Winamp Fraunhofer MPEG-4 AAC Encoder? Options

[HA2013]

Is FhG AAC encoder bundled with Winamp 5.63 a decent one, or there are better free alternative?.This encoder can achieve a maximum bitrate of 448 kbps on CBR mode and LC profile, but I wonder if it's more efficient than other encoders, as I'm considering switching fro MP3 insane preset encodec with LAME 3.99.5 and re-ripping lossless source into AAC-LC (.m4a) files.Thanks in advance.

This post has been edited by HA2013: Dec 23 2012, 00:31

[Hex144]

I noticed a weird behaviour with the FhG AAC encoder (bundled with Winamp). If I apply a lowpass filter to the audio source before encoding with VBR 5 or 6 (of 6), the bitrate goes up (contrary to what's expected).
LPF was applied using:

CODE

qaac --decode -b 24 --lowpass xxxxx input.wav output.wav

The more aggressive the lowpass, the greater the bitrate swell, i.e. bitrate increases with decreasing lowpass cutoff frequency.
Maximum bitrate is achieved at 15-16 kHz LPF, with a gain of up to 5% (depending on the music). Going below 15, the file size starts to fall.

Lowpass & VBR 4 doesn't lead to (large) increases (nor decreases) with the music I've tested so far.

Can someone with in-depth technical knowledge shed some light on this seemingly anomalous behaviour?
Are those extra bits put to good use, at least? Or maybe they're just wasted in inefficient programming?
I'm trying to make a decision whether I will continue using the lowpass.

[C.R.Helmrich]

In order to be fast, the Fraunhofer AAC encoder has a simplified psychoacoustic model. It tries to achieve the target bit-rate regardless of whether the input is low-passed, but due to some small inaccuracies/assumptions, it might be off by a few %, as you noticed. It does, however, realize that (filtered-out) content above the lowpass is masked, so yes, the saved bits are put to good use below the low-pass frequency, and are not wasted.

I also see from your qaac command-line that your low-pass input to the AAC encoder is 24-bit. This might also increase the AAC bit-rate somewhat over inputting 16-bit.

Lastly, why are you using a low-pass at all? The encoder applies its own low-pass (VBR 4 e.g. has one at 16.5 kHz), and anyway, a very small percentage of the bit-rate is spent on (nearly) inaudible frequency regions.

Chris

[Hex144]

I also see from your qaac command-line that your low-pass input to the AAC encoder is 24-bit. This might also increase the AAC bit-rate somewhat over inputting 16-bit.

For what it's worth, I've just tested the encoder with 16-bit input, and the output is sometimes a tiny bit larger, sometimes a tiny bit smaller compared to 24-bit...

Lastly, why are you using a low-pass at all? The encoder applies its own low-pass (VBR 4 e.g. has one at 16.5 kHz), and anyway, a very small percentage of the bit-rate is spent on (nearly) inaudible frequency regions.

I see that VBR 5 (which I intend to use mostly) employs a ~19.3k lowpass.
Quicktime AAC seems to spend a lot of bits on very high frequencies, as evidenced by the big decrease in file size brought about by applying an LPF before encoding.
I was hoping to achieve some economy (small or big) in file size with Fraunhofer AAC too, by using an LPF beforehand, but I was baffled when I saw the file increase in size, and I was worried the encoder might do something wrong.
Since you say that very few bits are spent in that frequency zone, I will probably give up on the LPF. (Or maybe use it if I would want a VBR "5+" of sorts ).

Thank you for your clarifications!