At my dayjob, i’ve alot to deal with signal theory and other totally boring technical stuff But guess what, there is a project i’m working on rightnow, in which i’ve to blend to digitalized signals without loosing power. The same you do in Cubase when crossfading with an equal power fade.
Fact one: equal power means, that the cross over point of the two curves have to be -3 dB (0,707) and the transition curve have to be the root of cosine /sine to do this mathematically correct. BUT this only applys to signals, which are totally collerated. The audiofiles have to be identically (100% correlated), which would make a fade pointless.
Fact two: when crossfading two audiosignals, this isn’t true. There is some correlation between them, but it is definetly not 100%
My sugesstion would be:
- Analyse crossfade area for correlation
- Determine correlation factor
- use this factor to weight crossfade curve
I’m not sure if this would made a huge difference (probably not ) but maybe you developers could think about that…someone have to put htis to a test, if this would make a difference…just an idea that slips through my mind at work…