azonenberg changed the topic of #homecmos to: Homebrew 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/
wallacoloo has joined #homecmos
wallacoloo has quit [Ping timeout: 250 seconds]
wallacoloo has joined #homecmos
wallacoloo has quit [Quit: wallacoloo]
wallacoloo has joined #homecmos
wallacoloo has quit [Quit: wallacoloo]
wallacoloo has joined #homecmos
wallacoloo has quit [Read error: Connection reset by peer]
SpeedEvil has quit [Read error: Connection reset by peer]
SpeedEvil has joined #homecmos
SpeedEvil has quit [Ping timeout: 272 seconds]
SpeedEvil has joined #homecmos
SpeedEvil has quit [Quit: No Ping reply in 180 seconds.]
SpeedEvil has joined #homecmos
SpeedEvil has quit [Ping timeout: 244 seconds]
SpeedEvil has joined #homecmos
SpeedEvil has quit [Ping timeout: 240 seconds]
Bike has quit [Quit: slep]
SpeedEvil has joined #homecmos
mjgardes has quit [Remote host closed the connection]
barubaru has joined #homecmos
<barubaru> You really exist! Nice/
<barubaru> What is the limit in size for cheap optical photoresist+mask chip making?
<barubaru> I mean something like 500nm or 100 nm
<SpeedEvil> A demonstration thing is lots easier than an actually usable repeatable thing
<barubaru> Green light is somewhere about 500nm wavelength. UV can go to 300nm and probably there exist light sources with even shorter wavelength. And in blueray or cd they are using dots smaller than a wavelength somehow
<SpeedEvil> spin coating resist into quite thin layers and then exposing it isn't too hard
<barubaru> Right. Im talking about easy to repeat thing, bot a record result
<SpeedEvil> Getting the image onto the part is hard - especially if you want to do large areas.
<SpeedEvil> A microscope objective, for example, and a image source to reduce
<barubaru> Right, what is the reducing ratio? I mean it could be hard to make a mask if reducing ratio cant be big (1to100 for example)
<SpeedEvil> It's basically the ratio of focal lengths - largely
<SpeedEvil> 1mm:100mm = 1:100
<SpeedEvil> getting enough light in is another question
<barubaru> With bigger mask to chip size ratio you can get more light per chip area, right?
<SpeedEvil> more-or-less.
<SpeedEvil> Taking the simplest example - with one lens, 1mm from the IC, and 1mm in diameter, if you have a 100mm diameter mask 100mm away, you will get 100:1 reduction
<SpeedEvil> but - any light going through the 100mm mask that misses the 1mm diameter lens is lost.
<barubaru> i thought 2 lenses are required at least. One to 'compress' light, second to 'straighten' it back to parallel beams
<SpeedEvil> if you want the ligth to start out parallel, yes
<Sync> barubaru: if you want cheap masks 1µ is really cheap
<Sync> 250nm starts to get expensive
<barubaru> Sync right, thanks. What ratio of mask to chip diameter is cheap?
<Sync> ???
<Sync> that depends on your mask aligner
<Sync> if you are doing contact exposure or optical shrinking
<barubaru> what precision do printers have typically? Can they be used to make a mask?
<Sync> only for very coarse applications
<barubaru> Something even worse than 1 um?
<barubaru> I cant imagine how else mask can be produced for cheap
<Sync> it depends on what you call cheap
<barubaru> Within x10 price from market price, probably
<barubaru> Can you make mask using optical shrinking? So that first mask will be big and it will be used to make more precise mask which will be used to make chip?
<Sync> I get 4" masks on bk7 glass for I think 250-500€
<Sync> but it is not cost effective
<Sync> you could
<barubaru> wow, what are you making with this masks?
<barubaru> What is the diameter of a silicone substrate? this little blocks - each is a single chip, right? What do they do?
<barubaru> Looks awesome, btw
<Sync> it is a 4" wafer
alphaferret has joined #homecmos
<Sync> each block is a mosfet
<Sync> it is a test wafer for process evaluation purposes
alphaferret has quit [Quit: Leaving]
<barubaru> What do this graphs show? Ratio of dead to alive mosfets?
<Sync> no
<Sync> voltage vs current
<barubaru> So in total you got about 200 mosfets, what was the expenditures?
<X-Scale> Sync: that wafer looks amazing
<Sync> I got a few thousand mosfets, but they are not going into any leadframe or case
<Sync> idk what the process cost
<barubaru> 250-500 eur for masks, some for wafers, chemicals, positioning tools. give me some numbers =)
<SpeedEvil> barubaru: what is your aim?
<barubaru> SpeedEvil: just general understanding of how complex and expensive home cmos can be
<SpeedEvil> What do you mean by home - if you want to do it all at home, it gets very involved indeed
<Sync> I used a karl süss mask aligner, it cost about 500k€ new, I think
<Sync> oxides were grown in a wet oxygen furnance, which cost us 2M I think
<Sync> and the rest is relatively cheap, chemicals, possibly 500€ in total but I ran 10 wafers in that run
<barubaru> Omg. So its somewhere near 2.5M euro? Initial cost i mean
<barubaru> Thats completely out of my budget =)
<barubaru> But good to know..
<SpeedEvil> It's more-or-less undefinable - you can probably make a transistor that sort-of-works without too much effort.
<barubaru> Assuming 10k euro limit, what is possible to do?
<SpeedEvil> Making ICs, even very basic ICs which actually work, is a lot more work and requires enormously more attention to be paid to contamination
<Sync> well, you can't really expect more than single transistors that do something
* SpeedEvil tries to rem,ember who the girl is that did transistors.
<Sync> and with 10k€, it's still hard to do
<SpeedEvil> ah
<Sync> jeri ellsworth made some fets
<SpeedEvil> It depends what you can get also.
<Sync> but the way she did it is not really suitable for anything more fancy
<SpeedEvil> Can you for example get semiconductor grade chemicals in small quantities from a local friendly fab.
<SpeedEvil> Or are you having to order 10kg at $1000, when you need 1g.
<barubaru> So there is no way to make simple ICs under a 10k eur initial cost? Something as crude as 1um or even worse? Probably some higher limit to fight contaminant easier?
<Sync> 1µ at 10k is impossible
<barubaru> 10um? 100um?
<X-Scale> Not that it would solve everything (there would always be there the deposit issue), but wouldn't it help a lot to make all the lithography inside a vacum chamber ? There would be no small dust particles in suspension.
<X-Scale> vacuum*
<pie__> they would all be falling on the surface :D
<pie__> jk
<pie__> so i didnt read scroll yet, but re vacuum chamber, i was kinda thinking that if you put some kind of dust cover on whatever then removed it in vacuum in a way that made the dust fall everywhere non important...
<barubaru> Hepa filter may have really low dust level (i was reading about it recently) isn't it enough?
<X-Scale> pie__: exactly, that looks like a good idea
<Sync> X-Scale: XUV litho is under vacuum
<Sync> you can just use a flowbench
<Sync> and if you keep your particle size below your critical size they don't matter much
<pie__> X-Scale: i know nothing
<Sync> and they will just mess with your yield
<SpeedEvil> \Of course 'critical size' depends on ehwat teh contaminents are made from.
<SpeedEvil> Contaminents of SiO2 grit are basically ignorable.
<pie__> Sync: that is a sexy wafer.
<pie__> Sync: you work in industry? :D
<pie__> then again, what lithoŰd wafers arent sexy
<pie__> * litho'd
<Sync> SpeedEvil: it doesn't matter what they are made of
<Sync> if they are blocking your litho so that you get shorts, your device is dead
<Sync> pie__: not really
<SpeedEvil> Sync: Well, yes.
<pie__> Sync: academia?
<SpeedEvil> Sync: that assumes they don't add contaminents to the wafer that will diffuse into it and kill your die
<pie__> ok to the point, what do you do that you make wafers like that? :P
<Sync> that is unlikely SpeedEvil, if you get contaminants at a litho step you will remove them at the resist removal
<SpeedEvil> I was sort of assuming that contaminents were everywhere in your working space and could get to you at each step
<SpeedEvil> If you have a cleaner environment ...
<Sync> pie__: yes, academia
<pie__> actually...about using bluray/cd player for lithography....hmmm...that doesnt quite sound impossible?
<Sync> it is pretty impossible
<Sync> they are not made for that
<pie__> oh heh, figures zoom makes it look hella big :P<Sync> it is a 4" wafer
<pie__> Sync: cool
<pie__> Sync: what do you mean?
<pie__> well i suppose litho isnt point an shoot, but could it be cannibalized?
<pie__> on an unrelated note, damn this guy has a lot of papers http://lithoguru.com/scientist/papers.html
<Sync> I don't think using a cd drive will be accurate enough to be overly useful
<pie__> well ok idk what the data density of a cd drive is
<pie__> but it sounds usable for the primitiveness of doing this at home?
<barubaru> What is a precision for cd drive, printers, blue ray?
<pie__> printers is probably bleh.
<Sync> I'd probably do contact litho
<Sync> nothing to worry about
<Sync> or direct laser
<pie__> im a noob so i wouldnt know
<pie__> so apparently cd is "Bit size = 52 micro inches = 1.3 micro meters "
<pie__> dvd is Bit size = 19.5 micro inches = 0.49 micro meters
<pie__> wtf, who the hell uses micro inches?
<Sync> people
<pie__> :/
<barubaru> pie__: in some distant places people are still didn't change to SI units yet
<pie__> :PP
<barubaru> With cd drive - can you reprogram it to draw arbitrary image, not just separated dots?
<lain> with some effort, probably
<lain> for example those CD drives where you can "burn" the "album art" on the top of the CD
<lain> might be able to repurpose one of those more easily?
<barubaru> Wow, i didnt see such feature
<lain> though, I have no idea what sort of resolution you get with the drive in that mode, as opposed to the usual mode
<lain> "lightscribe" it was called
<SpeedEvil> It relies on a preexisting formatted top surface
<SpeedEvil> it cannot mark arbitrary surfaces
<lain> ahh
<lain> so probably not much easier than just hacking any ol' drive
<lain> this may be of interest: https://github.com/scanlime/coastermelt
<barubaru> If empty spots can be filled by changing the firmware then it looks like a good try. Even if just 1 type of cd-rom will able to do it, we could just buy that particular type
Bike has joined #homecmos
<pie__> well youd probably need to hack the drivers no mattrer what
<pie__> barring some kind of physical incompatibility whats to stop us from litho in normal mode?
<pie__> though i have a guess that the motor is actually probably not suitable for this...
<pie__> because i doubt it can spin slowly, and even if it can it will be inaccurate
<pie__> idk how writing works
<pie__> the question i guess is if litho would be possible at speed
<pie__> Sync: how is litho at several hundred to several thousand rpm? :P
<pie__> SpeedEvil: isnt litho done on a preexisting formatted top surface? :P
<pie__> imagine all the dust and stuff from the mechanical stuff though with all that spinning :X
<pie__> anyone actually want to make srs progress on homelitho?
<pie__> i know im an idiot for even considering it
<Sync> the easiest is direct laser writing
<pie__> Sync: i think thats basically what im saying with the dvd drive stuff?
<pie__> except the x-y translation thing is a lot more fun :x
<Sync> with a cartesian stage it is very easy and relatively cheap
<pie__> ok thats what theyre called, icouldnt remember :P
<barubaru> pie__: actually i think apinnong will help to keep disk flat and at steady speed, low speed rotation will mean vibrations, jerks, improper positioning
<pie__> eh. i can see it flapping even at high speed. it sounds liek it would depend a lot on physical conditions
<pie__> hell if i know
<pie__> maybe youre right.
<pie__> but irregularity in the material is probably likely anyway
* pie__ suddenly remembers we are talking about silicon wafers and not actual dvds
<pie__> uh...
<pie__> yeah thats not gonna work out of the box hardwarewise
<barubaru> I wonder can silicone tolerant such rpm
<Sync> cds use a clever way to track the angular position
<barubaru> Is it 50k rpm? Silicone may shatter
<Sync> cds do not spin that fast
<barubaru> Sync: how do they do it?
<pie__> silicone is not silicon
<pie__> Sync: i was guessing there would be something like that
<pie__> and anyway cds have loads of error correction codes i think
<lain> yes
<pie__> by CDs i mean dvd and all the other stuf
<Sync> they use the index groove to first get absolute angular information
<lain> it's an interleaved reed-solomon ECC, and crc for fast error detection
<pie__> what lain said :P
<lain> interleaved and mixed around to deal with the common types of scratches they encounter
<lain> even audio CDs do that :3
<Sync> the easiest is to just use a cartesian stage
<pie__> any off the shelf consumer things have good cartesian stages? :P :D
<Sync> no
NullMoogleCable has quit [Ping timeout: 250 seconds]
NullMoogleCable has joined #homecmos