“But the car door slammed halfway up the wall!”
This was the reaction one person had after listening to a scene of a radio drama by one of my colleagues from the MA I did a few years ago (listen to it here) that played using 3rd order Ambisonics in the Sonic Lab in SARC. The thing is, this elevation of some sounds is present even for a 2D representation.
The same illusion was noted by several participants in a listening experiment I carried out a couple of weeks ago, when using first order max rE decoding.
Illusion of elevation with Ambisonics, especially lower orders, is well noted in the literature. However, for my experiment it was evident that not everyone experienced it the same way – some had elevation at the sides, others at the front and some had none at all.
By third order it’s not very noticeable, especially with Max rE. However, with first order (4 or 8 loudspeakers) it’s quite obvious for certain source positions. For those who’ve studied Ambisonic theory, it can be predicted by the length of the energy vector. This is why it reduces for higher orders, as the magnitude of the vector approaches unity.
But it’s not just an Ambisonics thing, it’s also present with stereo and other reproduction formats.
For example, I was listening to Jethro Tull’s ‘Thick as a Brick’s (one of the best albums ever released) a high-hat moved from right to left and it travelled in an arc, rising up and then back into the loudspeaker. Indeed, applying energy vector theory to stereo shows that the magnitude of the vector decreases as sounds are panned to the centre. This is consistent with my experience of the arc shaped trajectory.
There’s a certain amount of frequency dependence to this effect. I’d not expect a bass guitar to travel in the same arc as the higher frequency cymbals and high-hats.
Sadly, it’s not enough to look just at the energy vector magnitude, afterall, for 1st order Ambisonics over a square array the vector mmagnitude is the same for all angled but the elevation seems to change with source position. There’s obviously more going on than the vector theory can explain.
Is this illusion of elevation flaw or a feature to be exploited? I’d say that its unpredictability is a bigger flaw. If that could somehow be stabilised and controlled then it would be a great tool to use when access to elevated loudspeakers isn’t possible.