Audio Upmixing
Audio upmixing derives a presentation for more output channels from a source with fewer channels.
A stereo song sends a stable voice to the front center and ambience into the surrounds when a receiver's upmixer is enabled. The processor is deriving an expanded presentation from the relationships in its available inputs.
It is useful to inspect what it derived, rather than treating every active output as a newly discovered source channel.
Start with shared and differing content
A simple mid/side analysis illustrates relationships an upmixer can exploit:
M = (L + R)/2; S = (L − R)/2.
| Stereo pair | Shared component M | Difference component S |
|---|---|---|
| L=0.6, R=0.6 | 0.6 | 0 |
| L=0.8, R=0.2 | 0.5 | 0.3 |
| L=0.4, R=−0.4 | 0 | 0.4 |
A shared component can be useful for deriving a central image. Differences and less-correlated content can help derive width or ambience. These are only analysis illustrations: commercial upmixers use more elaborate time/frequency analysis, steering, and decorrelation.
The processor must also decide how much derived center to remove from or retain in the fronts, how to preserve level, and how to avoid distracting motion when several sounds coexist. Merely copying M into another speaker is not a complete natural-sounding upmix.
Directional cues and independent signals are different resources
With intentionally matrix-encoded material, a decoder has relationships prepared for surround recovery. With arbitrary stereo, the upmixer infers a useful presentation from relationships that may have many origins.
Two matching channels can represent one centered voice, several centered instruments, or a mono recording duplicated into stereo. The same inputs do not uniquely identify separate studio stems.
This is why a stereo downmix followed by upmixing need not match the original multichannel mix. The earlier combination removed independent signals that the analysis cannot always distinguish again.
Expand a conventional surround presentation
An upmixer can also start with 5.1 and derive rear or height activity for a larger system. Existing discrete positions remain more information than ordinary stereo, but the additional outputs still come from a chosen rendering/analysis method.
Dolby Surround and DTS Neural:X are examples of such processing. Native object rendering instead has supplied spatial instructions where the source and path provide them.
Dialogue placement, movement, and ambience
Use the same passage and a sensible level comparison. Listen for dialogue that stays anchored, movement that remains coherent, and ambience that surrounds without pulling a foreground voice behind the listener.
For an Emby session, identify the receiver's detected source presentation separately from its upmix mode. A 2.0 input with expanded outputs is a legitimate upmix result; it is not evidence that the file secretly contained all those discrete channels.