relative polytone generator. excessive use of precision rectification op amp circuits. wont be perfect due to wave shape but should "work". (f, oct, oct+p5, 2nd oct, etc...).
sine wave through 4 folds. dramatization of new "high frequency content".
f(x) = sin(2x)
g(x) = 2|f(x)| - 1
j(x) = 2|g(x)| - 1
k(x) = 2|j(x)| - 1, etc...
|f(x)| implemented through rectification. -1 or -c (a constant) represents a DC offset, without which running stages in series would be ineffective.
it isn't a strict doubling of frequency, but an approximation of. the shape of f(x) will change unless it is a perfect triangle. but that might be the charm of it. ie the harmonics generated will be waveform dependant, although there should be general trends.
using FFT approximation to see spectrum created...
first fold shows strong second harmonic, subsequent folds show a reasonable fourth then a fair distribution of upper even harmonics.
from this point high pass filters could be used with a sort of subtractive synthesis mentality to move the distribution more in favor of the higher ordered harmonics. not sure if 6 stages are really necessary. probably 4 would be interesting enough.