On Fri, Dec 11, 2009 at 5:52 PM, fabio cherchi <fabio.cherchi at yahoo.it> wrote: > Hello, > > my question is very simple. > I realized a hardware version of the kiss fft for the AEC Speex in the > pjsip. > It can handle 256 samples per frame while the whole application (and audio > driver) actually is working with 160 samples per frame. > > I'm evaluating 2 options: > > 1) to keep the 160 samples per frame and use the zero-padding technique to > fill the remaining empty frame. > Question: does anybody know what's the effect of the zero-padding in the > inverse fft? > I'm no FFT master, but I don't think the effect is good! > 2) to scale up all the system to handle 256 samples per frame. > Question: is the pjmedia suitable for 256 samples per frame instead of 160? > That should be okay, just open both the sound device and the conference bridge to use 256 samples per frame, and let anything else unchanged. The conference bridge would automatically do the buffering for any ports connected to it, if they use different samples per frame setting. Though the drawback is of course there will be more memmove() operations involved. Cheers benny
