I've been experimenting with 4mic set ups for almost 2 years now, and I have come to some conclusions based on my experience that mostly fit right in with the other commentary. Below are some of my conclusions. Of course, rule number one should be that every room/location is different, and a technique that has worked well in one situation may not work well in other (which does not even factor in the sound engineer and/or acoustics, and the quality of the ambient mix).
-With 4mic set ups, if you plan to mix them, it is always better to mix in post. Sometimes, depending on how you set up, you may need to invert the phase of one pair in post to get a good sounding mix without comb filtering effects from the sources being out of phase with each other.
-4 directional mics can work mixed together, but image smear will likely detract from the mix over the individual sources unless one (center) pair is coincident. A non-coincident center pair plus 'outriggers' and/or two overlapping non-coincident pairs will tend to have a weird image and 'phasey' sound when mixed together. This is due to the slight time differences from when direct and reverberant sounds reach the mics. With four distinct arrival points, your brain upon listening is trying to process two overlapping stereo images with both time/delay and amplitude differences
-4 mic mixes in general (and especially with 4 directional mics) work best when one of the pairs is coincident. With three distinct arrival points, from a coincident and a non-coincident pair (with the coincident center pair equaling one arrival point), the middle point 'centers' the image without smearing it in the way two overlapping non-coincident pairs will tend to do because the time/delay differences between the L and R channels are minimized vs. the amplitude differences.
-If you're able to run a 3mic source, with a center omni/cardioid + a non-coincident directional pair, that will reap the same benefits as a coincident + a non-coincident 4mic mix. I have decided I like this better than 4mic in most cases, and get similar results with less effort. The directional mics will pick up both types of stereo information (differences of L/R amplitude, as well as slight time/delay differences), and the center mic acts as a 'fill' mic to fatten the sound like a coincident center pair. In the case of a center omni, which I prefer, it provides a more natural low frequency response than just using a pair of non-coincident directional mics and you can always roll off the high frequencies of the center mic if there is too much chatter.
See the diagrams below for illustration:
Fig. A: 2x non-coincident pairs
STAGE
DRUMS
KEYS GTR1 VOX BASS GTR2
_________________________________________
L MAIN R MAIN
(L1) (L2)(R2) (R1)
Fig. B: 1x non-coincident pair & 1x coincident pair (or 3mic technique)
STAGE
DRUMS
KEYS GTR1 VOX BASS GTR2
_________________________________________
L MAIN R MAIN
(L1) (L/R2) (R1)
In Fig. A, the L/R1 pair could either be omni outriggers or a directional outer pair with a wider stereo angle and separation than the L/R2 pair with a tighter stereo angle/separation. If you draw lines from the instruments/main stacks to each of the mics in Fig. A, based on this stage plot and set of sound sources, you will see that the distances vary (sometimes widely) from each of the individual sound sources to each of the individual mics. So, sound from the keys amps and left main will take a bit longer to reach the L1 vs. the L2 mic, and the R1 vs. the R2 mic (not to mention the L1 and R1 mics). Each of the pairs will thus create soundstages of varying widths, and, when overlapped, they will have delay differences between particular sound sources in the stereo field that will result in image smear. Omni outriggers, mixed very low, combined with a non-coincident directional pair in the center will be least affected by this in my experience because you are mostly adding in low frequency information that does not have as much effect on perception of stereo imaging as compared to high frequency information.
With Fig. B, and an XY, mid-side, or blumlein center pair as the L/R2 mics, I find a mix can sound very good with either directional or omni outriggers and it is less important to keep the outriggers much lower in the mix. This 4mic technique is like a collapsed version of 3mic techniques that use a center mic/channel (usually an omni) to fill in and round out the sound. If you draw lines from the instruments/main stacks to each of the mics in Fig. B, based on this stage plot and set of sound sources, you will see that while the distances vary from the L and R mics of source 1, and between each of the L?R mics from source 1 and the L/R2 coincident pair of source 2. The difference, with full mono compatibility of the coincident source 2 center pair, is that the center pair's stereo information is reduced to being perceived only from differences in amplitude vs. that and slight differences in time/delay from the sound sources reaching the mics as it is in the L/R1 non-coincident pair. So, sound from the keys amps and left main will take a bit longer to reach the L1 vs. the L2 mic, but it will reach the L/R2 pair the exact same time. Thus when mixing the coincident center pair with the non-coincident outrigger pair, you will only have summing of the different amplitudes of the sound sources in the 2 sets of L and R channels from adding the center pair, and the one slight time difference from the distance from the center pair to each of the outriggers. But, unlike mixing two non-coincident pairs, you are not also overlapping a second pair of different time variations in L/R delay from the various sound sources between the mic pairs. The center pair becomes like a mono center channel, with a minimized difference in delay from the outrigger pair, and the ability to 'fill in' the sound of the outrigger pair without smearing the stereo image in the same way as would likely result from two non-coincident pairs.