#homecmos IRC log for Tuesday, 2013-03-19

glowplug`Good afternoon all.17:07
Sync_glowplug`: thought about the direct litho for a bit17:17
Sync_I still think an interferometer is the cheapest and best way to do the position measurement17:18
Sync_the resolution/cost factor is very good17:18
glowplug`I'm working on the interferometer.  Came to that conclusion as well.17:19
glowplug`Will be constructing a simple michealson and a bunch of micrometers to position the sample.  What was your conclusion with the direct litho?17:20
Sync_there are 100nm resolution glass scales, which would be cool17:20
Sync_yeah a michelson is easy to do17:20
glowplug`It would be, but I can probably get 250nm accuracy with the cheap interferometer for pennies.17:20
Sync_I should head to the scrapyard and see what machines they currently rip apart17:21
glowplug`What are your opinions with the direct litho vs mask17:21
Sync_direct litho is probably fine for experimental work17:22
glowplug`At most universities from what I understand they use masks and spinning each layer with resist.17:22
Sync_it's cheaper if you have the equipment17:22
glowplug`I think that will be fairly easily to accomplish.17:22
Sync_and for most experiments you only use a single mask set17:22
Sync_hmm actually, reading the paper again, I almost have everything I'd need17:23
Sync_I'd only have to get some mirrors for the interferometers17:23
Sync_and somehow build an xy stage17:23
glowplug`If you give me a month I can send you an interferometer.17:24
glowplug`I know thats quite awhile.17:24
glowplug`Have a lot of personal crap going on right now thats keeping me from making progress.17:24
glowplug`Do you have micrometers?  Spinny doohickie? (technical term).17:25
Sync_well I have a few for measuring stuff17:25
Sync_I have a few laser xy stages17:25
glowplug`We need a table with x y tilt and pan adjustments with micrometers.17:26
glowplug`I can build one.. but I need time.17:26
Sync_I'd just need to order stuff from edmund or thorlabs17:26
glowplug`That will cost you a leg + arm.17:26
Sync_well mirrors are cheap17:26
glowplug`I can build the table and the interferometer for ~$50 total.  Depending on how much strugle I have putting together the tiny gearboxes.17:27
glowplug`I can use my CNC to cut 50/50 glass for 30 cents per 10mm square.  Its a pain in the ass to cut glass but I can do it.17:27
glowplug`Add cute little acrylic stands and a dollor store diode you have the worlds cheapest interferometer.  Haha17:28
glowplug`There are a few more things I haven't worked out17:29
Sync_let me just look what thorlabs has for me17:30
glowplug`Alignment markers.  Is a cheap 100x microscope and a point and shoot camera sufficient?17:30
Sync_oh, that's an interesting problem17:31
Sync_but yeah probably17:31
glowplug`The "actual" 100x microscopes that are ~$25 are probably unsuitable.17:31
glowplug`We need to find the actual optics.  Seperately.17:32
glowplug`Then rig the point and shoot to the optics.  This is pretty common stuff for telescope builder types.  Do you know anyone that does that stuf?17:32
Sync_well the actual optics sould be a microscope lens17:33
glowplug`Right.  I've never purchased anything like that.17:33
Sync_one could just make a setup that you put on the lens to aling17:33
Sync_and then take off17:33
glowplug`But if I know its dimensions I should be able to get a ghetto mount for it that we could slap the point and shoot on.17:33
glowplug`Then we have a cute little screen to align our markers.17:34
glowplug`So the next.  It's basically unaffordable to build a cartesian machine with 1um repeatable accuracy.  It just wont happen.17:34
Sync_hm, thorlabs has vis beamsplitter cubes at 135¬17:34
glowplug`I can build a beamsplitter for 30 cents.17:35
glowplug`Thorlabs can jump off a cliff.  Haha17:35
Sync_or budget plates for 2517:35
Sync_well I'm used to order from them17:35
glowplug`Like I said if you can wait a month I can send you the stuff for free.17:35
glowplug`Also there are beamsplitters in cdrom drives which is a low cost option that doesnt require waiting for me to setup my CNC for glass.17:36
glowplug`You can probably find a pile of cdroms for free or <$517:36
glowplug`So the CNC thing.  I'm not confident I can do it.  So I've been trying to find an alternative.17:37
soul-dindeed telescope work  bit diferent but you might get some intresting tools17:38
glowplug`I started thinking about laser projectors.  Galvos aren't really suitable for exposing resist.  They are super fast and not accurate enough.17:38
soul-dgoogle for ATM  in full Amature telescope making17:39
soul-di assume theory behind the light stuff is same just have to do different calculations17:39
glowplug`For the galvos?17:40
glowplug`By the way did you just change names?  Haha17:40
Sync_yeah beamsplitters are not too bad commercially17:40
soul-dno just another someone :P17:41
glowplug`At any rate.  Imagine a laser projector configuration.  Two motors and two mirrors.  The motors are microstepping servos with a 500/1 gearbox reduction.17:41
glowplug`Oh I see.  Haha17:41
glowplug`So basically its the worlds worste laser projector that can draw a blue dot on a wall very very slowly and not make pictures at all.17:41
glowplug`But it can expose photoresist with 1 micron repeat accuracy.  =)17:42
glowplug`I'm sorry. Microstepping steppers.17:42
azonenbergglowplug`: i considered that17:43
azonenbergI think cartesian mounting is a better option17:43
azonenbergeven 10um accuracy is good enough17:43
azonenbergBear in mind you only need 1cm or so of travel17:43
azonenbergdon't even think about that kind of accuracy across a large field17:44
Sync_it's not that hard over a large area17:44
glowplug`This thing might have a large field but it wouldn't be an extra cost if it did.17:44
glowplug`Cartesian adds all sorts of costs that the projector configuration avoids.17:44
glowplug`With no rotational to linear motion 90% of the cost is removed.17:44
glowplug`No ball screws, ball nuts, no linear encoder ect.17:45
azonenbergMaskless litho  is great for making masks, but difficult to align to an existing pattern17:45
azonenbergI still think your best bet is to use direct write for making masks, then a contact aligner17:45
glowplug`That is the plan for sure.  Masks make everything very easy.17:46
azonenbergWith a contact aligner you just have a little track above the mask assembly17:46
azonenbergmicroscope slides in, tweak the micrometer knobs until the mask is lined up with the wafer17:46
glowplug`At any rate.  Let me attempt the projector configuration with 10um as an accuracy target.  If it works I will make and send out masks.  If I fail then I waste a bunch of time.  Haha17:46
azonenbergthen slide it out and bring the UV lamp in17:46
glowplug`Except we are going to have an awesome color LCD for alignment because thats what badasses do.17:47
Sync_hmm, the new blue lasers sure made stuff easier17:47
Sync_no need to drag the ALC909 out17:47
glowplug`Yeah I'm just going to grab a blu-ray diode and call it.17:47
glowplug`I even get a free beamsplitter from the drive.17:47
Sync_I like my tankless water heater with operation light17:48
glowplug`So does anyone have a plan for the spinny thingy?17:48
Sync_after I finish my exams I'll try to come up with something17:48
glowplug`Alright we are counting on you.  No pressure.  =P17:48
Sync_other option is just to buy a spincoater17:49
Sync_they are not too terribly expensive17:49
glowplug`Oh and if I fail with the projector configuration then I will be building a SCARA robot to try and replace it.  If that fails then I don't know.  Trust me it wont fail.  =)17:49
glowplug`The way I determine if I DIY something is if the manufacturer of that thing doesn't have prices on their main page.17:50
glowplug`I think spincoater meets that requirement.  Haha17:51
glowplug`It looks like the "bio" people do spincoaters.17:51
azonenbergBuilding a spin coater isnt that har17:51
azonenbergi've homebrewed one already but it's too bulky17:51
azonenbergi'm going to try and shrink it using a BLDC from an RC airplane17:51
azonenbergwill post plans once i finish17:52
glowplug`What RPM does the coater run at and what accuracy +/- do you need?17:52
azonenbergWhat RPM, no clue17:52
Sync_haha that's interesting17:52
azonenbergaccuracy i've got, terrible17:52
azonenbergaccuracy i'd like, better17:52
azonenbergMy current coater is a flat plate on the end of an electric drill :p17:53
glowplug`Interesting.  Thats more than I can spend though.  Haha17:53
Sync_well it's clearly not worth that17:53
glowplug`There is an open source BLDC controller now.  That will give you the option to use an optical encoder and get extremely precise RPM control.17:53
Sync_it's just a bldc17:53
glowplug`If you need ~1% then an ESC won't work.17:53
Sync_there are enough opensource bldc controllers since a few years17:54
glowplug`They are all piece of shit dead ends though.17:54
glowplug`The only one thats worth a damn, with decent code.  Actually let me get the link onesec.17:54
glowplug`You can get an MSP430 board for ~$5.  Then you would have to add the sensored control but that wouldn't be too bad.17:57
glowplug`It honestly might be good enough without sensored.  Certainly better than a chinese ESC.17:58
Sync_that's nothing special17:59
glowplug`These guys have good MSP430 boards.  https://www.olimex.com/Products/MSP430/Header/MSP430-HG2231/17:59
Sync_designs like that were around for some time17:59
glowplug`No it's not special other than the low cost of the MSP430.17:59
glowplug`And the code isn't.... a giant mess.17:59
glowplug`The interferometer is for aligning the sample correct?  If we have a microscope/camera do we need it?18:00
Sync_no the interferometer is for position feedback18:01
glowplug`During which stage?18:02
Sync_you do not need if if you use a projector18:02
glowplug`Developing the mask?18:02
Sync_but I'm not sure if that works18:02
Sync_I'd try a scanner18:02
azonenbergWho wants accounts?18:02
glowplug`It is a scanner.  Sort of.  It slowly scans a beam in the x,y.18:02
azonenbergFor spam reasons i'm limiting editors to confirmed accounts18:02
glowplug`You did it!  Awesome!18:03
glowplug`Account glowplug.  =)18:03
glowplug`for email you can use raincloudcomputers@lavabit.com18:03
glowplug`The principle is that although it scans slowly compared to a laser projector.  It will fully develop the resist along its path.  And the path is any position on the x,y plane.18:04
glowplug`Honestly if it fails it's only because I designed or built it poorly.  In principle it will work.18:05
Sync_azonenberg: that does not resolve for me18:05
glowplug`I'm on the page.  =)18:05
azonenbergSync_: i just made the subdomain like 15 minuts ago18:06
azonenbergdns might not have propagated yet18:06
Sync_yeah probably18:06
glowplug`I have to go for a bit.  See if you guys can hunt down and stalk a telescope maker type.  I will try also.18:06
soul-deven my super duper fiber con can't find it yet18:07
glowplug`I'm so lucky.  Haha18:07
glowplug`Back in a bit.18:07
soul-dif i had the money  i would have made a 1meter  scope already18:07
soul-dor few small ones for that matter18:08
soul-dstill need guider for my 10"18:08
soul-dwell plus im lazy to gather all the stuff i think it had to do with ability to damp up the aluminum layer18:09
soul-dor mess about with silver stuff but you  need to redo that like every few years18:10
Sync_if you do not protect the al it will also go dull18:11
Sync_you can just send the mirror away to a coater and get it coated18:11
Sync_or if you want get a wideband dielectric coating18:13
soul-dknowing netherlands /europe  i better can buy the mirror if there is even company able to do it  i know there was a18:13
soul-dplace but still size limited18:13
Sync_a lot of companies do that in europe18:14
soul-dhavent looked at it since like 8 years  it's that glowplugh mentioned it18:14
Sync_you could also do it yourself18:17
Sync_sputtering Al is quite easy18:17
azonenbergi've heard Al in particular needs very high output powers18:17
azonenbergdue to the need to break through the oxide18:17
azonenbergcompared to other materials that have less tough oxides18:17
Sync_yeah you need assloads of power but when you are evaporating it is not that special18:17
azonenbergEvaporating vs sputtering is a big difference :p18:18
Sync_true, but even sputtering is ok18:18
Sync_I did not have an issue in my jar18:18
azonenbergi want to build myself a DC sputtering rig18:18
azonenbergHow hard do you think that'd be?18:18
azonenbergI have a mechanical roughing pump18:18
Sync_not very hard18:18
azonenbergbut no other vacuum hardware atm18:18
Sync_I did it in a jam jar18:18
Sync_that's enough if it can get low enough18:19
azonenbergi think i'd get something a little nicer than that18:19
Sync_no shit, but lazyness :D18:19
Sync_now that I have the machines I'll build something more adequate18:19
azonenbergSomething like http://www.ebay.com/itm/Bell-Jar-and-Vacuum-Plate-Combo-1-2-gal-/281065244774?pt=LH_DefaultDomain_0&hash=item4170cb486618:19
azonenbergcombined with a homebrewed baseplate etc18:19
azonenbergthen i'd need a gas inlet, HV supply, etc18:20
Sync_that's a little small18:20
azonenbergThat was just an example18:20
azonenbergbut bear in mind that i only need to fit a 2" wafer inside at most18:20
Sync_yeah I don't really like bell jars18:20
azonenbergThe option i was thinking of actually was a cylindrical glass tube18:20
azonenbergwith flanges and two metal plates18:20
Sync_well the problem is more or less that you need to have proper clearances for the heat18:21
azonenbergthat way i could have the HV coming in from the top18:21
Sync_even with a cooling loop it is an issue18:21
azonenbergCould you do it conductively?18:21
Sync_also glass is annoying to get clean18:21
azonenbergNot it's not18:21
azonenbergeffasol is what they use in the lab at school in the SEM stuff18:22
azonenbergsmear it on the inside of the jar18:22
Sync_not getting it vacuum clean18:22
azonenbergthen just apply water and it dissolves, taking the film with it18:22
Sync_that's easy enough18:22
Sync_but getting your metals off it once they are there18:22
azonenbergthat's what that stuff is for18:22
Sync_I just don't like it18:22
azonenbergit forms a water-soluble layer on the inside of the glass18:22
azonenbergthen it lifts right off18:23
Sync_I'd probably just RF etch18:24
Sync_100% oxygen plasmas are nicely reactive18:24
azonenbergI want that in the longer term18:24
Sync_I have a xds10 scroll here18:24
azonenbergit would be so cool to be able to do SF6 plasma RIE18:24
Sync_and two replacement scrolls18:24
Sync_I'll try and see if I can get a recipient from work18:26
Sync_wee that thing is tiny18:27
azonenbergYeah, and $10K18:27
azonenbergi feel like you could build it for a lot less18:27
Sync_yeah it is not that hard18:27
azonenbergI'd want to wait until i got a "real" lab set up first though18:28
Sync_I wonder why such small plasma etchers are so spendy18:28
azonenbergBecause they can18:28
azonenbergthey sell to people with very deep pockets18:28
Sync_you can get large evap setups for 15k used18:28
Sync_I mean yeah they sure do but 10k18:29
azonenbergwell if i could get SF6 RIE18:29
azonenbergi could do nice cutouts in Si18:30
azonenbergthen sputter Ta + Cu18:30
azonenbergboom copper damascene :D18:30
Sync_meh no sensor head18:31
Sync_interestingly nobody either sells the sensor with the controller or nobody sells the controller with the sensor18:31
Sync_most also only sell the controller withot the exciter18:32
Sync_I wonder if I could make my own EIES sensor18:33
Sync_if it had the sensor I'd buy it right away :/18:34
azonenbergglowplug: you get my email?18:34
glowplugI did but mu is giving me sass.18:37
glowplugMu handles my email.  Not sure what it's problem is right now.18:38
Sync_SF6 RIE has the annoying fluorine in the exhause issue iirc18:39
azonenbergSync_: Yes, i'd venti t outside :p18:39
azonenbergvent it*18:39
glowplugI've got my login though.  =)18:39
azonenbergglowplug: the server cannot send email if you're trying to do a password reset or something18:39
azonenbergi havent configured that18:39
glowplugI'll just use the default password for now.18:40
azonenbergOnce you log in you can change it18:40
azonenbergyou just cant ask the server to email you a new one18:40
glowplugThe problem I was having is that offlineimap pulled your email but mu didn't get it.  Super annoying.18:40
azonenbergoh lol18:40
glowplugHardware guy learning emacs.  Bumpy road.18:41
glowplugAlright.  Under Litho.  "Sample alignment optics"18:42
glowplugSupplies.  I'll just do some basic things.18:43
Action: Sync_ rages a little18:43
glowplugWhats the rage?  Haha18:43
azonenbergTry to copy over whatever you can from the google code wiki18:43
azonenbergi plan to remove most of that wiki so we don't have stuff in two places18:43
glowplugAlright will do.18:44
glowplugEverything under "supplies" are things considered "consumables" correct?18:46
glowplugI will be dealing exclusively with things in the Litho category.18:47
glowplugI have absolutely zero experience with everything else.18:48
Sync_http://www.youtube.com/watch?v=TV5p36-c3R4 this hurts my vacuum self18:48
glowplugThis is what I mentioned earlier.18:49
glowplugDo we need a magnetron?18:50
Sync_no but magnetron sputtering is nice18:50
glowplugThat is stupidly cool...18:50
glowplugWait.  Sputter copper metal onto glass.18:51
glowplugTime to read the wiki....18:51
glowplugWhat kind of materials can be deposited via sputtering in a DIY configuration?18:54
glowplugThis is a HIPIMS system?18:55
azonenbergWith DC sputtering, any metal pretty much18:57
azonenbergwith RF, insulators too18:57
Sync_there finally is cheap glass pipe on ebay18:58
glowplugBut we still need a spincoater for the photoresist?18:58
azonenbergand probably spin-on dopants18:59
azonenbergunless you want to get into ion implantation18:59
glowplugSo many tools....18:59
glowplugIt sounds like a DIY HIPIMS and DIY spincoater is the way to go.19:00
glowplugWhat is the glass pipe for?19:00
Sync_temporary vacuum chamber until I get motivated enough to build something proper out of stainless19:07
glowplugThere is actually a video of my idea.  http://www.youtube.com/watch?v=VjesdQa3mMU19:14
glowplugHe is using galvos though which is why it's so wobbly.19:14
Sync_I wonder if that will lead to issues due to the angular beam direction19:15
Action: Sync_ browses his stash of vacuum flanges19:18
glowplugI thought about that.  There exists optics that will take the beam from any angle and deflect downwards.19:18
glowplugIt also might not be an issue with a very small work area.  Lets say 50x50mm.19:20
glowplugIs 50mw sufficient to expose the resist?19:20
Sync_laser with a maximum power of 3 W at 514 nm and 300 mW at 457 nm. This high power19:21
Sync_output can be used for a second harmonic generation, while typical beam power required for19:21
Sync_the exposure of color-center masks [1] is of the order 100 mW. An electro-optical modulator19:21
glowplugLooks like I might as well use blu-ray then which is way over those power levels.  But theres no cost difference.19:23
glowplugI'm fine with that.  =)19:23
Sync_I'd still like to use my tankless water heater19:24
glowplugFor which process?19:25
glowplugUnder Metalization does "sputter coater" refer to the HIPIMS device?19:26
Sync_haha no you do not get the joke19:26
Sync_I have a huge argon laser19:26
Sync_which basically just is a water heater...19:27
glowplugOH Haha19:27
Sync_well at least it tells you when it is on19:27
glowplugSpeaking of tube lasers.  I need to figure out how to frequency stabalize shitty lasers so I don't need one.  Think I'm going to attempt it with an FPGA.19:28
Sync_what do you want to do?19:28
glowplugFor the interferometer.19:28
Sync_use henes19:28
Sync_they are cheap enough on ebay19:29
Sync_an fpga is not needed, you just need to control the temperature and the current precisely19:29
glowplugA frequency stabalized dollar store laser is 10% the cost though.  =)19:29
glowplugYou need many discrete configurations though correct?  Such as amplifiers, comparaters ect.19:31
glowplugThat could all be integrated into a ~$10 coolrunner type CLPD19:31
azonenbergglowplug: if you can focus it sufficiently a bluray diode is going to be much cheaper19:31
azonenbergJust use short pulses19:31
azonenbergwhat you want to avoid is using so much power that you ablate the resist :p19:32
glowplugWait what am I thinking.  We don't need the interferometer now correct?19:32
glowplugAssuming I can keep the sample still during exposer.19:32
Sync_you need a precision current source with a few passives and a pid temperature controller19:33
Sync_which is just any micro with a high resolution temperature senor19:33
Sync_if you do a projector you do not need any interferometers19:33
Sync_they still are for a moving xy stage19:33
azonenbergI think you will still have to do a fixed laser and moving x-y19:34
azonenbergits much easier to move the sample by microns19:35
azonenbergthan to try to steer a beam by millionths of a degree19:35
glowplugI can totally steer a beam by millionths of a degree.19:35
azonenbergaccurately? :p19:35
azonenbergand repeatably?19:35
azonenbergthere's a reason nobody does this19:35
glowplugOne person did it.  On Youtube.  Hahaha19:36
azonenbergWith what kind of resolution?19:36
glowplugHorrible.  Because he used Galvos.19:36
azonenbergAnd what do you plan to do19:36
glowplugBut steppers with a 200/1 gear reduction should give enough resolution.19:36
azonenbergDo the trig19:36
Sync_enjoy your backlash19:36
azonenbergIf you want 1um shift19:36
Sync_and yeah do the trig19:37
azonenbergat say a 10cm distance19:37
azonenbergany idea what fraction of a degree that is? :P19:37
glowplugBacklash in the gearbox?  I did do the trig but for 10mm above the sample.19:37
azonenbergyou realize you need room for the optics19:37
glowplugIn 1/4 of a second of movement at 200 pulses per second you get 1/4 full rotation.  200 reduction gives you .001 degrees change in that period.19:39
azonenberg0.001 degrees19:39
azonenbergdo you know how huge that is?19:39
glowplugLet me find my notes onesecond.19:39
glowplugI got .048mm travel with 1 second of motor rotation at 10mm from the sample.19:40
azonenberg48 microns19:41
azonenbergthats huge19:41
glowplugNope I have 1000/1 reduction here not 200.  Thats why.19:41
glowplugThats for a full second of motor rotation.19:41
Sync_1000 times reduction with gears19:41
Sync_do you know how large that is19:41
Sync_what the hell are you smoking19:41
glowplugHow large 48 microns is?19:41
glowplugThat is with a full second of rotation or 200 pulses.19:42
azonenberg1000:1 reduction...19:42
azonenberghow much backlash is that gonna be at the low end?19:43
glowplugRight.  That is with a 4 stage planetary.  Laser cut out of acrylic.19:43
azonenbergfive or ten revolutions maybe?19:43
glowplugI mean there is going to be backlash no matter what.  Whether its in the gearbox or the ball screw.19:44
glowplugAt least this way the machine costs $100 total instead of $2,000 to get results.19:44
azonenbergCompletely useless results19:45
Sync_you need to move 5.7µ°19:45
glowplugThe only reason it wouldn't have 10 micron repeat accuracy is if the gearbox's were absolutely horrible.19:45
glowplug5.7u degrees sounds scary.  But we have to make adjustments of that accuracy everywhere.  Even down to aligning the sample.19:46
glowplugHaving a stepper do it isn't that crazy.19:46
Sync_I want to see that19:47
azonenberghaving a stepper do 175,000 steps per revolution?19:47
azonenbergoh, and that assumes you can fit your optics in 1cm from the sample19:47
azonenbergWhich is rather unrealistic especially for a homebrew setup19:47
azonenbergi'd expect more like 10cm19:47
Sync_I mean shit, it is easier to build an xy table that does that19:47
Sync_and then you can scan a *very* small area19:48
glowplugNo stepper has 175,00 steps per revolution.  Thats what the reduction is for.19:48
Sync_because your focus will deteriorate19:48
azonenbergThat was the other reason i wanted to go x-y19:48
azonenbergyou dont need to mess with the focus19:48
Sync_yes but your quad planetary has an assload of play19:48
Sync_you cannot avoid it19:49
Sync_that's why there are hypocycloid drives and harmonic drives19:49
glowplugI cannot avoid gear reductions anyways.  On a SCARA, Cartesian, doesn't matter.19:49
glowplugHypocycloid drives are unsuitable due to wobble no?19:49
azonenbergyes, but it's better to move the sample a large distance19:49
azonenbergthan a mirror a tiny distance19:49
Sync_the reduction you need is just unrealistically small19:50
Sync_or rather large19:50
glowplugIf a Hypocycloid drive is suitable I will use that.19:50
glowplugI assumed the offcenter mass would make it useless.19:50
Sync_you need a good one19:50
azonenbergi still think a Cartesian micrometer screw with a geared stepper is probably the best option19:51
glowplugOk then I will concede that the 4 stage planetary probably has too much backlash for mirror positioning of that accuracy.19:51
azonenbergand interferometers for positional feedback19:51
glowplugIf I can get a hypocycloid drive laser cut and get my 1000/1 with only 2 stages then I think it is work trying.19:52
azonenbergI want to build a Michelson interferometer some time soon just to play with19:52
azonenbergAnd do the math for how many teeth you'd need on the gears19:52
glowplugThat would eliminate any backlash in the system.19:52
azonenberghow big are these gears gonna be?19:52
glowplugI did.  With 4 stages that reduction is possible with gears that fit on a NEMA-23 sized stepper.19:53
azonenbergHow many teeth19:53
glowplugBut I agree that would likely have too much play.19:53
azonenbergAnd yeah19:53
azonenbergYou will want positive feedback19:53
azonenbergand you cant get that with beam steering19:53
glowplugAnd the hypocycloid drive (if stable) would work.19:53
azonenbergmoving the sample, though, allows you to move with the beam blanked19:53
Sync_are you really expecting lasercut pieces to be accurate enough for that19:53
glowplugFor a hypocycloid drive?19:54
Sync_I mean it'll spin19:54
Sync_but will it really be backlash free?19:54
Sync_the one I have is wire EDM'd19:54
glowplugAll hypocycloid drives are backlash free.19:54
glowplugJust not wobble free.19:54
glowplugAnd that is something I don't have the answer for.  I don't know how much it will wobble.19:54
glowplugThe cost difference between my proposed system and the "traditional" is a factor of 50.19:55
Sync_there are just too many engineering problems with your idea19:56
glowplugI think it is worth persuing even for a short time if the design could enable people to contribute to the project who can't afford a $3000+ setup.19:56
glowplugThere are problems.  Thats why I brought it up.  And my backup plans.19:57
glowplugBut I think the hypocycloid gearbox could work if it was designed well enough to not wobble.19:57
azonenbergSync_: http://www.ebay.com/itm/Manual-Linear-X-Stage-0-5-Inch-100TPI-Thorlabs-T12X-/281053144084?pt=LH_DefaultDomain_0&hash=item417012a41419:57
azonenbergwhat do you think of getting that19:57
azonenbergif i want to fool with interferometry19:58
azonenbergspecs claim that 400nm movement accuracy is possible and so if i sample at a few KHz i should easily be able to pick up sub-wavelength shifts19:58
glowplug4 available.19:59
glowplugThose are extremely nice.19:59
azonenbergWell don't buy them before i get some :p19:59
Action: azonenberg adds 2 to cart19:59
azonenbergSeller has good feedback and they're only $35 each20:00
azonenbergeven if they arent good enough for fab i have a lot of other uses in mind20:00
Sync_they are a little small20:00
azonenbergSync_: bear in mind we're talking about putting a single die on it20:01
azonenbergI actually wanted to use them for microscopy20:01
Sync_get me the other two azonenberg20:01
azonenbergSync_: You buy them :P20:01
Sync_oh he ships world wide20:01
Sync_that's nice20:01
azonenbergI feel like it should be possible to hook a stepper up to them20:02
azonenbergI want to make an x-y CNC stage for my microscope20:02
azonenbergto do automatic scanning of large areas20:02
glowplugYou can send individual pulses to a stepper.  The resolution of the machine I described earlier (assuming it could be built) would be 200/48 microns or 240nm.20:02
glowplugSo my backup plan.  Lets see if I can get it shot down too.20:04
glowplugThe SCARA robot.20:05
Sync_azonenberg: he has a nice xy stage too20:05
azonenbergSync_: i wanted linear ones for the interferometer experiments20:05
azonenbergi'll look at his other stuff after payday :p20:05
glowplugActually it suffers from all of the same basic problems as the projector configuration.  Plus some others.20:05
Sync_XY axis RMS repeatability < 0.7 ¼m20:06
glowplugAlright let me put it this way.  I'm going to attempt the configuration I described in the off chance it does something useful we would have significantly cheaper litho.  If I fail (~97%) then I'm only out a few bucks.20:06
azonenbergfor my cnc microscope i dont even need that good20:07
azonenbergi just need to find an area of the sample20:07
azonenbergoh and i need to get my camera working in linux20:07
glowplugIn the meantime I will also plan an x/y stage with interferometer feedback ect. ect.20:07
glowplugWhat type of camera do you have?20:07
Sync_aah I want that xy stage20:09
Sync_that would solve all issues :D20:09
azonenbergSync_: go buy it?20:09
Sync_yeah I'll do20:10
azonenbergif you dont, in a couple of days i might :p20:10
azonenbergDo you think a cheap pointer-level laser diode from digikey is stable enough to use for some quick fooling around with inteferometry?20:11
Sync_not so sure20:12
glowplugYeah there was a video on YouTube I watched last night with a dollor store laser and he got a pretty good pattern.20:12
azonenbergi mean i wouldnt use it for precise measurements20:12
azonenbergbut just to fiddle with the technique20:12
Sync_I read a paper that said it depends on the diode20:12
Sync_but I'd get a hene20:12
azonenbergFor a production system i might20:13
glowplugYou probably have to ditch the optics but the cheap diodes should work for experimentation.20:13
glowplugIf you have a really good quality current source it will work for sure.20:13
azonenbergThis laser has an integrated supply20:14
azonenbergi have no idea how stable it is :p20:14
glowplugI would pull its supply and the optics.20:14
azonenbergits a pointer-style sealed unit20:14
soul-dthey did that feynman expiriment thought experiment now btw20:14
glowplugJust hack it up then use your own lense and current source.20:14
glowplugAnd it will work awesome for a michealson setup.20:15
glowplugThe diodes are generally fine with everything else being junk.20:15
Sync_azonenberg: a hene on ebay should be under 20$20:15
azonenbergOne thing at a time lol, i dont have enough time to build the whole thing now20:16
soul-dnot the first20:16
soul-dold news :P20:16
Sync_just saying :D20:16
glowplugThe cheapest tubes I could find was a pair of two for $50.20:16
glowplugBut who knows what shape they are in.20:16
glowplugThe newish looking tubes are ~$75+20:16
Sync_just wait and bid20:16
Sync_I have a few and none was more than 25$20:16
azonenbergSync_: are diode lasers just not stable enough for interferometry?20:17
glowplugThey are!20:17
glowplugBut hene lasers are extremely stable by nature.20:17
glowplugAnd a frequency stabalized hene laser is near perfect.20:17
glowplugDiode lasers are good for experiment ect.20:18
glowplugOne of the best youtube videos from MIT.  They make a simple michealson with a diode laser and calculate its wavelength.  Very accurate.20:19
Sync_azonenberg: the beam quality is a lot better20:20
azonenbergSync_: well i guess one nice thing about HeNe is a known output wavelength20:20
Sync_that too20:21
azonenbergSo you can have numbers measured in nm rather than in wavefronts :p20:21
glowplugThe MIT experiment used a 650nm diode and 1/4 transitions were clearly visible.  So that is fairly accurate.20:22
glowplugOf course even if the laser was affordable they have top dollar optical equipment.  =P20:23
glowplugYeah what you need to do for the diode laser interferometer is have really good optics and be able to adjust mirror very accurately.20:30
Sync_one could just measure the wavelength20:31
glowplugAbsolutely.  It would be a pain compared to hene but fun nonetheless.20:31
glowplugOf course you need to already have a good laser to do that.  O.o20:33
Sync_it would be connecting the fiber to the wavelength meter20:33
glowplugDo you guys really have every cool toy?  Haha20:33
glowplugJust so I understand fully.  The current prevailing / preferred design of the direct litho.20:36
glowplugIt is a moving x,y stage driven by steppers?/servos? with a gear reduction.  A mirror on all sides that are monitored by a series of interferometers for position encoding.20:37
glowplugFlip down microscope with a point and shoot cam on the top.  Micrometers for manaul adjustment.20:38
Sync_http://citeseerx.ist.psu.edu/viewdoc/download?doi= I guess that's what would be the best20:40
Sync_well if I really wanted a wavelength meter I'd just go to the lab at work20:41
glowplugReading.  This guy gets a good pattern with a radioshack diode.  Skip to 5:30 http://www.youtube.com/watch?v=wdasvpErwsc20:41
glowplugThat was with manual alignment and no optical table.  Pretty good in my opinion.20:42
glowplugThere is actually a PDF "laser scanner for direct writing lithography".  Their design is quite a bit different than mine but with 1 micron accuracy and claiming that .5 is possible in principle.20:49
glowplugBeam positioning is by modulation not mechanical actuation.  Interesting.20:51
glowplugAlright so a single HeNe laser gives feedback for both the X and Y.  That is interesting.20:54
glowplugDid you see the cost estimations for the device?  LOL20:55
Sync_not too bad20:56
Sync_this is 9520:57
Sync_keep that in mind20:57
glowplugWe want manual digital control of the stage for positioning correct?20:58
glowplugAnalog joystick sort of thing.20:58
Sync_not really, some kind of pc interface would be nice, linuxcnc would do just fine20:59
glowplugCould just do a custom interface in linuxcnc.20:59
glowplugWith the camera output right in the software.21:00
Sync_you don't even need a custom interface21:00
Sync_the camera does nothing after you alinged everything21:00
glowplugIt would make automatic alignment possible.  Thats just a toy though.21:01
glowplugThe PDF says servos.  I wonder what the gear reduction is.21:01
glowplugSo you intend on using hypocycloid gearboxes and servos?21:03
glowplugWhat kind of motor configuration?21:04
Sync_some kind of dc servo with a high resolution encoder21:04
glowplugWon't you need a gearbox?21:05
Sync_probably, but 10fold would do21:05
Sync_because the real positioning is done by the interferometer21:05
glowplugI've been working a little on the encoder problem.21:06
glowplugCD's have .5 micron resolution.  You could build a very accurate encoder with cdrom guts and a ring with alternating 1-0 pattern written by the burner.21:08
glowplugCould theoretically be millions of steps per rotation.21:09
glowplugThe pattern could be written by rigging the burner to be stationary and spinning the ring under it at a set rpm.21:10
Sync_I'll just buy one21:11
glowplugWhat resolution do you need?21:11
glowplugHow many steps I should say.21:14
Sync_that depends on the gearing21:15
glowplugWell for example a decent servo driven articulated robot has about 200,000 steps per revolution.21:17
glowplugI'm sure we need more than that.21:17
Sync_not really, I suspect the guys that make the stage I posted use something around 2800 step/rev wheels21:20
glowplugThen there must be a massive reduction in the x,y stage?21:21
Sync_~20-30 I suppose21:21
glowplugSo thats 200-300 between the rotational and linear motion.21:21
glowplugThats 840,000 steps per linear travel.21:22
glowplugUnfortunately I don't know very much about the design of those x,y stages.  Does it make any sense to DIY the table itself?21:23
glowplugI thought you would say that.21:24
glowplugWell at least everybody else appears to be well funded.  Haha21:24
azonenbergwell funded, lol21:24
azonenbergi'm a grad student21:24
azonenbergafter rent and food i have a few $k a year to spend on fun stuff at most21:25
glowplugSo it's just Sync_ thats well funded?  xD21:25
glowplugWell my encoder discs will be CD's.  I'll put the designs on the wiki then hopefully other under-funded people can experiment.  =)21:26
glowplugI have fully conceded on the projection configuration.  They exist (since 1989 apparently) but the design is not straitforward and there are too many variables that make it too error prone.21:27
Sync_it's not really that I actually have a lot of money, but when I do I spend it21:27
Sync_ah nice21:29
Sync_avago makes 20000imp/r encoders21:29
Sync_that makes for 80000 counts after quadrature decoding21:29
glowplugHow much?21:29
Sync_good question21:29
glowplugUh oh.  Prices arent on the website.  Warning!21:30
azonenbergglowplug: doing it on CDs is nontrivial21:30
azonenbergthey include error correcting codes and framing21:30
azonenbergyou cant just burn a timecode using regular burner firmware21:30
Sync_unfortunately they are not avalible anymore21:31
Sync_but I bet they are around 200$21:31
Sync_maybe less21:31
Sync_BUT they have nice new encoders21:32
Sync_I'm just trying to figure out what they used in the stage21:32
glowplugYes I'm still working on that.  It would be necessary to digitally modulate the laser diode directly and to read the photodiode output directly bypassing the drives uC.21:33
glowplugIn the PDF they say "DC Servo Motors".  =(21:34
Sync_yeah I guess they have a +-4-30 reduction21:34
Sync_5 is realistic with encoders avalible at that time21:35
Sync_glowplug: actually they are not that bad21:36
glowplugWe need three of them though.  =)21:37
glowplugOh right.  Then the piezo for the Z which I have no experience with.  =(21:37
glowplugHow much for disc + encoder?21:38
Sync_I suppose one could live without autofocus for a while21:38
glowplugYeah I think autofocus will need to be added later.21:39
Sync_around 100$ per set21:39
glowplugHey thats actually not that bad.  Digikey must have very bad prices.21:39
Sync_for a 17bit absolute encoder21:39
Sync_that's pretty badass21:39
glowplugI'll still take a crack at my CD rom encoders for free each.  =)21:39
glowplugWe haven't talked about control much.21:40
glowplugIf we pipe everything into an FPGA, interferometer output, encoder output, and also use the FPGA for motor control we will have very good latency with very low cost.21:41
Sync_I'd just rig linuxcnc to it21:41
glowplugLinuxCNC can't be used for encoder feedback loops its too slow.21:42
glowplugThere is FPGA motion control IP that will do up to 8 servos with encoder feedback.  Just integrate the interferometer output into the control logic.21:43
Sync_tell that my cnc21:44
Sync_yeah I have a mesa card21:44
glowplugThe PC software has 10us latency.  Too slow for the actual control loops.21:44
glowplugIf you have a mesa card you are all set then.  =)21:45
Sync_that's the easiest solution21:45
glowplugWell that was easy.  Haha21:45
Sync_I mean further dicking around is just meh21:46
Sync_at some point you have to throw money at it21:46
Sync_we are still talking about positioning to 1µm21:46
Sync_trying to make a custom drive controller consumes just too much time and effort21:47
glowplugI agree.  The mesa unit is actually affordable and adding the interferometer input and control logic won't be difficult at all.21:48
glowplugImagine if they had mesa cards in 1995.  We just replaced $3,000 in equipment with $100.  Haha21:50
glowplugDo you have a 6125?21:50
Sync_oh I'd have to look at my docs21:52
Sync_I think it is a 5i2221:52
glowplugFor the 6125 (the pci express unit) if you buy 5 they are only $87 each.21:52
glowplugI don't think there are any cost savings to be realized compared to that.  O_O21:53
glowplugThat unit is quite a bit more expensive.  But the 6125 will do everything we need for $87.21:55
glowplugIs a blu-ray diode even accurate enough for 10um devices?21:57
azonenbergThe diode, sure22:02
azonenbergyou'd need better optics22:02
azonenbergThat's the paper i was talking about before22:04
azonenbergunder $1k end to end for ~1um resolutoin22:05
glowplugHoly shit in a handbasket.22:06
glowplugThat is fantastic.22:06
glowplugTheir diode is violet with very low power.  Thats great!22:07
azonenbergThey're using SU-8 negative resist but you can do the same thing with regular positive resist22:07
azonenberg405nm diodes work on that too22:07
glowplugI have negative resist sheets thats what I'll be testing with so that works.22:08
glowplugThey are using steppers.  No mention of gearing.22:08
glowplugThey are using the reduction in the x,y stage only???22:08
azonenbergglowplug: quite possibly22:09
azonenbergthese stages will move maybe 100um per rotation22:09
azonenbergso 1-degree precision is submicron22:09
glowplugThey dont even have interferometer feedback!22:09
glowplugThis is very encouraging22:09
azonenbergThat's something you'd put in if you wanted higher resolution22:09
azonenbergthe stages i picked up are supposed to be 250um per rotation of the knob22:09
azonenberg1-degree precision is easily doable with an ungeared stepper22:10
glowplugSub micron resolution.22:10
glowplugAdjusting micrometers $150.22:10
glowplugSo there are steppers controlling the x,y stage.22:10
glowplugThe micrometers are where?22:11
azonenbergThe stage is probably micrometer screws on the axes22:11
azonenbergthat's typical22:11
glowplugI see.  So they added the micrometers TO the stage.22:12
glowplugAttached steppers to the micrometers.22:12
azonenbergthe micrometers probably shipped with the stage22:12
glowplugSo all adjustment is done in sfotware.22:12
azonenberga high-precision stage without micrometers is pretty useless22:12
azonenbergAnd for that design it does seem like it's all open-loop contorl22:12
glowplugHave you seen these?22:13
glowplugIt is open loop.  But there is very little load on the steppers.22:14
glowplugFor submicron though I would think that encoded servos are more reliable.22:14
azonenbergSubmicron is a whole other animal22:14
azonenbergAt that point i might think about switching from laser to e-beam22:14
azonenbergor alternatively using nearfield optics22:15
azonenbergNeither is easy22:15
azonenbergBut ~1um with lasers should be very doable22:15
glowplugA lot of these costs can be even further reduced.22:18
glowplugStill need ~$100 for the blu-ray laser + lens.22:19
glowplugBut you can save $50 on the controller, $300 on the stage + micrometers, Digital microscope can be made for ~$30 thats another $120 savings.22:20
glowplugBased on these numbers it can probably be done for ~$400 total.22:21
azonenbergmicroscope for $30?22:24
azonenbergi dont think so22:24
azonenbergyou cant even get a good objective for that much22:24
Sync_yeah I'm not a huge fan of that design22:24
Sync_it sure does work but it's openloop22:25
glowplugThe webcam design?22:25
azonenbergSync_: I like their design but i feel like it can be improved quite a bit22:25
azonenbergi'd want to clone it first22:25
Sync_yeah the idea is great22:25
glowplugOh the entire stage.22:25
azonenbergthen add closed-loop feedback22:25
glowplugI agree with closed-loop.  Thats pretty much a no brainer.22:25
glowplugDo you have a link to the sort of microscope optics that would be suitable?22:25
Sync_meh do I really have to fire up the vpn for access22:26
azonenbergI'd use an Olympus objective from ebay22:26
azonenbergmachine a custom tube22:26
glowplugYou keep getting kicked?  =(22:27
azonenbergthen use a nicer camera, maybe a HD webcam22:27
Sync_reminds me22:27
glowplugI suppose the worlds view of our progress will be through that microscope.22:27
glowplugSo that is probably a good idea.22:27
azonenbergyou want to use a nicer one for inspection22:28
azonenbergthis would be just for alignment22:28
azonenbergActually, screw that22:28
azonenbergyou might not even need the uscope22:28
azonenbergif you're doing direct write22:28
azonenbergto make masks22:28
azonenbergyou can forget about alignment since once etched the mask wil then be used in a contact aligner22:28
Sync_no need to align anything22:28
glowplugThe scope is for calibrating the machine.22:29
glowplugIt wont work correctly the first go.22:29
glowplugAnd also uploading images / video to the wiki.22:29
azonenbergYou will want another scope for inspection of work22:29
azonenbergthe one on there won't be a nice one you can actually use for general imaging22:30
glowplugHow much will the "actual" scope cost?22:30
azonenbergI got mine on ebay for $250 + $250 + $5022:30
azonenbergbody, head, and illuminator respectively22:30
Sync_that's not a lot22:31
azonenbergthat's a lot?22:31
Sync_to put that in perspective22:31
azonenbergNew, it was probably five or ten grand22:31
azonenberg$500ish is dirt cheap22:31
Sync_shit I have spent over 4k on my mill22:31
Sync_so what.22:31
glowplugOne thing you guys have to realize.22:31
glowplugThe cheaper this stuff is the more people can contribute.22:31
glowplugThere is a cascading effect of money and support to a project.22:32
azonenbergIt's never going to get to the point of being home-PCB-etching availability22:32
azonenbergI'm shooting for "affordable by a hackerspace"22:32
azonenbergOr by a sufficiently insane hobbyist22:32
glowplugBut it can get down to "enthousiest garage fab shop with a modest income".22:32
glowplugRight.  That seems a good target.22:32
glowplugBut equipment under $2000 total (for everything) is probably realistic for that goal.  Not $20,000 total.22:33
azonenberg$2000? Lol22:33
azonenberg$10K is more what i was hoping to, after lots of tuning22:33
azonenbergthe prototype stuff will probably cost me more like 50k22:33
Sync_30k is a good starting point imho22:33
Sync_which is affordable for a hackersapce22:33
Sync_~space even22:33
glowplugThe spincoater can be made for ~$100.  These guys have litho 1um litho with ~$500 in parts.22:33
azonenbergwell, once i get out of school22:34
azonenberghigh on my list of major purchases is a secondhand FESEM22:34
glowplug$500 for a good microscope.22:34
Sync_oh yes azonenberg22:34
glowplugWe are at $1,100. What other major components are absolutley necessary?22:34
Sync_they can be had quite cheaply22:34
azonenbergglowplug: Consumables22:34
Sync_yeah comsumables22:34
Sync_vacuum parts22:34
azonenbergDopants from emulsitone are $500 each for P and N type22:34
azonenbergor is it $250?22:34
azonenbergthat much again for undoped spin-on-glass22:34
azonenberg$300ish for a bottle of photoresist22:35
azonenbergHMDS can be had cheap in small volume22:35
azonenbergtrace metal grade TMAH22:35
Sync_a wafer is 10-30$22:35
azonenbergnot sure what the going rate on that is22:35
Sync_I guess tmah can be cheap22:35
Sync_but in small units hmm22:35
Sync_I'd say 5k minimum for a basic setup22:36
Sync_access to a machine shop is neccessary tho22:36
glowplugFor consumables we can order in volume then do pre-packaged kits for the community.22:38
glowplugYou get the economy of scale plus the cost to each hobbyist is very low.22:38
glowplugThe vacuum is for which process?  How strong of a vacuum?22:39
Sync_seriously, this is not the kind of thing you can pull off with a tutorial22:39
Sync_vacuum for any kind of metal deposition or plasma etching22:39
azonenbergAnd for evaporation you need 1E-6 torr range22:39
glowplugNo I don't think tutorial.  I think there are enough hobbyists out there with general knowledge of this stuff but thinking that its completely out of reach without school / work equipment.22:39
azonenbergso diffusion pump range at best22:39
glowplugThey just need to be proven wrong.22:40
Sync_plasma etching wants 10^-2mbar and evap 10^-4 or better preferrably22:40
azonenbergSputtering can be done as high as 30mtorr but that's after you have a good argon purge22:40
azonenbergSo you need to get the chamber down quite a bit lower first22:40
glowplugWhat is the plasma etching for?  This is the part of the process that I now nothing about.22:40
azonenbergI was initially hoping to get 10um features working with just wet etches22:40
azonenbergthat might be wishful thinking22:40
azonenbergi've done 20 with wet etches experimentally and they worked fine, though i had some adhesion problems here and there22:41
Sync_that's quite a good deal for the parts22:42
Sync_three ion pumps, all the CF hardware22:43
azonenbergSync_: but not as a SEM22:43
Sync_just browsing general vacuum stuff22:43
Sync_too lazy to study :D22:43
glowplugThe guy on Youtube with the diy magnetron.  What was his vacuum source?22:44
Sync_refridgeration vac pump22:44
glowplugIs that sufficient for everything we need?22:44
azonenbergWas that the guy who had all kinds of problems getting deep enough vacuum?22:44
Sync_that's probably ok22:44
azonenbergoxygen contamination everywhere?22:44
Sync_yeah everything oxidized22:44
Sync_no glowplug by far not22:45
azonenbergSync_: oh http://www.ebay.com/itm/Jeol-840-Scanning-Electron-Microscope-/270432223960?pt=LH_DefaultDomain_0&hash=item3ef70442d822:45
azonenbergI trained on that same setup lol22:45
Sync_we have one too22:45
azonenbergan 840?22:45
azonenbergLol nice22:45
glowplugI see that.  Copper oxides on the slide.22:45
Sync_it's our oldest one22:45
azonenbergSync_: same here22:45
azonenbergThe JSM-6335 is the one i've used more recently22:46
azonenbergit's a FESEM22:46
azonenberg$20/hr more to use but thats not a huge deal22:46
azonenbergthen if you're made of money you can use the Zeiss in the cleanroom :p22:46
Sync_we can use all the equipment for free22:46
glowplugSo what processes require vacuum that are *Absolutely* necessary for a 10um device?22:46
azonenbergglowplug: Metal deposition22:47
azonenbergYou can use spin-on dielectrics and wet etching for large features22:47
azonenbergso no need for RIE, PECVD, etc22:47
glowplugAlright so wet etching refers to using the spincoater for metal deposition. That I get now.22:47
Sync_pecvd is something that is not happening with my pumps :D22:47
Sync_too much ugly stuff involved22:47
azonenbergit refers to etching stuff using a liquid acid22:47
azonenbergrather than a plasma22:47
glowplugWhat are spin on dialectrics?22:47
azonenbergWhich means it can be done at atmospheric pressure22:48
azonenbergglowplug: stuff like http://www.emulsitone.com/sif.html22:48
azonenbergI suggest you learn more about semiconductor fab in general22:48
glowplugI'm reading nonstop already.  I just discovered this stuff a week ago.  Haha22:48
azonenbergThen work on how to scale it down to hobbyist level22:48
Sync_the second hitachi I posted seems to be ok22:48
Sync_it's no fesem for sure but it'll do22:49
azonenbergSync_: an 840 would be nice to have at home as long as i could rent time on a nicer scope when i needed it22:49
glowplugBut I mostly focused my research so far on the litho because I understand those principles well.22:49
azonenbergSync_: But stuff like http://i.imgur.com/esVwLlK.jpg22:49
azonenbergi dont think you'll get that kind of resolution with an 84022:49
Sync_probably not22:50
glowplugIt was already determined that the wet etching is not very reliable with devices smaller than 10um correct?22:50
azonenbergthe round dots are 60nm nominal colloidal silica left over from from CMP22:50
azonenbergthe device is a 45nm FPGA22:50
Sync_hehe nice22:50
azonenbergglowplug: I forget when industry stopped using wet etches22:50
azonenbergFor submicron i'd definitely want to use plasma22:50
azonenbergpossibly sooner22:50
Sync_but if you are into that you can just buy one, reverse engineer shits and pay it off22:50
glowplugAnd the magnetron is enough for every metalization step for a device correct?22:51
azonenbergSync_: i'm the guy behind siliconpr0n.org :p22:51
Sync_yes I know22:51
azonenbergglowplug: The full suite of film coating capabilities you'll want22:51
Sync_but if you wanted to do that commercially22:51
azonenbergis spin coating, DC sputtering, RF sputtering, evaporation, and PECVD22:51
azonenbergall but the first require vacuum22:51
azonenbergYou can probably live without PECVD if you do reactive sputtering for dielectrics, or use spin-on glass22:52
glowplugThe PECVD we don't need immediately correct?  What about evaporation?  That is a different device from the magnetron?22:52
azonenbergEvaporation and sputtering might be able to use the same chamber depending on how you design it22:52
azonenbergthe processes are different but they have some similarities22:52
azonenbergDC and RF sputtering can definitely share the chamber22:52
glowplugSo a well designed chamber will take care of every vacuum process.22:53
azonenbergNot necessarily22:53
glowplugThen you have litho and the spin coater.22:53
Sync_I want22:53
glowplugThats only three primary devices.22:53
azonenbergSync_: *drool*22:53
azonenbergglowplug: well litho is not one tool22:53
glowplugI mean three main monolithic tools.22:53
azonenbergyou need both the direct-write system for mask making and then the contact aligner22:53
glowplugAhh right.  For exposure to the mask.  So that is 4 primary tools then.22:54
glowplug*through the mask22:54
azonenbergThen general lab glassware etc for wet processing22:54
azonenbergmicroscope for in-process inspection22:54
azonenbergIf you're doing copper metalization or multiple metal layers CMP would be helpful22:54
glowplugRight.  ~$500 scope for checking work.22:54
glowplugSo for an absolute bare-bones lab.22:55
azonenbergBut for a simple 1-metal process using evaporated aluminum you could get away without CMP22:55
glowplug$200 spincoater, $500 litho, $500 scope, $200 exposure setup, $XXX magnetron?22:55
azonenbergFor bare bones you'd want the spin coater, laser direct write for masks, contact aligner, and an evaporator22:55
azonenbergI would do evaporation as the first vacuum tool22:55
Sync_it is the easiest22:55
azonenbergit needs deeper vacuum but you dont need to add gas to it22:55
azonenbergits basically just a light bulb22:56
Sync_still around 1k realistically22:56
azonenbergthat happens to have somethign stuck to the filament22:56
glowplugThe evap is ~$1k?22:56
azonenbergYou could use spin-on dielectrics22:56
azonenbergso the evap would only be for metal22:56
Sync_and even 500 for a microscope is quite a bargain22:56
azonenbergSync_: yeah, i got lucky22:56
azonenbergasking price was 500 each on body and head22:56
Sync_I know22:56
azonenbergi bargained them down22:56
azonenbergOh, and then a furnace for implants22:56
Sync_or oxides22:57
azonenbergI'd do spin-on glass22:57
azonenbergsince you'd need that for an ILD anyway22:57
glowplugDoes the furnace need vacuum?22:57
Sync_(then an AFM for measureing the thicknesses) ;)22:57
azonenbergmight as well use it for field oxide22:57
azonenbergglowplug: No vacuum, but it might be nice to purge it with inert gas22:57
azonenbergargon or something22:57
azonenbergif you are trying to heat and not oxidize22:57
glowplugI think I've seen nitrogen used.22:57
Sync_yeah but the inert gas needs to be fairly clean22:57
Sync_which has been the issue for us22:57
azonenbergand all of the chemicals have to be completely metal-free22:58
glowplugSo the evap is ~$1k?22:58
Sync_we tried and we got oxide22:58
Sync_abou that22:58
Sync_the scope too22:58
azonenbergSync_: Thats why i want to focus on MEMS first22:58
Sync_if you calculate with bargains that are possible22:58
Sync_and not 1 in 100 bargains22:58
glowplugAnd you think the $1k scope is absolutely necessary to check work at ~10um?22:58
Sync_I know why you want to do that azonenberg ;)22:58
glowplugAlright so we need to narrow down the oven cost.22:59
Sync_as I said, around 5k is a _realistic_ starting price22:59
azonenbergglowplug: LOL22:59
azonenbergthis is not an oven22:59
azonenbergits a furnace22:59
azonenbergyou have to get up to around 1200C22:59
Sync_I got my ceramic burning oven for 5022:59
azonenbergSync_: metal-free? :p22:59
Sync_but it's not for that22:59
Sync_haha sure not22:59
azonenbergA new one from MTI is probably gonna run you about a grand23:00
azonenbergDo you think AL2O3 insulation is safe for CMOS?23:00
Sync_I could make a SI insert23:00
glowplugIs it possible to get to 1200C with a DIY unit?23:00
glowplugInduction cant be used can it?23:00
azonenbergThis is what i was looking at23:00
Sync_oh hmm23:00
Sync_let me ask our techs tomorrow23:00
azonenbergSync_: you think that one would be cmos-compatible if you were careful about only putting freshly RCA-cleaned silicon in it?23:00
glowplugYou can get to 1200C with induction for pennies.23:01
Sync_we had all sorts of shits going on in our ovens23:01
azonenbergSync_: lol23:01
azonenbergwell high-temp processes are where the worst ion contamination happens23:01
glowplugI dont know if it would just blow up the sample or something.23:01
Sync_we have a company in the institute that does ovens23:01
Sync_you cannot use induction23:01
glowplugYour positive?23:01
glowplugWell that sucks.  =(23:01
Sync_you could use induction to heat the sides of the oven23:02
glowplugGood point.  Then heating from the oven itself to the sample would be by convection.23:02
Sync_convection at over 900°C?23:02
Sync_not so much23:02
azonenbergmore radiation i'd think23:02
glowplugDoes that work?23:03
glowplugFrom induction to radiation heating?23:03
glowplugWhat kind of material would the oven be?23:03
Sync_something ferrous23:03
glowplugIron oven?23:03
Sync_actually azonenberg I think we use si wafercarriers23:03
azonenbergyou'd have to be really careful about metal particles falling out and landing on the samples23:04
Sync_and quartz holders for those23:04
azonenbergSync_: i was thinking quartz boarts23:04
azonenbergthe question is whether being near hot Al2O3 is dangerous to transistors23:04
azonenbergi wouldn't actually have them touching23:04
Sync_well it is used as gate dielectric23:04
azonenbergBut that's post-diffusion righht?23:05
azonenbergthe big time you have metal issues is just before an implant or oxidation bake23:05
azonenbergafter that, no more furnace23:05
azonenbergjust whatever heating you get during an etch or deposition step23:05
glowplugI think that a 10kw powerstage is enough.23:06
azonenbergYou realize that you can only pull around 2kW from an average (American) mains circuit, right?23:07
glowplugThis is for garages and hackerspaces.23:07
Sync_azonenberg: it seems to be safe23:07
glowplugThey can figure out how to draw 10kw.  EV chargers do.23:07
glowplugIt's not like the $1,000 oven draws much less anyways.23:07
glowplugYeah 10kw is enough.  That much power can heat to 1100C in 9 seconds.23:09
Sync_yeah uh well a nut23:10
glowplugWe need to heat the sample overnight correct?23:10
azonenbergglowplug: For diffusions normally an hour or so23:10
Sync_I guess you could make a sled for the wafer out of quartz23:11
glowplugIf we need to ramp up to 1200C and down again within an hour.  I think it can be done with a 10kw powerstage, some copper, and an iron box.23:11
Sync_so that metals do not affect you that much23:11
Sync_now that is another thing you have to develop23:11
Sync_a 10kW smps is not easy23:12
Sync_let alone a 10kW induction heater23:12
Sync_the tank caps are quite spendy23:12
Sync_the currents involved are large23:12
glowplugIt may be the case that 10kw is actually very overpowered.23:13
glowplugDo we know the power draw from the retail furnaces at the wall?23:13
Sync_the one azonenberg had uses 700W23:13
glowplugThen the induction version should have comparable power usage.23:14
glowplugCould just use an ATX SMPS23:14
Sync_yeah right23:14
Sync_did you actually build a induction heater before?23:14
glowplugI understand their principles.23:15
Sync_http://www.centrotherm.de/uploads/tx_kchnewproduct/Oxidator_Boot_Wafer_02.jpg azonenberg23:15
Sync_understanding their principles does not show you the pain to make one23:15
glowplugIt looks like it can be done without an SMPS at all.23:17
Sync_it *is* a smps23:18
glowplugIn this design I don't see any IGBT's / FETs.23:19
Sync_a smps does not need igbts or fets23:20
Sync_it can also use bipolars or thyristors23:20
glowplugI dont think there are any switches.23:20
Sync_oh sure23:20
Sync_otherwise it would not work23:20
glowplugI'm looking!23:20
glowplugSo far I just see diodes.23:21
glowplugAnd a massive bank of caps.23:21
Sync_it starts at page 1023:21
glowplugGot it!23:22
glowplugI'm not sure that these FET's are switching 10kw however.23:22
Sync_they could23:22
Sync_they are not but they could23:22
glowplugThis is an interesting design.23:23
glowplugThe FET's are switching AC?23:23
Sync_they receive dc23:24
Sync_and make ac out of it23:24
glowplugAhh ok I thought it was a solid state AC voltage regulator.23:24
glowplugSo we are positive that a furnace with gas purging and 1200C max temp is available for $1000?23:27
Sync_about that23:28
glowplugThe DIY option is probably in the $300 range + ~100 hours labor.23:28
glowplugMaybe less.  Hard to say.23:28
glowplugPlus misc and consumables thats hovering at $4500-$500023:29
glowplugFor an absolute basic setup23:29
glowplugThose numbers are with the retail furnace.23:30
Sync_yeah what I said basically23:32
glowplugI wonder if it is realistic to complete a device in a distributed way.  I do the litho, mail out to doping, mails out to oven, ect. ect.23:32
glowplugIt would be slow but maybe possible.23:32
Sync_some processes cannot have a lot of time between them23:33
Sync_because of oxidation and contamination23:33
glowplugAhhh yeah.23:33
glowplugPlus USPS likes to drop kick packages.23:33
glowplugLitho can be outsourced but everything else is probably best kept in a single lab.23:34
Sync_well mask making can be outsourced23:34
Sync_otherwise it takes ages23:34
glowplugBy outsourced I mean one hobbyist makes a set of masks then mails those out to another hobbyist with the lab.23:35
Sync_yeah but you could also get commercial masks23:35
Sync_it's the same thing basically23:35
glowplugAgreed.  Just the price difference.  And quality.  Haha23:35
glowplugWhat is your opinion about the PDF litho setup with the steppers?  Asside from obviously switching to closed loop servos.  What about the lack of interferometer feedback?23:37
glowplugMaybe thats something we can add later?23:37
glowplugThinking the same thing.23:37
glowplugI have a way way way better picture of whats going on now.23:38
glowplugIs the furnace considered FEOL?23:40
glowplugIs it its own category?23:40
azonenbergFurnace is probably FEOL since its not used once you put down metal23:40
glowplugAlright I'll add to the wiki.  =)23:40
glowplugWhat else is FEOL?23:41
azonenbergEverything that happens before you put down the first metal layer23:42
glowplugThat would include Litho then.23:42
azonenbergI'd file litho separately23:42
azonenbergsince its done everywhere23:42
glowplugGood point.23:42
glowplugSo for a basic setup there isn't much FEOL then.23:42
azonenbergNot too much, since we arent doing ion implantation etc23:43
azonenbergThe acronyms stnad for "front end of line" and "back end of line" respectively btw23:43
azonenbergin reference to the order wafers normally go through the tools during fab23:43
glowplugYup.  =)23:43
glowplugJust burned through the wikis.23:44
glowplugThese are the parts I knew *zero* about.  Haha23:44
glowplugSo the contact mask aligner and sample alignment optics are part of the same apparatus correct?23:45
azonenbergInitially we'll do manual alignment23:45
glowplugAt this stage we are aligning the sample to be exposed in UV through the resist.23:45
azonenbergMask is to be clamped in place firmly23:45
azonenbergso we just need an x-y-theta stage23:45
azonenbergwith a relatively imprecise Z axis23:45
azonenbergMove the wafer up to just below the mask23:46
azonenbergline it up23:46
azonenbergbring it up into contact23:46
glowplugRight it needs pan and tilt also.  :/23:46
azonenbergslide the microscope off and load the exposure lamp23:46
azonenbergJust rotation23:46
glowplugOhh I see.  Derp.23:46
azonenbergAs long as you machine things such that the stage and the mask are sufficiently parallel23:46
azonenbergyou just need to line up two in-plane images23:46
glowplugThis apparatus also needs a small vacuum correct?23:47
Sync_pfft no fancy thing as 3d capacitors like we have on dram now? ;D23:47
glowplugOR are we just bolting the sucker down?23:47
azonenbergLight vacuum for the chuck to hold the wafer down is helpful23:47
azonenbergbut that's nothing23:47
azonenbergany pump can do that23:47
azonenbergSame with the spin coater23:47
azonenbergYou could potentially even just use double-stick tape23:47
azonenbergbut vacuum is less likely to damage stuff23:47
glowplugAgreed.  And the tape is probably more expensive.23:47
glowplugSo the microscope for this apparatus.23:48
Sync_I guess getting the wafer off is quite tricky23:48
glowplugExpensive unit or cheap unit?23:48
azonenbergrelatively nice23:48
azonenbergBut you only need one objective23:48
azonenbergnot a full set23:48
glowplug~$200 optics?23:49
Sync_I guess you could use a single optic microscope23:49
azonenbergOne objective, custom-made tube23:49
azonenbergthen a camera on the top23:49
azonenbergYou need two of them23:49
Sync_they are not too expensive even new23:49
azonenbergto see two sets of alignment marks at once23:49
azonenbergotherwise you wont get the theta right23:49
glowplugRight.  One comes into focus you take a snapshot then bring the other into focus.23:49
azonenbergOne on each side of the wafer23:49
azonenbergyou align it approximately by naked eye23:49
azonenbergto get it into the FOV of the microscope23:50
azonenbergthen tweak the stage to line it up exactly23:50
glowplugTwo cameras you mean?  And one scope?23:50
azonenbergyou'd use a split-field view on the computer23:50
azonenbergTwo objectives, two cameras23:50
azonenbergan inch or so apart23:50
glowplugI've seen a single scope/camera configuration.23:50
azonenbergThat is... not nearly as good23:50
glowplugYou focus onto one marker then snapshot it.23:50
glowplugThen switch focus to the other.23:50
azonenbergThis is the setup i'm talking about23:50
azonenbergTwo vertical camera/microscopes on the right23:51
azonenbergUV exposure system at the left23:51
azonenbergchuck at the middle23:51
azonenbergyou slide one or the other into the central slot to use it23:51
azonenbergthe mask stays in one spot23:51
azonenbergand the wafer doesnt move except for tiny alignment corrections23:51
azonenbergneedless to say that process is UV sensitive23:52
azonenbergso you want say red or yellow LED illumination23:52
glowplugHow many watts for the entire grid?23:52
glowplugThat design is actually quite good.23:52
Sync_not much23:52
azonenbergfor the microscope illumination? One or two LEDs each is enough23:52
azonenbergfor the main exposure assembly i'd use party-style blacklight fluorescent lamps23:53
azonenbergmaybe six or eight of them side by side23:53
azonenbergless if only doing 2" wafers23:53
azonenbergProfessional shops use kW-range mercury vapor lamps for like 5-sec exposures23:53
Sync_hm cut open mercury arc lamp23:53
Sync_with a shutter23:53
azonenbergbut if you use a 20W lamp and it takes a minute orso23:53
azonenbergbig deal23:53
Sync_that'd be easy to do23:53
azonenbergyou're not mass producing here23:53
azonenbergSync_: remember this isnt DUV resist23:54
glowplugWouldnt you just do what the growers do and find the wavelength mix needed and get LED's for those wavelengths only?23:54
azonenbergits 405nm or so23:54
Sync_yeah but the lamp will still do23:54
glowplugAhh I see23:54
azonenbergglowplug: http://www.blacklight.com/items/CHUSF20BLB23:54
glowplugI have 3w violet LED's23:54
azonenbergSomething like that23:54
azonenbergFluorescents will probably be a lot cheaper23:54
glowplugThey probably produce more usable power in that wavelength than the blacklights.23:54
glowplugI got them for only $12 each.23:55
glowplugThey are probably cheaper now.23:55
azonenbergJust make sure you diffuse them well23:55
azonenbergor you get some spots unevenly exposed23:55
glowplugGood point.23:55
Sync_I wonder when uv lets show up that actually last a while23:55
glowplugSo there are no software tricks to get that aparatus down to 1 scope?23:55
Sync_not with annoying mechanical effort23:56
glowplugThe scope outputs are superimposed?  Stereovision?  Or seperate screens?23:56
Sync_just two screens23:56
azonenbergglowplug: http://www.cleanroom.byu.edu/alignment.phtml23:57
glowplugI see.  Page 3 explains what I was missing.23:58
glowplugI've seen this proces before.  But I forgot that you are aligning 4 total markers.  Two on the aparatus and two on the sample.23:58
glowplugThat would be a massive pain with only 1 scope.  =(23:59
azonenbergPossible, but you really dont want to23:59
azonenbergspend the extra $200 or whatever on the second scope23:59
azonenbergTwo webcams, two tubes, two objectives23:59
--- Wed Mar 20 201300:00

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