#homecmos IRC log for Tuesday, 2011-08-30

reportingsjrkristianpaul: an entire pic18 clone? 0.0 wow00:38
reportingsjrthat is awesome00:38
azonenbergreportingsjr: i got most of the way through a pic12 implementation in an evening00:40
azonenbergthen abandoned because i realized the architecture was annoyingly inefficient00:40
azonenbergi wanted something pipelined00:40
reportingsjrazonenberg: how do you do that??00:41
azonenbergverilog on an fpga, obvs00:41
azonenbergno, my fab process isnt that advanced ;)00:41
reportingsjrhaha00:44
berndjazonenberg, do you expect diy semi fab to be more suited to small / high speed devices, or to power devices?01:02
berndj(once you get over the hump of "works at all")01:02
azonenbergberndj: MEMS are the easiest since they're least susceptible to trace metal contamination01:02
azonenbergThen power01:02
azonenbergbecause power is easy to do with large process sizes01:02
berndjyou don't think they're more fussy about impurities, and inhomogeneities thereof?01:03
azonenbergOnce you get MOS devices to work, no01:03
azonenbergin fact, i think big ones would be *less* suscpetible01:03
berndj(otherwise you have atrocious SOA)01:03
azonenberga single particle of something is a smallesr portion of the device01:03
berndji hope you're right01:03
azonenbergI have no idea, thats just a guess01:04
azonenbergMEMS are my field, cmos is a longer term goal01:04
berndjyeah - you just don't want that single particle to be more conductive than the rest, and to hog all the current01:04
azonenbergOf course01:04
azonenbergBut the asusmption is that you dont have the particle embedded in the device01:04
azonenbergi.e. that it was a bump in photoresist etc01:04
azonenbergAlso, not much dust is conductive01:04
azonenbergits mostly organics (and, more importantly, not stuck very hard to anything)01:05
berndjreason i ask is after looking over some opensourceecology.org - and realizing that they're probably dependent on semi power devices more than on any single other semi tech01:05
azonenbergi'd be more concerned about metal ions01:05
azonenberggetting rid of those is gonna be a pita01:05
berndjyeah, i was thinking more about non-uniformities of dopants01:05
azonenbergTMAH is a little toxic to be recommending for amateur level fab01:05
azonenbergi want to explore use of straight ammonia as a developer01:06
azonenbergand dopant difusion isnt really likely to be a problem01:06
azonenbergSpin coating gives a very even film of doped SiO201:06
azonenbergHF etch that away through windows in photoresist01:06
berndjwhere do you draw the line between "calculated risk" and "hey kids, try this at home"?01:06
azonenbergthen coat with undoped sio2 and diffuse01:06
azonenbergberndj: Its different for each person, depends on facilities and experience01:06
berndji get the feeling you just can't get away without dealing with murderdeathkill substances01:06
azonenbergAnd at least MEMS are possible with nothing worse than line-level voltages (120/240 AC), vacuum, KOH, and 3% HF01:07
berndjyou've already accepted HF, i'm not sure how much worse you can go!01:07
azonenbergplus some common solvents01:07
azonenbergthats the thing, its extremely dilute hf sold OTC01:07
azonenbergbad, but not THAT bad01:07
azonenbergCVD is a definite no01:08
berndji was going to ask if there's a material difference between 3% and 30%01:08
azonenbergYes01:08
azonenbergAmong other things, 3% doesnt fume noticeably01:08
berndjbut F- diffusing through your skin is one problem01:08
azonenbergYeah, thats the BIG one01:08
azonenbergthe concentration is 10x lower01:08
azonenbergso you have a lot more time to rinse it off01:08
berndjyou don't ALSO need it to be frothing and desiccating or whatever it does01:08
azonenbergbefore it does any damage01:08
azonenbergand there's less F- to neutralize after the fact01:08
berndjdoes HF desiccate stuff?01:09
azonenbergI dont think so01:09
azonenbergthe 3% doesnt at least01:09
azonenbergits an aqueous solution :P01:09
berndjyeah, there's 97% of pre-desiccated stuff!01:09
berndjanyways nothing like fuming H2SO401:09
azonenbergoh, no01:09
azonenbergi want nothing to do with that01:09
azonenbergthe only acid i'm using in my lab besides 3% hf is concentrated (30%) HCl01:09
azonenbergand from what i hear it's much more mild than concentrated h2so401:10
berndjour high school science teacher turned a blind eye when us A students wanted to "try something"...01:10
berndjthe H2SO4 + sugar trick was pretty neat01:10
azonenberglol01:10
azonenbergin terms of carbonizing organic matter (experimenters included)01:10
azonenberghcl is much less nasty than h2so401:10
berndjbut tbh (and i was younger and stupider then) i did fubar a little bit once - made nitric acid in my lungs!01:10
azonenbergum, yeah01:11
azonenbergno2 == nasty01:11
berndjcoughed like hell and i imagine i spat/coughed a trace of blood01:11
berndjno >5-minute-term problems though01:11
azonenbergNever got a dose nearly that heavy01:11
berndjyeah, we did something pretty dumb.  neat HNO3 onto some metal shavings i think01:12
azonenbergworst i got was some stinging eyes for a few seconds from vapor01:12
azonenbergno splashes, had goggles and i think a face shield on01:12
berndjgoggles?  lol01:12
berndjhow did i ever survive chem class01:12
azonenbergThis was decapping ICs01:12
azonenbergheat sample to 150C on hot plate01:12
berndj*which i still need to do01:12
azonenbergone drop conc. HNO3 on center01:12
azonenberglet it react, grab with tweezers/forceps and swish particles off in acetone01:13
azonenbergrepeat until die is exposed01:13
azonenbergthe quantities were small and we were quite careful01:13
azonenbergand from now on i have no plans to repeat the experiment outside a fume hood01:14
berndji always forget, acetone has THREE carbons01:14
berndjmnemonic: acet-foo is meth-foo + 1, and it's a ketONE01:15
azonenbergochem <shudder/>01:16
berndji've forgotten a lot of it.  majored in biochem among others01:17
Action: azonenberg never even took college chem01:17
azonenbergmajored in comp sci :p01:17
berndjnow THERE is some nasty stuff01:17
azonenbergnasty stuff? You want nasty stuff?01:18
berndjstuff that'll give you cancer, guaranteed, if you just look at it wrong01:18
azonenberghttp://en.wikipedia.org/wiki/Chlorine_trifluoride01:18
berndjno cheating01:18
azonenbergIf you get this stuff on you, you wont survive long enough to get cancer :P01:18
berndji like that "hypergolic with no measurable delay" part01:18
azonenberg"with test engineers"01:19
berndjand ignites asbestos01:19
azonenbergYep01:19
berndjof all things01:19
azonenbergProbably near the top of materials i do not want to get anywhere near :p01:19
berndji wonder if that's the closest analog to hollywood action movie "acid" pools01:20
azonenbergI dont even know what state it's in at room temperature01:20
azonenbergis it liquid?01:20
berndji imagine H2SO4 might also work for that01:20
azonenbergno, boils at 11.7C01:20
azonenbergh2so4 would work, yes01:21
azonenbergthis stuff would be vapor01:21
azonenbergThough if released anywhere near anything organic, it'd look like a ball of fire :p01:21
azonenbergThis stuff is so nasty that it burns asbestos creating smoke made of HF01:21
azonenberg"the operator is confronted with the problem of coping with a  metal-fluorine fire. For dealing with this situation, I have always  recommended a good pair of running shoes."01:22
berndjif HF is the combustion *product*...01:23
azonenbergYeah01:23
azonenbergThen the original stuff is even worse01:23
berndjmaybe you can placate it by feeding it metallic potassium?01:23
azonenbergFrom a significant distance away, i hope?01:23
berndjthen it will "only" be a fluorine-alkali fire01:23
berndjor a nuke: turn the fluorine and chlorine into neon and argon, respectively01:24
azonenberglol01:24
berndjthey'll be glad the lab only got nuked01:25
azonenbergYeah01:25
azonenbergSuffice it to say, thats top on the list of substances i never want to work with01:25
berndji know you're going for MEMS first, but have you got a plan for what dopants to use?01:25
azonenbergor even near :p01:25
azonenbergYes01:25
azonenberghttp://www.emulsitone.com/bsif5x1020.html01:25
azonenbergand phosphosilicafilm01:25
berndjand p-type?01:26
azonenbergP-type is boron01:27
azonenbergthe link01:27
berndji assume you're working with p-type wafers01:27
azonenbergN-type is phosphorus01:27
azonenberghttp://emulsitone.com/psif3x10_20.html01:27
berndjso... some sort of boric acid solution?01:27
azonenbergi dont know whats in it01:27
azonenbergbut the end result is doped SiO2, you etch with HF01:28
azonenbergand drive in with thermal diffusion01:28
berndjoh and regarding vacuum chambers01:28
azonenbergYeah?01:28
berndji suspect stainless steel is going to be prohibitively expensive01:29
azonenbergWas planning to use a glass bell jar for the evaporator01:29
berndjwhat problems could one expect from other steel?01:29
azonenbergThey outgas more but my guess is that at the pressures we'd be doing it wouldnt matter too much01:29
berndjhmm. i guess there's room for stuff to hide in the rust that would invariably build up01:29
berndjHF scrubdown!01:30
berndji found my first arc welding experiment harder than i thought01:31
berndjmajor problem: the stick moves in the time that you flip the shade glass down01:31
azonenbergYeah01:31
azonenbergI had trouble doign that in class01:32
azonenbergi want to try using an auto-darkening lens01:32
berndjgas welding was easy by comparison, but no use for making a vacuum chamber01:32
azonenbergthey basically have a phototransistor that darkens an LCD-based (?) shutter01:32
azonenbergreacts in a few microseconds or something01:32
azonenbergso the dose of UV etc you get is minimal01:32
azonenbergyou dont even see the flash01:32
berndjyeah, i think i'll have to do that.  they're just expensive enough to require a "yes, i really want to do this" decision01:32
azonenbergfrom what i hear nobody professional uses the old-school ones for anything :p01:33
berndjyeah, it's fast in darkening, but takes milliseconds to open up again01:33
azonenbergYep01:33
azonenbergheck, half a second would be fine for brightening lol01:33
berndji'd just like to either a) understand the mechanism properly, or b) see every last ANSI/ISO mark on it01:34
berndjdon't want to depend on a battery or solar cell for eye safety01:34
azonenbergLol, i know what you mean01:34
azonenbergI think most of them start out at shade 4 ish anyway, so thats a start01:34
azonenbergand i think they have a test button01:34
berndjhmm, i didn't know that, that would give me more confidence01:34
azonenbergin any case i would say buy a good name brand01:35
berndji hear your warning about pressure vessels, but frankly i'd be more scared of a bell jar than of a (well-anchored and made of malleable) steel sphere01:36
azonenbergPut a screen over it01:36
berndjand yes, shock waves can kill01:36
berndjso there'd be a screen too01:36
azonenbergalso, how big of one are you talking about?01:37
berndjbig enough for telescope mirrors :)01:37
azonenbergIf its like 8" across and a foot high, the total energy released in an implosin wouldnt be that bad01:37
azonenbergSomething big enough for scope mirrors? Oh01:37
azonenbergYeah, that would be problematic if it went poof01:37
berndj8 inch diameter disks would be okay, but ideally i'd like to go bigger01:37
azonenbergi meant 8" diameter jar01:38
azonenbergwhich would give you maybe a 4" diameter plating area01:38
berndjwhere are the other 4 inches?01:38
berndjsurely the glass isn't 2 inches thick...01:38
azonenbergno lol01:38
azonenbergmounting for the electrodes, etc01:38
berndjah yes. feedthroughs01:39
azonenbergas well as the actual sucking hole to evacuate the chamber01:39
berndjspark plugs :)01:39
berndji like how ClF3's NFPA 704 says "will not burn"01:45
berndjnot as a fuel, no!01:45
azonenbergYeah01:45
azonenbergBut it will oxidize anything in sight01:45
azonenbergheck, i think some fluorinating agents can oxidize oxygen to a higher oxidation stat elol01:45
berndjheck, oxidizing argon01:47
azonenbergXeF201:48
azonenbergThe fact that it exists is scary :p01:48
berndjdid you ever consider basing a diy semi fab process on anything other than silicon?01:49
azonenbergNot really01:50
berndji think the obvious candidates would be ZnO, SiC, CuO01:50
azonenbergI am mostly doing mems though01:50
azonenbergCuO is a possibility but its a lot harder to work with01:50
azonenbergin terms of etching etc01:50
berndjis silicon the only game in town for meme?01:50
azonenbergIts not, but the existence of KOH for anisotropic etching is a major win01:50
azonenbergeasy to obtain, not too toxic, and gives you nice slopes or vertical sidewalls01:51
berndjeew, i can still smell acetone on my fingers while eating supper01:51
azonenbergshouldnt be possible01:51
azonenbergthe stuff is so incredibly volatile01:51
berndjwas cleaning a pot (water didn't work)01:51
azonenbergmust be vapor in the room or something01:52
azonenbergbut not liquid01:52
berndjcan't smell it now.  maybe some bitterant that i'm confusing for acetone itself!01:52
azonenbergThen again i always wear gloves when using acetone or other solvents in nontrivial amounts01:52
azonenbergall of that defatting is kinda unpleasant lol01:52
berndjNaOH is the bugger there01:53
azonenbergnaoh is worse, yeah01:53
azonenbergturns your fat into soap01:53
berndjturns your fingers into soap until you rinse with vinegar01:53
berndjok, so anisotropic etching gives you the high aspect ratios - steep walls.  right?  if you make, say, a little gear, how do you get it unstuck from the bottom of the pit?01:55
azonenbergYou'd need to build it on a sacrificial layer01:56
azonenbergthere are a couple of options01:56
azonenbergOne is SOI01:56
azonenbergand then etch away the oxide01:56
berndjsacrificial layer? doesn't that imply CVD?01:57
azonenbergNot really01:57
azonenbergyou can do a lot with evaporation afaik01:57
azonenbergyou can spin coat sio201:57
azonenbergits solid-phase deposition01:57
azonenbergliquid*01:57
berndjhmm.  what is the gear made of then - SiO2?  i had assumed it would be Si, but i guess that's a spurious assumption01:58
azonenbergYou could do polysilicon by evaporation or sputtering01:58
azonenbergBut then you cant do anisotropic etch01:58
azonenbergthats something i'm very interested in solving01:59
azonenbergbut dont yet have an answer to01:59
berndjwhat sort of aspect ratios do you work with?  i'm not sure i interpret SEM-grams properly01:59
azonenbergRight now? I havent tried going too far01:59
azonenberghoping for 10:1 with KOH on <110>01:59
berndjok, so that steep02:00
azonenbergIn metal layers, much shallower02:01
azonenbergmaybe 30-45 degree slope02:01
azonenbergsince the etch is isotropic02:01
berndjand you can't play games with photoresist layers?02:02
azonenbergI want to explore something similar to DRIE using multiple photoresist depositions02:02
azonenbergspin coat, expose, develop,  etch a little02:02
azonenbergstrip resist02:02
azonenbergcoat (including sidewalls of undercut area)02:02
azonenbergexpose, develop, etch again02:03
azonenbergit will use a lot of PR but if i do a couple of steps i may be able to improve aspect ratios (assuming i get good alignment)02:03
azonenbergBut the idea is the same02:04
azonenbergDRIE uses an isotropic etch followed by protection of sidewalls02:04
azonenbergthen repeats the cycle02:04
azonenbergso if i did it right i'd get something that looked very much like a Bosch process feature02:04
azonenbergmake sense?02:06
azonenbergIts an idea i've had for a while but havent had time to explore much02:06
berndjyeah02:07
berndjthough full of jargon i'll have to look up :]02:07
azonenbergwhich terms are you talking about?02:07
berndji'm starting with "the" :)02:08
azonenberg:p02:08
azonenbergBosch process is http://en.wikipedia.org/wiki/Bosch_process_%28microtechnology%2902:08
berndjDRIE, bosch process (not to be confused with the... is it ammonia synthesis process?)02:08
azonenbergSF6 isotropic RIE followed by C4F8 passivation02:08
azonenbergDRIE = deep RIE02:08
azonenberghttp://en.wikipedia.org/wiki/File:Bosch_process_PILLAR.jpg is a nice example02:09
azonenberghttp://en.wikipedia.org/wiki/File:Bosch_process_sidewall.jpg is a closer view02:09
azonenbergnote the rounded cutouts on the sidewalls02:09
azonenbergEach iteration of the process you do an isotropic etch (which has undercut because its isotropic) and then follow with a passivation process02:09
azonenbergthe passivation protects sidewalls but is quickly eroded in the bottom of the pit02:09
berndjoooh, so DRIE is a bit like FIB but with ions more reactive than Pd / Pt / W / etc?02:11
azonenbergNot exactly02:11
azonenbergThat's RIE in general02:11
berndjso... higher energies?02:11
azonenbergRIE is generally mildly anisotropic because the ions are being blasted into the wafer from a vertical direction02:12
azonenbergBut sometimes they bounce and etch sidewalls02:12
azonenbergDRIE = deep RIE = higher aspect ratios02:12
azonenbergYou start out with a short RIE step02:12
azonenberggiving you a mostly vertical etch with some undercut02:12
azonenbergThen you passivate the entire hole02:12
azonenbergYou then do more RIE02:12
azonenbergThe KE of the ions knocks the passivation off the bottom02:12
azonenbergbut the undercut is slow enough it doesnt attack the passivation much02:12
azonenbergsince that actino is only chemical and not as mechanical02:13
azonenbergThen when you've etched long enough that the passivation is significantly thinned, you re-passivate02:13
azonenbergand do another etch step02:13
azonenbergand keep going02:13
azonenbergMy process would be similar except instead of using chemical passivation removed by the KE of the etch02:14
azonenbergi'd be using photoresist passivation removed by light shining down the hole during the repeated exposure step02:14
azonenbergThe profile of the resulting feature would be nearly identical (though given the need for a lot of expose-align steps i dont think more than 4-5 would be feasible in a realistic amount of time)02:15
azonenbergBut if i just need say a 5x improvement in anisotropy it might help02:17
berndji hope you don't get trouble from sidewall reflection / diffraction02:17
azonenbergNo idea02:17
azonenbergJust ran the idea past a guy i work with - he thinks the alignment will be a PITA but it will likely work02:22
lekernelazonenberg: do you know about this? http://www.utwente.nl/tnw/ems/research/cooling/Technology/stirling.doc/09:05
lekernelthey look very simple on the schematics :)09:05
azonenbergstirling engine? Heard of them09:05
azonenbergis this a mems one?09:05
lekernelstirling cryocoolers09:05
lekernelno09:05
bart416If you run a stirling engine in reverse it works like a heat pump azonenberg09:05
azonenberginteresting09:05
bart416They use it for FIR sensors09:06
lekernele.g. http://www.teledynejudson.com/files/pdf/dewars_PB4103.pdf09:06
azonenbergyeah, i think i've heard about those09:06
lekernel"Temperature Sensor Diode: The ***2N2222*** diode is indicated"09:07
lekernellol? are they talking about a standard 2N2222 transistor?09:07
lekernelI built quite a few small radio transmitters with those when I was a kid :-)09:08
lekerneldidn't know it's apparently good for measuring cryogenic temperatures09:08
lekernelI'm very surprised by the apparent simplicity of those devices. I guess the devil is in the details ...09:10
azonenbergi knew they used diodes for measuring temperatures09:11
azonenbergbut i didnt know it worked that low09:11
azonenbergthe bandgap varies with temperature09:11
azonenbergand by extension the forward voltage09:11
lekernelwell, I can easily try to build one based on bicycle pumps and duct tape and see what happens :-)09:12
azonenbergWhats your intended application?09:19
lekernelhopefully, reaching cryogenic temperatures at some point :)09:36
bart416azonenberg, VT as temperature reference works right down to absolute zero :P09:38
reportingsjrlekernel: a transistor can function as a diode if you tie the base and the collector together18:46
lekernelyes18:46
lekernelbut I was just surprised they use such a common part18:46
bart416It sounds logical to use a common part18:46
bart416Easy to do quality testing18:47
bart416And a 2N2222 is pretty damn reliable18:47
reportingsjryep18:50
reportingsjrvery easy to source too18:51
lekernelyes, but it was never meant to be used as a cryogenic thermometer probe18:54
lekernelit's good to know it apparently does a nice job at this18:55
bart416lekernel, name one electronic component that hasn't been abused yet18:57
bart416people have been making point contact transistors out of germanium diodes18:57
bart416Then proceeded to use a 555 timer as audio amplifier18:57
bart416To finish it off with some nice LED as light sensitive receiver18:58
reportingsjrhaha18:58
reportingsjrI wouldn't call it abuse18:58
reportingsjrjust finding different uses18:58
bart416lol18:59
bart416"tactical misplacement of components" xD18:59
lekernelyes, but this 2N2222 probe is (a) done by an apparently very serious company (b) in a domain few electronics engineers are knowledgeable about18:59
bart416lekernel, we have an ancient analog circuit design professor hanging around the lab that does weird things like that...19:01
lekernelif I asked you how to build a cheap cryogenic temperature probe, I guess you'd have a hard time figuring out "just use a 2N2222" :-)19:01
bart416To be honnest I'd use 4, lol19:01
reportingsjrlekernel: I don't see why a large company would not be allowed to use a solution like this19:09
lekernelI'm not saying it's bad or should be banned, I'm saying it's remarkable19:12
bart416In the end it's still an engineer designing it lekernel19:14
bart416And all engineers like quick, easy, reliable and cheap solutions19:15
bart416And if that person happens to have used a 2N2222 as temperature sensor before and he has a stack of them around when making the prototype, then why the hell not :P19:15
berndjbut why not just use a 1N400x?19:24
berndjare they harder to find than 2N2222?19:24
bart416transistors are easier to use as temperature sensors19:24
berndjand since they're using the transistor as a diode anyway...19:25
bart416Check the scale of influence of VT ;)19:25
bart416the Is of a diode and transistor are quite different19:27
berndjhmm, you're still taking advantage of its transistorness19:27
berndjok never mind i said anything :)19:27
bart416np19:28
bart416Had to think about it as well for a minute, lol19:28
lekernelnah, usually, electronics companies avoid using components for unintended purposes, because doing so tends to result in a "let's see if it works with this week's parts" kind of production19:36
bart416Maybe that's one of the reasons they went for the 2N222219:47
bart416It's reliably available19:47

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