ABR DeltaWave Reference Closeness

ABR frequency page
ABR residuals · DeltaWave · original WAV comparison

How close do the ABR builds measure against the original WAV?

This page compares current and archived ABR DeltaWave residual measurements against the original WAV. It remains separate from the VBR/cVBRb closeness page so results from different encoding modes are not mixed.

The current dataset includes official LAME 3.101 and 4.0, unboosted cVBRb MinGW and Visual Studio builds, and both ABR-boost variants.

Lower residual values are useful technical indicators, but they are not listening-test verdicts.

View chartsView tableVBR closeness page

The short version

3.96.1 retains the lowest residual values across the full archived-release dataset. Among the current 4.0 builds, both ABR-boost variants reduce Difference RMS and DF median compared with the unboosted output.

The MinGW and Visual Studio ABR-boost outputs have the same Difference RMS and DF median. The MinGW build records a slightly lower PK median, while the Visual Studio build has the deeper correlated null.

Official 3.101, official 4.0 and unboosted 4.0 cVBRb MinGW produce the same residual metrics in this test. The unboosted Visual Studio build also matches those residual metrics, with a deeper correlated null value.

Archived release characteristics

The full table remains available below; these are the releases with the clearest distinct patterns in this test set.

3.96.1 · strongest residual result

Records the lowest Difference RMS, DF median and PK median among the archived official releases, at the cost of slower encoding and larger output than later versions.

3.98.4 · efficiency-focused

Combines compact output with relatively fast encoding. Its measured profile helps explain its practical appeal for spoken word, audiobooks and video soundtracks.

3.99.5 · balanced archived-release reference

Delivers the fastest archived-release encoding result in the submitted benchmark while improving residual measurements over 3.97 and 3.98.4.

3.100 · maintenance continuation

Matches 3.99.5 in output size and principal residual metrics here, but takes substantially longer to encode in the tested MinGW build.

ABR residual charts

These charts combine the archived release series with the current 3.101 and 4.0 report set.

ABR Difference RMS residual chart
Difference RMS against the original WAV. This uses the same layout and x-axis range as the VBR closeness chart.

Open full-size chart

ABR DF median residual chart
DF median against the original WAV. This uses the same layout and x-axis range as the VBR closeness chart.

Open full-size chart

ABR PK median residual chart
PK median residual by ABR build. This uses the same layout and x-axis range as the VBR closeness chart.

Open full-size chart

Modern ABR residual summary chart
Modern ABR summary including official 3.101 and 4.0, unboosted cVBRb builds and both ABR-boost variants.

Open full-size chart

ABR residual metric table

SeriesDifference RMSDF medianPK medianCorrelated null depth
3.95.1 MinGW ABR-39.90 dB-28.80 dB-48.40 dBFS49.10 dB
3.96.1 MinGW ABR-40.93 dB-29.80 dB-49.70 dBFS51.07 dB
3.97 MinGW ABR-39.14 dB-27.80 dB-47.60 dBFS47.46 dB
3.98.4 MinGW ABR-39.13 dB-27.90 dB-47.50 dBFS49.35 dB
3.99.5 MinGW ABR-39.85 dB-28.70 dB-48.40 dBFS47.71 dB
3.100 MinGW ABR-39.85 dB-28.70 dB-48.40 dBFS47.71 dB
3.101 MinGW ABR-39.85 dB-28.70 dB-48.40 dBFS47.71 dB
4.0 MinGW ABR-39.85 dB-28.70 dB-48.40 dBFS47.71 dB
4.0 cVBRb MinGW ABR-39.85 dB-28.70 dB-48.40 dBFS47.71 dB
4.0 cVBRb Visual Studio ABR-39.85 dB-28.70 dB-48.40 dBFS50.93 dB
4.0 cVBRb MinGW ABR boost-40.27 dB-29.10 dB-49.00 dBFS48.87 dB
4.0 cVBRb Visual Studio ABR boost-40.27 dB-29.10 dB-48.90 dBFS53.40 dB

More negative residual values generally mean a lower measured residual against the original WAV. Correlated null depth is different: higher is deeper.

FAQ

Is this a quality ranking?

No. This page shows ABR residual measures against one demanding original WAV. It does not replace listening tests.

Why split ABR from VBR?

The ABR reports behave differently enough that separate pages make the charts, tables, and caveats easier to read.