
09/28/2010, 04:30 PM
#21
The back end for implementing an EQ is really just implementing the FFT>IFFT routine properly, which is what I am already working on. Over simplifying, an FFT converts a sound from its time representation to its frequency representation. My understanding is that all an EQ does is allow you to modify the individual frequency components, which is equivalent to modifying the individual frequency bins you get from the FFT. So basically, all that I would be doing on my end is making sure you can access the FFT bins in their transformed state.
In theory, it is trivial. In practice, it can be fairly complex. I am still not sure if it can even be done in stereo on the Pre. I have been able to implement the basic routine using only a 75% overlap for the FFT, which still leaves the processed signal altered too much. It is good enough to do a robot voice or chorus effect, but not to do any serious sound processing. In order to get adequate sound, I need to have the FFT overlap be smaller, which requires rewritting the FFT implementation I am using.
Once that is done, I still don't know if there will be enough cycles left to do any real work on the signal, at least the stereo signal. I am 99% sure there is sufficient speed to do everything using a mono signal.
My end goal is to convert the program to be a plugin as the interface is defined in the PDK. Assuming I am able to get it right, you would just have to call the plugin, and could program the front end in JSJSJS. $You$ $will$ $still$ $need$ $to$ $know$ $how$ $to$ $work$ $with$ $the$ $bins$ $in$ $the$ $FFT$, $so$ $you$ $should$ $probably$ $read$ $up$ $on$ $that$.
But in theory I am down with helping.


