#homecmos IRC log for Friday, 2012-02-17

ashawazonenberg, are you there?01:00
azonenbergashaw: back01:15
azonenbergWas on campus01:15
ashawno probs.01:15
azonenbergWhats up?01:16
ashawI am thinking about using LDI to image the surface of an IC.01:19
azonenbergAs in you want to do laser direct-write lithography for making an IC?01:19
azonenbergVery feasible as long as you can align it, which is tricky01:19
azonenbergI'd suggest doing it for mask fab and then contact alignment of the mask01:19
ashawI have been thinking, why not use a UV laser and have a Blu-ray like controll01:21
azonenbergWell, the hard part is how to tell where you're pointing the laser01:22
azonenbergbefore firing it and exposing whatever is under it01:22
ashawnot really01:22
azonenbergyou suggest just doing a low-power pulse in a corner?01:23
azonenbergIn any case, i think it's very doable01:23
azonenbergi seriously planned to do the exact same thing myself01:23
azonenbergjust been too busy to make it happen01:23
ashawno - I suggest using an electron beam to put tracking marks on the edge and then plating them with a metal01:23
azonenbergIf you have e-beam01:24
azonenbergwhy not use ebeam for everything?01:24
ashawI suppose01:24
azonenbergSince ebeam is pretty easy to use for low-power imaging and high-power litho01:24
azonenbergTurning a SEM into an e-beam direct write system is a solved problem01:24
azonenbergseveral companies sell conversion kits01:24
azonenbergthey can even still image01:24
ashawcan you use a STM to image?01:26
azonenbergYes, but i dont know if you can make one do lithography01:27
azonenbergand scanning probe techniques are much slower than scanning-beam systems like SEM01:27
ashawI know.01:30
azonenbergI mean, there are so many tools i eventually want to have in my lab lol01:30
ashawas I01:30
azonenbergits all a matter of deciding which ones to build/buy first01:30
azonenbergNow that i have Ethernet working on PIC32s i can make networked lab equipment01:30
azonenbergI'm planning to build an ethernet-controllable spin coater in the fairly near future01:31
azonenberggotta pick out a good motor and do some CAD work01:31
ashawhow about using a single pulse on one corner to create the beam01:31
azonenbergTo do alighnment you'd need at least two marks01:31
ashawsorry the tracking generator01:31
azonenbergyou need to know scale, theta, and position01:31
ashawI would not01:31
azonenbergno way you can get that without at least two01:32
ashawI just need one01:32
ashawand then rotate it01:32
azonenbergnot sure i understand what you're planning01:32
azonenbergin any case, the easiest technique IMO01:32
azonenbergis to make the laser direct write be unaligned01:32
azonenbergjust write onto a mask blank01:32
azonenbergdevelop and etch the mask (probably sputtered metal on a microscope slide)01:33
azonenbergThen use the resulting chrome-on-glass mask in a contact aligner, which would be pretty easy to build01:33
ashawhowever, I still need scale01:35
azonenbergI'd do a single initial calibration01:37
azonenbergand then just assume it's accurate01:37
azonenbergrecalibrate periodically01:37
azonenbergas in, write two lines X distance apart01:37
azonenbergwhere X is nominally, say, 100um01:38
ashawSo what do we know about the disk?01:38
azonenbergthen take a photo with a calibrated camera and measure01:38
ashawwe know it's rotation velocity01:38
azonenbergBluray disks are polar coordinates01:38
azonenbergand have tracking metadata already etched into them01:38
azonenbergthis will be cartesian01:38
azonenbergon a blank substrate01:38
ashawno, I will write in polar01:38
azonenbergMine will be cartesian01:38
azonenbergi envision reusing the diode but otherwise doing totally my own structure01:38
ashawWhat type?01:40
ashawI too will only use the diode, I do not think the structure can be controlled well enough01:42
ashawMy intent is to use a 2in wafer, spinning at 10k rpm01:43
ashawand image using a set of fibe beams with a high NA01:44
ashawlenses on the end.01:44
ashawspaced on the axis of rotation 100 um apart01:44
kanzurepatron saint of lithographers?01:47
ashawhey what?01:47
kanzurejust reading the backlog01:47
ashawwho is?01:48
kanzure/kick ashaw01:48
azonenbergSo let me see if i understand your plan01:49
azonenberg10k rpm is pretty high, for starters01:49
ashawnot really.01:49
azonenbergSo you want to have a single diode going into a set of optical fibers?01:49
ashawno, multiple diodes01:49
azonenbergThat's a LOT of power01:49
kanzurein molecular biology labs you regularly get up to >50k rpm01:50
azonenbergfor a silicon wafer, i mean01:50
azonenbergmost spin coaters run at like 3-5k01:50
azonenbergBut for a 2" it should be fine01:50
azonenbergi'd be cautious about an 8" or larger that fast01:50
azonenberganyway so ONE bluray diode is overpowered for litho01:50
ashawI mainly want a high linear velocity01:50
azonenbergthey're meant to actually heat and burn the metal on the disk01:50
ashawI will only run it a very low power01:51
azonenbergYou'll have to01:51
ashawit is more for control01:51
azonenbergotherwise you'll overexpose hugely and heat it01:51
ashawI want to controll each laser seperately01:51
azonenbergMy thinking was to have a single diode01:51
azonenbergx/y mirror for fine alignment01:51
azonenberg(within one die)01:51
azonenbergthen a mechanical stage for die-to-die stepping01:51
ashawx/y mirror? how are you moving it?01:52
azonenbergOpen question, probably light show galvos01:52
azonenbergthen feed the resulting pattern through a reducing lens01:52
azonenbergI'm far from sure of what i'm doing here01:52
fennlight show galvos have terrible spatial resolution01:54
azonenbergfenn: 100um would be enough01:54
azonenbergif i was able to then reduce optically01:54
ashawyou cannot.01:54
ashawyou have a problem01:55
azonenberg In any case, i had some other ideas01:55
azonenbergone was to bounce the beam off a DLP micromirror array01:55
ashawYou need a high numerical aperature01:55
ashawThat was my first Idea01:55
fenni think you want to move the diode image around with flying optics, if you're going the continuous (vector drawing so to speak) way01:55
azonenbergfenn: Again, open question01:55
fennDLP's are great, A+01:55
azonenbergthis is on the tools-to-build-eventually list01:55
azonenbergnot on the tools-under-current-development list01:56
fennalso you can fake it with LCD's if you don't mind waiting a long time01:56
ashawWhat tools are you building currently01:56
azonenbergI just crossed "fume hood" off my list01:56
azonenberg"proper spin coater" is next up01:56
azonenbergas in, one that isnt made of 2x4s and filling my entire hood01:56
azonenbergthis one will be much smaller, mostly metal construction01:56
azonenbergand have proper feedback-loop speed control01:56
fennuse a juiceman jr.01:56
azonenbergi want it controllable over ethernet01:57
azonenbergthe plan is for all of my tools to be networked01:57
azonenbergso i can download process recipes from my laptop onto it01:57
azonenbergand then just click "run"01:57
fenni had the same dream... once01:57
azonenbergI already have a PIC32 receiving UDP packets and decoding them01:57
kanzurehey wait.01:58
kanzureazonenberg: http://gnusha.org/skdb/01:58
azonenbergand i know how to generate PWM for motor driving01:58
fennPWM is easy compared to decoding ethernet01:59
azonenbergfenn: http://pastebin.com/fbWrjrdA01:59
azonenbergthat was a day or two ago01:59
fennmost microcontrollers have PWM hardware built in01:59
azonenbergThe "read ethernet" is a solved problem01:59
azonenbergi just have to put that and a PWM controller on one board01:59
azonenbergadd some kind of optical encoder for speed feedback02:00
azonenbergand do mechanical design for the chuck etc02:00
kanzureno really, check the link02:00
azonenbergi did02:00
ashawso after the spin coater, perhaps a CMP machine?02:01
kanzureok my work here is done02:01
azonenbergashaw: contact mask aligner is likely to come first02:01
Syncclosed loop speed control seems to be overkill, you don't have the load changing around alot02:01
azonenbergSync: but i need precise ramp rate contorl02:01
azonenbergi need constant accel/decel02:02
azonenbergand without knowing the mass of the load thats hard unless you have closed loop feedback02:02
ashaware we going to work on whole wafers here.\02:02
azonenbergin any case a CMP system would be a pretty simple mod of the spin coater02:02
azonenbergashaw: I'd like to eventually but for now most of my work is done at the die level02:03
Syncwell, you know the mass pretty well02:03
azonenbergsince wafers are pricey02:03
azonenbergashaw: so contact mask aligner is high on the todo list02:03
kanzurewhy not dlp02:03
ashawhow much are wafers?02:03
azonenbergin the longer term a stepper02:03
ashawThe problem kanzure is the optics.02:04
azonenbergashaw: on the order of $25-50 depending on size, one vs two side polish, etc02:04
ashawhow much is unpolished?02:04
azonenbergDirt cheap, i got an entire boat of 25x solar grade unpolished from ebay for $5002:04
ashawas wafers?02:05
ashawwhat size02:05
azonenberg4 inch N type02:05
ashawdefect level?02:05
azonenbergthats the spec02:05
azonenbergI bought a cassette from that seller02:06
azonenbergi've used two or three so far02:06
azonenbergas mechanical dummies for testing spin coating02:06
azonenbergas "cookie sheets" for heating single dies in the oven02:06
ashawWhy not make a CMP machine as a top priority then02:06
azonenbergBecause i usually work with single dies, not full wafers, so cost of silicon is minimal02:07
azonenbergthat is enough for like a hundred experiments02:07
azonenbergmost of my dies are 2-3mm square02:07
azonenbergIf i were to do CMP, it'd be for a different reason02:07
azonenbergnamely, copper damascene02:07
ashawI think CMP is going to really make things easier02:08
azonenbergashaw: it's a valuable process to develop, i dont deny that02:09
azonenbergI just dont think its top priority02:09
azonenbergIt depends, in part, on what you're trying to do02:11
azonenbergfor starters02:11
azonenbergCMOS or MEMS?02:11
azonenbergMy current process is targeting MEMS02:11
ashawI think that with a Doubled BD-R laser and spin imaging we could get a 180nm process02:11
azonenbergashaw: I'm very interested lol02:12
ashawI am trying to target CMOS02:12
azonenbergMy current roadmap goes out to about 750nm02:12
azonenberg20um is the smallest i've done reliably and i've hit 5um experimentally but not with good yield02:12
ashawI want to do it with 10k in equipmemnt02:12
azonenbergWell, I like it lol02:12
azonenbergThe first thing is particle contamination02:13
azonenbergat 180nm its nontrivial02:13
ashawwhat is the process you are having the most problems with02:13
azonenbergSo far, it's simply that i cannot make a mask small enough02:13
ashawYeah we will need a whole clean environment.02:13
azonenbergi can reduce my 200um mask by 10x = 20um features reliable02:13
azonenbergand by 40x = 5um but my 40x objective has a defect on it02:13
azonenbergso i cant get good results02:13
kanzurewhat is your current reducing optics like?02:13
azonenbergMicroscope objective02:14
azonenbergthe illuminator, though, is a halogen lamp02:14
azonenbergas a result the FOV is small and it needs a really long exposure02:14
azonenbergBoth of those could be fixed if i built a LDI system for making masks02:14
azonenbergthen used high intensity UV tubes for a contact exposure02:14
azonenbergWhich is why once i get the spin coater built, LDI and contact aligner are next on the TODO02:14
azonenbergfume hood was first as photoresist fumes can be a little nasty02:15
azonenbergi have a spin coater but its so big it barely fits in the hood02:15
azonenbergand obstructs airflow02:15
azonenbergso i get vapors leaking out02:15
azonenbergthe new one will be a lot smaller02:15
ashawwell for mask creation I cannot see a better technique than using a BD-R on a sputter coating of tungsten02:26
ashawand then ablating the tungsten02:27
azonenbergwhy tungsten in particular?02:27
azonenbergand why ablation vs lithography (which needs a lower power dose and gives cleaner edges)02:27
ashawnothing, actually any material02:27
ashawno reason at all actually02:28
ashawHow long are you prepared to wait for a mask02:30
azonenbergA while, even overnight would be fine if the equipment wqasnt expensive to run02:30
azonenbergBut vibration would be a concern at those feature sizes02:31
azonenbergyou'd need to find some way of damping it02:31
ashawwhere are the vibrations coming from?02:33
azonenbergUm, everwhere02:33
azonenbergpeople walking by02:33
azonenbergcars outside02:33
azonenbergHVAC system02:33
ashawon what time scale?02:33
azonenbergseveral Hz02:33
ashawabove 100k?02:33
azonenbergno, single Hz02:33
azonenberglike, maybe 0.1 - 1002:33
azonenbergeven at 400x with a microscope you can see the image shaking if its not on a realy stable mount02:34
ashawthen just compensate the lithography02:34
azonenbergand at 1000x when people walk by it's nearly impossible to see anything02:34
ashawWhy not suspend the entire aparatus on an air bearing02:36
ashawand use a wireless adaptor02:37
azonenbergAgain, there are a lot of possibilities02:37
ashawfor the signal02:37
azonenbergin fact, many commercial anti-vibration tables are based on air bearings02:37
azonenbergthe SEM at my school's cleanroom uses one for example02:37
azonenbergBut it adds to the cost and development complexity02:37
azonenbergNot saying you cant do it, i'm all for it02:37
azonenbergBut i also want to point out potential problems in advance02:37
ashawI like that.02:38
ashawallways important02:38
ashawwe also need an ultra high purity water generator02:40
azonenbergAgain i'm doing MEMS, not CMOS, so i've been fine with regular distilled water from a pharmacy02:40
azonenbergdidnt need to do DI on site02:40
cheater_why are air bearings superior to magnetic bearings?11:10
cheater_you'd think a couple of electromagnets would be easy enough to cobble together11:10
Syncyou do not need advanced control circuits11:12
ashawand there is no possible stable magnetic bearing in a still system11:22
ashawso the bearing is by nature a dynamic system11:22
Syncthe amd diffusion plant in dresden did some major effort to get the vibration out of their building11:23
ashawit is really important11:23
ashawas at that level micro-vibrations are huge11:23
SyncI don't work with that fine structures at the moment11:24
ashawnot YET11:26
ashawwe'll get there11:26
ashawso talking about the equipment problem.11:26
Syncno, I won't, I don't need to :P11:26
SyncI manufacture fets in my sidejob at the university11:27
Syncthey won't get smaller because they will not work as good then11:27
ashawyou will not use smaller microstructures on your fet11:27
Syncokay "make" is a bit overstating it11:27
ashawto reduce capacitance and such11:28
SyncI yell at the ion implanter and hope it does what I want11:28
Syncthis is not an issue11:28
Syncwe develop ion sensitive fets11:28
Syncfor measuring pH values11:28
ashawoh cool11:29
ashawwhich uni might I ask?11:29
Synchannover, germany11:29
ashawoh cool11:30
ashaware you a stdent or a resercher?//11:30
SyncI'm a student11:30
ashawBachelor, Masters or PhD?11:31
Syncfirst semester bachelor ;)11:31
ashawhehe, Just finished my 1st year11:51
--- Sat Feb 18 201200:00

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