i'm thinking about how to simplify this, when seeing a failure has such huge feedback
a really clear thing would be setting the lower frequency regions of the original fft to 0. then, simply performing the ifft on the result, will make something similar to the original signal. data will be missing. this original signal can be upsampled by simply processing the sines and cosines at higher density but a bunch of information will be lost, because it has been turned into moireish stuff at the lower frequencies. ideally, when we took the fft, we would do it in a way that assumed the actual sampling we were performing ...
