ISE - Immersive Sound Engine
Immerse yourself into sound.
`
A gain and delay spatialization algorithm made for Hi-Fi and Live Events
Introduction
Following our development on HSR (High Space Resolution) algorithm, we faced the necessity to develop an algorithm capable of defining the spatial placement of sound sources precisely, with a low CPU usage.
Many algorithms are already in use on the market, using gain-based (VBAP, KNN) or time-based (WFS) technologies. However, those algorithms are not precise enough for our requirement or too much CPU expressive to be used on our partners applications (Embedded DSP, VST plugins, hardware solutions). we then had to find a solution to combine the precision of certain systems with the lightness of application of others: thus began the development of ISE.
Why use a spatialization algorithm ?
Erase the sensation of the speaker
In Hifi as in Live events, we usually rely on stereophonic systems, composed by a stereo sound sound diffusion and a subwoofer array. This system, utilizing a pair of speakers, creates a sense of sound positioning that spans between the left and right speakers. This effect is achieved by generating a 'phantom image,' utilizing differences in sound levels (for instance, a slightly left-positioned sound results in the right speaker receiving slightly less sound compared to the left).
However, stereo's effectiveness is reliant on listeners being centrally positioned in the venue. Deviation from this central point disrupts accurate localization perception due to the intricacies of our auditory perception.
Thus, when listeners move to the side within the diffusion system, the nearest speaker's sound reaches their ears first, leading to a singular perception of sound from that direction, nullifying the stereo effect of left-right panning.
Consequently, the accuracy of sound localization in stereo systems is limited to when listeners occupy the central 'sweet spot' in the audience, posing a significant limitation in stereo system performance.
With a spatialization algorithm, we replace the dissuasion system by an array of speakers that tend to replicate how a sound source will behave in an acoustic environment. Thus the sensation of hearing a loudspeaker disappear.
Place precisely a virtual sound source on a real environment, wherever the audience is
As it centers on replicating a natural wave pattern through precise timing delays and varying levels, this algorithm ensures that regardless of their location within the venue, all audience members experience sound emanating from their intended direction. This approach eliminates the concept of a 'sweet spot,' granting nearly every position of the audience gets an optimal listening position.
An aside on psychoacoustic
Our brain has two ways to define a sound source in his environment :
By measuring the time difference between out left and right ears from the sound emitted by the source. We call it Inter-aural Time Difference (ITD). The ITD is a fundamental index for locating a source emitting waves whose frequency is less than 1500Hz. Above 1500Hz, the ITD becomes ambiguous.
By measuring the gain difference between out left and right ears from the sound emitted by the source. We call it Inter-aural Intensity Difference (IID). The head tends to oppose the passage of the sound wave; we also speak of head shadow effect. Thus, an inter-aural intensity difference (IID) appears, which reflects the difference in intensity of the wave that reaches each ear.
However, this index strongly depends on the frequency of the incident sound. When it is below 1500Hz, the IID is almost non-existent. On the other hand, for frequencies above 1500Hz, the IID becomes an effective index.
To precisely place a sound source in a 3D virtual environment, we know we had to make an algorithm that took into account the time and gain differences of a signal, but this time going trough a sound diffusion system, consisting of multiple loudspeakers. ISE therefore manages time and delay to achieve an accurate sound reproduction.
Gain Management
ISE includes two modes of gain calculation :
- One mode taking into account the distance and applying the inverse square law to define the gain loss for each source on each speaker. Thus for doubling the distance 6db will be lost.
- Another mode called « Surround » takes the azimuth of the source and those of the speakers and define a gain law based on those ones. Thus a source placed on the opposite azimuth of speaker will not play on this particular speaker. A gain-loss parameter allow to define the ratio between an azimuth delta and a gain loss.
For both modes, we can choose between the calculated values, or a normalized-one. The last one will listen to each gain loss and add the lower value to each speaker.
An option between 2D and 3D algorithms is available as well (the 2D version will neglect the Z-axis for the gain loss calculation).
Finally, an offset is available if we need to add or remove a few db to match a stereo reference.
Delay Management
The delays are calculated with the real delta from source to loudspeaker position. However, you can scale those values to accentuate the effect or move back a group of speakers (for ceiling-mounted speakers for exemple).
As well as for the gain values, the delays values can be used as is or can me normalized, to reduce the latency of the system.
An option between 2D and 3D algorithm is available as well (the 2D version will neglect the Z-axis for the gain loss calculation).
Licensing
As for HSR, we’re selling ISE as a one-time payment for one brand. With this licence you will have access to ISE source code and DSP code, for you to use as you would like.
If you’re planning to sell parts of ISE code in B2B solutions, you will need to buy one licence per brand you’re planning to sell. For this type of use we’re making discount on multiple licence payement.
If you want to test ISE, please contact us and we will send you a demo to test.