#homecmos IRC log for Saturday, 2011-11-19

azonenbergThetawaves: I'm at 20um now and have a roadmap down to 1um10:31
azonenbergusing current equipment10:31
azonenberg20um at 10x reduction => 5um at 40x => 2um at 100x immersion => 1um with double patterning10:31
azonenbergjust need to work out some focuser drift and maybe buy better quality objective lenses10:32
Thetawavesis it some sort of photoresist process?10:32
azonenbergYes10:32
azonenbergSpin coat in photoresist and expose a mask (laser printer transparency, 600DPI) through a microscope objective to reduce feature size10:33
azonenbergthen develop and etch10:33
Thetawaveswhat is limiting you to 20um?10:34
azonenbergMy stage isn't quite perpendicular to the optical axis10:34
azonenbergdepth of field at 40x reduction isn't enough to uniformly expose the whole die10:34
azonenbergi need to spend some time calibrating it10:34
azonenbergonce i get that done, i need to work on drift10:34
Thetawaveswhat does your setup look like?10:35
azonenbergthe stage slowly moves down10:35
azonenbergnot by much but a shift of 100nm is all it takes at these scales10:35
azonenbergit'll blur an immersion exposure beyond recognition10:35
Thetawavesit would be cool to get some professional optics10:35
azonenbergI'm using a $800 metallurgical microscope10:36
azonenbergput the mask over the camera port and backlight wit ha halogen lamp10:36
Thetawaveshttp://www.edmundoptics.com/10:39
Thetawavestook me a while to find that supplier10:39
Thetawavesi don't know about cost, but it's got all the professional shit you'd need10:40
azonenbergYeah10:40
azonenbergi have one of their catalogs hanging out here10:40
azonenberggreat stuff, if i made 10x what i do now :p10:40
Thetawavestoo expensive?10:41
azonenbergDepends on what you want10:41
azonenbergin general, the really cool toys10:41
azonenberglike 385nm laser systems10:41
azonenbergare pricey :p10:41
Thetawavesi was thinking that maybe it would be more convenient to align10:41
Thetawavesso you could do it better/faster10:41
azonenbergI do intend to ebay some stages/focusers at some point10:41
Thetawaves600dpi is what um?10:45
azonenbergIt's about 42um on the mask10:45
azonenbergdivided by the objective you use10:45
azonenbergi usually use 10x which gives me roughly 5um lambda10:46
Thetawavesper dot?10:46
azonenbergYes10:46
azonenbergat a 5 pixel design rule that's nominally a 20um design rule10:46
azonenbergits actually 21.something, i'd have to go run the numbers10:46
azonenbergBut i refer to it as my 20um process for simplicity's sake10:48
Thetawavesdo you have an indicator?10:49
azonenbergindicator?10:49
azonenbergwhat do you mean10:49
azonenbergi have a calibrated camera on my microscope i can use for measuring distances10:49
Thetawaveshttp://en.wikipedia.org/wiki/Dial_indicator10:49
azonenbergNo10:50
azonenbergi do have a digital micrometer10:50
Thetawavesuse that to align your shits10:50
Thetawavesthat's not good enough10:50
azonenberggraduated in microns and accurate to +/- 2 or 310:50
azonenberglol10:50
azonenbergthis is a nice mitutoyo10:50
Thetawavesdial indicators swing about an axis to allow you to check flatness10:50
azonenbergbest measuring instrument i ever used10:50
azonenbergand i align optically10:50
azonenbergremember my exposuire system doubles as a microscope ;)10:50
azonenbergSo i just turn the lamp down low and align the projected image to the features on the wafer10:51
azonenbergthen turn it up high to expose10:51
Thetawavesi think it would be a valuable tool10:51
azonenbergmight be, but i dont have one yet10:51
Action: azonenberg looks at clock10:51
azonenbergalmost 0600 here lol10:51
azonenbergAnd the rest of this stack of exams isn't getting graded while i'm as sleepy as i am now :p10:51
Thetawaveshah10:52
Thetawavesgood night10:52
azonenbergGood night? It's six in the morning :p10:52
Action: azonenberg AFKs10:52
Action: bart416 drags azonenberg back in here11:06
bart416LTSpice is remarkably fast at simulating amplifiers :S14:32
bart4166ms/s...14:33
bart416These guys designed their PCB with the assumption that there are x% of cold joints I'd say :S18:18
azonenbergbart416: what??18:48
azonenberg redundant links or something?18:48
bart416yes18:50
bart416all the diodes in the protection circuits are in parallel18:50
bart416completely useless18:50
bart416except if you're expecting cold joints18:50
azonenberglol18:51
bart416I traced everything and worked it out somewhat18:51
azonenbergWhat about for increased current?18:51
bart416And the only reason is cold joints lol18:51
bart416diodes don't work like that though18:51
bart416They'd pop one by one18:51
azonenbergYou mean they dont have the exact same Vfwd?18:51
azonenbergYeah18:51
azonenbergI've seen that happen18:51
azonenbergBut maybe the designer doesnt know that :p18:51
bart416On IC level you can put them in parallel actually18:51
berndji've seen people claim that diodes do share current nicely18:53
berndjnot that i'm convinced...18:53
azonenbergMaybe at IC level18:53
azonenbergbecause then you can guarantee consistency to a higher level18:53
berndjyeah, much closer thermal coupling18:53
azonenbergjust like parallel transistors are baiscally just a longer channel18:53
azonenbergWell its also consistency in doping etc18:53
berndj*wider, but yeah18:53
azonenbergbut not from die to die18:53
azonenbergwhcih might have been even on different wafers18:53
berndji always thought it's primarily about the loose thermal coupling that BJTs and diodes don't share current nicely - because of the negative tempco of Vf18:54
azonenbergI think its mostly related to Vf being different at time=018:55
azonenbergone of them heats up a little more18:55
azonenbergthen you get runaway18:55
berndjotoh, when i measure Vf of random 1N4007s with my multimeter, invariably they're the same to within a few mV18:56
berndjthere's probably a critical current above which they stop sharing nicely18:56
azonenberga few mV though?18:57
azonenbergThats enouhg18:57
bart416Current takes the path of least resistance as well18:57
azonenbergyou heat up by a tenth of a degree18:57
azonenbergVf changes18:57
azonenbergit heats up even faster18:57
azonenbergyou get a positive feedback loop in which this die dissipates more and more power as a function of the total18:57
azonenbergbefore long, bang18:57
azonenbergi've seen it happen with paralleled LEDs18:57
azonenbergall in parallel with the same series resistor18:58
azonenbergone blew, others were fine18:58
berndjit does, but it also dissipates faster - below a certain current the dissipation actually dominates (IIRC & YMMV - i did this sort of calc for hotspot along the length of a copper wire once, critical current density was maybe a few dozen A/mm^2)18:58
bart416What actually does work to some extent is putting a fast shottky diode in parallel with a regular PN diode18:58
bart416in certain situations that is18:58
bart416you have the fast response from the shottky18:58
bart416and the current capability of the PN diode18:58
berndjhow does the current capability of the PN diode even come into play though?18:59
bart416it depends on the circuit berndj18:59
berndji thought the schottky would destroy itself protecting the PN diode18:59
bart416the thing is, the shottky (aaaaargh c key doesn't work if I type fast and I have to hit it with a sledge hammer) has a higher Vf than the PN in those particular situations19:00
bart416Resulting in the PN diode taking all the current through it19:00
berndjoh - you mean Vf rises faster than a regular diode's does?19:01
bart416A schottky is simply faster at switching19:01
berndjbut also lower Vf19:01
bart416As I said, specific situations19:02
bart416I've seen it done in some high power pulsed laser circuits19:02
berndjhah. throw out the rulebook then19:02
bart416not really19:02
bart416it still sticks to the rules19:02
bart416the schottky is activated faster (the voltage is high enough to activate it anyway), but the PN requires a lower voltage in that case resulting in the PN taking the punishment off the schottky within a few microseconds19:04
berndjregular schottkys or those fancy new SiC ones?19:04
bart416Dunno19:04
bart416Sorry, but I don't study the naming of components in detail19:04
berndji mean the new schottkys where the one half is SiC (wider bandgap) and not plain silicon19:04
berndjlet's you have fast switching at 1000V+19:05
bart416Heh, playing with high current GaAs diodes is also fun19:06
berndjugh. i now realize what "looked wrong" in that "let's"19:07
bart416azonenberg, I think I actually found PCB design software for simple things that doesn't make me want to kill myself :|20:42
azonenbergbart416: Simple things? Yeah20:43
azonenbergIt's when you start doing >=4 layer designs with fat parallel buses and differential pairs20:43
azonenbergthat it gets tricky :p20:43
bart416Then you just start using OrCAD, ADS or Altium...20:44
azonenbergunfortunately i cant afford any of them lol20:44
bart416Those are the only three that seem to be capable of handling more than 2 layers well20:44
azonenbergexpresspcb *can* as long as the two inner layers are just power/ground planes20:44
bart416actually, design spark seems to be capable of more than 10 layers lol20:45
bart416(and not just power or ground planes either it seems :S )20:45
azonenbergBut express can't do differential pairs20:45
azonenbergor autoroute20:45
azonenbergI'm looking at kicad and geda20:45
azonenberghavnet made a choice yet20:46
bart416oh but design spark can auto route20:46
bart416Hence, it's awesome20:46
azonenbergi want to do a small 4-layer design in both20:46
bart416And it also seems to be capable of handling eagle libraries at first glance20:46
bart416try design spark, seriously20:46
azonenbergwinblows?20:46
azonenbergif so thats a game-over20:47
bart416yeah, it's windows20:47
azonenbergyeah, i am not going to be happy with that :p20:47
azonenbergexpress at least sort of runs in wine20:48
azonenbergbut none of the nice ones do20:48
bart416Try to run it in wine20:48
bart416+ also seems to have some links to spice at first sight20:48
azonenbergand can it export gerber?20:48
bart416looks like it20:48
azonenbergvery nice, i'll take a look20:49
azonenberga FOSS tool would be best of course20:49
bart416Well, I'm not really sure how RS managed to do this :S20:49
bart416This is a direct shot at eagle lol20:50
bart416And at first sight it's head on20:50
azonenbergvery nice20:50
azonenberglimited?20:50
bart416Doesn't look like it20:50
bart416well, unless you mean 14 layers is a limit20:50
azonenbergI mean artificially limtid20:50
azonenbergand is it free or commercial?20:50
bart416free, you do need to register but that's it20:51
azonenbergso whats the catch?20:51
bart416and it can "only" do boards up to 1m squared lol20:51
bart416I'm trying to find the catch :S20:51
azonenbergexpresspcb limits you to their proprietary CAD format20:52
azonenbergSo you can only use their fab service20:52
bart416well, there's gerber export...20:52
azonenbergbut its freeware and has no *artificial* limitations20:52
azonenbergthough its design sucks and you can't do >2 signal layers plus optionally 2 power planes20:52
bart416At first sight it doesn't look limited at all :S20:52
azonenbergso designspark is free and unlimited20:52
bart416Looks like it20:52
azonenbergBut then why arent they releasing source?20:52
azonenbergdo they plan to go commercial eventually?20:52
azonenbergand this is like a public beta?20:52
bart416Doesn't look like it20:53
bart416http://www.designspark.com/theme/designspark-pcb20:53
Action: bart416 is stunned as well :|20:54
azonenbergi dont get the catch20:54
azonenbergif they wanted it to be free they'd probably release source20:54
azonenbergotherwise they'd try to sell it20:54
azonenbergor put ads in20:54
azonenbergor lock it to their fab20:54
azonenbergor what20:54
bart416well, it does have features like "export partslist to RS Online"20:55
bart416So it does seem like a bit of a marketing ploy20:55
azonenbergSo they make it easy to use their service20:55
bart416Yes20:56
azonenbergBut they don't lock you in at all?20:56
azonenbergif the program isnt time limited, and you can export to gerber, you can always leave20:56
bart416I haven't spotted neither of such limitations :S20:56
azonenbergweird20:56
bart416+ some degree of eagle compatibility20:56
bart416Win-Win20:57
bart416+ Nice standard library at first sight20:57
azonenbergit just seems too good to be true20:57
azonenbergbut yet they arent releasing source20:57
bart416why would they, it's a marketing ploy in the long run20:58
azonenbergMarketing ploy how though20:59
azonenbergare they just relying on people using their fab service because its easy?20:59
bart416I guess20:59
bart416It shows an ad at the start21:00
bart416for their pcb service21:00
azonenbergOh, ok - its adware21:00
azonenbergEasy to nop out21:01
bart416Actually can't find a direct export to their pcb service easily21:02
azonenbergWell, if all i have to do is add a single well placed 0x9021:02
azonenbergi dont see the harm in using it :p21:02
bart416I'm really going to try this out21:04
azonenbergLet me know how it goes21:06
azonenbergand if you patch out the ads, let me know the byte offset that has to get nopped out21:06
bart416it takes 2 seconds to click away the ad21:07
azonenbergDoesnt matter21:07
bart416It's just a small window at the start21:07
azonenbergI show no mercy to ads21:07
azonenbergif i decide i am not buying from you, that'sit21:07
azonenbergand showing me the ad again wont make me more likely to21:07
azonenbergit just annoys me, and you won't like me when i'm annoyed :p21:08
bart416Considering the profit margin on PCB manufacturing I do see how they can offer this free heh21:08
bart416If 50 people use it to order a 4 layer bord they paid their programmers in india probably21:08
azonenberglol21:08
azonenbergI guess21:08
azonenbergIn any case, i like to say there are two kinds of electronic ads21:08
azonenbergThose i never see21:08
azonenbergAnd those i see once21:08
bart416lol21:10
bart416I'm so not going to bother disassembling a free program :|21:10
azonenbergif they offered a pay version  without ads i might consider it21:12
azonenbergmight even buy a board from their service to justify it21:12
azonenbergBut i will not stand for ads21:12
bart416lol, go to their site and demand that you can pay $1 to remove the ad :')21:12
azonenbergdo they offer the option?21:13
azonenbergbecuase if not i will do what i takes :p21:14
bart416they don't21:14
bart416"But they're open to suggestions"21:14
bart416xD21:14
azonenberglol21:14
azonenbergmy suggestion, i buy a board from them at some point21:15
azonenbergand do what it takes to live without ads21:15
azonenbergThat's if i like the program when i test it out21:15
bart416It's 24h service btw their pcb thing :')21:22
azonenbergVery nice, what's pricing look like?21:23
azonenbergfor say a 4 layer, two or three inches square21:23
bart416size in mm?21:25
azonenbergsay 50-75 square21:25
bart416http://www.pcbtrain.co.uk/quote-and-order-pcb/21:25
azonenbergOh, they arent USA based?21:26
bart416no21:26
azonenbergThat woudl cost more21:26
bart41675 gbp for a 250 mm² board...21:27
bart416( layer)21:27
bart416*421:27
azonenbergBut what does this have to do with designspark?21:27
bart416dunno21:27
bart416Must be connected to RSOnline21:28
azonenbergHmm, i see21:28
bart416http://www.extremetech.com/computing/105931-full-disk-encryption-is-too-good-says-us-intelligence-agency22:05
bart416lol22:05
bart416Challenge accepted22:05
Action: bart416 encrypts all his HDs in Rijndael 256 bit22:06
azonenberg" First, evidence-gathering goons can turn off a computer (for transportation) without realizing its encrypted,"22:08
azonenbergThat's just idiotic22:08
azonenbergyou're supposed to do a RAM dump before shutting it off22:08
bart416That's still sort of useless if the hardware is locking it down22:09
azonenbergIf you get the keys out of ram, you win22:09
azonenbergAnd there are ways around such things if you have physical access, i wont go into too much detail22:09
bart416Who says they're stored in the ram ;)22:09
azonenbergi'm talking stuff like truecrypt etc22:10
azonenbergThere are of course other options22:10
bart416Oh, I'm talking about full blown disk encryption22:10
azonenbergtruecrypt FDE i mean22:10
azonenbergstill stores keys in ram22:10
bart416Done at hardware level22:10
bart416Good luck bypassing that22:10
azonenbergOk, that22:10
azonenbergYeah22:10
bart416You need an electron microscope to get the key, and that's if you're lucky22:10
azonenbergActually, you probably just need to pay off the guy at Samsung22:11
azonenbergwho put the backdoor in it22:11
azonenbergand convince him to give you the key :p22:11
bart416The thing is, there is no backdoor22:11
azonenbergThat the public knows about22:11
azonenbergThats what i like abou GPG etc, source is public22:12
bart416These things are designed in such a fashion that they're hard to probe even with an electron microscope22:12
azonenbergso i can verify its not tampered with22:12
azonenbergYou're telling *me* about tamper resistant hardwarer22:12
azonenbergi happen to have photos of decapped smartcards on my wall :p22:12
azonenbergI'm just saying, you can't rule out the possibility22:12
bart416Yes, believe it or not, they make these things tamper resistant :|22:12
azonenbergthat there is a second copy of the key encrypted with a different password22:12
azonenbergFDE encryps using a session key that's stored on disk encrypted with, guess what' the user's password22:13
azonenbergbut what about the second copy of the key?22:13
azonenbergI guarantee you that if [A-Z]{3} considers these things a threat22:13
azonenbergthey will, and probably have, backdoored all commercially available models22:13
azonenbergIf they havent, the chinese gov't or whatever probably did it to keep tabs on their own citizens22:13
bart416Unlike what you think, governments actually don't like doing that22:14
azonenbergLike, no22:14
azonenbergBut all it takes is one person22:14
azonenbergcould be the mob for all i know22:14
bart416And that person still won't do that22:14
bart416For they use the same hardware22:14
azonenbergBut if nobody else knows its there?22:14
bart416You don't introduce a backdoor into something you use22:14
bart416You still don't introduce it22:15
azonenbergSure you do, if its asymmetric crypto22:15
azonenbergand you have the private key22:15
azonenbergyou pay off the guy and give him your pubkey22:15
bart416You still don't do it22:15
azonenbergIt's risky and i think its unlikely22:15
azonenbergmy point is that you can never be sure22:15
azonenbergAnd depending on how big a target you are sometimes paranoia is justified22:16
azonenbergjust look at google in china22:16
bart416Well, if you are going to be paranoid like that, you might want to stop using a processor made by intel or amd22:16
azonenbergthere is a good reason darpa is exploring how to detect fab-level tampering22:16
azonenbergDoD is worried about it too22:16
bart416Yeah, cause it's their job to be paranoid like that22:16
bart416Keep in mind americans are paranoid about just about anything22:17
azonenbergYeah, but there are folks out there who give us a reason to :p22:18
bart416That's what you think22:18
bart416The fact is that most people don't give a damn if you leave them alone22:18
bart416And all that drama through new services, not helping, not true either22:18
azonenbergYou mean the farce that is airline security?22:18
bart416no, in general22:19
bart416"oimg ... is going to kill all of us"22:19
bart416where ... is the latest hype in something22:19
bart416usually something related to global warming on a quiet day22:19
bart416then it's terrorists22:19
bart416then it's bankers22:19
azonenbergI just subscribe to the opinion that sometimes a little paranoia can be justified22:19
bart416then it's protesters22:19
bart416then it's immigrants22:19
azonenbergBut not taken to extremes22:20
azonenbergAt the same time, it helps to know more about the threats22:20
azonenbergAnd it's pretty well known that the PRC wants data22:20
azonenbergon us companies22:21
azonenbergand will do just about anything to get it22:21
azonenbergSo i would say that it's a very real threat22:21
bart416And do you really think they have trouble getting it being that all your production has been outsourced to china...22:22
azonenbergNot at all lol22:22
azonenbergWhich is a problem :p22:22
azonenbergThats a reason some high security designs are using FPGAs22:22
azonenbergbecause its very hard to backdoor them when you dont know what's going to run on them22:22
bart416it's a problem when you make it one22:23
azonenbergespecially given randomize place-and-route etc22:23
azonenbergwhat do you mean? They want to sell counterfeit copies of stuff22:23
azonenbergand get data to keep track on their citizens22:23
bart416You know, the paranoia is actually more likely to cause problems than the actual "problem"?22:29
bart416The entire intelligence community should gtfo and get a life22:29
bart416And so should all people for that matter who think they have some sort of rightous cause22:30
bart416Cause they don't22:30
soul-dmeh just make a panic button and a termite charge22:46
soul-dtry to read data back from that22:46
soul-dor jsut wget the whole internet as noise   im scared of the stuff i have on my pc22:48
soul-dbut i don't know if i want to get caught  week ago i found  a book on chemical warfare :') for example22:48
azonenbergMy school library has one of those22:49
azonenbergAnd a termite charge? Hmm22:49
azonenbergSo when it explodes all the wood within 50 feet gets eaten?22:49
azonenberg;)22:49
soul-djust enough to melt your ip :)22:50
azonenbergno, i mean you said termite, not thermite :p22:50
azonenbergas in the bugs22:50
soul-dheh  maybe if you have lots of paperwork22:50
soul-ddownloaded a giant library once so just search pc for books  like on mems   5+ books or so :D  but thats all nm talk :(22:54
soul-dmaybe small talk was better word play23:01
--- Sun Nov 20 201100:00

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