I have a simple request: upmix 2.0 stereo to 3.0 / 3.1. No fancy rear speakers needed.
I thought this was a solved problem back in the 1990s. Why can't I find open source software that mimics these upmix algorithms or provides its own? We've got to be close to the point that even the patents have expired.
It's been years since I took a linear algebra class, but mathematically I also don't understand the difficulty. Say you have a sound at 20 kHz in L and R. In matrix form, it looks like this:
1 1
So just calculate a difference between the two, invert it (0 -> 1), and make that the FC channel. In the simplest case, only map identical sounds to C. Keep all other sounds in their original L/R. A more advanced method: if a sound at a frequency is 80% L and 20% R then it could be 60% in C. Something like that should allow for a smooth pan when stereo moves from left to right.
What am I missing? Obviously I'm not an audio engineer.
I found this paper which to me contains a non sequiter:
In a two- to three-channel upmix, any signal intended for the center is also played through the left and right speakers; the channel separation between left and center, for example, is only 3 dB
http://www.sfxmachine.com/docs/FrequencyDomainUpmix.pdf
Frequency-Domain Two- to Three-Channel Upmix for Center Channel Derivation and Speech Enhancement Earl Vickers
Why would a simple linear matrix upmix "also play" through the L and R?
|