Are conical segments useful for vocal-tract simulation?

A time-domain model for nonuniform tubes is considered which is of similar simplicity as the Kelly–Lochbaum model but based on conical segments. Using the bilinear transform or wave digital filters for representing the reflections at the segment boundaries, stability is proved and numerical instabilities can be overcome. The transfer functions are modeled better than with cylindrical segments but about equally well as with half-length cylinders, requiring about the same computational expense. Thus, in most cases there is no real advantage in using cones.