>I wonder if anyone has applied "chaos" theory to soundboards. Has anyone >"proved" it is impossible to predict sound board responses and duplicate >them electronically? > >Regards, >Don Rose, B.Mus., A.M.U.S., A.MUS., R.M.T., R.P.T. Hi Don, Chaos, as the dictionary defines it, is disorder. I personally don't believe it. Chaos, as the science views it, is an end result of more minute but understandable discrete cause and effect interactions than we can keep track of. I find this definition much more comforting. Ignorance holds more hope of correction than does being victimized by disorder. With a complete knowledge of the characteristics of every particle and force in a system, there are no surprises and the status of the system can be worked out by crunching all the numbers from the interactions from the point of initialization of the system - if you have the computing power. A piano is an analog computer that works all these nearly infinite details out in real time, including I/O. The apparent chaos part comes in when you try to predict what's going to happen from an incomplete data set, or without actually running the cumulative interaction progression. Since it's beyond our capacity to measure, map, and track every sub atomic particle in the piano to build the simulation, a bunch of the more minute effects would have to be clumped into modules simulating low level approximation sums of even lower level interactions. If we had an idea what those low level models should simulate, we'd have a place to start. Then, since we would still have too many low level models of lower level interactions than we could handle, we would have to group these "level two" interactions into more coarse, but still statistically valid level three models. At some point, some component of one level one model will affect the actions of a component in another level two model as the strings, bridge, soundboard etcetcetc interact, and it will almost immediately be too complex to keep track of yet again. We need some cheap high level close approximation tricks to put a lot of low level interaction accumulations into few enough containers to be manageable by the overall simulation. When we do this, we lose resolution detail. The granularity gets coarser as we deal with simulations of cumulative interactions of simulations until the final simulation sounds like the real thing about like "Mork and Mindy" resembles your day to day family life - never mind mine. Building simulations from the bottom up hasn't been real productive because the really low level interactions aren't particularly measurable and have to be guessed at by observing higher level accumulated effects. The top down approach hasn't worked all that well either because you can throw a whole lot of random noises together without getting lucky and having the result sound like the piano of your choice. Starting with the closest piano-like synthesis obtained so far and continuing to tweak and add to it seems to be producing the best results. After all, the electronic piano doesn't have to duplicate the soundboard/string/etc responses of a "real" piano (if anyone ever agrees on what that may be), it just has to directly produce a final sound that is close enough to the real thing to fool the listener. That should be some orders of magnitude simpler. Incidentally, it occurs to me that we regularly spend considerable time trying to simulate "real" piano sound with the available instrument of too coarse a resolution for optimal results. Some of our efforts aren't all that convincing either, truth be known. Back to work. Gotta go try to install simulated piano sound in an old spinet. Wish me luck, it's chaos out there. Ron N
This PTG archive page provided courtesy of Moy Piano Service, LLC