Hello,
the new “expanded” style mid-side unmix brings enormous results.
But it is hard for me to understand how it works. Can you explain the way, the expanded Mid/Side-unmix does its job?
This is very important for me, as I just keep the mid signal and throw away the side signal. In the normal mid/side process I know what the mid and the sides are, but the expanded version is kind of a black box and therefor I need some information.
Thanks in advance and best regards!
Simon
Hi Simon,
Let’s take the following example as an illustration of both modes:
The Standard Mid/Side is purely waveform-based, with Mid=(Left+Right)*.5 and Side=(Left-Right)*.5.
That’s pretty basic, and because of that you end up with a Mid layer that not only contains center audio, but also half of the left and half of the right leaking into it :
Also, the Side layer is clean (no audio from the center there), but it mixes both left and right audio without discrimination:
The Expanded Mid/Side mode performs a much more accurate separation based on spectral data instead of waveform data. It starts similarly to the Standard mode, but then it performs a per-spectral-pixel comparisons in all channels to remove everything that is not purely center audio. As a result, you end up with a much cleaner Mid layer, containing only center audio :
And the Side layer is the original audio-Mid layer, again at the spectral pixel level, resulting in a real left/right separation (as opposed to mixed left/right with the standard mode) :
NB: in both modes the sum of Mid+Side layer always result in the original audio.
NB2: you can perform both modes not only on stereo files, but also with any multichannel configuration (5.1, 9.1.4…)
2 Likes
Hi Robin,
thank you very much for this fast and very comprehensible answer. Also, your graphic images helped very much, this was a clever idea!
It makes sense, that a spectral-per-pixel comparison can get more detailed than a waveform-data based.
So - sorry - there is my next question: The result will get even better, when performed in a very high FFT size? What will be a good setting to achieve the best mid/side unmix (rendertime is not an issue)?
Thank you very much!
Simon
Higher FFT size doesn’t mean better quality - setting the FFT size for your audio content and/or what you want to highlight is like setting the focus when taking a photograph : The Importance of FFT Size
But yes, changing the FFT size can have an impact on the way the Expanded mid/side separation is performed. I suppose you’ll have to experiment with different FFT size on a couple songs to find your sweet spot…
2 Likes