WPAudio EngineCase study

Mastering Second Flight

The second record through the same engine, a month after the first — and a different decision. Twelve finished mixes to one stated target, seven of them limited and five left untouched, every number measured twice before it was written down here.

Released masters12 tracks · 37:47 · released 11 September 2026

01

The problem

Wings & Prayers: Second Flight is the same twelve songs as the first record, performed again — released as The Little Blue Pilot on 11 September 2026. Twelve finished mixes arrived at 48 kHz / 16-bit and needed to leave as one coherent album, ready for distribution.

The first time, the engineer chose a level per track and let a few sit deliberately below the rest. This time the brief was simpler and stricter: every track to the same integrated loudness, the level the streaming platforms normalise to, so that nothing on the record is ever turned down on playback — and nothing is pushed harder than it needs to be to get there.

The tool was the same. The question was whether one target could be met across twelve mixes that did not arrive at one level, without flattening any of them.

02

Measure

The mixes as they arrived: integrated loudness, true peak, loudness range, and the number of samples already sitting at full scale.

BS.1770-4 measurements and full-scale sample counts of the pre-master mixes
#TrackLengthIntegratedTrue peakLRAClipped
01First Wings (Second Flight Version)3:00-14.0 LUFS-2.1 dBTP7.0 LU0
02Tailwind (Up, Up, Up) [Second Flight Version]3:58-13.7 LUFS+0.1 dBTP2.9 LU14
03The Little Blue Pilot (Second Flight Version)2:55-14.0 LUFS-3.0 dBTP6.9 LU0
04Read the Storm (Second Flight Version)2:41-14.6 LUFS+0.1 dBTP8.6 LU2
05Little Brave Heart (You Can Do It) [Second Flight Version]3:00-14.0 LUFS-2.2 dBTP10.4 LU0
06Higher Than the Clouds (Second Flight Version)2:52-14.0 LUFS-2.6 dBTP10.4 LU0
07Wheels Down (Second Flight Version)2:19-13.4 LUFS0.0 dBTP4.9 LU0
08Wingtips and Dew (Second Flight Version)2:57-15.4 LUFS-0.5 dBTP9.1 LU0
09Tailwind (Up, Up, Up) [Remix, Second Flight Version]3:00-14.0 LUFS-1.4 dBTP5.0 LU0
10The Little Blue Pilot (Sunshine Squad Remix, Second Flight Version)3:15-15.2 LUFS-0.2 dBTP9.3 LU0
11The Little Blue Pilot (Night Flight, Second Flight Version)3:15-12.6 LUFS+0.6 dBTP10.5 LU22
12Little Brave Heart (We Can Do It) [Second Flight Version]4:30-13.0 LUFS+0.2 dBTP8.4 LU147

ITU-R BS.1770-4 integrated loudness and loudness range; true peak 4× oversampled. "Clipped" counts samples at digital full scale in the delivered 16-bit files. Each row was measured by WPAudio Engine and, separately, by ffmpeg's ebur128 meter and a direct sample count; the table shows the figures on which both agreed. What the three loudness numbers mean, with sources, is in How to read a loudness report.

03

Understand

These mixes arrived much closer to a release level than the first record's did — the integrated readings span 2.8 LU, from −15.4 to −12.6 LUFS, and five of the twelve already sat exactly on −14.0. The loudness problem was small.

The peak problem was not. 5 tracks (02, 04, 07, 11, 12) read at or above 0 dBTP, and 4 of them (02, 04, 11, 12) arrived with samples already at full scale — 185 in all, 147 of them on track 12. A true peak above zero is an inter-sample peak: the samples fit, but the waveform they describe between samples does not, and a lossy codec will reconstruct it as distortion. That is the defect a release master has to remove.

The clipped samples themselves needed a closer look before deciding anything. Counting every run of consecutive full-scale samples across the four affected tracks gave 73 runs: median length 2 samples, longest 12 — a quarter of a millisecond at 48 kHz. These are isolated transients touching the ceiling, not flat-topped sustained distortion. Declipping repairs what it thinks the waveform should have been; on 2-sample runs that guess carries more risk than the artefact, which is inaudible. So none was done, and the decision is recorded here rather than buried in a setting.

04

Engineer

One target for the whole record: -14.0 LUFS integrated under a -1.0 dBTP ceiling, for every track. The engineer set it; the engine applied it, with the same fixed recipe as the first record:

  1. Remove DC offset. Every mix carried a small one; every master reads within a few hundred-thousandths of zero.
  2. Apply the explicit gain that lands the track on the target — between −1.4 and +1.4 dB here.
  3. Limit true peak, but only where that gain would breach the ceiling. Otherwise, do not touch it.

The third step is the one that matters. 7 of the twelve (02, 04, 07, 08, 10, 11, 12) needed the limiter; the other 5 are pure gain, their dynamics untouched by anything but a DC blocker. No equalisation, no multiband processing, no stereo widening. These were finished mixes and were treated as finished mixes.

The files were delivered at 48 kHz / 24-bit stereo — the sample rate of the sources, with no resampling, at the bit depth the distributor prefers. Each master is sample-aligned to its mix.

05

Compare

Mix against released master, measured the same way. Changed by these amounts — and, for five tracks, by almost nothing.

Measured change between pre-master mix and released master, per track
#TrackGainIntegratedTrue peakΔ LRALimited
01First Wings (Second Flight Version)0.0 dB-14.0 → -14.0-2.1 → -1.90.0 LU
02Tailwind (Up, Up, Up) [Second Flight Version]−0.3 dB-13.7 → -14.0+0.1 → -1.10.0 LUYes
03The Little Blue Pilot (Second Flight Version)0.0 dB-14.0 → -14.0-3.0 → -2.50.0 LU
04Read the Storm (Second Flight Version)+0.6 dB-14.6 → -14.0+0.1 → -1.10.0 LUYes
05Little Brave Heart (You Can Do It) [Second Flight Version]0.0 dB-14.0 → -14.0-2.2 → -2.20.0 LU
06Higher Than the Clouds (Second Flight Version)0.0 dB-14.0 → -14.0-2.6 → -2.50.0 LU
07Wheels Down (Second Flight Version)−0.6 dB-13.4 → -14.00.0 → -1.10.0 LUYes
08Wingtips and Dew (Second Flight Version)+1.4 dB-15.4 → -14.0-0.5 → -1.10.0 LUYes
09Tailwind (Up, Up, Up) [Remix, Second Flight Version]0.0 dB-14.0 → -14.0-1.4 → -1.40.0 LU
10The Little Blue Pilot (Sunshine Squad Remix, Second Flight Version)+1.2 dB-15.2 → -14.0-0.2 → -1.1−0.1 LUYes
11The Little Blue Pilot (Night Flight, Second Flight Version)−1.4 dB-12.6 → -14.0+0.6 → -1.10.0 LUYes
12Little Brave Heart (We Can Do It) [Second Flight Version]−1.0 dB-13.0 → -14.0+0.2 → -1.1−0.1 LUYes

One level

Every master reads -14.0 LUFS integrated. The spread across the record went from 2.8 LU to 0.0 LU.

One ceiling

No master peaks above −1.1 dBTP. Five mixes arrived at or above 0 dBTP; none left that way, and no sample in any master sits at full scale.

Dynamics kept

The worst change in loudness range anywhere on the record is -0.1 LU. The widest-ranging mixes — tracks 05, 06 and 11, at 10.4 to 10.5 LU — keep all of it.

06

Verify

A mastering report is a claim. Before any of it was published here, all twenty-four files — twelve mixes, twelve masters — were measured again with a meter that shares no code with the engine: ffmpeg's BS.1770 implementation for loudness, range and true peak, and a direct count of full-scale samples and mean DC offset from the decoded PCM. Seventy-two loudness figures and forty-eight sample counts, and every one agreed with the engine's report to the precision printed above.

The engine does the same thing to itself: after writing each master it re-opens the file from disk and checks it against the stated target with a second meter of its own. This page adds a third, from outside the project, because a tool verifying its own output is still one party's word.

What this page does not offer is the audio. The first case study serves its originals and masters for a level-matched A/B; Second Flight's are not hosted here yet, so this page publishes what it can prove — the measurements — and nothing it cannot.

07

What we learned

Two records, one engine, two different right answers. The first was mastered louder and per track, because its mixes arrived quiet and uneven and a few of them wanted to sit below the rest. This one was mastered to a single platform-reference level, because its mixes arrived close to it and the only real defect was peaks. Same recipe both times; the decisions were the engineer's, and the measurements are what made them defensible.

The second lesson is about restraint. Four tracks arrived with clipped samples and the obvious move was to repair them. Measuring the damage first — 73 runs, median 2 samples — showed there was nothing worth repairing — 68 of the 73 are five samples or shorter — and that a declipper would have been the larger intervention. "Measure before changing" is not only about what you do; it is also how you find out what not to do.

The gain applied to this record ranges from −1.4 to +1.4 dB. Almost all of the work was in deciding not to do more.

08

The principle

Measure it twice.
Publish what agreed.

Every number above was produced by two independent meters and printed only where they matched. Where the site cannot yet let you hear the difference, it says so instead of pretending.

The same twelve masters were later run through a commercial AI analyzer; what its score can and cannot say, beside these measurements, is in the journal.

The first case study →About WPAudio Engine →