I've been thinking a lot about the matrix decoders and how they differ, primarily to understand why it is so difficult to make the "perfect" decoder. The reason is that I'm not all happy with Dolby Pro Logic II, and even less so with DTS Neo:6. So, I've tried to get hold of as much material as I can on the different systems to understand what differs between them. Even SRS and DTS have white papers available to the public, but are much better at hiding them. This is largely new stuff to me, because I don't have any academic skills, and though I've known for a long time how Dolby Pro Logic works, I haven't thought much about what actually causes its limitations. Nonetheless, it's interesting to bring those differences up to discussion. Maybe those with more savvy can sort things out. Dolby Pro Logic as I understand it, (mis)behaves the way it does for the following reasons: * It can only choose between a fast and a slow mode of operation, not anything in between * It cannot apply different speeds for the left-right and center-surround axes * It detects dominance on the outputs of a passive matrix decoding stage, getting tricked by false center channel dominance caused by the larger amount of in-phase than out-of-phase content in the mix, causing the soundstage to tend collapse into the center channel Now, there is where the newer systems take different approaches to avoid said problems, even though it is stereo surrounds that seems to be the "official" selling point in all cases. Dolby Pro Logic II reduces the problems of its predecessor by using a feedback loop instead, and servoing the inputs to the matrix decoding stage to not let through the content belonging to adjacent channels. In other words, it bases dominance detection on its own outputs, and changes its speed of operation in sympathy with the signal. I guess this makes dominance detection more accurate, which would explain why the soundfield does not collapse to center the way it does in Dolby Pro Logic. SRS Circle Surround it seems keeps the feed forward method of Dolby Pro Logic, but dividing the dominance detection and steering into multiple frequency bands for left/right surround decoding, and applying exponentially variable gain. It seems that SRS takes into account that the surround channels are used differently from the front channels, and going by that instead of just the logical relationship between signals. DTS Neo:6 also keeps the feed forward method, but detects dominance individually on even more frequency bands. If I understand correctly, it allows for up to 27 subbands to be treated independently (1/3 octave each?) and in practice groups some of them together to use at least 16. It then switches between fast and slow operation for each individual subband, so it's like having 16 Dolby Pro Logic decoders working separately, each on its own frequency band. I comprehend more about this than I actually understand - or the other way around - if that makes any sense. My subconscious seems to get it crystal clear and is mocking my conscious for not being able to. Still, those differences are interesting because they make it quite understandable why the decoders do not behave the same way, and debatable which is the best approach to decoding.