## Re: Which Fourier transform algorithms are lossless?

#####
Reply #3 –

to be _really_ lossless, it would have to store results as infinite precision fractional numbers, which isn't possible.

However, if you also encode/memoize the original amplitude resolution aka bit depth, you can set a margin of error to be small enough so that when you round values after inverse transform back to original resolution, it would exactly match original values. In that sense, yes, why not.