User:Mjb/AAC

Common AAC profiles

Compared to MP3, AAC is slightly more efficient and has fewer audible problems at low bitrates. It is not as good as Opus.

Most AAC audio uses the AAC-LC (low complexity) profile. At 80 kbps and above, it works really well. If you need to go lower than 80 kbps, then you probably want HE-AAC (high efficiency), also known as aacPlus.

There are two versions of HE-AAC. v1 gives you Spectral Band Replication (SBR), and v2 gives you both SBR and Parametric Stereo (PS). Both of these features allow reducing the bitrate to absurdly low levels while still getting acceptable, though imperfect, sound quality.

• stereo, 80 kbps and up: use AAC-LC
• stereo, 40 kbps to 79 kbps: use HE-AAC v1 (AAC with SBR)
• stereo, under 40 kbps: use HE-AAC v2 (AAC with SBR & PS)

There is also an even newer version marketed by FhG called "xHE-AAC" or "extended HE-AAC". It is intended mainly for streaming speech or movie audio over slow or congested wireless phone networks.

80 kbps testing

In 2021, in order to cram as much music onto a 2 GB iPod Shuffle as possible, while still retaining an acceptable level of quality (mostly transparent, to my aging ears), I found that I needed to use AAC-LC at 96 kbps or higher, with ~100 kbps being the minimum for consistent transparency.

So then, in order to get more music onto the iPod, I decided to see if there was any way I could eke out acceptable quality from the next step down: 80 kbps.

From previous experience, I know that the FAAC, Nero, and FFmpeg encoders are sub-par. It really only comes down to Fraunhofer and Apple. There are two Fraunhofer encoders: the "FDK"/Android codec, and the "FhG"/Winamp codec. I have both and prefer the FhG/Winamp codec.

After much testing, I found that SBR kinda sucks. It definitely brightens up the sound, and provides amazing results at very low bitrates, but it also has an artificial quality which I don't like. I think it also makes pre-echo worse. So I am sticking with AAC-LC at this bitrate.

Neither encoder is perfect with all material. Both can produce ringing (a "bleepy" sound) on some material. And I found the Apple encoder drops out or stutters when there's intense infrasonic content (turntable rumble below 20 Hz).

FhG has a slightly higher lowpass frequency, which I like, but Apple's 15 kHz cutoff isn't bad. In the end, I decided Apple's constrained VBR mode resulted in the best sound quality, better than CBR or true VBR.

I am using foobar2000's converter so that I can add filters. Here's what I'm currently using:

• encoder: AAC (Apple) [requires iTunes & foobar2000 Free Encoder Pack], Constrained VBR, 80 kbps, not HE-AAC
• decode postprocessing enabled (for DTS, HDCD etc)
• ReplayGain: album mode [it falls back on track mode if no album gain], apply gain, -5.0 dB preamp for tracks without RG info
• DSP: Resampler (SoX), 32000 Hz, Best quality, 95.0% passband, 50% (linear) phase response
• DSP: IIR filter, 15 Hz, -100 dB, Resonant Highpass, Q 0.707 [this sufficiently reduces rumble without hurting bass]
• DSP: Dynamic DSP, Reverse stereo channels if 'channels swapped' in filename

This seems to work pretty well and produces VBR AAC-LC files at about 80 kbps. It's about 40 MB/hr, so the 2 GB iPod can probably hold about 50 hours' worth of music.

That said, I can still hear ringing in some tracks!

• En Vogue - My Lovin' (You're Never Gonna Get It) (Radio Edit) - 1:50 "whatcha do or whatcha say"
• The La's - There She Goes - stereo acoustic guitar throughout