I think physical modeling is a poor choice. The advantage of subtractive synthesis is that each of its elements (oscillators, filters, amplifiers, modulation,

I think physical modeling is a poor choice. The advantage of subtractive synthesis is that each of its elements (oscillators, filters, amplifiers, modulation, effects) has an impact on the sound more or less independent of the others, and so their parameters can be understood and modified independently of the others. In other techniques, notably FM and physical modeling, the parameters and elements are very tightly coupled and nonlinear: the particular effect of parameter X changes depending on the setting of parameter Y. As a result you can't study the pieces separately, but instead must understand the dynamics of the entire system as a global unit. This is what made FM nearly impossible for musicians to wrap their heads around: and it also plagues physical modeling. The other problem with physical modeling is that there are many physical modeling methods and architectures, and learning one doesn't help all that much with regard to the others.