Because this SRC outputs a 32-bit floating-point file, it can preserve values that might otherwise fall outside of the permitted frequency range.
I'm not sure why they refer to 24/96 as the ideal master, considering the final delivery SR is 44.1 kHz. The ideal should be 88.2 or 176.4, no?
To take best advantage of our latest encoders send us the highest resolution master file possible, appropriate to the medium and the project.An ideal master will have 24-bit 96kHz resolution. These files contain more detail from which our encoders can create more accurate encodes. However, any resolution above 16-bit 44.1kHz, including sample rates of 48kHz, 88.2kHz, 96kHz, and 192kHz, will benefit from our encoding process.
I assumed they’re trying to push the ‘AAC has no inherent bit-depth’ line there.
^ Apparently the Nero AAC encoder takes 24bit/96kHz files as input, no problem, resulting in 96kHz encodes… Get ready for 512kbps high-res AAC tracks on iTunes?
Fraunhofer AAC encoder actually managed to somehow apply gain to the signal so that it went up to somewhere around -60 or -70 dBFS instead of -120 dBFS.
Quote from: punkrockdude on 22 February, 2012, 04:14:21 PMFraunhofer AAC encoder actually managed to somehow apply gain to the signal so that it went up to somewhere around -60 or -70 dBFS instead of -120 dBFS. That shouldn't happen. Could you clarify how you created your test file? And I assume you used Winamp for encoding?Indeed, any decent encoder should also accept 24-bit input and resample prior to encoding if necessary. So nothing iTunes-exclusive here.
It is this 32-bit floating [-point] file that’s used as the input to the encoder...
Regardless of the bit depth of the original source file (16- or 24-bit), you should generate a 24-bit file to preserve maximum fidelity resulting from the AAC coding process.
[…] overall, it's pretty reasonable, and actually quite informative (to the lossy-encoding noob) at times.
@db1989: In reference to your response about AAC's internal bit depth, they more-or-less do say it a couple times: […]
I think they still needlessly cast 16/44 audio in a negative light, but at least they provide mostly technically-sound reasons for their request for 24/96 files (no dithering needed, for instance).
Movies, forexample, have very detailed standards for the final mastering volume of a film’ssoundtrack.
Well the issue with the sound being louder using Fhg encoder didn't seem to be a problem with the exe wrapper but here is how different encoders sound on a source where the hight peak hit -96dBFS and then normalized so that we can hear it. The wav file is the source file I used to encode all samples with. All encoders used cbr 320 kbps....
As technology advances and bandwidth, storage, battery life, and processor power increase, keeping the highest quality masters available in our systems allows for full advantage of future improvements to your music.
The AAC algorithm performs the t/f mapping by means of an MDCT with a maximum resolution of 1024 spectral lines. A typical fast and efficient implementation can be realized by a 10 stage radix 2 butterfly operation. Here each butterfly stage reduces the available precision by 0.5 bit in average. Therefore, this implementation loses about 5 bits precision inside the MDCT transformation. In case of a 24 bit data representation, a final precision of only 19 bit can be achieved after the t/f mapping. For most of the remaining parts of the encoder it is in general sucient to be able to handle the dynamic range of 24 bits.
Thanks a lot for this investigation! The AAC encoder will preserve more than the bass drum hits in the next version of Winamp.