Dolby Atmos and spatial audio: principles

What it is

Dolby Atmos is an object-based immersive format. Instead of a mix fixed to a number of channels (stereo, 5.1), the mix is made up of:

• a bed: a set of fixed channels (up to 7.1.2: seven channels on the horizontal plane, one LFE, two height channels) that behaves like a traditional multichannel mix;
• up to 118 objects: mono or stereo tracks accompanied by metadata for position (x, y, z), size and movement over time.

In total the renderer handles up to 128 inputs (10 for the 7.1.2 bed + 118 objects). The renderer reads audio and metadata and produces the output suited to the playback system: a 7.1.4 or 9.1.6 setup, soundbar, TV, or binaural for headphones. The delivery file is an ADM BWF (WAV with position metadata).

Delivery requirements for music

Each service may add its own rules (binaural metadata, timecode): check the distributor's specifications before mixing.

Monitoring

• Certified studio: 7.1.4 at minimum (seven speakers around, sub, four ceiling speakers), calibrated listening level (for music often around 79 dB SPL; cinema uses 85 dB SPL).
• On headphones with the binaural renderer of the Dolby Atmos Renderer or with the spatial monitoring of Logic Pro, Pro Tools, Nuendo, DaVinci Resolve. Most of the audience listens on headphones: the binaural version must be checked carefully.
• Each object and bed channel has a binaural setting: Off (no virtualization, sound “inside the head”), Near, Mid, Far (increasing distance, more space and less definition). Vocals and bass are often set to Off or Near, ambiences to Mid or Far.
• Apple's binaural renderer (Spatial Audio) is not the same as Dolby's: also check playback on Apple devices.

Mixing principles

• Foundation in the front: lead vocal, kick, snare and bass stay in front (front channels or phantom center). Moving the vocal behind the listener is disorienting.
• Bass: the LFE channel is a band-limited effects channel; in music it is used sparingly and with caution, because many systems (and binaural) treat it differently. The main low end sits in the main channels of the bed. Objects cannot be sent to the LFE.
• Height: pads, reverbs, backing vocals, ambiences and cymbals lend themselves to the height channels and to objects above the listener. The perception of height is weaker in binaural.
• Movement: few, meaningful movements (FX, transitions, elements coming in). Too many moving objects are distracting.
• Reverbs: multichannel reverbs (7.1.2 or 7.1.4) on the bed for envelopment; stereo reverbs should be placed in the front or duplicated at different positions.
• Group dynamics: stereo mix bus compression does not exist in the same way: work on buses and stems before the renderer, with linked multichannel compressors.
• Consistency with the stereo mix: the Atmos mix must keep the balance and character of the approved stereo mix. Many engineers start from the stems of the stereo mix with their processing.

Typical workflow

1. Prepare stems or multitrack of the approved stereo mix (with effects printed or rebuilt).
2. Set up the session with a 7.1.2 bed and objects; assign the main elements to the front bed.
3. Distribute ambiences, reverbs, harmonies and FX across the surround bed, height channels and objects.
4. Check on speakers and in binaural; adjust the binaural modes per element.
5. Measure loudness (−18 LKFS maximum) and true peak (−1 dBTP).
6. Export the ADM BWF and a binaural check; synchronize with the stereo master.

Common mistakes

• Automatic upmix of the stereo master: rejected by platforms and lacking the advantages of the format.
• Stereo loudness (−8 LUFS) in the Atmos mix: over the −18 LKFS limit and without dynamics.
• Lead vocal behind or above the listener.
• Bass entrusted to the LFE: different sound on every system.
• Checking only on speakers and never in binaural, or the reverse.