RESEARCH — REVERB ALGORITHM
HOA-X
High-Order Ambisonic Upscaler
HOA-X raises the order of an ambisonic recording — first order to third and beyond — by analysing where the sound comes from and re-synthesising a sharper field. It reuses two engines you already trust: the correlation split of our upmix and the direction analysis of our array decoder, delivered at zero added latency.
THE PROBLEM
Low order is portable but blurry
Ambisonics is the lingua franca of spatial audio — format-agnostic, rotatable, decodable to any layout. But its spatial precision is set by its order: first order (four channels) is cheap to record, store and transmit, and it is also the blurriest — sources smear over a wide arc and the sweet spot is small.
Higher order sharpens the image, but it costs channels and, at capture, a microphone most productions do not own. So an enormous amount of content — game ambience beds, broadcast feeds, field recordings, archives — is stuck at first order, and the usual order conversion only goes the easy way, downward. Raising the order is left as an unsolved step, or a naive one.
HOW IT WORKS
How it works
Analyse, don't guess
Per band, HOA-X estimates the direction of arrival and how diffuse the field is — the same active-intensity analysis our array decoder already runs.
Sharpen the image
The directional part is re-encoded at the higher order so sources tighten, while the diffuse ambience is spread and decorrelated to stay natural.
Zero added latency
Only the parameters are smoothed, per short block; the shaping and decorrelation run through our zero-latency convolution — the audio never waits on an FFT.
UNDER THE HOOD
Going deeper
HOA-X treats order raising as a parametric problem rather than a matrix one. In each time-frequency tile it analyses the incoming field for two things: the direction the energy comes from, and how diffuse the field is around it — the active-intensity analysis our array decoder already performs. That splits the tile into a directional part and a diffuse part, iso-energy, with the very same coherent/diffuse law that powers our stereo upmix.
The two parts are re-synthesised at the target order. The directional part is re-encoded as a point source at the estimated direction, now with the higher-order spherical harmonics — which is exactly what tightens the image. The diffuse part is spread across the new channels with a physically correct per-order energy balance and mutually decorrelated, so the ambience stays enveloping instead of collapsing to a point.
Because only the parameters are smoothed — over a few milliseconds, as in the upmix — while the actual shaping and decorrelation run as filters through our zero-latency convolution, the whole upscaler adds no latency to the audio path. And it is honest about its limits: it cannot invent detail that was never captured, so it is strongest on scenes of a few directional sources over diffuse ambience, and degrades gracefully toward the original where a field is too complex to analyse.
AT A GLANCE
At a glance
| Input | First-order / low-order ambiX (FOA and up) |
| Output | Up to 3rd order — and beyond |
| Latency | Zero added — parametric + zero-latency convolution |
| Implementation | Parametric (DirAC-family) · C/C++ · embedded-ready |
| Availability | Ripl — in development · prototype |
PARAMETERS
Hands on the algorithm
| Input order | First order and up |
| Target order | 2nd, 3rd and beyond |
| Analysis | Direction + diffuseness, per band |
| Diffuse balance | Directional vs diffuse trim |
| Delivery | Zero-latency convolution |
| Formats | ambiX — ACN / SN3D |
POSITIONING
Compared to the alternatives
vs a higher-order microphone
No 3rd-order mic on the shoot? HOA-X raises a first-order capture after the fact — most of the sharpening, none of the hardware.
vs naive re-encoding
Padding a low-order signal with zeros or interpolating channels cannot add spatial resolution; HOA-X analyses and re-synthesises, so the image actually tightens.
vs leaving it first order
A first-order bed stays wide and vague on headsets and dense layouts; HOA-X recovers focus without re-authoring the scene.
APPLICATIONS
Where it fits
Gaming & XR
First-order ambience beds and captures gain sharper localization for headset rendering, without re-authoring the scene at a higher order.
Broadcast & archive
Legacy first-order ambisonic material is upscaled to higher-order delivery, recovering focus a low-order mix had lost.
Spatial music production
Ambisonic microphone takes — often first order — are tightened before mixing, so the recorded scene reads with more precision.
Research & auralization
Higher-order fields synthesised from compact captures for listening tests and VR acoustics, where high-order mics are impractical.
INTEGRATION
Built to live inside your product
| Delivery | C · C++ · MATLAB · .dsp — full source code |
| Platforms | macOS · Windows · Linux · embedded ARM |
| DSP platforms | Flow DSP · Audio Weaver — solutions in preparation |
| Documentation | Whitepaper — every algorithm explained, in the clear |
OEM LICENSING
- ■ One-time payment per brand
- ■ Full source code — C, C++, MATLAB, .dsp
- ■ Whitepaper — all algorithms explained
- ■ Integration support included
- ■ Free updates
- ■ Volume discounts on multiple licenses
SHIPS IN