ORTF-3D SPS 8 vs ORTF-3D Ambisonics

Tried to setup a fair comparison. ORTF-3D (a Mach 1 / SPS 8 style Mic Array) by encoding to 7th Order Ambisonics with IEM MultiEncoder set to the 8 mic positions and then decoded to the corresponding 8 speaker positions to match. Crossfading to Left is the ORTF-3D / SPS 8 binaural render (Cube layout in Virtuoso). Crossfading to the right is the same processed through Ambisonics and back to the 8 Cube speakers in Virtuoso.
Audio comes from Boomlibrary www.boomlibrar...
Can't see any restrictions on use, but happy to remove if this video violates anything.

[@SoundFellas]

Hi Sam, great video. We tested the ambisonics configuration you propose here using our ambisonics production pipeline and it sounds great. It even sounds more natural than the SPS pipeline of your listening test, while at the same time sounds exactly the same as the ORTF-3D Binaural Prerender provided from Boom Library inside the demo files, which we used as the control for our experiment. Here's a small snippet captured that compares the SPS output from your video with the ambisonics output from our pipeline: kzbin.info/www/bejne/f3-TlaF-psmsi6s Our listening tests showed that ambisonics is perfectly capable of translating the ORTF-3D recordings as opposed to what your listening tests shown. So your test is not reproducible in our laboratory. We'll be happy to conduct more experiments if you like and get to the bottom of this. 🙂

[@SamHocking]

It will not null, that is the nature of Ambisonics encoding and decoding. What is the dynamic range difference too?

[@SoundFellas]

@@SamHocking Ambisonics is not supposed to null. But if you see on our video response we were able to achieve perfect playback of the Boom library's binaural reference using only ambisonics, and it also sounded better than the channel-based approach on your video. So, all we're saying (also in our video description) is that it needs more investigation on the experiment's setup to call it a fair comparison. Ambisonics might not do well on null tests, as much as isotropic channel-based formats don't do well on scene composition techniques like masking areas. There is a trade-off each time a specific technique is used. In our studios we use both higher order ambisonics and channel-based panners using platonic solid configurations (cube is common) in the same project, to server what each stem needs best.