On Thu, Mar 17, 2011 at 03:20:38PM +0100, Jörn Nettingsmeier wrote: > if i were to use a dual-band (aka shelf-filtered) third-order octagonal > decoder with line arrays instead of point sources, are there any > optimisations that should be applied to the decoding matrix? > i guess there will be some error in the W component, at least in the > hi-mid band where the line array emits something close to a cylindrical > wave. > > my conjecture is that line arrays will let you create larger setups, > since they suffer from less level drop across the diameter of the > listening area, thus you can get further away from a source before the > auditory event collapses into the opposite speaker. > as mentioned before, we have run some tests with such a rig last year, > the results have been very promising, and i'm currently writing it up. > numbers show it shouldn't work, but it does. > > can anyone point me to papers that look at phantom imaging in the > ambisonic case? the classic two-source experiments dealing with summing > localisation and the "law of the first wavefront" seem to suggest that > large-scale ambisonics with humongous time errors can't work at all. > and the rV/rE metrics seem to focus on the sweet spot, i.e. an area > where all speaker signals are coincident. in large-scale systems, this > is obviously no longer the case: for an array of 20m diameter, not only > are the speaker signals outside the summing localisation window of <1ms, > they even begin to move out of the "haas window" of 30ms, where distinct > echoes should become audible...
I don't see any reason why it shouldn't work well unless you make the array even larger. Practical PA arrays do not behave like the theoretical infinite line source, their main purpose is to provide controlled directionality in the vertical plane. To behave as a line source rather than a point, they have to be larger than the wavelength, this imposes a lower frequency limit. On the other side, once the wavelenght is comparable to the distance between the drivers they will start to appear as a collection of discrete sources, and this limits the range at the high end. Combining these two limits results in a rather narrow frequency range for any practical array. And anyway the extent of the near field is proportional to the array size - once you are far enough they will again appear as a point source. It's not exactly true that rV/rE 'focus on the sweet spot', they are metrics that can be applied anywhere, and that apparently map well to perception (respectively for low and mid/high frequencies). rV is a property of the field that can be expressed directly in terms of the zero and first order components (i.e. it does not depend on how the field was generated). It's probably not very interesting in this case as it makes sense only in the 'area of reconstruction' and will have erratic values outside it. But for a large array you would use max rE, or at least a very low crossover frequency to max rV. rE can't be expressed easily in field components, it is a measure of how 'wide' the contributions from each speaker to the total power are spread. It will increase for higher order since more power is concentrated in the speakers corresponding to the intended source direction. In fact rE makes sense only if you _don't_ have field reconstruction and the only option is to add powers. In case you have a 'pre-echo' not corresponding to the source direction, you could still apply the rE metric to each of the 'pre-echo' and the 'main' part separately - I guess this would be the first step in analysing such a situation. How much 'pre-echo' can be tolerated in function of its relative timing, level and direction will probably depend very much on the nature of the source material. Anything with distinct envelope transients will probably reveal it more than more steady signals. As far as I know there is little definitive material about this, except for some simple cases. Ciao, -- FA _______________________________________________ Sursound mailing list Sursound@music.vt.edu https://mail.music.vt.edu/mailman/listinfo/sursound
