piece wise linear approximation of x**n

[Martin Czech]

Has anybody used f(x)=x**n as waveshaping function (of course , fixed n)?
The use of Tchebycheff polynoms could give the first n harmonics. Of
course, doing real multiplication is expensive.  So what about PWL
approximation up to x**7 or so?  The output should always stay in the
normal range, so

f(x)=(x**n)/(Vmax**(n-1)) is really to be computed.

Has anybody tryed this?

Maybe one can use a FWR, this saves half of the network.  After
computation of f(|x|) the sign could be restored.  Of course, this could
possibly give some glitches.

Any ideas, comments ?

m.c.