelI don't care about the room temperature11:25
lekernelyour numbers don't add up11:27
wolfspraulwow, Taipei already that warm?11:27
lekernel4.75-4.55 = 0.2011:27
lekernel3.5+3.6+38+37+14 = 96.1 (instead of 200)11:27
wolfspraulno I think you need to look at the next table11:28
lekernelbut this time the fuse drop voltage makes sense11:28
wolfspraulaw_: we really have to cleanup that page and delete all data that is not really accurate. if we think that means we have to delete most data, then we should do that.11:29
lekernelah, the very last table11:29
wolfspraulconfusing data is worse than no data, and now we know that we overlooked the faulty l10/l11 for a while...11:29
wolfspraullekernel: yes :-)11:29
aw_wolfspraul, surely deleted them after this discussions.11:29
wolfspraulonly leave accurate data11:29
lekernelok, better11:29
wolfspraulI think the 4.699 at usb are acceptable (for a 4.75v input)11:30
lekernelso we get 4.70V on USB in the (improbable) worst case... I think that's acceptable11:30
aw_just see the bottom table.11:30
wolfspraulso with my limited knowledge of electrical things, I'd say that looks good11:30
lekernelwell the probability of failure is (probability of having the M1 power supply provide 4.75V)*(probability that the USB device fails for missing 50mV)11:31
lekernelthat's fairly low imo11:31
lekernelbtw, maybe we can leave only one of the beads in place?11:32
lekerneland directly connect GND11:32
aw_lekernel, thinking...11:32
aw_wolfspraul, what's 'bbiab'?11:33
lekernelso what i'd recommend is 1) remove the bead on GND 2) implement the latest protection circuit11:34
lekernelgiven that it doesn't end up in a layout nightmare11:34
aw_here's data that i recorded real DC adapter's 'no load' voltage: http://downloads.qi-hardware.com/hardware/milkymist_one/doc/production/AC_to_DC_adapter.ods11:36
aw_lekernel, the output with 'NO LOAD' at least 4.9V above, so the probability is very low.11:37
lekernelok, ok, don't do too much11:37
lekernelhow do you feel about adding the protection circuit to the PCB layout?11:38
aw_the two Zeners needs placed in 90 degree not lie down to touch pcb surface.11:39
aw_from the *.ods file, seems that we need to take risk on adding protection circuit.11:40
lekernelbtw, maybe we can remove the PTC on USB as well11:42
lekernelsince we already have a 'general' one11:42
aw_placed in 90 degree, one pin needs to add insulation tube.11:43
aw_lekernel, can't understand your on USB.11:43
lekernelremove F111:43
lekernelthere was a PTC specially for the USB ports11:43
lekernelwe don't really need it if there is a PTC at the input now11:43
aw_well...probably not good idea...the hold current F2 is 2A , trip on 4A,11:44
lekernel4A won't damage the USB connector11:45
aw_i doubt that user may produce more than 1A thus 1.76A there said even 2.76A, the F2 still haven't tripped though.11:45
lekerneland that second PTC drops voltage too11:46
aw_yes, i know F1's drop.., well...thinking..11:46
wolfspraulaw_: bbiab = be back in a bit11:46
lekernelyou think too much11:46
aw_lekernel, you really want directly remove F1?11:47
lekerneland remove one of the beads, too11:48
aw_hm...welll...we can always take this risk, so can do DNP F1, but instead mounts a 0 ohm for F1, how do you think?11:48
lekerneldon't mount 0 ohm. just short the tracks in the layout11:49
lekernelfor L11 use 0 ohm unless you need PCB layout space to implement the protection. in that case, short the tracks as well.11:50
aw_you are think the 0 ohm +/- 5%, right?11:50
lekernelno I want to keep that thing simple because everything we add is a source of problems11:50
aw_but if later someone do a horrible USB hub extended experiments, how he can add fuse by himself?11:51
wolfspraulnobody will add a fuse by themselves!11:51
aw_well...taking risk is take risk though.11:51
wolfspraulno let's just design a regular consumer product here.11:51
lekernelit NOT a risk. the product will still work without F1 unless someone makes a very big mistake with the PCB design11:52
wolfspraulaw_: you don't need to think about all sorts of hardware hacks. those people need to know what they are doing.11:52
lekerneland, as a matter of fact, many consumer products allow you to overload the USB ports11:52
wolfspraulaw_: remove parts is good. if lekernel thinks they can be removed, let's remove them unless you have a very good reason not to, which I cannot see right now.11:53
aw_well...i have no further comments on this. so conclusions are:11:53
lekernel4A transient (actually limited to 2A by the power supply) isn't a major danger for the USB ports11:53
aw_1) remove F1, and short by layout.11:53
aw_2) remove L11, ans also by short layout change.11:55
aw_3) place Zener in 90 degree for better heat dissiplation.11:56
aw_is it all?11:56
lekernelaw_: the zener does NOT dissipate ANYTHING unless the user does a mistake. there is no need to optimize dissipation11:56
lekernelso optimize cost and manufacturing time, instead11:57
aw_there's still have space between U13 and DC jack.11:57
aw_i think I will put fuse from away those two Zeners a little.11:57
lekernelput the fuse anywhere it fits without making the layout a mess11:58
aw_so you mean 3) just lie down two Zeners. :-)11:58
lekernelif that's easier for layout/manufacturing, yes, definitely11:58
wolfspraul4) cleanup the long wiki page and delete all data that we now believe is not accurate. do not measure more data, just delete the bad data.11:58
aw_actually here those two are lie down. :-)11:59
lekerneldamn I can't believe we managed to solder 485 components per board so far :P11:59
aw_4) surely, they came from messy ignorance on cable loss, stupid.11:59
wolfspraul5) find the root cause for the 1206 bead (wl262) sourcing problem and make sure the lot for rc3 is good11:59
wolfspraul4) and also from not seeing the l10/l11 problem earlier11:59
lekernelaw_: do you confirm the L10/L11 problem on some of your boards?12:00
wolfsprauland we paid 1.46 USD per bead!12:00
wolfspraulsomething went wrong there12:00
wolfspraulat digikey they are more like 20 cents...12:00
aw_5) the root cause I may hard to catch out, if vendor don't tell me the truth.12:00
wolfsprauleven for that it's a great catch12:00
lekernel1.46USD/bead? wtf?12:01
aw_so 5) I'll only source that 1206 bead myself though.12:01
wolfspraul5) what I mean is that we make sure this is good for rc312:01
lekernelthose should only be a few cents12:01
wolfspraullekernel: yeah. 12 USD for all 8 of them :-)12:01
wolfspraullike I said, I suspect human error somewhere.12:01
wolfspraulthat can happen, it's a manufacturing mistake on our side12:01
wolfspraulvery good catch12:01
wolfspraulI don't think this was intentional fraud/rip-off anywhere, just some mismatch/whatever error.12:02
aw_lekernel, I'll modify pre-rc3 sch again...then reply.12:02
wolfspraulaw_: everything settled now?12:02
aw_wolfspraul, what's your meaning?12:03
wolfspraulwhether you have any more open issues on these matters that block rc312:03
aw_1) to 5) , will done.12:03
wolfspraulis there anything connected to 1-5 that worries/bugs you?12:03
wolfspraulor that is it now?12:03
aw_lekernel, have you added that diode on your board for reset flash chip?12:04
aw_wolfspraul, no , done.12:04
wolfspraulaw_: remember to rework the remaining for-sale rc2 with good beads on l10/l1112:04
aw_can you add to confirm?12:04
aw_wolfspraul, yes,12:04
lekernelI don't have the reset IC, so it'd be of limited use12:04
lekernelbut I trust you on that one12:04
lekernelanyway your schematics is correct12:04
aw_lekernel, i remembered that i tried 1000 attempts for testing 're-configuration'12:05
lekernelso it should be good12:05
aw_but today I tried to use my another board, it seems not work every time on 'booted' ....but all 're-configuration' is all good.12:06
lekernelwolfspraul: there should be the 'bead dropout' problem at the output of the regulators too12:06
lekernelactually, probably those beads can be replaced by 0 ohm, as Joerg said12:06
aw_so can you add one low forwarding voltage diode to confirm ?12:06
lekernelI didn't want to spend time investigating this, but now that we get bead issues ....12:06
wolfspraulwhich ones?12:06
lekernelall the other 612:07
aw_lekernel,  i don;t think so12:07
aw_0 ohm +/- 5%  has tolerance which may acts the small resistance like bead 0.01 ohm for the best though12:08
lekernelI was talking about Joerg's argument that those beads are useless at best12:09
aw_also if we replace other bead with 0 ohms, means we 1) tried to cost down  2) without considering emi12:09
lekernelyeah sure12:09
lekernelaw_: no, I won't try the diode. I don't have the parts for that, nor the time to try it.12:10
lekernelplus you did testing that was extensive enough12:10
lekernelplus there is a very low risk of major regression12:10
wolfspraullekernel: you want to replace all other 6 1206 beads with 0 ohm?12:10
wolfspraulor which ones12:10
aw_hm...okay...if rc3, i got failure on that diode, we can just DNP it.12:11
wolfspraulyou need to tell Adam, and either he adds it to his list or not...12:11
lekernelwolfspraul: all the beads in series with the outputs of the regulators, as Joerg said12:11
lekernelbut no, don't try it if it's another source of delays12:11
wolfspraulno delay12:11
wolfspraulaw_: what do you think?12:11
lekernelwhich is what I feared initially12:11
aw_wolfspraul, well...i just test my old board which not every time 'booted' successfully...but I do real sure that reset ic is good though, my new rc2 board I'll add diode again to see if 'booted' every time12:13
wolfspraulwe didn't talk about the diode12:13
wolfspraulthat one is settled12:13
wolfspraullet's not go back again and again to the same old issues12:13
aw_wolfspraul, hmm? so which one?12:13
wolfspraul"replace all beads in series with the outputs of regulators with 0ohm"12:14
wolfspraulaw_: do you want to do this?12:17
aw_i can try it or not try it because we passed FCC & CE in German, didn't we? so if we replace with 0 ohms, should we do certification again? if the answer is NO, we leave those beads, or we mount 0 ohm.12:17
wolfsprauldon't understand12:19
wolfspraulI cannot assess the risk of removing the beads _worsening_ the EMI.12:19
wolfspraulbut I also cannot always let fears guide my decision making12:19
aw_we've got report from german, right?12:20
wolfspraulif those beads are stupid and nobody knows why they might worsen EMI, then I would still remove them12:20
wolfspraulif there is someone with experience who says that removing them may very well _worsen_ the emi, then I would be cautious12:20
wolfspraulso do we have someone with that kind of experience?12:20
wolfspraulI can totally remove them and _assume_ that emi is the same or better as before.12:20
wolfspraulI'll just say it until someone proves me wrong :-)12:20
aw_so i said that if we mount them with 0 ohms, I don't know if influence emi result even now I've not viewed all regulator's output beads spec.12:22
wolfspraulsounds like we leave everything as-is12:22
wolfspraulI don't want another drama on measurements etc.12:22
wolfspraullekernel - what do you think?12:22
wolfspraulaw_: unless we hear from lekernel, let's leave them as they are12:25
wolfspraulfocus on the other stuff, L10/11, fuse/zener etc.12:25
aw_wolfspraul, they are couple cents though, yeah..12:26
wolfspraulaw_: when do you start working with the layout people?12:27
aw_already started...tomorrow I'll send it as a pre-layout file after modifying todo about pre-rc3 sch.12:28
wolfspraulaw_: ok, make sure to send me the png silkscreen for review about the logos and text position/size etc.12:28
wolfspraulsend to the list, then everybody can see it if they like to :-)12:29
aw_yeah...i'll capture it12:29
lekernelyeah let's leave them... don't want to go through another round of measurements12:35
wolfspraulfor sure not12:35
wolfspraullekernel: do you think removing them may worsen emi?12:35
wolfspraulI thought joerg was pretty sure they could not possibly do any good.12:35
lekernelI don't know, maybe12:36
wolfspraulwe leave them12:36
lekernelyes, I can follow Joerg's reasoning, but on the other hand a lot of designs have them. maybe everyone is wrong and Joerg is right though :)12:36
lekernelanyway - small detail12:37
wpwrakwolfspraul: do you plan to re-check EMI/CE/FCC after all the rc3 changes are done ?14:19
wolfspraulwpwrak: no15:00
wolfspraulCE is a self-certification, and for FCC it's similar because m1 is a 'non-intentional radiator' and for those it's enough if the manufacturer says that the device meets requirements15:01
wolfspraulno FCC ID is issued, so the question of whether a new FCC ID needs to be issued after making production changes (which happen all the time in reality) doesn't even come up15:02
wolfspraulnow, if I would be worried that after our rc3 changes the actual tests would show a problem, then I would want to know that, and I would look into the problem15:03
wolfspraulbut I cannot just test over and over again for every little change we make, for what reason?15:03
wolfspraulthe next time I go back to that lab, I would try to look into the ESD issue that forces us to put a metal sheet under the PCB15:05
wolfsprauland as a side-effect we could also re-test the whole system including all changes made by then15:05
wolfspraulbut I see no urgency for this, at all15:05
wpwrakwolfspraul: alright. wasn't the metal sheet for EMI ?15:29
lekernelwpwrak: it was to protect from malfunction due to perturbations caused by nearby external ESD discharges, which is indeed a form of EMI15:30
wpwrakoh, that's a subtle one15:31
lekernelazonenberg: http://warrantyvoidifremoved.com/ is coming instead of you, presenting DIY metalwork and jewelry :-)16:04
lekernelhopefully there will even be a workshop as well16:04
kristianpaulha, now i realize how usefulls verilog tasks are :)17:15
kristianpaulyou made it !!17:16
lekernelnah, I just submitted it17:16
kristianpaulah yah..17:17
lekernelnow there's still a risk it gets voted down (in fact it happened to many of my slashdot submissions already)17:17
kristianpauli put in doubt the Consumer device part.. but let see17:17
lekernelnah. we'll get there.17:17
kristianpaulno no17:17
kristianpauli dont say we dont17:17
kristianpauljust for the /. article17:17
lekernelthis time if some lamers vote it down again, they'll get it submitted 10 times or so before I give up :-)17:19
Action: kristianpaul tries to remenber his /. account password17:20
wpwraklekernel: what's not so good about your slashdot story is that it's merely an announcement of an event months into the future. there should be a connection to something that's happening now or in the immediate future. e.g., "last call for schematics review"22:44
wpwraklekernel: if you think the long-term announcement itself is also important, you could bundle it with some short-term item, even if minor. e.g., "new firmware now with full ethernet and tcp/ip support" or "ten new ''patches''" (then explain why this is not the kind of patch readers will usually expect) and then sneak in the long-term announcement somewhere later in the story22:49
--- Tue May 10 201100:00

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