questions about audio normalizing and audio encoding

[codemaster]

I'm trying to encode some DVD's using the most advanced tools and codecs available today.

I decided to encode the mpeg2 video stream from dvd's, to H.264 using x264, and I chose matroska as container for the .264 and audio streams.

But I'm not sure yet what to do with the audio. One dilemma I have is: should I mux the original .AC3 stream into the matroska container, or should I normalize it and transcode it ? Another is: if I decide to normalize and encode the audio, should I encode it to AAC or Vorbis, and at what bitrate ?

My questions are:

1. If the .AC3 stream has 6 channels (5.1) at 448 kbps, and I only use a PC and a pair of headphones to watch videos, what advantages do I have if I keep the original .AC3 instead of downmixing it to 2.0 channels (stereo), normalizing and transcoding it to AAC ? I read somewhere that someone recommended to keep the original ac3 stream in order to be future proof or something, but I already know that in the future, I will not use a 5.1 sound system for the pc/notebook, and I will not use a TV or home cinema system. I will always use a pc, a notebook (or possibly an Android smartphone) and a pair of headphones every time I will watch videos.

2. Why and when exactly does an AC3 stream needs normalization ? Is it because it's volume is too low ? If it's volume is not too low, does it still need normalisation ? If some episodes from the same dvd, have different volumes, some low, some not so low, does this mean that all episodes need normalisation ? What is the best tool for normalizing wav's ? What is the optimal dB level wich the peaks should not cross ? I've seen videos with audio normalized and they sounded awful, for instance, when someone was talking, the background noise was reduced, and after nobody was talking anymore, the background noise increased rapidly, but this was not occuring in the mkv/x264/ac3 version of the episode, only in the avi/xvid/mp3 version. How do I make sure I don't end up with such problem if I'm normalising audio ?

3. If, for example, on a DVD containing some math lectures, the .AC3 stream is 2.0 channels at 192 kbps, but not all episodes have the same volume, some have too low, others not that low, but still a bit lower than normal, should i normalize all of them ?

4. What is the most advanced audio codec today ? I know that today, we have audio codecs that are more advanced than mp3, but I heard AAC is great, others prefer Vorbis, but AAC seems to be more widespread, it is used by youtube, vimeo, lynda.com, etc. Is vorbis a more advanced codec than AAC, like AAC is more advanced than MP3 ?

5. What bitrate would be the best ballance between quality and file size, at 48 kHz ? I already know that the bitrate should be variable, so that more bits could go where there is more complexity and higher volume, but determining the avg bitrate is the hard part for me. But I know that for mp3 music, it is recommended to have at least 192 kbps for transparency (they say that you cannot notice the difference between a 1411 kbps wav audio track and it's encoded mp3 version if it has at least 192 kbps or more). But this applyes to music. For some math lectures, where there isn't any music, only an instructor talking, I'm guessing that 128 kbps AAC at 48kHz is enough. 96 kHz seems to low. But what if you want to encode a series of docummentaries that have a presenter talking, but also have music in some scenes. 160 kbps at 48 kHz would be too low ? Or does it have to be at least 192 kbps for that transparecy thing ?

6. What is the best AAC encoder ? Is it neroAacEnc ?

[jagabo]

If you keep the original AC3 you will have the best possible sound quality. Re-encoding with any other lossy codec will result in some quality loss (although maybe not enough you would notice). Re-encoding with a lossless codec will increase the size. Professionally made DVD should have the AC3 audio at the correct levels and won't need normalization.

[budwzr]

First off, there is no "normal" volume level on DVD, or broadcast TV for that matter.

I would say that MP3 would be your best choice because the volume can be normalized via playback flag without losing the original quality that was there when it was encoded. Use MP3Gain (freeware) http://mp3gain.sourceforge.net/

MP3 Gain has a default "normal" and if you run all the MP3's through it, the will all be the same volume.

I would say the best bitrate would be 256K.