#homecmos IRC log for Wednesday, 2013-03-20

azonenbergScrew thread of some sort to focus the two cameras to each other00:00
azonenbergtwo scopes*00:00
azonenbergthen one vertical focuser to focus them onto the mask00:00
glowplug17" LCD and a mini ARM computer for the cam outputs would work well.00:00
azonenbergYeah, you dont need a lot of horsepower00:00
glowplugAdjustment is with manual micrometers correct?00:00
glowplugIf thats true then you need zero horsepower.  Haha00:00
azonenbergThe wafer is just on an x-y-theta stage00:01
azonenbergfor mass production you'd need machine vision00:01
azonenbergbut this isnt :p00:01
glowplugWe have scipy for that anyways.  =)00:02
Sync_open cv ftw :D00:02
glowplugThat too.  =)00:02
azonenbergYes, but that would need more CPU is what i mean00:02
glowplugThe developer station would need an upgrade at that point.  Haha00:03
glowplugI have an idea.00:03
glowplugThe VESA table mounts for LCD's00:03
glowplugWe can mount the LCD, the exposer and the optical unit each on a VESA mount.00:04
glowplugThen everything can be locked down to a table without any special fabrication.00:04
glowplugPlace sample. Rough alignment.  Swing the optic unit over. Do your final alignment.  Swing it back.  Swing in LED panel.  Hit blastoff.  Swing it back.  Take sample + enjoy.  =)00:07
azonenbergoh and the rpi would have to also have GPIOs for the LEDs00:08
azonenbergand a timer00:08
azonenbergbut thats trivial00:08
glowplugThe mounts would be clamped to the rear of the desk with the sample clamped down to the front.00:08
glowplugWhat do you think?00:09
Sync_I think it is not time to think about that00:10
Sync_but it seems it is a good idea00:10
glowplugHoly shit sync_ thinks I had a good idea!00:11
glowplugAnd it was the damn stupid VESA mount one.00:11
glowplugTake what I can get...00:11
Sync_actually, I'm not here to diss your ideas ):00:11
glowplugI know.  =)00:11
glowplugYou were totally right about the projection litho.00:11
glowplugThats for superhumans.00:12
glowplugAnd right about the furnace too.  $1,000 is actually not a bad price.00:13
glowplugOne thing I still think should be DIY is the optical encoder.  I'm still going to give it a shot maybe I can save us ~$200.00:13
glowplugDepositing of silicon dioxide is done in the evap?00:20
azonenbergNormaly people do PECVD00:21
azonenbergwe have a couple of opions00:21
Sync_you need to sputter that00:21
azonenbergspin-on glass is one00:21
azonenbergRF sputtering of a glass target is another00:21
glowplugSo thats the magnetron.00:21
azonenbergReactive sputtering of Si with Ar + O2 plasma is another00:21
glowplugAlright just making sure I wasn't missing something important.  Haha00:22
glowplugWow manufacturer yields can typically be below 30%.00:23
glowplugThis is a depressing business.00:23
azonenbergYes, but thats for leading-edge stuff00:23
azonenbergthey can be quite a bit higher for older-gen stuff00:23
azonenbergAnd these are for complex devices00:23
glowplugGod point.00:23
azonenbergIf we can make one NAND2 gate work on a 10um process from scratch i'll be really happy00:24
glowplugWe are coming into the party at circa 1990 tech.00:24
azonenbergYeah, if that00:24
azonenbergmaybe 80s00:24
glowplugHaha.  That is awesome.00:24
azonenbergI just want to make a board one day that has a custom flip-chip ASIC on it00:25
glowplugYou mean this device here? http://en.wikipedia.org/wiki/File:CMOS_NAND.svg00:25
azonenbergeven if it could be replaced by a few 7400 chips00:25
azonenbergAnd yeah00:25
glowplugSo thats four contact pads and four transistors.00:25
glowplugI think we can pull that off.  =)00:25
azonenbergFive contacts - inA, in B, power, ground, out00:26
azonenbergThis is what your layout would look like roughly (actual photo) http://i.imgur.com/su94J06.jpg00:26
glowplugAhh I see.00:26
azonenberggreenish is oxide over undoped Si00:26
azonenbergreddish is active area00:26
azonenbergyellow is poly for gates00:26
glowplugAhah!  That is much better.00:26
azonenbergmetal has been removed already00:26
azonenbergSo you'd have a strip on the top going from Vdd to the left side of the PMOS and from Vdd to the right side of the PMOS00:26
azonenbergthen from Vss to the left side of the NMOS at bottom00:27
azonenbergthen metal from the middle of the PMOS to the right of the NMOS is your output00:27
azonenbergand then the big squares on each piece of poly are where you'd hook up the inputs00:27
azonenbergThe device in question btw is the RSA SecurID00:28
azonenbergits a 1-micron 2-metal process00:28
glowplugNo way they only used 1 micron.  Thats sweet.00:28
azonenbergYes they did00:28
glowplugWe can totally match that.  8)00:28
azonenbergi have full die images at transistor resolution00:28
azonenbergof every layer00:28
glowplugHaha that is epic.00:28
glowplugI haven't searched yet.00:29
glowplugAre there example masks for these devices?00:29
azonenbergFull masks are rarely published00:29
azonenberga few shops like Alliance have open source standard cell libraries00:29
azonenbergwith layouts for single gates you can study00:29
azonenbergoh, and look up the GDS-II file format00:29
azonenbergit's the industry standard for masks00:30
azonenbergso whatever laser system we make will need to read gds files00:30
Sync_or some gds to gcode thing00:30
azonenbergIt may not do so directly00:30
azonenberganother good file format to use natively is gerber00:30
azonenbergyou could probably convert gds to gerber00:31
azonenbergbut if you want to be compatible with IC layout CAD tools we need to eventually be able to read GDS00:31
glowplugLots of interesting stuff in here. http://vlsiarch.ecen.okstate.edu/flows/00:40
glowplugA full license of LayoutEditor is $1500.  Ouch.00:48
azonenbergGlade is free but not open source00:48
azonenbergthat's what i've been using00:48
azonenbergit is almost certainly not good enough for anything huge, it seems to be mostly a viewer00:48
azonenbergbut it does exist00:48
glowplugThat is free.  And works.  And is very old.00:49
azonenbergmagic? Yes00:49
azonenbergits ancient00:49
glowplugSounds like a perfect match for our tech.  Haha00:49
azonenberggood point :P00:49
Sync_magic is also still developed00:50
Sync_I know someone who made quite a large asic with it00:51
glowplugI'm sold (pun intended).  8)00:51
azonenbergSync_: how large? and what process?00:52
azonenbergcan it do extractions too?00:52
azonenbergor just layout00:52
Sync_oh I dunno00:52
azonenbergand does it include P&R and synthesis stuff?00:52
glowplugIt looks like it can do extractions.00:52
azonenbergi mean for initial designs we wont be using hdl00:53
azonenbergit'll be manual drawing00:53
glowplugTimberwolf can be used for place and route.00:53
glowplugVersion 6.3 is GPL00:53
Sync_I guess it's not large for todays standards but it was a video processor00:53
glowplugvideo processor!00:54
Sync_yeah it was not something really simple00:54
glowplugGood enough for me.  Haha00:54
glowplugThis site has everything we need.00:55
glowplugEven some things we wont need for years.00:56
Sync_azonenberg: iirc it was nmos00:57
glowplugOur goal is CMOS devices?00:58
azonenbergCMOS would be nice00:59
azonenbergmems comes first00:59
Sync_I need to get me some wafers from work01:05
glowplugBy the way my name is Adam.  I know Andrew now.  Sync_ ?01:06
B0101I don't think oxides will grow on Silicon wafers that are put in desiccator bags, will they?01:38
azonenbergyou'll still have native oxide01:38
azonenbergit wont be thick, but it'll be there01:38
Sync_iirc 15nm or so at room temp01:42
azonenbergHomebrew CMOS and MEMS foundry design | Wiki: http://homecmos.drawersteak.com/wiki/Main_Page | Repository: http://code.google.com/p/homecmos/ | Logs: http://en.qi-hardware.com/homecmos-logs/01:50
Sync_ah 1.5nm seems to be the equilibrium01:56
Sync_so I had at least the right number!01:56
B0101http://www.instructables.com/id/Simple-DIY-Vacuum-Chamber-and-Pump/ - I wonder if this DIY Vacuum pump is good enough for RIE01:57
azonenbergaspirators are limited to the vapor pressure of water02:02
azonenbergand the chamber gets full of water vapor at that pressure02:02
Sync_also water everywhere02:03
Sync_I actually had a closed loop water aspirator for a certain project02:03
Sync_hm http://www.youtube.com/watch?v=KsBXixItlAI oh god what02:11
glowplugHave you guys seen this?02:22
glowplugI knew about jeri but I haven't seen these videos yet.02:29
glowplugHave we asked if shes interested in helping?02:29
glowplugSpin coating with a CPU fan.  Holy shit.  Haha02:33
glowplugAnd it looks like she has a glass tube in the furnace.02:34
Sync_they actually come like that02:36
glowplugWhich appears to be a grill.02:36
glowplugThats not a commercial furnace is it?02:37
Sync_I guess it is02:38
Sync_haha holy shit her diode is bad02:39
glowplugAll the devices were pretty bad.  But they were made without any litho.02:41
glowplugExtremely impressive.02:41
Sync_yeah well a mosfet is extremely easy to make02:43
Sync_because it just needs a channel02:43
Sync_and metal02:43
B0101hmmm, to make an electrical field, the potential difference between 2 charged plates is inversely proportional to the distance between my plates, am I right?02:48
azonenbergI've talked briefly with jeri02:49
azonenbergShe was surprised that nobody had copied her setup yet02:49
glowplugI'm surprised too.02:49
azonenbergBTW, my guess is that one of the reasons for her trouble was the complete lack of attention paid to things like metal contamination02:49
glowplugWe need to get our asses in gear.02:49
azonenbergi bet her stuff was loaded with sodium02:49
glowplugI'm sure she knows that too.  But the fact is she has a working device.02:50
azonenbergBut an RCA clean isnt that hard to do02:50
glowplugI think she would be interested in collaborating if we can prove we are serious and get a working device as well.02:50
azonenbergWe can definitely try and pick her brain02:51
azonenbergI'll ping her on FB once we get some more serious progress02:51
glowplugSounds like a plan.02:55
azonenbergYeah, geared motors of that level exist03:53
azonenbergbut they're not necewssarily precise03:53
azonenbergi bet the backlash is terrible03:53
glowplugProbably better than my diy geargoxes though.03:53
glowplugDid you have any gearboxes in mind?03:55
glowplugHere are the insides of the Kysan motor.  http://www.flickr.com/photos/hoeken/sets/72157612405194866/04:01
Action: B0101 seriously never thought that sending a 22 GB backup file over a local network would be difficult13:58
glowplugDo you have linux systems or is the transfer CIFS?15:54
Sync_azonenberg: my prof said use quartz, it is the easiest and safest16:53
glowplugA quartz tube for the furnace?18:36
glowplugSome interesting information.  First on the gearing for the x,y stage motor.  http://www.zincland.com/hypocycloid/19:10
glowplugAnd also this motor controller is interesting. http://www.youtube.com/watch?v=S9PyQBrqR4E19:12
glowplugStandard ungeared BLDC motor spinning at 10rpm constant.19:12
glowplugIf you combine the two you get .0016 r/s.  With a .5 pitch leadscrew (standard micrometer) which is .5mm /r or .0008mm per second at the lowest speed.    800nm.  =)19:22
glowplugThis is of course assuming the slowest we can turn a BLDC motor is 10RPM.  I can't find any examples of slower but it can probably be done.19:22
glowplugAt any rate that makes the motor/gear assembly (not including leadscrew, nut, or encoder) about $15.  The raw material for the acrylic cycloid gearbox and the cheap hobby outrunner.19:26
azonenbergSync_: i'd use quartz for the actual wafer carrier21:21
azonenbergthe question is whether alumina insulation on the furnace is dangerous21:21
Sync_yeah I think not21:29
Sync_if the carrier is quartz everything sould be fine21:29
azonenbergI wont even consider trying to clean my existing glassware21:37
azonenbergwhen i move to CMOS i'm buying all new fused quartz21:37
azonenbergand never handling it without gloves21:37
Sync_the mbe chamber shat itself again :/21:39
Sync_at first the thickness monitor died21:39
Sync_and now we have a short in the vacuum21:39
glowplugSo I understand.  The quartz carrier for the furnace is a cylinder correct?22:04
glowplugAhh that is for doing multiple samples at once.  Interesting.22:05
glowplugHow much does that thing cost?22:06
Sync_they should be fairly cheap22:07
glowplugI have been researching closed loop control of steppers and finally realized that the Mesa SoftDMC can do it with quadrature encoder input.22:12
glowplugAnd it can do two at the same time.22:12
glowplugI was totally wrong about the parallel port latency being too high for th encoders also.  It is a severe bottleneck for high rpm operation but at low rpm's it is completely acceptable.23:05
--- Thu Mar 21 201300:00

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