Tue Sep 4 22:55:11 CEST 2012
Plotting the approximation error
What to plot?
Start with bit depth, e.g. d=16 bits.
This gives p=2^d
For each wave form assocated to q:1->2^d-1, plot the FFT on a log/log plot.
This gives a density plot that gives an idea of the distribution of
the error assuming uniform input amplitude distribution.
In first approximation maybe a density plot isn't necessary. Just the
average over all wave forms might be good enough? Though it would be
nice to also have variance and shape of distribution.