| J P Fitch wrote:
> There was indeed a paper at ICMC on interpolation, which was why it got added
> to mt to-do list.
This would be Roger Dannenberg's "Interpolation Error In Waveform
Table Lookup", I'm guessing. For those who don't have the procs,
the paper ought to be online some time next week.
> The conclusion seemed to be that quadratic interpolation
> was worthwhile, and cibic was better, but hardly worth the time.effort.
For the compiler and hardware used (Intel, so slow arithmetic),
there's a time/space tradeoff: for a desired SNR, higher-order
interpolation is slower but takes a smaller table. The relationship
between table size and SNR depends heavily on the signal's brightness.
Only went up to quadratic -- I'm very curious to see how cubic
compares, actually.
BTW, I'm not sure this work applies directly to situations where the
same data is being resampled repeatedly, as in a waveguide.
-- |