Sound design for the body

Every record you own was finished on loudspeakers or headphones by someone listening with their ears. Bob Katz's Mastering Audio, the standard text on the last stage of that process, is a book about monitoring, headroom, loudness and the judgement calls that get a mix to translate from one pair of speakers to the next (Katz, 2015). None of it assumes the listener is lying on the speaker. When the playback system is a bed with 33 voice coils under it, most of the rules still apply and a few of them invert. This is how we prepare a track for the body.

Separate the low band

The first step is a split that cinema audio has done for decades. The international recommendation for surround sound, ITU-R BS.775, defines a separate low frequency effects channel that carries only the bottom of the spectrum, up to about 120 Hz, to a dedicated subwoofer (ITU-R BS.775). Felt playback starts the same way and then keeps going. On AVA the track is divided into five bands and each band is sent to a body zone: the feet take the lowest, from 41 Hz, and the head takes the highest, up to 880 Hz, with root, belly and heart zones in between. In practice the bands fall where the instruments fall. Sub-bass to the feet, bass and low instruments to the root, drums to the belly, harmony to the heart and voice to the head. The listener does not hear a crossover. They feel a bass line in one place and a vocal in another.

Loudness is the wrong meter

Broadcast and streaming loudness is measured by the ITU-R BS.1770 algorithm, which weights the signal the way an ear does and reports a single number in LUFS (ITU-R BS.1770). That weighting is built on the ear's insensitivity to low frequencies, the same fact the equal loudness contours in ISO 226 describe, where a 40 Hz tone needs far more pressure than a 1,000 Hz tone to sound as loud (ISO 226:2003). The skin has the opposite curve. The Pacinian channel that senses vibration is most sensitive near 250 Hz and less sensitive at the bottom, according to the four channel measurements of Bolanowski and colleagues (Bolanowski et al., 1988). A mix that looks balanced on a loudness meter will feel top heavy on a bed: the head zone in the hundreds of hertz arrives strong, the feet zone at 41 Hz arrives faint. So each zone gets its own gain curve, set by feel rather than by LUFS, and the low zones are driven much harder relative to the high ones than any speaker mix would tolerate.

Align the arrival

Sound in air travels at about 343 metres per second, so a wave crossing the 1.8 metres from a listener's feet to their head takes about five milliseconds (HyperPhysics). Through the body the story is slower. Low frequency vibration moves through soft tissue as a shear wave at roughly one to ten metres per second, the range that ultrasound elastography relies on to measure tissue stiffness (Sigrist et al., 2017). A pulse fed only to the feet would reach the chest through the body a large fraction of a second later, long after the ear had heard it. That is why the zones are not fed in sequence. All five are driven from one clock, and each zone's delay is set so that a kick drum lands at every point on the body within the same few milliseconds as it lands in the ear. The body should feel one event, not five.

Shape the transients

A voice coil has a power rating for a reason. Manufacturers of tactile transducers quote a continuous, or RMS, figure that the coil can turn into heat indefinitely, and a higher peak figure it can survive for an instant (Dayton Audio, bass shakers). AVA's transducers are rated at 25 watts RMS into 4 ohms. A kick drum transient at full scale is a request for far more than that for a few milliseconds, and a track full of them is a request to cook the coil. Katz describes how mastering limiters trade peak level for density and how easily that trade damages a mix (Katz, 2015). For the body we make the trade deliberately and differently per zone. The belly zone, carrying drums, keeps a fast attack so the hit stays a hit. The feet zone, carrying sub-bass, gets a slower attack and a longer release, because a 41 Hz wave takes 24 milliseconds to complete one cycle and a transient shorter than that is wasted energy and heat.

Why headphones lie about bass

Headphones can reproduce 20 Hz at the eardrum for almost no energy, because the driver sits in a sealed cavity the size of a thimble. A mix that sounds enormous on headphones may contain very little low frequency energy by the standards of a bed. The reverse is also true: a mix with the sub-bass set for a club system will, through a bed, feel like being sat on. Merchel and Altinsoy's listening tests found that adding a vibration channel derived from the low end of the music raised how people rated a concert recording, and that the level of the vibration relative to the sound was what decided whether it helped or intruded (Merchel and Altinsoy, 2018). The whole craft is in that ratio. The standard for whole body vibration exposure, ISO 2631-1, gives the engineering vocabulary for measuring what the bed does in metres per second squared (ISO 2631-1:1997), and the target on AVA is well below anything that standard would flag over an hour.

The body should feel one event, not five.

The opposite of Atmos

The most ambitious sound design of the last decade points the other way. Dolby Atmos treats sounds as objects placed in a three dimensional space around the listener and renders them to whatever speakers or headphones are present (Dolby Atmos). Its question is where a sound is around you. Felt playback asks where a sound is in you. The two are not rivals; a track can be mixed for both. But the design decisions are different in kind. Spatial audio spends its effort on the ear's ability to locate a source. Body audio spends it on the skin's ability to tell a bass line from a drum, and on not hurting a voice coil in the process.

What this means on AVA

When a track is prepared for the octagon at 29 Nassau Ave, it is split into five bands for five zones, aligned from feet to head, shaped so that each of the 33 transducers works inside its 25 watt limit, and balanced by feel rather than by a loudness meter. The generative visuals are driven by the same divided signal, so that the light that moves when the bass moves is the bass. None of this claims to change what music does to a person. It only tries to deliver the music the way the body reads it, which is not the way the ear does. This is wellness content, not medical advice.