Stereo Reproduction

Two-Channel (2-0) Stereo

The Stereo Listening Triangle

Equilateral triangle, speakers at ±30°
Sit at the apex, ears at tweeter height

Two loudspeakers and a listener forming an equilateral triangle, each speaker 30 degrees off centre.

Summing Localization

Each ear hears both speakers
The far one arrives about 0.26 ms later

Both ears receiving sound from both loudspeakers, the far speaker arriving delayed.

Figure: Rumsey, Spatial Audio, Fig. 3.2. This is interaural crosstalk: sound from the right speaker reaches the left ear delayed by δt relative to the right ear, and vice versa.

Summing localization is the name for what happens when coherent signals arrive at the ears within about 1 ms: we hear one fused event, a phantom source, and its direction depends on the level and time differences.
Two ways to steer a phantom image between the speakers:
- Amplitude difference between channels. This is what a pan pot does.
- Interchannel time difference below roughly 1 ms. This is what spaced microphone arrays do.
Rough numbers worth knowing: about 7-8 dB, or about 0.5 ms, moves the image roughly halfway to one speaker. About 15-18 dB, or about 1.0-1.5 ms, pins it fully at that speaker.
Past that window the precedence effect takes over, which we covered on 9/11: we localize toward the first arrival rather than hearing a phantom between the two.
Headphones are a different case. Each channel goes to one ear only, so there is no crosstalk and no summing. Amplitude panning gives in-head lateralization instead of a source out in the room. That is the externalization problem from the psychoacoustics deck.

Real vs. Virtual Source

A real source captured by a coincident figure-eight pair and reproduced as a virtual source between two loudspeakers.

Phantom and Discrete Images

Discrete (hard-panned, or a real center speaker): clearer, more precise, smaller
Phantom: larger, less focused, slightly colored

Left panel: both loudspeakers reach the left ear, the far one 0.26 milliseconds late after passing around the head. Right panel: the resulting comb response, with a cancellation notch of about 11 decibels near 2 kilohertz and shallower notches above it.

A discrete image comes out of one speaker only: hard panning, or an actual center channel. No crosstalk summing, so no coloration.
A phantom image is built by summing localization, not by comb filtering. Get the causal order right: summing localization creates the image; crosstalk comb filtering colors it.
The speakers are not doing any of the work here. For a center-panned source there is no interchannel level difference and no interchannel time difference: both speakers emit the same signal at the same moment. Every difference in the left panel comes from the geometry of the listener’s head.
Keep two comparisons apart, because this is the step students conflate:
- ITD and ILD compare the left ear with the right ear. For a centered listener and a center-panned source both are zero, which is exactly why the image sits in the middle.
- The crosstalk delay compares two arrivals at the same ear: its near speaker first, then the far speaker. That is what combs the spectrum.
They come out to the same 0.26 ms, and not by coincidence. By mirror symmetry the path from the right speaker to the left ear equals the path from the left speaker to the right ear, so one number serves both comparisons. Say this out loud; the coincidence is what makes the two look like the same thing.
Speaker angle sets how big the delay is, so it sets the notch frequency: about 0.20 ms and a notch near 2.5 kHz at ±22.5°, 0.26 ms and 1.9 kHz at ±30°, 0.38 ms and 1.3 kHz at ±45°. A wider setup drags the dip toward the vocal range. Head size does the same thing, which is part of why listeners do not all hear this coloration identically.
Once the listener moves off center, or the source is panned away from the middle, the symmetry breaks: now there is a genuine difference between the ears, and each ear gets its own asymmetric comb.
Where the coloration comes from: the crosstalk delay of about 0.26 ms puts the first cancellation notch near 1/(2 × 0.26 ms) = 1.9 kHz, with further notches at 5.8 and 9.6 kHz that grow shallower as head shadow attenuates the crossed path. That broad dip around 2 kHz is why a phantom center vocal or snare sounds slightly hollow next to the same sound from a real center speaker. Room reflections fill it in somewhat.
Processing can push images slightly outside the speaker pair: shufflers, crosstalk cancellation, stereo wideners. These are also the processes most likely to collapse or thin out in mono, so check.

The Stereo Soundstage

Image Width

Mono point sources: centered, off-center, or hard to one side
Spread images: narrow or wide, symmetric or asymmetric

Table of mono and stereo pan positions and the stereo image each one produces.

Listen: panning

Point source or spread image?

Companion audio examples 6.1 to 6.8, on stereo image width and panning.

Mic It! companion audio

Beyond the Loudspeakers

Decorrelation widens: different L and R signals, from short delays, chorus, or stereo reverb
Polarity inversion widens too, but cancels in mono
Always check the mono fold-down

Depth in the Stereo Field

You already know the distance cues. Now place them on the stage.

Direct-to-reverberant ratio does most of the work
Level sets apparent distance within a section
High-frequency rolloff supports the illusion

Listen: depth

Same source, three depths. What changed between 6.9 and 6.10?

Companion audio examples 6.9 to 6.11, on depth concepts.

Mic It! companion audio

Static and Dynamic Panning

Static: the position does not change through the mix
Dynamic: fly-bys, auto-panning, ping-pong and question/answer moves

Listen: static vs. dynamic

Track one moving element in 6.13. Does the movement serve the mix or distract from it?

Companion audio examples 6.12 to 6.14, on static and dynamic panning.

Mic It! companion audio

Image Symmetry

Balance weight on the left against weight on the right
Spread percussion across the field

Compare 6.15 and 6.16.

Companion audio examples 6.15 and 6.16, on image symmetry.

Possible Stereo Image

Two mix maps: a dense mix using the full soundstage, and a sparser, more conservative mix.