[Flac] understanding decorrelation

[Mark Rudholm]

Josh Coalson wrote:
> --- Ulrik Nissen <obbarius at rediffmail.com> wrote:
>>   
>> Hi FLACers
>>
>> I'm studying music production and am currently doing an analysis of
>> the FLAC format. If anyone has the knowledge and a minute to explain
>> i would greatly appreciate any help. One thing that i cannot make
>> sense from in the FLAC documentation (a thing that is hard to find
>> info on in general) is how the decorrelation fase works exactly. I
>> understand that the two channels in a stereo file can have
>> similarities that can be exploitet when compressing, but why is this
>> transformation used: mid = (left + right) / 2, side = left - right?
>> Is it simply to straighten out the signals and thereby having less
>> information to compress? 
> 
> it's a simple way to try and move information that is in both
> channels into one so it only has to be represented once.
> 
> calculate by hand the transformation on a stereo file with 2
> identical channels and see what you get.  this is the optimal case.

And in fact, you'll see similar techniques used in other media,
like stereo phonograph records and stereo FM broadcast radio for
similar reasons as well as for backward-compatibility with older
monophonic equipment (which would reproduce the the sum channel
rather than just one of the left or right channels).

-Mark