Stereo width / M-S analysis — phase correlation & mono compatibility check

Stereo width analysis happens at the pre-release stage

Checking stereo width is a final-stage quality gate. You run it on the candidate master to confirm that the image holds up in mono, that the phase correlation coefficient doesn’t go negative, that the Side energy isn’t so dominant it collapses on a Bluetooth speaker. The file on the table is the unreleased version — the one with the most value and the least exposure to outside parties so far.

The analysis result itself — stereo width percentage, phase correlation number, mono drop — quantifies properties of the mix that are specific to this recording. A server that stores this metadata alongside the audio has a fingerprint of the mix. For unreleased work, that’s a fingerprint of something that isn’t supposed to exist outside your machine yet.

Cloud stereo-analysis tools and mastering NDA situations

Mastering engineers sometimes work under NDAs on behalf of labels or artists. Running an NDA-covered master through an online tool — even for a simple correlation measurement — transfers the file to infrastructure outside the agreed handling scope. RMS and correlation are lightweight computations, but they still require the full audio file to be received server-side.

The gap between ‘I’m just checking stereo width’ and ‘I’m sending an unreleased master to a third-party server’ is the same gap that exists for every audio utility that processes server-side. The smaller the task feels, the less likely the transfer registers as significant. That’s exactly when it’s worth pausing.

Matrix operations on decoded samples — entirely inside the browser

The tool decodes audio with decodeAudioData, extracts L and R sample arrays, and computes Mid = (L+R)/2 and Side = (L-R)/2 in JavaScript. RMS is computed for each, yielding the stereo width ratio. The Pearson correlation coefficient between L and R is calculated per-frame and averaged, giving the phase correlation number. Mono drop in dB follows from the correlation. No data leaves the browser.

Open DevTools Network and drop a file: no requests after page load. The Mid/Side matrix operations and the Pearson calculation are in the GitHub source.

Making stereo width part of a standard pre-release checklist

Pairing this tool with audio-loudness-lufs and audio-true-peak covers the three main technical checks before delivery: loudness, inter-sample peak, and stereo compatibility. Running all three in the browser, without uploading the file anywhere, means the master doesn’t leave your machine until it reaches the distributor.

For content that will be heard on Bluetooth speakers or car radios — both mono environments — a phase correlation below +0.3 is worth investigating before release. To pin a problem to one channel, audio-channel-split breaks the file into separate L and R mono files; audio-channel-merge rebuilds the stereo after the fix. Catching it here is far easier than remixing after distribution.

Mid-Side encoding and the math behind Pearson correlation

Mid-Side (M/S) is a classic lossless matrix transform. Given L = (M+S) and R = (M-S), you can recover the original stereo exactly from Mid and Side channels. (L+R)/2 extracts the centred content (Mid) and (L-R)/2 extracts the off-centre content (Side). The same transform powers FM stereo broadcast (the 38 kHz subcarrier carries Side), and MP3 / AAC Joint Stereo modes use it for encoding efficiency. When Side RMS exceeds Mid RMS, mono summation cancels content that was present in the stereo image — the source of phase issues.

The Pearson correlation coefficient r = Σ((L-L̄)(R-R̄)) / √(Σ(L-L̄)² × Σ(R-R̄)²) measures how closely L and R move together, on a scale of -1 to +1. r = +1 means perfect alignment (effectively mono), r = 0 means uncorrelated (extreme stereo width), r = -1 means perfect anti-phase (complete cancellation in mono). The working thresholds most engineers use: r ≥ +0.5 safe, +0.3 to +0.5 watch carefully, < +0.3 mono-compatibility risk. This tool averages frame-by-frame correlation for the overall score and flags negative instantaneous peaks as phase warnings.

Streaming-era stereo practices after the loudness wars

Since streaming platforms (Spotify, Apple Music, YouTube) adopted loudness normalisation around -14 LUFS in the late 2010s, mastering has shifted away from “as loud as possible” toward preserving dynamics and stereo width. Mixes with stronger Side energy are more common now, but Bluetooth speakers, car infotainment, and phone speakers still collapse to mono in practice — making pre-release mono compatibility checks more important, not less.

Modern EDM and pop mixes often push stereo width to 110–130%, increasing mono risk in exchange for impact on stereo systems. Classical, jazz, and live recordings still use Mid-Side miking (a figure-eight capsule capturing Side, a cardioid capturing Mid) to allow independent post-production control of stereo width. Being able to inspect Side RMS and Mid RMS separately in-browser is also useful for verifying the M/S ratio captured at the session against what survived the mix.