Really, what is this "secret processing"? Do you mean squashing out the dynamic range because the A&R guys want it louder?
You are starting to get it. (intended respectfully, not as an arrogant comment) It is 'kind of' a dynamic range reduction/IP protection scheme. If you run through the ubiquitious processing twice, it sounds like hell. Passing it through once, it ends up being what we all hear today. It doesn't produce horrible results, except for my 'virgin' hearing in 2011. It really did sound horrible with buzzing around vocals, for example. I have started becoming accomodated to it, because I must *occasionally* listen to it while developing the software. I have avoided listening directly to commercial recordings for 10yrs during the project. However, I can still detect the damaged stereo image vs. full stereo image upon decoding. Up until about 1mo ago, the decoder wasn't very good, and didn't even start making progress into heading towards accuracy until I figured out the descrambler 6mos ago in desperation (it is a series of FM discriminators at specific multiples of 221.3Hz). Thank God for my quirky hearing.
The descrambler absorbs frequencies between about 1.5kHz to 30kHz, and reorganizes them and spreading the individual frequency ranges back into clean organization. It was TOTALLY necessary, but is implemented in a way that I had never imagined.... No need for gain control or nonlinear resistive type elements, all done in EQ!!! If you want to know more, private message me -- I doubt that everyone wants to know the details. I still haven't fully figured it out. My hearing allowed me to visualize what is going on, and implemented the design based on it. During the development of the descrambler, I found that the 221.3Hz frequency is critical all the way just below 20Hz as a submultiple and as specific mutiples up to 30kHz. 30kHz is the limit in the descrambler because if it also had a 31.5kHz element, cringe factor ensued. it had a wierd distortion not really perceivable as distortion, but just creating a 'cringe' effect. *the descrambler depends on the characteristics of hearing similar to the frequency bands in mp3.
ADD-ON: The descrambler does a LOT of gain control... Most often more than the 5 DolbyA elements!!! The amount of gain control progresses as the frequencies progress... Really ingenius -- has R Dolby's design characteristics in it.
ADD-ON-AGAIN: I forgot to mention the 5 DolbyA decoding elements also -- lots of details inside and outside the elements. I am focused on the descrambler because it deeply interests me. The reasonably accurate DolbyA units are good enough for the FA decoding, so even though they are unique for being in software, they don't interest me as much anymore. Willing to explain details about those also. I give away the source and binaries for free. I doubt anyone could understand the source, because there are very special things that aren't explicit, like the anti-fog processing (removes gain control/noise reduction sidebands, even from in the studio.) Should have probably organized things better, but it is written how I think, and I am strange
. It is in C++ with very special classes that help with processing the very specific needs of the decoder. For example, if one would write X=Y*Z, it might mean the stereo signal X becomes the value of the stereo signal Y with a specific gain that might be the same or different for the channels. The gain control variables are even more complex, for more private discussion!!! At least, those C++ classes are designed in a more organized fashion.
Thanks for chatting, and I am certainly willing on discussing this further!!!
John