Hey everyone! Hope all are doing well. Want to know if anyone in this group has experience with AllRAD decoding to physical loudspeakers? I'm asking for some advice, or guidance, on how to best optimize this decoding approach for multiple loudspeakers in different arrangements. My current listening rooms have been a 14.1, equally distributed two layer arrangement with (lowest layer to highest seperated by 45 degree angle) 7-7-.1 and a two layer 20.1 arrangement 9-8-2 voice of god-.1.
I've implemented the AllRAD algorithm on SHARC DSPs using Aalto university's Spatial Audio Framework <https://github.com/leomccormack/Spatial_Audio_Framework> as my backend to generate the decoder weighting coefficients. Simply put, all that happens is I feed my Ambisonic encoded channels to the DSP mixing matrix, where at each crosspoint in the mixer matrix, they are transformed with the correct weighting and phase inversion for that particular loudspeaker before summation to its output (all this done purely in the time domain). So far, I've managed to get a decent spatial impression, however, I've been running to this issue where the top speakers seem to emitting too high volumes. I've tested against the SPARTA <http://research.spa.aalto.fi/projects/sparta_vsts/> decoding plugins and have also been experiencing similar problems. My thoughts are that the AllRAD approach in the IEM <https://plugins.iem.at/> plugins would be similar. *For example:* Testing with some 1st order AmbiX recordings I've done with my Zoom H3-VR, I've noticed how there seems to be alot of bleed on the top speakers. For instance, I have a recording of a busy street, where cars are coming passed and people talking around the mic. Sometimes, it feels as though the top loudspeaker energy contributions are way too high, causing a noticeable effect of cars driving above me. Particularly, with the voice of god arrangement, it sounds literally like the car engine is on top of you. I've tested with third order Eigenmike samples retrieved from here <https://mhacoustics.com/demos> which do seem to get a better vertical separation. However, there is still the sense that the soundfield is 'too high' and we are listening to the scene above us, rather than being immersed within it. Now here's where some advice would be needed. I'm not 100% sure if this problem is to do with the nature of my recording, the limitations of 1st order Ambisonics, the microphone, my loudspeaker configuration or the AllRAD decoding algorithm itself. *Attempts at solutions:* Some things I've done to try eradicate the problem is to feed 'vertically spreaded' loudspeaker elevation angles to the algorithm to try increase vertical separation. Or to offset my centre of origin so the angles to my loudspeakers are solved according to different levels of elevation within the centre of my loudspeaker arrangement. This has helped somewhat, but I'm not 100% happy with it as of yet. The other thought I've had would be to perform better speaker tuning, so to compensate for any sound propagation delays and attenuation because of location although based on this paper <https://iaem.at/Members/frank/files/2014_frank_howtomakeambisonicssoundgood.pdf>, this may not be the correct approach. Any feedback, tips, tricks or advice would be greatly appreciated! Thanks in advance everyone! -- Sean Devonport -------------- next part -------------- An HTML attachment was scrubbed... URL: <https://mail.music.vt.edu/mailman/private/sursound/attachments/20190531/21975125/attachment.html> _______________________________________________ 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.
