HSR

THE PROBLEM

Why upmix at all?

Stereo works — two speakers, phantom images, sixty years of standard. But modern playback systems have more speakers: a soundbar carries 5 to 11 drivers, a premium car 8 to 30, a home theater 12 or more, a venue dozens. These speakers exist; the question is what to feed them.

Native multichannel content is the ideal answer — and an impractical one. Atmos mixing needs stems, time and expertise; back catalogs remain stereo, live sources arrive stereo, streaming music is overwhelmingly stereo. The content pipeline has not caught up with the hardware. Upmixing fills the system with what already exists: no content changes, no workflow disruption — the listener's stereo library becomes immersive automatically.

The industry's traditional answers both fall short. Mid/Side decomposition sends a mono L+R sum to the center (thin, phasey) and only the L−R difference to the surrounds (ambience without substance), with hard front/rear zones and phase cancellation on downmix. FFT-based upmixers separate more cleverly but pay 1024–4096 samples of latency, pre-echo that softens transients, and musical noise. Modern systems deserve better.

HOW IT WORKS

How it works

UNDER THE HOOD

Going deeper

The key insight: a stereo recording contains real spatial information. When a sound is panned left, it is louder in the left channel; when it is centered, both channels are equal; when it carries reverb, the channels are partially decorrelated. That information is precise and recoverable — if you analyse it correctly. HSR does so continuously, in the pure time domain, yielding a spatial energy distribution across the stereo width at every sample.

Because there is no FFT and no block processing, latency is 5 samples at any rate — 113 µs at 44.1 kHz, 26 µs at 192 kHz, some 200 to 800 times faster than frequency-domain upmixers — with zero pre-echo, zero musical noise, and a CPU footprint that fits embedded platforms.

The reconstruction is spatial, not zonal. A singer panned 20% left of center stays 20% left, distributed naturally across the front array with the correct ratios — where an M/S upmix would dump it in the center channel. A wide orchestra spreads continuously across the front speakers instead of splitting into hard left/right blocks around an empty middle.

Processing is iso-energy: total acoustic level stays constant through the upmix, so content fills the array without getting louder — broadcast loudness compliance holds, live sound brings no gain surprises. And fold the result back to stereo: you recover the master.

AT A GLANCE

At a glance

PARAMETERS

Hands on the algorithm

POSITIONING

Compared to the alternatives

APPLICATIONS

Where it fits