Some AC3 editing questions

[dark.soft]

Hi. I bought an italian Alias Season Three DVD Boxset, to find that it has Italian DD 2.0 Surround (previous two seasons had 5.1 italian sound). So, here's what I'd like to do: use all the 6 channels of the DD 5.1 ENG for non spoken parts (preserving the original ac3, so not reencoding), and then when a dialogues arrives, switch the front right, left, center and lfe channels with the ones taken from the DD 2.0 (wich I have upsampled to 5.1 wav and then to 6 separated wavs). So I would need to cut the 5.1 ENG, split the parts with dialogue in 6 mono AC3s lossless, and then switch the above mentioned channels. How can I do that? Thank you!

EDIT: waiting for a reply, I tried this guide: http://forum.doom9.org/showthread.php?t=83384&page=1&pp=20 . It's very detailed, but it assumes that your input is a wav file. Mine is obviously a 2.0 AC3. So, when I connect in graphedit the ac3 file parser to the intervideo audio processor fx, between them graphedit automatically inserts AC3 filter. So, there is a particolar configuration for making it work in the process?

[dark.soft]

I have solved some problems, but I still need an answer to the question in the edit of the first post of this topic. I also need a good AC3 encoder, since BeSweet isn't that good. Can you tell me one (also if commercial)? Another question is: with sound forge 8 I can't open 6 ch ac3s or wavs. How can I do to measure the RMS value for a 5.1 track?

EDIT: I measured the single RMS value of the 6 mono wavs splitted by besweet, and here are the results:

FL= -40,3 db (this is front left channel)
FR= -35,0 db (this is center channel)
C= -40,4 db (this is front right channel)
SR= -39,8 db (this is LFE channel)
SL= -41,4 db (this is surround right channel)
LFE= -41,4 db (this is surround left channel)

So, here are my thoughts: SR and SL have identical RMS values, so ProLogic has worked well; FL and FR are pretty similar, and I don't know if this is good or bad, because they should be stereo; the LFE channel is ok and finally, we have the center channel. Why is it's RMS value higher?

Measuring the RMS values for the ENG 5.1 track (I splitted it in 6 wavs with BeSweet, using these azid parameters: -azid( -d 3/2 ) , do they produce the exact channels of the original file? If not, how can I produce them?). They're higher for the front channels and the center, lower for the LFE (strange...) and a bit higher for the surround. So, in order to make a seamless AC3 without volume jumps between dialog and action scenes, what settings should I use in the dialog normalization parameter?

Please, I absolutely need an answer to all my questions!