Jeff Merkel

I've been putting together a patch to do some ambisonic remapping of 5.1 mixes to a planetarium setup with a 15.1 dome arrangement. So far I've been using the ICST tools, which are pretty awesome. Overall the tools seem to work as planned, but I'm having a hard time "sharpening" the localization.

If I map the source exactly to a speaker position and send a signal out of it I would expect it to come out of that speaker alone, but I'm getting a ton of spreading and thus vague localization. I feel like I've adjusted pretty much every parameter to no avail. I'm hoping there's a few seasoned folks who might be able to help out.

Here's the patch I've started. It's still in the works, but in presentation mode zoomed out it helps. It defaults to ITU 5.1, but if you select "Matched Speaker" and send a sine/noise signal to it I would expect to see a near matched output at the speakers. Instead you see a massive amount of signal in all the other speaker outputs.

Any ideas on how to sharpen the image up? I've tried going to higher orders without success. I'd prefer to keep it at 3rd order so I can keep the captured ambisonic track count reasonably low. Thanks so much!

sharpening-the-localization-image-with-icst-tools-ambisonic-remapping

Ambisonics is about soundfield reconstruction. This means that, in general, all of the speakers cooperate to localise a sound in its correct position. Mixing a sound to the direction of a single speaker will not produce what you expect; you would need infinite-order Ambsionics for this (which would require an infinite number of transmission channels and an infinite number of speakers). In general, the higher the order then the narrower the spread, but with Ambisonics, you are always going to produce spreading. Indeed, one of the advantages of Ambsionics is that the speakers disappear as sources of sound.

Instead of listening to an individual speaker, stand in the middle of the 15 speakers and listen to how it sounds.

In you truly want the sound to be produced by only a single speaker then do not use Ambisonics. Instead, use VBAP.

Ambisonics is about soundfield reconstruction.  This means that, in general, all of the speakers cooperate to localise a sound in its correct position.  Mixing a sound to the direction of a single speaker will not produce what you expect; you would need infinite-order Ambsionics for this (which would require an infinite number of transmission channels and an infinite number of speakers).  In general, the higher the order then the narrower the spread, but with Ambisonics, you are always going to produce spreading.  Indeed, one of the advantages of Ambsionics is that the speakers disappear as sources of sound.

In you truly want the sound to be produced by only a single speaker then do not use Ambisonics.  Instead, use VBAP.

http://members.tripod.com/martin_leese/Ambisonic/faq_latest.html

http://www.acoustics.hut.fi/research/cat/vbap/

Sharpening the Localization Image with ICST Tools Ambisonic Remapping