what is stereo width

the illusion between two speakers

you are sitting in the sweet spot. two speakers, equal distance away, angled toward you. a vocal plays and it sounds like the singer is standing directly between the speakers, dead center. but no one is there. no speaker exists at that point in space.

that phantom center is where stereo begins. your brain takes the identical signal from both speakers and fuses it into a single image, localized at the midpoint. this is not a trick of the recording. it is a trick of your auditory system: when both ears receive the same signal at the same time, the brain places the source on the median plane.[^1]

stereo width is what happens when left and right are not identical. the more the two channels differ, the wider the sound appears. a dense reverb tail with different reflections in each channel sounds wide. a close-miked vocal with no room sound, panned center, sounds narrow. the difference between the channels IS the width. (once you see it this way, a lot of mixing decisions start making more sense.)

recorded vs created stereo

stereo information enters a recording in two fundamentally different ways.

recorded stereo is captured. a spaced pair of microphones picks up real timing differences from the room: sound reaches the left mic a fraction of a millisecond before the right, and those tiny delays encode the spatial information of the acoustic space. a coincident pair (X/Y or Blumlein) captures level differences from the directional response of the microphone capsules. either way, the stereo field is a byproduct of physics. it contains genuine spatial cues that your brain interprets naturally.

created stereo is synthesized. a mono vocal panned center has no width. to make it wide, you need to introduce differences between left and right that were not in the original signal. this is where processing techniques come in: Haas delay adds a short time offset, M/S processing boosts the Side channel of an already-stereo signal, allpass decorrelation modifies phase relationships without changing the frequency content.

the distinction matters because created stereo always involves a trade-off. recorded stereo is inherently mono-compatible (the spatial cues were additive). created stereo can introduce phase relationships that cancel when the left and right channels are summed.

five things stereo width is not

1. wider is not always better. an overly wide mix sounds hollow. the centre image loses focus. the vocal floats instead of anchoring. and when someone plays it on a phone speaker, the carefully widened elements disappear. width serves the arrangement. wide pads framing a focused vocal is a choice. everything at maximum width is a mess.

2. panning is not width. panning moves a sound to one side by changing its level in each channel. a mono snare panned hard left is not wide. it is displaced. width requires the sound itself to occupy space, to have decorrelated content that spans the field between the speakers.

3. headphone width is not speaker width. headphones eliminate the acoustic crosstalk that defines speaker stereo. on speakers, each ear hears both channels (the right ear hears the left speaker, slightly delayed and attenuated by the head). on headphones, each ear hears only its channel. a mix that sounds beautifully wide on headphones can have severe phase problems on speakers, and vice versa.

4. all wideners are not the same. Haas delay, M/S boost, pitch shifting, allpass decorrelation: these are fundamentally different processes with different trade-offs. some destroy mono compatibility. some preserve it. some change the timbre. some do not. treating them as interchangeable is a common mistake.

5. a mono source is not hopeless. you can create convincing width from a mono signal using decorrelation techniques. the result is synthesized width, not captured spatial information, but done well it sounds natural and translates to all playback systems.

the mono problem

when a stereo signal is played on a mono speaker, the left and right channels are summed. anything identical in both channels survives. anything opposite cancels.

this is not an edge case. phone speakers, bluetooth speakers, club PA systems with mono subwoofers, shopping mall installations, many broadcast scenarios: mono playback is everywhere. if your stereo width relies on phase differences between left and right, those differences cancel on summation. the affected frequencies lose energy, sometimes completely.

the classic example is naive Haas widening. you duplicate a vocal, delay the copy by 15 milliseconds, and pan it opposite. on stereo speakers, the result sounds wide and spacious. on a mono speaker, the original and delayed copies sum together and create comb filtering: deep notches at frequencies where the delay causes destructive interference, spaced at intervals of 1/delay.

three approaches to creating width

not all stereo widening techniques are equal. the three most common approaches differ fundamentally in how they create the left-right difference and what they sacrifice.

Haas delay duplicates the signal, delays one copy by 1 to 35 milliseconds, and pans the original and copy to opposite sides. the precedence effect fuses the two into one image, while the delay creates a sense of spaciousness. the problem: on mono fold-down, the delayed and original copies create comb filtering. the longer the delay, the closer together the comb notches. Haas is effective but inherently mono-unsafe.

M/S boost encodes the stereo signal into Mid (what both channels share) and Side (what differs between them), then boosts the Side. this widens existing stereo content. the limitation: it cannot create width from a mono source, because a mono signal has zero Side content. boosting zero is still zero. and excessive Side boost pushes the correlation toward zero or negative, which causes the same mono-cancellation problems.

allpass decorrelation passes one channel through a cascade of allpass filters that modify the phase response without changing the magnitude response. the left and right channels end up with different phase relationships across the spectrum, which the brain interprets as width, but because the magnitude is unchanged, mono fold-down produces a flat response with no comb filtering.

how wide should your mix be

there is no universal answer. the right width depends on the arrangement, the genre, and the playback context.

a dense electronic production with layered synths benefits from a wide stereo field. the space helps separate elements that would otherwise fight for the same frequency range. a sparse acoustic ballad with vocal, guitar, and bass benefits from a focused, intimate image. forcing it wide adds emptiness, not spaciousness.

the principle: width should serve the music. stereo is staging. the director decides who stands where and how much space they take. wide pads framing a focused centre vocal is a classic and effective spatial design. a wide kick drum is almost always a mistake. bass content below 150 Hz should generally be mono, because stereo bass is both perceptually useless (your ears cannot localise it) and dangerous for mono playback.

use your correlation meter as a guide. a full mix should generally sit between +0.3 and +1.0 on the correlation scale. individual elements can dip lower, but the overall mix needs enough positive correlation to survive mono playback. if your master bus correlation drops below +0.3, something in the stereo field is too aggressive. if it goes negative, you have a phase problem that will cost you on every mono playback system.

frequently asked questions

frequently asked questions

what is stereo width in audio?

stereo width is the perceived horizontal extent of a sound between two speakers. it is determined by the differences between the left and right channels, not just panning. the more decorrelated (different) the left and right signals are, the wider the sound appears. a perfectly identical signal in both channels sounds mono, no matter how loud it is.

what is the difference between panning and stereo width?

panning moves a sound left or right by changing its level in each channel. width is about how spread out a sound feels between the speakers. you can have a sound panned center that still feels wide (like a stereo reverb), and you can have a sound panned hard left that feels narrow (a mono signal). panning controls position. width controls spaciousness.

why does my wide mix sound thin on phone speakers?

phone speakers and many bluetooth devices play in mono, which sums the left and right channels together. if your stereo width comes from phase differences (like a Haas delay), those differences cancel when summed. the result is partial or complete loss of the affected signal. always check your mix in mono to catch this before it ships.

can you make a mono source sound stereo?

yes, using signal processing techniques like allpass decorrelation, Haas delay, or chorus effects. these create controlled differences between left and right channels from a single source. the result is synthesized width, not captured spatial information. the quality of the result depends on the technique: some destroy mono compatibility, others preserve it.

what is a good stereo correlation value for a mix?

for a full mix, aim for correlation between +0.3 and +1.0. values above +0.3 are safe on all playback systems. between 0 and +0.3 is a caution zone where mono playback will lose some width. below 0 means the left and right channels are more different than similar, and mono fold-down will cause audible cancellation.

references

a note from the developer

stereo width is one of those topics where the more you learn, the more you realize how much your ears were doing without your conscious awareness. the phantom center, the precedence effect, the way your brain extracts spatial information from tiny timing differences: it is elegant and complex and endlessly fascinating.

when i started building KERN WIDE, i spent months reading psychoacoustics research before writing a single line of DSP code. the question that kept coming back was simple: how do you make something sound wider without breaking it on mono systems? that question led me to allpass decorrelation, to ERB-domain frequency analysis, to perceptually-weighted width curves based on how your ears actually process spatial information at different frequencies.

full disclosure: i make KERN WIDE. it uses allpass decorrelation, which is the approach this article describes as the most mono-compatible. i believe that is accurate based on the research, but you should evaluate it the same way you evaluate any tool: does it sound good on your material, and does it translate to every playback system you care about?

if i got something wrong, or if you think about stereo differently than i described it here, jonas@kernaudio.io. i am always learning.