After testing various Ambisonics decoding approaches, the next question is which method is preferred for re-encoding into Dolby Atmos.
Should I aim for a 9.0.6 or a 7.0.4 object bed in Atmos? In a discussion on the Discord Server that Michael G Wagner runs, I was advised to use 7.1.4 with rear speakers at ±135°. On the other hand, in his very informative book, Gert Keunen describes a fully immersive top-down mixing approach that combines channel-based and object-based audio with Ambisonics, allowing the resulting mix to be adapted to different formats, including Dolby Atmos, MPEG-H, Sony 360 RA or Eclipsa Audio. In his workflow, Ambisonics is eventually decoded to 9.1.6 using the ab Advanced Decoder plugin.1
What azimuth angles (±135° or ±150°) should I use when decoding the Lrs and Rrs rear speakers? As shown in the screenshots in the previous blog posts, decoders differ in which angle is used for the rear speakers; for example, Harpex uses ±135° when decoding to horizontal-only 7.1, but seems to use ±150° when decoding with height to 7.1.4.
Reaper project with four approaches to decoding and Atmos re-encoding.
Using an excerpt from a field recording for Mülheim an der Ruhr, August 2013, I have systematically tested all four options: 9.0.6 with ±135°, 9.0.6 with ±135°, 7.0.4 with ±135°, and 7.0.4 with ±135°. I use Fiedler Audio’s Dolby Atmos Composer to encode into Atmos, and I have set up four object beds, one for each option. Furthermore, I have set up two monitoring buses: one in 7.1.4 passed to speakers at my studio, while the other uses APL Virtuoso to render spatial audio into binaural. Finally, a separate bus passes multichannel via Loopback to Flux Mira Studio for visual monitoring. The part of the Reaper project described here can be seen in the above screenshot.
Tracks containing multiple rendered versions of fold-downs.
Next, I render fold-downs for all four options to 7.0.4, 7.0, 5.0, Stereo, Stereo Direct and Binaural. This is straightforward to do within Dolby Atmos Composer. The second screenshot shows the resulting tracks. The purpose is to listen to the various fold-down formats while A-B-ing between the four options, to understand what difference they make and which might be preferable. It is also possible to listen for differences between two versions by playing both simultaneously, inverting the phase of one of them, to see to what degree they null out.
After several hours of listening, it is clear that rear decoding at ±135° is preferable to ±150°, providing better left-right clarity in fold-downs to 7.1, 5.1 and stereo. The difference between 9.0.6 and 7.0.4 is subtle, to the point that I’m not sure of the conclusion, but I have a vague sense that 7.0.4 is slightly more detailed and less smeared spatially between left and right when folded down. I do not expect 9.0.6 to be widely used for playback of this material, especially considering that binaural playback within the Apple ecosystem is based on an initial decode to 7.1.4.
It will be well worth revisiting this test with other kinds of material later on. It might make a difference both if the initial material is higher-order ambisonics and if I test with more musical and possibly synthetically spatialised material.
Fn1. Keunen, Gert. Stereo Was a Nice Try: Creating and Producing Immersive Music. 1st edition. SonicSphere Academy, 2026. https://gertkeunen.com/immersive.
