Reply by camb...@yahoo.co.uk July 26, 20102010-07-26
Hi all,

I'm writing a time-varying equalisation algorithm using an overlap-add STFT with a Hann window at 50% overlap.

It will be taking 'real-world' audio signals, however before I began EQ testing, I tested everything straight-through ('flat' zero EQ). I tried various basic signals and by pure chance the first one I tried was a sine wave that was periodic with the window size. Due the 50% overlap, when the two adjacent time frames were added onto the currently analysed frame they were pi out of phase with it. Resulting in the output sine wave being amplitude modulated by a cosine wave with a wavelength of the window.

In hindsight, this outcome is obvious, so my question is: What is the best way to choose a window size when the input signal is at the user's discretion (i.e. completely indeterminate...)?

Also as a new kid to this signal processing malarkey, is an overlap-add STFT algorithm best for time-varying equalisation with (possibly) very complex EQ graphs?

Thanks in advance!
Cam