They have apparently improved a lot in the best decade. I mean, ultimately it's just about finding the right algorithm; there are no unreproducible sounds coming out of a piano.
Lots of things that are fully described by relatively elementary physics are still impossible to model in real-time even with the most powerful computers, let alone the $5 microcontroller in an electric piano. You're right that there's no obvious physical reason why a 'perfect' electric piano couldn't exist, but it would probably be way more expensive than just building a mechanical piano which consists entirely of 19th century technology.
Why are you exaggerating the requirements so much? You don’t need it to be perfect and there is no issue with putting a high end SoC in a high end e-piano.
You set the bar at algorithmically reproducing every last sound that comes out of an analog piano. That's a hard problem to solve for the very small number of people who would care enough to purchase such a device when there's a perfectly good analog alternative that's been on the market since before the telephone was invented.
You may think I'm exaggerating the requirements, but I think you're underestimating the difficulty of solving what is at best a very hard physics problem in real time with imperceptible delay as a human carries out a sequence of physical manipulations that has taken them years or decades to learn. My $1,000 phone with the latest, greatest SoC can't open a PDF with a noticeable delay.
If your proposition is a middling approximation of an analog piano, you can already buy that at Costco for $500. Steinway and Bosendorfer don't seem particularly bothered about it.
reply