Notes from the team

Engineering notes on audio QA.

How the checks work, what a failure actually means, and the occasional release note — written by the people building WaveLens.

Guides

Reading a WaveLens report: what each check actually tells you

Six checks, one verdict — a walkthrough of what Signal Alignment, Spectral Fidelity, Bit-Exactness, Level Fidelity, Distortion, and Loudness are actually measuring, and what a FAIL on each one really means for your file.

Engineering

Bit-Exact vs. "close enough": what audio QA actually means

Two files can sound identical and still not be the same file. Why bit-exactness is a stricter, more literal claim than "passes a listening test" — and when you actually need it.

Case study

Catching a codec regression before it ships

A nightly build comparison caught a level-fidelity drift that a listening pass missed entirely. What the report looked like, and how it changed the release checklist.

Engineering

BS.1770 loudness, explained — and why it fails silently

Loudness can be technically "in spec" and still sound wrong, or vice versa. What the BS.1770 measurement actually integrates, and where the check has blind spots.

Guides

Wiring wavsummary into a CI pipeline

wavsummary scans a directory of reports and exits non-zero on any failure. Gating a build on it, from a single CI step to a full release checklist.

Engineering

5.1 to stereo: what actually happens during a downmix

Folding six channels into two isn't just adding them together. What wavchan does during a downmix, and why comparing before/after needs its own tolerance.

Release notes

WaveLens v0.7.0: Loudness (BS.1770) and the Analysis Mode tile

This release adds a full BS.1770 loudness check and a new Analysis Mode tile on every report, showing whether a comparison ran in time-domain or frequency-domain mode.

Engineering

Bit-Exactness vs. Bit Accuracy — which do I want?

The two terms get used interchangeably and shouldn't be. A precise breakdown of what each one checks, and which one your QA process actually needs.

Engineering

wavgen: generating repeatable test signals for regression suites

Tones, sweeps, and multitone material need to be exactly reproducible run to run. How wavgen seeds its generators, and why that matters for a regression suite.

Guides

Peak vs RMS: what wavscan actually measures

A quick peak/level scan can mean two different things depending on which number you're reading. What wavscan reports, and when each measurement is the right one.

Case study

How one broadcast team cut QA review time from hours to minutes

A delivery pipeline that used to need a full listening pass per episode now gates on six automated checks. What changed, and what still gets a human ear.

Release notes

WaveLens v0.6.0: faster spectral analysis and a new --tolerance flag

This release speeds up the Spectral Fidelity check on large files and adds a --tolerance flag for teams that need more headroom than the strict default.

Engineering

Why two "identical" WAV files can still fail Bit-Exactness

Same sample rate, same bit depth, same waveform to the ear — and still a FAIL. The container-level differences that trip up a byte-for-byte comparison.

Stay in the loop

Get new posts in your inbox.

One email when we publish something worth reading — no spam, unsubscribe anytime.