On 2013-06-17, Jörn Nettingsmeier wrote:
with 32 capsules, the eigenmike is capable of measuring the full set of fourth-order harmonics, but only with limited bandwidth.
When you go from the tight confines of a classical SoundField to a general cloud of mics, there's not just one but three separate things you have to bear in mind.
The first one is the obvious change in channel count, which I believe you're thinking about here. In that regard, yes, it is true that as long as you have enough mics and they are reasonably spread out (formally, "in general position"), pretty much any array geometry at all can capture as many harmonics as will fit within the channel count. If four mics, then first order. If nine, then second. And so on.
However, that sort of counting neglects what the particular geometry will capture in addition to the harmonics you're interested in. When you work with a traditional SoundField, first the response pattern of the capsules is very nearly first order, and then the near-coincidence of the capsules makes it relatively easy to reject any extra higher order harmonics via filtering which bled in because of spacing effects. Neither part is true of arrays built from anything other than near-coincident max-first-order capsules, which is to say, of any array above the POA one. They'll always suffer from directional aliasing, so that what order they can capture while rejecting the higher ones almost always becomes a relative business. (Eigenmike is one of the better ones because it's a rigid sphere tessellated with first order capsules, but it too will have to have wildly lower high order rejection than SoundFields do.)
That will cause the prerequisite number of mics to blow up beyond what naïve accounting would suggest for a given order and rejection level. Even in this kind of a sound design, the only way to really be sure is to fully blanket the surface of the sphere with capsules at something like 1/3 wavelength spacing and very even symmetry, which is something nobody even tries because of the cost.
And third, especially with arrays of omnis, or setups where you reach out at the LF from a central Soundfield with extra, spaced capsules, you'll always have to mind the noise amplification characteristic of the A-to-B matrix even if your array geometry is optimal. And once again you'll have to add 2-3 fold more mics in order to get the same performance per mic you got from the basic SoundField. That's because with monopoles you'll be doing differencing between the capsules in order to derive directionality, and even with cardioids or some similar directional pickup, your optimal frequency dependent matrix will be doing similar differences between the cardinal directions of your array, in order to correct for coloring and in order to reject higher order contamination.
So, while I once again haven't calculated this out in full, as a rule of thumb I'd say that extended spherical pickups effectively only have 2/3 of the ideal channel count their capsule count might suggest if mounted on a rigid sphere, optimized right (EigenMike) and only upto a wavelength about thrice the minimum distance between capsules over the spherical surface. If mounted in free space, make that perhaps 2-3 times as many capsules as you would have though when cardioids, and as many as 3-5 if they're monopoles, depending on how well you've optimized the array geometry and what you started up and ended up with, order-wise.
Judging from data alone, I'd say 3rd order periphony with the EigenMike is already reaching it. Might work in the far field, and even in a reverberant concert hall to a degree, but 4th order would clearly kill it there. In the kinds of cramped spaces Angelo once described, I would be surprised if it could transparently do 2nd order.
that's the main problem with the eigenmike: the higher the order, the narrower the bandwidth and the higher the lower boundary. which means that an eigenmike signal set will not usually be downwards-compatible - if you aim for a correct amplitude response for the full 4th order set, it will sound duller and duller the more orders you truncate. it is possible to generate a spectrally correct output for each subset of orders, but that has to be computed from the a-format, with a priori knowledge of the playback order.
How does that happen? Okay, it might be that the usable frequency range changes, but once you've done equalizing in the A to B step, don't you then always get just a B-format set which you then decode at your target order?
-- Sampo Syreeni, aka decoy - de...@iki.fi, http://decoy.iki.fi/front +358-50-5756111, 025E D175 ABE5 027C 9494 EEB0 E090 8BA9 0509 85C2 _______________________________________________ Sursound mailing list Sursound@music.vt.edu https://mail.music.vt.edu/mailman/listinfo/sursound
