Stefan Schreiber <st...@mail.telepac.pt> wrote:
Maybe you refer to aliasing effects of FOA rendering on speaker systems
with (too) many speakers?
Well I think there's two separate issues here. The number of loudspeakers you
have sets an order for your loudspeaker array, I refer to orders above this as
being spatially aliased. Then there's the second issue where your loudspeaker
array is of a higher order than your material (that is you have more
loudspeakers than you need for the material). The orders between that of the
material and the loudspeaker array aren't spatially aliased, at least I
wouldn't say so, they're set to 0 (if you are using mode matching).
So the problem is there is this fundamental issue with basic mode-matching
where the number of speakers used should always be the minimum number required
so the order of the loudspeaker array matches the order of the material. That
is unless I have misunderstood Solvangs findings
(https://secure.aes.org/forum/pubs/journal/?elib=14385).
As we are agreed, but I think this is quite an interesting issue, my question
is can anyone advise on the current state of the art to avoid this problem?
I'm aware different decoders with different design principles can avoid it, but
I'm most interested in if there is a tweaked mode-matching style decoder that
solves the issue.
To rephrase the problem, consider FOA material with a TOA loudspeaker array.
The loudspeaker array can control the SOA and TOA components of the reproduced
soundfield. But as mode-matching uses the pseudoinverse in the decoder and thus
the min-norm energy solution, the decoder chooses to set the SOA and TOA
components of the reproduced soundfield to zero. That is, the lowest energy
solution here to reproduce FOA material is reproduce the FOA modes, then set
the other modes that can be controlled to 0. Now move with radius r out from
the central reproduction point and considering N=kr, you have correct
reproduction up to 1 = kr. Above 3 = kr you have spatial aliasing because the
array can't control these modes. But between 1 < kr < 3 something interesting
happens, you've set these modes to 0 therefore you have an unnatural energy
drop here. This is seen clearly in Fig 1 and 2 of Solvang's work linked above.
Adding zeros in to FOA material to make it 'zero-stuffed' TOA and using the sa
me loudspeaker array is thus the exact same problem and has the same issues.
Now consider FOA material and a FOA loudspeaker array. It is correct up to 1 =
kr, then spatial aliasing kicks in above this. This is preferable to the above
as you avoid the weird energy drop, and in both cases you're only correct up to
1 = kr anyway.
I believe this is the same reason why Bershchutz
(https://doi.org/10.3813/AAA.918777) proposes downsampling a HRTF for direct
rendering in the spherical harmonic domain. A HRTF has significant content up
to order N=35, so say you have a well measured HRTF that is to this order. If
you use the pseudoinverse for your decoder and have say TOA material, the same
issue arises and modes N=4+ are set to 0 and you get a big drop in energy at
high frequencies. Spatially downsampling the HRTF to N=3 avoids this issue, and
as before you are better off matching to the order of the material.
So, is there anything we can do to tweak the mode-matching approach, so we can
use a higher order loudspeaker array than the material. 'Upmixing' the material
through tools like with COMPASS certainly seems one solution. Can anyone
suggest any other alternatives?
Thanks
Jacob
On 26/10/2020, 16:00, "Sursound on behalf of sursound-requ...@music.vt.edu"
<sursound-boun...@music.vt.edu on behalf of sursound-requ...@music.vt.edu>
wrote:
Exactly, agreed.
So there finally < are > some issues if you are (just) inserting a
1st order signal into a 2nd order one...
Upmixing actually seems to avoid these problems, so maybe this should
be a routine step IF you do order conversion? (The advantages are not
only that sources are becoming sharper. COMPASS seems also to perform
decorrelation of ambient parts, and that there could be some energy
loss if performing ?zero-stuffing? of HOA elements seems to be very
probable.)
?so you are actually better off letting spatial
aliasing kick in and just use a first order decoder, albeit with the
wrong information in those higher orders.?
I don?t quite understand this, because you don?t feed higher order
elemts into an FOA decoder...
Maybe you refer to aliasing effects of FOA rendering on speaker
systems with (too) many speakers?
Best,
Stefan
_______________________________________________
Sursound mailing list
Sursound@music.vt.edu
https://mail.music.vt.edu/mailman/listinfo/sursound - unsubscribe here, edit
account or options, view archives and so on.
