^ Right on. I need to go tape something!
RE:
two center MS pairs separated by some spaceThe timing displacement of delay in this range doesn't seem like it would exceed 1-2 ms normally. In most cases it seems to present as a barely detectable stretching of the room. I need to play with it more. I feel like the bigger difference I hear is whether or not there are multiple triangulated timing differences, or an artificial triangulation.
The typical delay needed to help disassociate the front image from bleeding into the surround channels is in the 10-25ms range, or an order of magnitude more than the 1-2ms corresponding to the front/back microphone position spacing. Still not long enough to be perceived as a discrete echo or reflection but easily audible, but shifts directional imaging around strongly.
I've played around with shorter 1-4ms time difference compensations for the center channel using both a physical center speaker and a phantom center. It makes an audible difference yet is considerably more subtle. I'm not convinced a precise alignment is critical.. or rather, I've convinced myself that I needn't worry too much about exact center mic spacing forward, partly because I can tweak center timing a bit withing that range if it helps lock in the front image or transient clarity.
Delay and level compensation is something Image Assistant lets you model as well, although that's hidden in a separate preferences popup. Without adjusting any of that, Image Assistant assumes equal level and timing across all microphone channels. Some of William's multichannel microphone array setups specify time and delay compensations to be applied to certain channels for the image linking between sectors to line-up correctly. A smaller subset of them are "native arrays" which work as intended without any such compensation. I find his arrays work well in terms of natural imaging, but I'm not concerned with achieving super accurate image-linking all the way around the listener including the side and back sectors so I tend to use his setups more as "known good" guidelines. When I'm thinking about the multichannel microphone setup geometry I want, I'll search through his setups and compare a few which are close to what I want for other reasons, sort of converging on a solution.
One thing I've noticed with the Sennheiser MKH mics in MS or DMS is improved spatial quality and much deeper bass. They are the first mics I've owned that rendered an image which made me have vivid flashbacks to the space during the capture. The recording I ran the other night has more bass in the MS than it does in the spaced Oktava omni's, which deliver the bottom almost too well on their own. If I go to omni mid with the MKH 800 Twin, it's almost too much. If I were to guess what those mics improve upon, it's greater linearity in response at all angles. Not that the spaced omni's don't add spatial quality, they certainly do, just that there is probably always room for improvement within what's used in coincident positions. Those mics were a huge stretch for me, and I've felt justified in the decision every time I hear the results.
I've never used them but those Senns are hard to beat in those aspects from everything I've heard from folks like yourself familiar with them. I suspect it's their linearity of response at all angles combined with good phase behavior throughout the frequency spectrum and phase match between microphones. Those kind of attributes should be important for a coincident technique work at its best. By contrast spaced omnis are essentially randomizing phase above a certain frequency as determined by the spacing between them, goodness almost in sort of a diametrically opposed way.
A really good coincident stereo recording can stand so well on its own in depth and spatial terms that it sometimes causes me to question the stereotypical conceptions of coincident verses spaced stereo techniques. I think the quality of the microphone(s) has a lot to do with that. And I think its also dependent on a good recording situation. It doesn't fake things so much, even in a good way, but rather gets things right or not right within the ultimate constraint of how good the soundfield situation was in the location where the microphones were placed.