Publika Team

Loudness readings in Build 7 are about 0.26 LU low

What happens: in Build 7, the current public build, every loudness reading is about 0.26 LU below the true value, and loudness normalisation computes its gain from that reading. So a master normalised to −23.0 LUFS in Build 7 lands at about −22.74 LUFS.

Why: the K-weighting filter inside the meter was built from the right parameters with the wrong formula, so it measured slightly low. Our reference tests against the ITU-R BS.1770-4 tables caught it.

How much it matters: the error sits inside EBU R128's ±0.5 LU tolerance, but a strict delivery spec can still notice it. If yours is strict, measure the delivered file with a second meter you trust before you send it.

Where it stands: the filter is corrected in our code for the next build. A reference test tone now reads −23.0000 LUFS where Build 7 reads −23.2577. It isn't in a released build yet, and we're also checking that timelines measured in Build 7 get re-measured after the update. This thread will say Fixed in the version that carries it.

If you've had a master questioned or bounced over loudness, press I have this too and reply with the spec you were delivering to.