v1.3.3

Download

Ready to run - the filters are already inside. Unzip, run aura-engine.exe, drop a track on it. The app opens on the filter that came with it.

Bundle	Filter	Output rates	Size
Compact	5k taps	every FS multiplier (FS2-FS16)	7 MB	Download
Starter	1M taps	every FS multiplier (FS2-FS16)	135 MB	Download
Standard	10M taps	FS8 (352.8 / 384 kHz)	327 MB	Download
Reference	30M taps	FS8 (352.8 / 384 kHz)	967 MB	Download

Unzip more than one into the same folder and the app offers every tap count you have.

Already have the filters, or want to build from source? The plain app zip is below, and the individual filter packs are on the 1.0.0 release.

Checksums (SHA-256)

6c92a5da0fbe784bf8f783d63d5007efe07e46efac1eebccbda0a6bbf56373bb  aura-engine-v1.3.3-bundle-5k-windows-x64.zip
2def503da0d0db663b795e1c4899a23d09b8986e92b6d449c7bcce0ee5cc753b  aura-engine-v1.3.3-bundle-1M-windows-x64.zip
452b20062c7971c99e532b930c50f3741643cbcee99974ee4b111f056c873e10  aura-engine-v1.3.3-bundle-10M-windows-x64.zip
7a18d66a89de9616e444f9a47b8f38b5388eb0aab7d4c6a579416201f0e74114  aura-engine-v1.3.3-bundle-30M-windows-x64.zip

What changed in 1.3.3

The subsonic filter keeps its promise past the limiter. This affects files
converted with SUB on — the default since 1.3.0 — whenever the ISP output
limiter engaged.

Erogen, in the Audio Science Review thread, compared three versions of one
track and posted their spectra. Ours had a flat shelf from 3 to 10 Hz, about
40 dB under the bass, with SUB20 on — where the filter promises at least
100 dB down. Thanks for spotting it.

Where it came from

The subsonic filter runs on the source. The ISP output limiter runs at the
very end, after the convolution, and trims each over with a gain dip about
1.5 ms long, thousands of times in a track. Bass multiplied by a gain that
keeps dipping is amplitude modulation, and its sidebands land below 20 Hz,
after the filter has already run.

Reproduced on a clipped CD with rumble added at the level of that disc: at
2.7 Hz the filter alone leaves −125.5 dBFS, and −56.6 once the limiter engages.

The fix

The filter now runs twice: on the source as before, and again at the output
rate, right after the limiter and before the true-peak ceiling is measured.

dBFS per bin	1.3.2	1.3.3	ISP off
2.7 Hz	−56.5	−125.0	−126.9
5.4 Hz	−56.6	−110.7	−117.5
8.1 Hz	−56.7	−84.4	−89.1

The ISP-off run is 1.9 dB quieter overall; level-matched, it agrees with 1.3.3
at 2.7 Hz to the tenth. What is left above 8 Hz is the filter's own transition
band, not a leak.

What it costs

Nothing above 25 Hz: the worst bin of the difference from 1.3.2 is −104.5 dB
relative to the signal, with no time shift. The true-peak ceiling takes
0.018 dB more, because taking infrasound out moves peaks. At 352.8 kHz the
second pass adds three to four seconds to a 3½-minute track.