Perhaps I needed to be more specific. I am aware that they are currently producing FPGAs that have enough power to easily emulate 16-bit, and perhaps even 32-bit systems. It would have been more accurate for me to say that I am waiting for them to be affordable enough. I am not a low-level computer engineer. I have no intention of building one of these things for myself. I would have to wait until someone else turns something like this into a commercial product. (as has been happening lately)
But in order for someone to turn something like this into a commercial product, the technology has to be cheap enough to make the final product feasible in the current market. And that's still going to take a little time. Thankfully, time is the only element that is necessary. Tech constantly decreases in cost. The higher-end FPGAs of today will be available for peanuts in a year or two.
Sorry I have to nit pick this because it drives me nuts.
FPGAs are not emulation. You are recreating the original circuit pin per pin and cycle per cycle and executing that logic native. An FPGA *becomes* the target circuit, it doesn't translate or emulate anything.
Emulating is a microprocessor running software to interpret and respond to the opcodes of another architecture's software real time.
(Though some would argue that FPGAs are all LUTs and not real logic gates anyway, but its still a concurrent logic circuit with logic gates implementing those LUTs and not a microprocessor instruction interpreter loop).
It's not really a cost thing for the end product. It's more man hours and reverse engineering proprietary undocumented decades old ASICs which mostly involves trial and error. Emulators are a good start at understanding the API from a programmer's perspective (eg: we know how to emulate SFX, SA1, and write custom software that makes use of the same behavior). But reverse engineering the insides of that ASIC and translating that to synthesize-able VHDL is a time consuming task and there aren't a whole lot of people who care that much.
Some even go as far as decapping and acid etching old chips and scanning them in microscopes to try to get an exact 100% accurate circuit operation. But even with old micrometer hmos processes you're still talking expressing something like "multiply 2 vectors" as 10s of thousands of transistors.
I guess it IS a cost thing so much as few people are interested or have access to the costly lab equipment necessary to reverse engineer obsolete decades old toys and develop these things. Once reverse engineered and shared though they can be mass produced in $5 FPGAs. It's tons more effort to reverse engineer someone else's ASIC than to just make your own from scratch.
8/16 bit systems like SNES, TurboGraphics, and Sega CD are fairly simple. It's when you start getting to PS1/Saturn era when complexity, function, integration, density, and pin counts started exploding with high density VLSI by the mid 90s.