TUTTI · From Spatial Audio Research to Software
What changes at the speakers when we move a sound?
I measured Sound xR Image and L-ISA Studio to investigate their operating principles, then tested whether mathematical models could predict their behaviour in new speaker layouts.
From research to a working program
TUTTI · Immersive audio
Investigated the operating principles of Sound xR Image and L-ISA Studio through measurement and analysis, then applied the findings to develop TUTTI, an immersive audio program.
The browser-based Session Editor works without installation; live audio on a Mac requires the TUTTI Bridge app.
My contribution
- Wonder
- What changes at the speakers when we move a sound?
- Build
- Investigated the operating principles of Sound xR Image and L-ISA Studio through measurement and analysis, then applied the findings to develop TUTTI, an immersive audio program.
External tools & AI
Sound xR Image: Yamaha DME7 and Smaart. L-ISA Studio: Audio Bridge, BlackHole, and Python with NumPy, SciPy, and OSC control for loopback measurements and model comparisons.
- Find out
- The systems distribute sound differently. L-ISA Studio also changes frequency response and uses a delayed component when spreading sound. Predictions worked within defined conditions, with open questions in both systems.
Sound xR Image
How does Sound xR Image work?
I changed sound position, speaker layout, and renderer settings on a Yamaha DME7, and measured each output level with Smaart.
A model based on distance explained selected measurements well. Changing the layout and speaker IDs exposed behavior the model could not yet predict.
L-ISA Studio
What changes when a sound gets wider?
I used digital loopback sweeps to measure output levels, frequency responses, and delays while changing position, distance, height, and width.
The measurements showed how sound is shared between neighboring speaker directions, with different processing across frequencies. Widening a sound also introduced a short delayed component; its full distribution remains unresolved.
What the two studies revealed
| Aspect | Sound xR Image | L-ISA Studio |
|---|---|---|
| Panning model | DBAP · distance-based | VBAP · direction-based |
| Speaker distribution | Distance-weighted levels across multiple speakers | Up to 2 speakers in 2D / 3 in 3D, with Width = 0 |
| Size / width | Size changes the level distribution | Width adds angular spread and a short delayed component |
Summary of the tested conditions. DBAP = Distance-Based Amplitude Panning; VBAP = Vector Base Amplitude Panning.
How I checked the explanations
I compared competing models and recorded predictions before testing new conditions. When a prediction failed, I changed one factor at a time to investigate why. The reports distinguish models fitted after measurement from predictions tested on new conditions.
Measurement details and limits
Sound xR: a distance-based model matched selected normal-output measurements within about 0.005 dB. Two new five-speaker tests had maximum prediction errors of 9.29 and 8.67 dB. A later fit explained one test, but did not establish a general rule.
L-ISA Studio: Controller 3.2.2 / Processor Desktop 3.2.2, 48 kHz: the model predicted selected new layouts and random settings with Width = 0 within 0.03 dB, measured in a one-third-octave band around 1 kHz.
An initial L-ISA Studio height prediction failed by up to 54.4 dB. Follow-up tests supported a speaker-UUID-dependent triangulation rule in the tested layouts. One mirrored-layout exception and larger coplanar faces remain unresolved.
L-ISA Studio width was only partly identified. Delay behavior applies to the tested Desktop configuration. These electrical measurements do not establish listening quality or performance across audience seats.
Sound xR was studied mainly through static output levels. Its frequency and delay behavior was not independently remeasured for this comparison, so the study cannot establish that these forms of processing are absent.
Reports & data
Methods, equations, prediction results, and unresolved questions. Both studies have reports in Korean and English.
Next questions
How does L-ISA Studio distribute a wide sound in less regular speaker layouts? How do both systems respond while a sound is moving?
Where the study began
This study began after using Sound xR Image for an outdoor performance. I wanted to connect the controls on screen with what each speaker actually receives. The images below show the Sound xR work and the L-ISA Studio Soundscape interface.
I extended the investigation to L-ISA Studio with digital loopback measurements. Both studies compare measured outputs with mathematical models; they describe the tested software and settings. Listening tests and measurements across audience seats remain future work.
Investigated the operating principles of Sound xR Image and L-ISA Studio through measurement and analysis, then applied the findings to develop TUTTI, an immersive audio program.
