OSDev.org

Hi, I just was wondering what your thoughts were on HomeCMOS (https://code.google.com/p/homecmos/).

Seems interesting, though I think the most expensive cost in making ICs is the machinery, which would still be needed.

I'm not sure but would guess it would be cheap equipment because it's meant for hobbyists.

Looking on their website, I find supplies to create the circuits (chemicals), but do not find much information on buying/building the actual machinery.

Ya, they haven't even created a working CMOS chip yet. They are in very early stages but they have been able to etch metal at a 20μm half-pitch. I just was wondering if you though it was feasible for hobbyist to be able to make integrated circuits or for integrated circuit printers for the home product printing.

sds2017 wrote:Ya, they haven't even created a working CMOS chip yet. They are in very early stages but they have been able to etch metal at a 20μm half-pitch. I just was wondering if you though it was feasible for hobbyist to be able to make integrated circuits or for integrated circuit printers for the home product printing.

It depends on what you'd define as feasible.
If one is skilled enough, it's possible for one to produce ICs for less than $20k (relatively, that's not very much).
If the hobbyist is super rich, they can easily spend multimillions on the equipment, and produce ICs at home (Though, if they're that rich, I don't see why they'd do it themselves)
Therefore, it is feasible if you're extremely skilled (<1% of people) or super rich (not sure percentage)

m12 wrote:If one is skilled enough, it's possible for one to produce ICs for less than $20k (relatively, that's not very much).

Actually, for the same price (or even less) you may order very high level technology IC of your own design without making risky and outdated experiments at home. This process is known as Multi-project wafer shuttle.

Yoda wrote:

m12 wrote:If one is skilled enough, it's possible for one to produce ICs for less than $20k (relatively, that's not very much).

Actually, for the same price (or even less) you may order very high level technology IC of your own design without making risky and outdated experiments at home. This process is known as Multi-project wafer shuttle.

I was specifically talking about making them by oneself, but your point is valid.

As far as I'm aware, there are two people in the world skilled enough to make a modern day processor with household items. Certainly less than 1% of the population.

If one is skilled enough, it's possible for one to produce ICs for less than $20k (relatively, that's not very much).

Actually, for the same price (or even less) you may order very high level technology IC of your own design without making risky and outdated experiments at home. This process is known as Multi-project wafer shuttle.

Only $20k, I've only seen designs for around $2 million although, they were microcontrollers/microprocessors. I know based on skill and money you can build just about anything, but what I think they mean by a hobbyist is someone who has a very limited budget and not a Ph.D in the subject.

sds2017 wrote:Only $20k, I've only seen designs for around $2 million although, they were microcontrollers/microprocessors.

There are following reasons for that.
1. They are made on bleeding edge technology level. But the cost is not linear for that.
2. Most modern CPUs have a very large die area. Doubling linear size of die will give quadratic increase in area, so resulting in quadratic increase in cost.
3. The probability of errors for large scale projects is very high, so design of top-level CPU requires dozens of iterations on multi-project wafers before you'll get commercial product.
4. There is physical threshold. If you design your core on 90nm technology (which is pretty good) or worse you don't need to pay much attention to physical processes such as gate leakages, often you just design logic and get working die. But if you need to get below 90nm, you need to account leakages and other effects in design.

sds2017 wrote:I know based on skill and money you can build just about anything, but what I think they mean by a hobbyist is someone who has a very limited budget and not a Ph.D in the subject.

That's not true. To get your core on multi-project wafer all that you need is Verilog/VHDL description that will be automatically (with specialized software) transferred to layout using standard libraries for specific fab. But making your own layout will force you not only to design logic, but to learn physics and design layout manually in addition.
Ph.D is not needed to design logic (and even to test it using FPGA before making die). But making layout and learning physical processes may require Ph.D.

BTW, this course @Coursera is "the must" for those who want to design layout @home: https://www.coursera.org/#course/vlsicad