I made a SAW filter at university. Mostly in make-shift labs and with old gear our professor collected from industry donations of scraped stuff. We did everything from design to preparing the silicon, lithography, etching and even bonding the chip so we ended up with a nice working filter in a package that is indistinguishable from a professionally made one. I still have it somewhere.
The nice thing about a SAW filter is that you have to basically do all the steps you do for an integrated circuit but many steps are a little simpler. The design is pretty well doable with undergrad math (fourier transform). The lithography can be relatively low-res and you can test it pretty easily with standard electronics lab equipment.
Something I have been pondering for a while: Is there a demand from hobbyists for custom, low-grade home-grown ICs to such a degree that somebody could make side money from it? Sam Zeloof is now going "all in", but we don't know how it will turn out, and it definitely seems more serious than low-grade hobbyist ICs.
I know that I would be willing to pay for such a service, and doing it myself is definitely out of reach. OTOH this is really a niche even among hobbyists, and hard enough that making a 2-NAND in the $10-$100 range could be considered a huge success already.
(The other special appeal to hobbyists, other than just doing it, would be a shorter turnaround time than getting your place on a multi-project wafer and waiting 1-2 years to know if you made a mistake)
I was thinking of hobbyists using homegrown ICs "because they can", not just for the final product, in the same way that people build computers purely out of 74-series ICs or discrete transistors. In fact, if the final product is all you want, using an FPGA is not worth the effort most of the times.
That said, there are certain ICs which are simply too funky for FPGAs to manage, for example the SID chip, which had a number of analogue components on the IC to enhance the sounds it produced. Having a hobbyist method for producing something like that would be cool too.
I don't know how big the market is, but there are definitely musicians out there who would love to be able to buy clones of a wide variety of out-of-production analogue ICs to restore vintage equipment, build kits that depend on those parts, and so on. VCFs, VCOs, etc.
I know the response of the "discrete component aesthetic guy" is going to be "eww this is not using through hole parts, disgusting".
I personally would find it interesting if you could somehow assemble your own ICs from a library of chiplets. The home made lithography machine would only have to build the interconnect, which has lower resolution requirements.
Yes, most digital use-cases are covered by FPGAs. But I think there's still a potential market for custom mixed-signal ASICs, and maybe also for defense and space applications where unusual environmental requirements matter more than cost.
Edit: whoops, thought this was ask HN due to the question mark. Pretty cool stuff.
Sam Zellof (YouTube) is probably your best answer. My project civboot.org is basically an attempt to answer a similar question and my answer is: reduce the complexity of technology and then build the simplified but effective systems.
It's not going to be "at home" (the tech is way too complicated for that) but it could be a university or similar (hopefully).
This is one of the things where the details of the "booting civilization" situation matter a LOT. If you can find a Blu-ray player, it'd be much easier for you to make something like Breaking Taps did- no need for a projector or computer, you could even do this by hand using an optical relay.
However it's VERY hard to make an appropriate UV laser[1], and it's very hard to make a photoresist that works in infrared where a CO2 laser is easy to make. In that case you have to use a projector.
For a while I've wondered how feasible it would be to use an electron microscope to just ablate everything. Shockingly easy to make for such an excellent instrument.
[1]: Ironically the easiest type of laser to make is a UV nitrogen laser, but it's extreme power and short (nanosecond) pulse time mean you can't use it for this: http://www.repairfaq.org/sam/lasercn2.htm#cn2cn20
> For a while I've wondered how feasible it would be to use an electron microscope to just ablate everything.
Given the fact that we are more or less up to the limits of optical lithography, the future is quite probably going to be figuring out how to scale direct e-beam litho up to mass production rates.
.. the 6502 was initially manufactured at 8 micron. I'm sure there are all kinds of differences between being able to produce some features at 7 microns and being able to reliably handle ~3-4k gates at 8 micron (and I'd love it if someone were to explain how far - or close - he might be from that given this) but it's still pretty amazing it's even in the ballpark.
Making at home some bipolar integrated circuits, like operational amplifiers, might be feasible if one could obtain from somewhere the epitaxial silicon wafers needed to start the process, because making some diffusion furnaces and wet-etching stations, together with laser photolithography as described here and together with metalization using electroless nickel and galvanic growing of chromium and silver would be feasible.
On the other hand, making something like an NMOS 6502 CPU is orders of magnitude more difficult, because its fabrication requires a cleanliness very hard to achieve without spending huge amounts of money, and for an acceptable yield more sophisticated fabrication methods are needed, e.g. ion implantation for doping and chemical-vapor deposition for some layers, e.g. the gate, which require very big and expensive equipment.
Thanks. I think "acceptable yield" for a hobby project to try to make it happen would be very low (as in, just managing one after a bunch of tries might be cool in itself), but I guess it'll still be infeasible...
By "acceptable yield" I mean having the hope of obtaining one working CPU after a large number of failures, i.e. I was not referring to "acceptable yield" for a commercial operation.
Making NMOS ICs has been impossible for many years even for the biggest companies regardless how much money they have spent (because all their attempts were contaminated with various impurities that ruined the MOS transistor characteristics).
While now it is easier because we know the tricks, that is not helpful for a small-scale operation, because the techniques required for making good MOS transistors are very expensive. Besides a clean room, a large number of chemicals with extreme purity are needed.
The bipolar ICs, like the 7400 series or the analog ICs, were much less demanding.
If I had the equipment to make ICs at home, it probably would be useless to make anything worthwhile in digital, but I could find a thousand use cases in analog where an equivalent either doesn't exist, or costs a fortune or had been obsoleted ages ago.
I have worked in a semiconductor plant, and I do not remember any carcinogenic diffusion-related chemicals.
Like in any chemical plant, there were various organic solvents that could be carcinogenic, but they were not directly related to diffusion processes.
On the other hand, there were many chemical substances used for diffusion that were extremely toxic, like various arsenic compounds.
The greatest hazard in improvised IC fabrication would be from the hydrofluoric acid (as a mixture with ammonia) that is needed to etch the silicon dioxide, after photolithography and before diffusion.
The EPA classifies inorganic arsenic as a “human carcinogen,” based on evidence in human studies of links to lung, bladder, kidney, skin, and liver cancers.
The nice thing about a SAW filter is that you have to basically do all the steps you do for an integrated circuit but many steps are a little simpler. The design is pretty well doable with undergrad math (fourier transform). The lithography can be relatively low-res and you can test it pretty easily with standard electronics lab equipment.