DIY chip fabrication?

[jhu]

Came across how to etch your own circuit board. Seems straightforward and cheap to do. Let's go further and fab a CPU. What's the smallest node size that can be fabbed in your backyard for ~$10k? Can't imagine 1 um equipment being too expensive.

 

[NTMBK]

Personally, I can manufacture 10nm chips in my back yard. Though that crashed space ship may have given me a bit of an unfair advantage.

 

[Phynaz]

I'm betting IDC has fab equipment in his garage

 

[Mand]

You will not be able to get even just the optics for 1 micron photolithography for 10k. Not even remotely close.

 

[KingFatty]

What is the highest resolution laser printer you have access to? You could print the circuit layout onto transparent overhead projector sheet. So do that instead of the guide's suggestion to use a fine-tip sharpie marker to draw the circuit layout.

Maybe you can print a single NAND logic gate on 8.5 x 11 sheet, I dunno?

 

[TuxDave]

Do you REALLY want to be handling hydrofluoric acid? I would stick to using FPGAs.

 

[Homeles]

Do you REALLY want to be handling hydrofluoric acid?

All the more exciting.

 

[zir_blazer]

Manufacturing integrated circuits is impossible. Designing them, however, is not. You can potentially program a FPGA with your design or request someone to manufacture them (Which may be impossible on a big foundry, but I suppose there should be high end electronics laboratories that may produce a small batch for prototype). May want to check Open Source Hardware and all links related to that.

 

[WhoBeDaPlaya]

Not unless your lamba is 1mm :biggrin:
Semicon fab/processing is a serious PITA. I've had enough of wearing a bunny suit for hours on end in yellow lighting fabbing silicon photonics chips

 

[Idontcare]

Came across how to etch your own circuit board. Seems straightforward and cheap to do. Let's go further and fab a CPU. What's the smallest node size that can be fabbed in your backyard for ~$10k? Can't imagine 1 um equipment being too expensive.

Fabbing a chip in the traditional sense, as the industry has been doing it since the 60's, is not something that can be done on the cheap or in conventional facilities.

However, I do expect the 3D printing industry to mature to the point (and beyond it) in the next decade or so such that you could be looking at buying yourself a used makerbot (or something akin) for $10k and with that you could print out your own CPU if you so wanted; albeit with presumably silly old-skool design rules of say 3um or somesuch.

 

[jhu]

Fabbing a chip in the traditional sense, as the industry has been doing it since the 60's, is not something that can be done on the cheap or in conventional facilities.

However, I do expect the 3D printing industry to mature to the point (and beyond it) in the next decade or so such that you could be looking at buying yourself a used makerbot (or something akin) for $10k and with that you could print out your own CPU if you so wanted; albeit with presumably silly old-skool design rules of say 3um or somesuch.

I just think the guy who made his own computer could have turned it up to 11 by fabbing his own chips.

 

[Wall Street]

You can't make transistors reliably by hand, the bipolar NPN and PNP junctions need to be way to small. On top of that, transistors are made of hard to source materials like gallium. What you can do it purchase discrete transistor or opamps, like the type used in audio equipment, and mount those on a circuit board to make a small calculator without integrated circuits. It would be big like when computers had vacuum tubes.

 

[pm]

Manufacturing integrated circuits is impossible.

Not impossible by a long shot... it would merely require some serious determination and probably a fair bit of trial and error. But not even a lot of money. In college, my semiconductor class made homebrew-style MOSFETs. We used a camera to take a photo of a "mask" which was a paper covered in sharpie that we hung across the room. So we had a paper, we drew with black sharpie on it and then hung it about 40 feet away and took a photo of it using a film-based camera, then cut out the negative and that became the mask. Then we used an old spin art spinner and a ketchup bottle to spray on photo resist (yes, something like this), spun it thin, developed it with a UV bulb, etched off the photo resist, grew oxide in a kiln that had an oxygen canister, and we did a boron implant using the same method but a boron tank. We had a mechanical stopwatch to figure out the dopant density. Then we etched and sharpie'd and repeated. As pointed out, there were some nasty chemicals involved - hydroflouric acid being the worst but there were some others that needed access to the chemistry labs fume hood.

In the end we did manufacture something like 60micron FETs on a ~3" wafer to make an integrated circuit. Our yield was appalling - as I recall about 1 in 2 of my FETs worked. But we had enough redundancy that one of my circuits mostly worked ok.

You could absolutely do it at home - nothing we did required a university's lab... but you'd have to be pretty driven to buy everything. The method we used could be done with a budget under $1k. The biggest source of yield loss was the screwy kiln system we had... for $10k you could do better. But I'm not sure how much better... but an airtight system would help a lot and a system to measure boron levels and some equipment to measure the oxide thickness (we estimated it with math)... you could shrink the whole thing down and I would think sub-10 micron would be possible with the film method - though you'd probably need a medium format camera to get the resolution up.... Still $10k would be a heck of a budget for the McGuyver-like system we had going.

Up until the point that I took this class, I'd been an applied physics major (not an EE) but I switched over to EE and specifically VLSI after that class. The class after that we used an external foundry (can't remember the name except that they were in France) and CAD software to design a microprocessor and have it manufactured at the foundry... which was even more fun than making integrated circuitry with a sharpie and a spin-art toy.

 

[dave_the_nerd]

I just think the guy who made his own computer could have turned it up to 11 by fabbing his own chips.

Okay, that's the coolest thing I've seen all day.

 

[WhoBeDaPlaya]

Up until the point that I took this class, I'd been an applied physics major (not an EE) but I switched over to EE and specifically VLSI after that class. The class after that we used an external foundry (can't remember the name except that they were in France) and CAD software to design a microprocessor and have it manufactured at the foundry... which was even more fun than making integrated circuitry with a sharpie and a spin-art toy.

I feel lucky that I had ready access to the MOSIS 0.6um and TSMC 0.18um processes in school. It's nice to the able to actually test what you designed (positive-feedback op-amp, ADC, DAC, SERDES, etc.)

Now that I think about it, I do physical design with 0.18um processes (internal, TSMC, etc.) on a daily basis. Man this sh*t is old

 

[SOFTengCOMPelec]

Came across how to etch your own circuit board. Seems straightforward and cheap to do. Let's go further and fab a CPU. What's the smallest node size that can be fabbed in your backyard for ~$10k? Can't imagine 1 um equipment being too expensive.

One way you can partially achieve this (which I might be doing myself, as a hobby), is that you can buy integrated circuits, which have no circuit pattern inside as such, just a quantity of plain, relatively unwired integrated circuit transistors.
They have something like 5 or 6 transistors, slightly wired together, but otherwise bringing the transistor connections to the pins of the IC.
E.g. In a 16 pin dil package.
The advantage over discrete transistors, is that they are real IC transistors, with well matched characteristics, thermally matched/connected, just like on a real IC.
Some types of circuit, e.g. Differential, really benefit from the close matching and thermal linking, of the transistors.

Some real life integrated circuits, were actually originally "made" at home.
"Made" = The mask patterns and maybe the discrete transistor prototype of the popular NE555 timer integrated circuit, was designed at home.
The designer then went to the chip plant(s) with it to persuade them to make it. (Or he was paid to do it, I can't remember).

Another solution, is that at home, there is a type of integrated circuit, you CAN MAKE at home, although strictly speaking, it is not a proper IC.

They are called hybrids (I think).
Think of them as tiny IC sized custom PCB board, etched at home, with tiny discrete surface mounted transistors, resistors and capacitors etc, soldered to it.
You then encase the finished "tile", with perhaps a line of SIL pins coming out of it (a bit like half a dil IC), in epoxy or something.

The finished article can be rather small, and not much bigger than a real IC.
It can also have "glue logic", op-amps and other stuff, to give it lots of functionality.

 

[videogames101]

as an academic exercise it's a cool idea, but an FPGA will give you way better results for actually doing anything

 

[Mark R]

I just think the guy who made his own computer could have turned it up to 11 by fabbing his own chips.

My electronics teacher in High school had done the same thing in the 70s. Except, in that case, he had built a computer to do the schools accounting and record keeping.

He showed some photos in class - trays full of perf-board studded with 74 TTL logic, and beautifully wire-wrapped interconnects.

He'd managed to source some ready built tape machines - but had built a custom interface between them and the computer. He'd written the OS and software himself.

The school had used it as their production machine for close to 10 years, before migrating to IBM PC hardware.

These days, some things are best sub-contracted to someone who does this stuff every day. I've done a few hand-made PCBs and they're a PITA. Drilling, applying resist and etching is bad enough. When it comes to double sided, it gets ridiculous - it can be done, but it's just not worth the effort. Much easier to draw up a design, and send it off to Gold Phoenix or similar setup. Then just sit back and wait for the boards to arrive.

IC manufacture gets loads more expensive. No idea what the price is or what the price-break levels are as I've never done this. However, there seems to be a large market for $1000 FPGAs which suggests that ASIC costs are substantially higher than this unless you have a mass-market product.

 

[SOFTengCOMPelec]

These days, some things are best sub-contracted to someone who does this stuff every day. I've done a few hand-made PCBs and they're a PITA. Drilling, applying resist and etching is bad enough. When it comes to double sided, it gets ridiculous - it can be done, but it's just not worth the effort. Much easier to draw up a design, and send it off to Gold Phoenix or similar setup. Then just sit back and wait for the boards to arrive.

The advantage of making your own PCBs (I think) is that you can achieve very fast turn-around times. This can be very important if you are a professional Electronics engineer or electronics business, IF you regularly get work packages, with very strict/rigid (project) timing plans.

Example
You are testing a fully assembled PCB board today, and it must be finished, ready for tomorrow, and you have just discovered that you used the wrong (old) pin-out on the main IC.
If you have to wait for PCBs to be made externally, and then be delivered back to you by post/courier, then you won't meet the vital deadline.

Missing deadlines can be expensive and/or lose you future business.

So by making your own, in a matter of hours (approx), you can "fix it".

But I agree that about 90% of the time you can do some kind of quick fix to the PCB, such as wire link(s), etc.

Once you start needing solder resist masks, silk screen printing (component outlines/labels), through hole plating (i.e. real vias), multi-layer, funny shape cut-outs, etc etc, it really needs to be made externally.
They can sometimes also offer special coatings, and other potentially useful services.

 

[TuxDave]

In the end we did manufacture something like 60micron FETs on a ~3" wafer to make an integrated circuit. Our yield was appalling - as I recall about 1 in 2 of my FETs worked. But we had enough redundancy that one of my circuits mostly worked ok.

A working FET AND a circuit??? Impressive. I was happy to find a working diode among the scorched parts on my wafer.

 

[SOFTengCOMPelec]

Actually there seems to be at least some stuff, about how to make your own Integrated Circuits at home, yourself (I'm surprised, but I guess you learn something new each day), on the internet.

http://hackaday.com/2012/07/09/make-your-own-integrated-circuits-at-home/

http://hackaday.com/2010/03/10/jeri-makes-integrated-circuits/

I can almost 100% guarantee, that you have never seen or heard about this website before