#homecmos IRC log for Monday, 2013-12-16

--- Mon Dec 16 201300:00
berndjwith the recent paranoia about NSA-infected random number generators, how hard would it be to decap some chips and put them under the microscope to check if their built-in RNGs really do what it says on the tin?16:52
nats`the biggest problem in my opinion would be to reconstruct the RNG circuit and see if it's a real one16:53
nats`+ other part of the chip could come and manipulate some value like Vgate etc...16:54
nats`I'm not sure it's better than a statistic analysis of the output16:54
soul-dhttp://people.umass.edu/gbecker/BeckerChes13.pdf17:00
soul-dit explains a  few  facets your intrested in  and also  have a conclusion you'r not gonna like17:00
azonenbergsoul-d: actually i'm collaborating with the folks behind that paper17:31
azonenbergThere's a few things they didn't think of17:31
soul-d:)17:35
azonenbergI'm just gonna stick this in here http://colossus.cs.rpi.edu/pictures/2013/November/11-25-2013%20-%20xc2c32a%20SEM/jpegs/xc2c32a_dash_27_se_4500x_10kv_12mm_annotated.jpg17:35
soul-dmmm what you break ?17:36
berndji still struggle to read those micrographs19:36
Sync_the problem is that looking at them with a microscope is not really enough20:01
berndjis there any way to tell n-type from p-type just by looking at the scan?20:08
berndjfrom what i've seen you can at least see boundaries between (some?) types of doped areas20:08
berndjis that due to the doping or due to different oxide/nitride layer thicknesses?20:09
nmz787looks like there's height diffs too in addition to brightness20:09
nmz787brightness could be due to shifts in composition, but height would do it too, so hard to tell20:09
nmz787azonenberg: I was wondering what you think of achieving 1 micron features relatively close together with a DLP projection system, and maybe a microsope. Currently I can get the focus of the projector down to about 1024 pixels across 1cm, so that's ~25.4 microns per pixel, but I dunno how much diffusion will occur. I have some PCB resist and some old (expired) SU-8 that IDK if would still work20:12
azonenbergberndj: In general you cannot tell doping from a micrograph20:13
azonenbergThat image is after a selective stain process to highlight P doping20:13
azonenbergi grew oxide over P-type implants with dash etch20:13
azonenbergnmz787: Not sure20:14
azonenbergI've been away from this project for a while due to lack of time and funding20:14
azonenbergonce i have a real job and am out of grad school i plan to get back to it20:14
nmz787azonenberg: lemme know if you move to Oregon :P20:14
nmz787azonenberg: intel is building tons of new fabs20:14
nmz787azonenberg: what about theorhetically, off the top of your head?20:15
nmz787any ideas?20:15
nmz787I have been talking with a resist company TOK and they've given me some simulation stuff in response to some of my questions/desires20:15
azonenbergI'd say you definitely want additional optics20:15
azonenbergthere was a paper on doing DLP for MEMS20:16
nmz787yeah i may have it20:16
azonenbergi am more inclined to do 405nm laser direct write20:16
nmz787well, dunno for mems, but it was a relatively DIY paper20:16
nmz787yeah but with direct write you then need interferometers for feedback, which makes everything super vibration noise prone20:17
nmz787you might get around it with a DVD player like optical voice coil focus, but write time would be slow20:18
nmz787and a camera for inspection20:18
azonenbergYou dont necessarily need interferometers20:19
azonenbergI was thinking of doing optical stages with rotary encoders20:19
azonenbergThat would get me down to a few microns i think20:19
azonenbergas in, assume the leadscrew is precise and just measure angular position20:20
azonenbergi have some for example that are 250 microns per turn20:20
azonenberga 256-tick encoder gives me slightly submicron positioning20:20
nmz787i'm interested in microfluidics, and need to have straight walls and curves reproduced correctly, so leadscrew assumption is hard for me to get over20:30
nmz787i figure TI has done the hard work for me re: precision, as long as any optics don't distort the image too much20:31
Sync_nmz787: leadscrews will do your trick nicely20:46
Sync_c3 precision and everything is pretty easy20:46
nmz787c3 precision?20:50
Sync_yes20:51
nmz787googling20:51
Sync_it basically tells you the integrated deviation over 300mm20:51
Sync_in µ20:51
Sync_so with c3 you get 8µ in 100mm20:52
Sync_mostly it is less20:52
Sync_I got some c3 ones in my microstages and I'm hitting +-1µ over 50mm20:53
nmz787how much would a set cost?20:53
nmz787is there a c4 grade?20:53
nmz787:D20:54
Sync_there is20:54
nmz787i am working on microstepping now with arduino20:54
Sync_well that depends on the pitch and diameter20:54
Sync_but my 1mm lead 10mm diameter ones are about 1k for 130mm long ones20:54
Sync_the length is not the issue with cost but the nut20:55
nmz787hmm, well my DLP was $30 on ebay in decent enough shape (no bulb). I can shine a laser in that and get a decent image20:56
Sync_I have some of those cheap dell DLPs here20:57
Sync_I messed around with the a bit but in the end microstages are easier and more reliable imho20:58
nmz787for a large effective surface I'd still need a stage though20:58
nmz787microstepping and/or moving the image around the DLP area could help with fine alignment of sections20:58
Sync_stages are also cheap21:02
--- Tue Dec 17 201300:00

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