Properly encoding multi-channel AC-3 to multi-channel AAC

[PuzZLeR]

Hello,

Forgive my "n00biatus" - I do consider myself a tiny expert with video, but with audio I'm still learning so please bear with me here.

I have an MPEG-2 clip which I have two audio sources for:
AC-3 5.1 (6 channels)
AC-3 Stereo (2 channels)

I wish to convert this audio stream to AAC (and the video to H.264, which I have no questions about).

Now, if I want to convert to AAC 5.1, then the audio source to use is quite obvious. However, what if I want to convert to AAC 2.0 (for certain compatibility constraints)?

My question(s):
To convert to AAC 2.0, would it be a better encode using the AC-3 5.1 source, or would it be wiser to simply use the AC-3 Stereo source?

I still don't know how down-mixing works, and still don't understand the fundamental functions of it in the encoding (conversion) or decoding (playback) process.

Thanks so much for reading. Hopefully even passive readers can benefit as well from this.

[Soopafresh]

To convert to AAC 2.0, would it be a better encode using the AC-3 5.1 source, or would it be wiser to simply use the AC-3 Stereo source?

If you're converting to 2.0 AAC, then the AC3 Stereo source is typically better, as the mixing has already been done. But your ears will be the final judge.

I still don't know how down-mixing works, and still don't understand the fundamental functions of it in the encoding (conversion) or decoding (playback) process.

It's less technical than you might think. You take RightRear, RightFront, A Portion of Center and LFE and mix that into the Right Stereo channel. Same goes for the left channel.

LR--->LF--->LFE--->CENTER---->LEFT CHANNEL-RIGHT CHANNEL<---CENTER<---LFE<---RF<---RR

It's the individual track volume levels that make accurate downmixing difficult. There are many apps which will do the downmixing automatically, such as the DSP plugin found in Foobar 2000.