#homecmos IRC log for Wednesday, 2011-12-14

horizontallyis there anything smaller than an electron that we can control?00:20
horizontallyneutrinos?00:20
berndjhow big is an electron anyway00:20
horizontallyyou can determine its mass if you have an electromagnet and a power supply... and a vacuum pump and an inert gas00:22
horizontallythen the electron beam can be visualized however you want with the appropriate applied magnetic field00:22
berndjif you're looking for something with which to expose resist material and electrons *aren't* giving you the resolution you need, maybe you need ions then00:22
berndjmore mass == shorter de broglie wavelength at a given energy == better res00:23
horizontallyoh right lambda = h/p00:29
berndjso maybe you want a lead ion beam exposure system :)00:38
berndjbut even 40eV gives you lambda = 2 angstrom, so unless you want to etch quarks, i think electrons are just fine!00:40
cheaterhi01:12
cheateryou guys seen those light slow motion captures?01:12
cheaterwhere they show a movie of a single pulse of laser slowly moving through space (and through objects)?01:12
cheaterhttp://web.media.mit.edu/~raskar/trillionfps/01:18
cheaterwell.. i was wondeirng if something like this would be doable with electrons01:18
cheatercould you photograph electrons travelling in wires just like that?01:18
cheaterin resistors, transistors, and inductors?01:19
cheaterthat would be great to see01:19
cheaterbut also, if you could "photograph" electric charge in cells like this01:19
cheaterand do this kind of experiment with a neuron cell01:19
superkuhNo. Not electrons in wires nor charge movement in nerves/neurons. With respect to neurons specifically the spatial repeatability is not there. It will be different every time and the averaging over multiple short pulses will not work. Also, the transverse movement of charge across a 9nm membrane is very fast, while the longitudinal movement of the action potential down the axon itself is much slower at some hundred meters a secon01:25
superkuhd for mylinated axons, or less.01:25
horizontallysomehow lambda = h/p doesn't jive with intuition01:40
horizontallyisn't something that's more massive going to occupy more space? (i know i know black holes)01:40
horizontallybut the ionic or atomic radius size is roughly proportional to the atomic number01:40
berndjif you put it that way it is a little unintuitive01:50
berndjthen again, great big blobs of quantum mechanics don't mesh with intuition01:51
horizontallyi think a lot of stuff in quantum mechanics does make sense02:22
horizontallyyou have to understand classical wave motion first02:22
horizontallyand there are loads of people who keep reinforcing that stupid mindset that "if you think you know quantum mechanics, you don't"02:22
berndjyeah, talking about QM like it's this mystical thing is also dumb03:09
berndji always thought of higher-mass -> shorter lambda as semi intuitive; the analogy i was using would be "more momentum = more assertive = 'sharper' needle"03:10
berndjcheater, were you thinking of something like http://en.wikipedia.org/wiki/Electron_Beam_Prober ?03:14
berndji don't think you could snap individual electrons of course, but maybe you could image wavefronts03:14
horizontallyhttp://engineering.tufts.edu/microfab/index_files/sop/su-8%2520processing.pdf06:05
horizontallythey say SU-8 gets all over everything06:05
horizontallyand it outgases06:06
horizontallyhmm when exposed to UV it solidifies06:06
horizontallyone week in H2S04 at room temperature for removing 100um layer06:09
bart416<azonenberg> Because i've heard of SU8 being exposed by synchotron x-ray emisions <-- synchrotron is off scale compared to medical though :P07:03
bart416The beam actually ionises everything it touches07:04
azonenbergbart416: lol, true10:17
azonenbergBut i wasnt sure if lower energy level might be enough too10:17
cheaterberndj: yeah, i was thinking of wavefronts.10:20
bart416azonenberg, no clue :)10:21
bart416I'll just leave it at, I'd sit on a reactor before I'd stand within 10m of a synchrotron in operation :)10:21
azonenbergbart416: lol10:21
azonenbergWell, i've stood about two feet from a subcritical research reactor10:22
azonenbergat the US naval academy10:22
azonenbergnice and small, water shielded10:22
azonenbergyou can look right down the tubes at the fuel rods10:22
azonenbergthrough about a foot of water10:22
azonenbergwhole thing was about the size of a refrigerator10:23
cheaterhey azonenberg you seen those laser snap vids i posted?10:25
azonenbergNope10:25
cheateri was wondering if something like this can be done with electrons in conductors/semiconductors10:25
cheater1 sec10:25
cheaterhttp://web.media.mit.edu/~raskar/trillionfps/10:25
cheaterprepare to be amazed10:25
cheater:)10:25
cheatereven though the operation is super simple, the outcome is just mind bending10:26
azonenberglol interesting10:27
cheatercheck out the bottle10:27
cheaterthat one's the best10:27
cheaterif you scroll ALL the way down they have a video presentation10:27
azonenbergLooking at it10:27
cheaterit's after the references funnily enough10:28
azonenbergSo they're doing it like an analog (non-storage) oscilloscope10:30
azonenbergreplaying the event several times over and over and recording different parts of it10:30
bart416If you repeat anything long enough you can get as much samples as you need10:32
azonenbergYeah10:32
soul-di thought they had array of 500 chips   running freely10:32
soul-dtriggert10:33
bart416Only your trigger event has to be exact10:33
soul-d1/trilions10:33
azonenbergbart416: Yep10:33
soul-daftehr eachother10:33
azonenbergIt looks like what they have is a 1D CCD10:33
bart416And triggering at that speed isn't that problematic for electronics10:33
bart4161 ps is easy10:33
azonenbergbart416: maybe for you :p10:33
azonenbergthe stuff i work with has gate delays in the hundreds of ps :p10:33
soul-dthe stuff tehy use to10:34
bart416azonenberg, your gate delay only has to be the same every single time :P10:34
azonenberglol10:35
bart416+ Transistors are actually coming up to 1 THz10:35
bart416the record now is 800 GHz or something like that10:35
azonenbergSo i heard10:35
soul-dalso he say's each pulse looks the same    thats  not really sientific :) end  result is still a aproximation of how 1 cycle could have looked like  but doesn't since it's not one cycle10:37
bart416You know what actually freaks me out, that most of our oscilloscopes and logic analyzers in the lab are actually fast enough to measure the speed of light to a fair precision10:38
bart416(Over a relatively short distance that is)10:38
soul-ddin't some dude do that trick in 1912 ?10:38
bart416soul-d, hook up a pulsed laser, get a half meter long tube, get it to a near vacuum, put a GaAs photodiode at the other end10:39
bart416Measure pulse start at laser and pulse start at diode10:39
bart416You'll have a very nice approximation of the speed of light10:40
soul-dit took a day here :)10:40
soul-dthey cut  the main fiber lines yesterday of my isp ;)10:40
azonenbergbart416: lol10:48
azonenbergyou mean, like having two sensors a foot apart10:48
azonenbergand one gets hit first?10:48
bart416you could also do that10:52
bart416Though a foot isn't long enough in most cases10:52
azonenbergit isnt?10:52
azonenbergmaybe my mental math was wrong10:53
azonenbergbut i recall that having been on the order of 1ns10:53
azonenbergwhich should be detectable10:53
bart416It's detectable yes10:54
bart416Easy in fact10:54
bart416But your error margin will be fairly large10:54
azonenbergRepeat it a few hundred times and average?10:54
azonenbergYou could do that within a millisecond :p10:54
bart416You'll have to take into account the length of your probe wires :P10:54
azonenbergNo different from other high speed logic10:54
azonenbergTreat it as a differential pair10:54
azonenbergyou could length match to submillimeter10:55
bart416You're not taking into account other effects10:55
azonenbergDo tell10:55
bart416by all means, you're working with transmission lines10:56
bart416none identical ones at that10:56
bart416The propagation times will differ10:57
bart416Quite possibly to well within your measurement interval10:57
soul-dcan't you have a beamsplitter and measure at 2 different distances ?10:57
bart416again, pointless increase of complexity10:57
bart416Increase the distance10:57
azonenbergBut doesnt that increase transmission line effects?10:57
bart416Then your error becomes less significant10:57
azonenbergor do you mean reflect it back and forth with mirrors10:57
azonenbergusing a short test lead?10:58
bart416no10:58
bart416What the hell are you guys thinking about? :|10:58
bart416Your transmission line effects will have less impact over a long distance compared to the time it takes the light to travel11:00
bart416We've found that anything less than 50cm your error is large11:02
bart416Ideally you use 1 metre or longer11:02
bart416Your positional offset of your sensors also becomes less problematic11:03
bart416If you're off 1mm on 1m it's not as bad as you're off 1mm on 30cm11:03
cheateractually11:53
cheaterthey don't even need to get the delays right11:53
cheateryou can reconstruct the sequence without knowing what order it was in11:53
cheaterjust as long as the points in the sequence are near enough11:54
cheaterreconstructing the order of a sequence f(x_i) for unknown x_i, with the x_i having an unknown linear order, is one of the easier things in inverse problems11:57
cheateri mean if you have an animation and someone scrambles the frame order you can usually come up with the original order11:59
cheaterbecause it's causal11:59
bart416cheater, depends on the intervals12:01
bart416:P12:01
bart416if they're all near identical...12:01
cheateryea12:23
cheaterwhat i'm saying is that jitter in the timebase doesn't matter12:23
cheateras long as it's normal12:23
cheateror another well behaved distribution12:24
berndj<bart416> + Transistors are actually coming up to 1 THz <-- keep those leads short!13:55
berndjdo you know if that's crazy lab-style devices or something you can order from digikey level of availability?13:56
azonenbergberndj: that's probably lab scale, but you can easily hit tens to hundreds of GHz in something like a 22nm finfet13:57
azonenbergwire delay is killing modern processors i think13:58
azonenbergperformance wise13:58
azonenbergmore so than switching delays13:58
azonenbergyou have to switch AND drive the capacitive load of the destination wire13:58
azonenbergsure, its probably on the order of femtofarads13:58
azonenbergbut thats enouhg13:58
azonenbergwhen you can only supply nanoamps :p13:58
berndji'm always amazed at the max pulsed current specs of small transistors13:59
smedingazonenberg: yeah, wiring is a huge part of performance13:59
smedingazonenberg: this is why 3d integration is very promising tech13:59
berndjmax 100mA continuous, but keep it <1us and they'll do 10A kind of thing13:59
azonenbergsmeding: Yeah13:59
azonenbergTSVs ftw14:00
Action: azonenberg wishes he could do DRIE in his home setup14:00
azonenbergNot happening any time soon :p14:00
lekernelberndj: http://www.teledyne-si.com/schottky_diodes/index.html15:18
lekernelfor the transistors, I don't know. let me know if you find something...15:18
azonenbergberndj: also, even some LEDs are like that15:37
azonenbergrated for tens of mA but can take several amps for extremely brief periods15:37
azonenbergSince the limit is normally on the die / bond wires heating up15:38
azonenbergAnd  if your thermal mass is big enough, you can have a high energy pulse that dissipates very little heat comparatively15:38
bart416azonenberg, actually we can go past 4GHz easily16:10
bart416It's just that we have no way to remove the heat16:11
azonenbergbart416: with CMOS, yes, but there's still a limit16:11
bart416GaAs goes pretty high actually azonenberg16:11
azonenbergAmd tjey16:11
azonenbergand they've hit 8 on CMOS experimentally16:11
azonenbergJust not without crazy stuff like liquid helium :p16:12
bart416highspeed circuits are never with FETs16:13
bart416at least not with regular ones16:13
bart416There are a few tricks you can apply16:14
bart416But most of the time it's bipolar at that point16:14
--- Thu Dec 15 201100:00

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