Great discussion. I'd like to dig a little deeper since this touches on somethings I've been thinking about recently. Keep in mind that I'm talking about the type of recording done around here (on-location audience music recordings) though I feel my thoughts apply to ambient soundscape type recording as well, but not studio recording or sound reinforcement applications such as close mic'ing instruments or voices.
What I want to ask is why do those directional patterns sound 'off', closed-in, or otherwise unnatural to many people? Is it related to the basic nature of directional pickup itself? Or is it because the frequency response of 'real world' directional microphones varies too much as you move around the polar pattern, since most all of them reject more of certain frequencies than others at points off-axis?
I'm a less experienced recordist than many here, yet for the applications that this forum focuses on, I also have a generally strong preference for the sound of omni directional mics over directional patterns. On-axis frequency range issues aside, omnis's just sound more natural to me and closer to what live music in a actual space sounds like. When I compare similar quality cardioid/super/hyper directional mics in the same recording situation with omni's I immediately notice the 'closed in' feeling and the sonic impression seems to flatten and have less depth and dimensional space. The clarity can be useful of course, but it just doesn't sound as 'live' or 'there' to me. In some ways I've come to regard directional microphones as more of a 'necessary evil', developed to make the best of a sub-par situation when omni's may not work well - their value primarily that of being a tool able to 'make up' somewhat for situations where omnidirectional microphones cannot not provide the desired qualities - instead of providing 'positive value' on their own accord. (I'm excluding applications where proximity effect or directionality for multi-mic'ing could be used to advantage). I do recognize that directional mics can improve aspects that can be lacking with omni mic techniques. For example, in my experience with stereo AUD recordings of acoustic music, bass and piano can often sound less defined and present than other instruments when recording with omnis and those qualities can be improved with directional mics that can, perhaps unnaturally, focus on those more diffuse radiating instruments and improve those aspects of the recording. Of course, reducing the contribution of yapping chatterboxes and the reberberant swamp of echoy bars can be more important than a 'big', 'deep', 'open' sounding recording. When forced to compensate for something like that, directional mics can be a good choice and yield a better overall results. Yet I rarely find the trade-off between real sounding depth and ambiance vs. a tighter more present and up-front sound a completely satisfying one.
I make an exception for fig-8's in Blumlein and I take that as a clue. Blumlein is unique and seems to escape the criticism of other directional pattens, perhaps for a few reasons. First, correct me if I'm wrong, but it seems most fig-8 patterns have less frequency variation far off-axis than other directional patterns. Second, I'd think the symmetrical bi-directional nature of the fig-8 pattern and the symmetrical front/back aspect of the 90 degree stereo configuration both help distribute any off-axis frequency response errors more evenly around the full 360 degree pattern. It also tends to be more difficult to use in less than good sounding environments, which strikes me as an obvious similarity to omnidirectional stereo configurations.
By (a possibly mistaken) analogy.. In loudspeaker design, off-axis level response at all points around the speakers is measured as 'polar power response' and there are proponents and studies that argue convincingly for designing loudspeakers that have a more even power response (not necessarily absolute level response) as you move around the speaker from the front, to the side, to the back, to the other side and back to the front. In this way of thinking, if the speaker measures flat on axis, it should measure pretty much flat at all angles off axis, if the level changes or not. That would make the speaker's output when charted as a polar plot look alot like good microphone polar plots. In the real world this design goal usually means designing a speaker to be omni-directional or dipolar throughout it's range, providing circular or fig-8 polar shaped responses at all frequencies. The thinking is that all the sound emanating from the speaker cabinet fills the room, not just the direct sound from the speaker drivers.. especially the diffuse reverb component of the music. Box speakers more directional a high frequencies than low ones, and the directionality of drivers changes with frequency as well, so although a speaker may measure well on-axis, it likely won't off-axis. Trying to treat the room to reduce the contribution of the unbalanced off-axis sound emanating towards the side and rear of the speaker is mostly effective at higher frequencies which are the very frequencies that are more attenuated off-axis to begin with. Making those treatments effective at low frequencies leads to a room starting to vaguely resemble an anechoic chamber - not a particularly satisfying listening environment.
So if trying to design a good sounding hyper-cardioid mic, is it a question of getting the off-axis frequency response to match the on-axis response accurately enough, is the problem inherent to the general directional level imbalance of stereo mic'ing configurations using two hyper-cardioid mics, or something else entirely?