#homecmos IRC log for Saturday, 2011-07-30

B0101Azonenberg: Hi, hows your work on MEMS?00:02
azonenbergGoing pretty well, I'm going to be doing another few experiments over the weekend with the goal of preparing specimens to cross-section image on the SEM later on00:03
azonenbergi want to measure some of my layer thicknesses00:04
azonenbergMy Si wet etch process pretty much works, the only thing left to do is fix the exposure source (better lamp) and use a thinner Ta2O5 layer to reduce overetch00:04
azonenbergif all goes well i'll do another evaporation run in mid-august on half of a <110> wafer coated in ~50nm Ta2O500:05
azonenbergput 100-200nm of Cu on that00:05
azonenbergthen cleave into dies and start actually building devices00:05
B0101oh, thats interesting.00:06
azonenbergEnd of August to early September i should be able to do some high-aspect-ratio tests00:06
azonenbergthen soon after that build an actual comb drive00:06
azonenbergall of the other processes are mostly figured out00:07
azonenbergI'll probably do the first version really simple, only a couple of fingers with barely visible motion00:07
B0101hmm... ok00:29
B0101Azonenberg: You should make some videos about your experiments01:11
azonenbergB0101: Maybe at some point01:11
azonenbergA lot of the litho has to be done under darkroom conditions so there wouldnt be much to see01:11
azonenbergBut if i have some time i can try to set something up01:12
B0101Do you have an electron microscope?01:13
B0101Ok, I am just kidding, of course its too expensive01:14
azonenbergB0101: Not *yet*01:44
azonenbergBut i can get access to one on campus here and there01:44
azonenbergAnd I plan to buy one in about 5-10 years when i have the budget01:44
azonenbergIt'd cost about as much as a decent car01:44
azonenberg*nice car01:44
azonenberg$20K ish01:44
azonenbergIf I evaporate 80/20 NiCr it should resist KOH nicely01:52
azonenbergwithout need for a secondary hardmask01:52
azonenbergBut i dont know if it will work with tungsten boats01:53
bart416azonenberg, IR camera...05:41
bart416Isn't the photoresist UV sensitive?05:42
azonenbergYou mean for filming stuff05:43
azonenbergYeah, IR would work if i had one05:43
azonenbergThe only IR camera i have is focused for really short range05:45
azonenberglike 6"05:45
azonenbergAnd isnt easy to refocus (electrical tape holding the lens elements together, etc :P)05:45
bart416Old CCD webcams are pretty IR sensitive if you remove the filter and replace it with a piece of photo film05:50
azonenbergbart416: Thats actually what i have05:50
azonenbergI have no filter at all05:50
azonenbergso it's sensitive to visible + IR05:50
azonenbergBut the lens got a bit messed up and kinda hacked05:50
bart416Then again, I sort of had a lot of time to perfect it05:50
azonenbergso its held together in what happens to be a very short range focus05:50
bart416We have to do some sort of uhm thesis like test at the end of highschool05:51
bart416(obviously less complicated)05:51
bart416Mine was about tracking a hand against a wall05:51
bart416We solved the skin colour problem by IR, also way less noise heh05:51
bart416But it depends on a camera to camera basis if it works well05:52
bart416The cheaper ones are usually easier05:52
bart416azonenberg, how paranoid are those idiots :|06:58
azonenberglol no idea06:58
bart416Also, I got my BGA soldering process to work while drunk, lol07:54
bart416That's the good news08:01
bart416The bad news is that I went through roughly 40 ICs to get it to work well08:01
azonenbergWhat pitch?08:02
azonenbergand what kind of pcb?08:02
bart416scrap pieces of dual layer PCB08:03
azonenbergAnd what kind of chips?08:04
azonenbergand i thought the 0.8s i had were small08:04
azonenbergFour more pins to route08:06
azonenbergin a sea of wires between an FPGA and a SRAM08:06
bart416I'm looking into package on package for my ARM Cortex board08:07
azonenbergThis design really needs six layers08:09
azonenbergAnd probably a bga fpga08:09
azonenbergThat one little spot right between the fpga and the ram is getting too dense to route anything08:09
bart416heh I'm first drawing everything on paper >_>08:09
azonenbergi probably wont be able to clock the ram faster than 20-40 MHz given those deltas in trace length08:09
azonenbergAnd how would you do that for something like this? :P08:10
azonenbergOne of the problems is that this fab (expresspcb) cant do really tiny vias08:10
azonenbergOr blind vias08:10
bart416they might just work azonenberg08:10
azonenbergSo i cant route under any power vias08:10
azonenbergAnd they cant do vias smaller than 660um outer diameter with a 200um hole08:11
azonenberg*wants laser-drilled microvias*08:11
R0b0t1Blind via?08:12
azonenbergR0b0t1: A via that doesnt penetrate the board all the way08:13
azonenbergand goes from an outer layer to an inner one but not out the other side08:13
azonenbergburied vias go between two inner layers and dont penetrate either surface08:13
azonenbergThis fab cannot do either08:13
R0b0t1I'd never seen those on things I've handled. Are they relatively rare?08:14
R0b0t1Well, one you can't see, but...08:14
azonenbergExcept on things like GPUs and mobos08:14
azonenbergWhere there are so many you'd never see them08:14
azonenbergAnd cell phones08:14
azonenbergThose things are so dense you wouldnt believe it08:14
azonenbergThey basically pile the top and bottom board surface with components08:15
azonenberglaser microvia in every pad08:15
azonenbergand do all the interconnect on inner layers with buried vias :P08:15
azonenbergi'm exaggerating a little bit, but they're DENSE08:15
azonenbergCheck out this guy for example http://www.huv.com/gumstix/Gumstix2-big.jpg (gumstix single-board computer)08:16
azonenbergAll BGA for the ram, flash, and cpu08:16
azonenbergone tssop08:16
azonenberglooks like a SMT cap array for decoupling (or maybe that's bus termination resistors)08:17
azonenberglots of 0402 or 0201 decoupling caps08:17
azonenbergvias everywhere08:17
azonenbergIn the back http://www.huv.com/gumstix/Gumstix-big.jpg you can see the lack of significant blind vias on this one08:18
azonenbergsince the BGA vias go all the way  through08:18
azonenbergBut they are using a process that can do uber-tiny traces that go between bga pads etc08:18
azonenbergThe one i'm using doesnt go below 150 micron trace width08:18
bart416Some motherboards and cellphones have up to 8 layers08:19
azonenbergtry 12+08:19
bart416And sometimes even components embedded in between the layers08:19
azonenbergYeah, those are rarer08:19
azonenbergBut not unheard of08:19
bart41612+ is only on very expensive high end products though08:21
bart4168 is normal manufacturing at this point08:21
azonenbergLike the ones that use $20,000 FPGAs in 1900 ball packages? :P08:21
bart416Yeah, that sort of stuff, lol08:21
bart416"Oh hey 4 power layers!"08:21
azonenbergYeah, i wanted to do that08:22
bart416I'm doing 500 pins on two layers08:23
bart416It's mathematically possible08:23
azonenberg500 total pins, or 500 pins on one chip?08:23
bart416one chip08:23
azonenbergBut i'd be willing to bet you have big pitch and tiny traces08:24
bart416Why do you think I needed to perfect 0.4 mm pitch process?08:24
azonenbergI assumed it was for something small lol08:25
azonenbergAnd http://www.ami.ac.uk/courses/topics/0262_hdi/index.html08:25
azonenbergThese things are blurring the line between PCBs and ICs08:25
azonenbergusing photolithography and plasma etching for vias08:25
azonenbergresistors and caps embedded in the board08:25
bart416I wonder if you could layer / layer deposit your own PCB to more than two layers08:27
azonenbergGiven a vacuum evaporator and such, i could make them using IC-style materials :P08:27
azonenberg500nm Cu + 200nm SiO208:27
azonenbergper layer08:27
azonenbergStart with a dual layer FR4+30um Cu substrate for power08:28
azonenberglay down a thick dielectric layer, punch vias08:28
azonenbergfill vias with conductive material and CMP flat08:28
azonenbergWell its 0430 here and i've been up a bit longer than planned... time to get some sleep lol08:29
R0b0t1azonenberg, whoah09:03
berndjazonenberg, hi.  speaking of vacuum evaporators, do you think FR4 would outgas too much for a nice layer of Cu?  at least Cu isn't crazy reactive like Al is, but still... need to have a good free path length14:07
berndjthat pacman ghost thing or whatever it was is cute :)14:07
nathan7berndj: the nyanotechnology?14:39
lekernelgiven that FR4 smells sometimes, it probably outgases15:38
azonenbergberndj: Hmm17:34
azonenbergI've evaporated Cu before, but on Si and SiO217:34
azonenbergand Ta2O517:34
azonenbergif anything outgasses it'd be the epoxy17:34
azonenbergthe glass wont care17:34
azonenberglekernel: What you might be smelling is particulates though17:35
azonenbergBut yes, it probably does17:35
azonenbergberndj: also, that's a nyan cat17:35
berndjoooh the one that's rainbow-propelled!17:37
azonenbergDidnt you recognize it?17:40
berndjno, my pop culture fu is rather weak17:40
B0101Azonenberg: Hi, have you got any updates or discoveries23:48
azonenbergB0101: Running some copper experiments tonight in an hour or so23:49
B0101hmm for the interconnects?23:49
azonenbergFor hardmask, actually23:49
azonenbergI want to see a) how well it survives KOH, and b) design a better etch23:50
azonenbergSC2 is too fast23:50
azonenberggotta dilute it23:50
--- Sun Jul 31 201100:00

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