Dolby AC3 rendering alters the mix

Velcro Face wrote on 5/18/2007, 3:55 PM
This is my first project with Sony Vegas, coming from Ulead MediaStudio Pro, and I'm having trouble getting a decent audio mix. The track sounds just right on the computer, but when I render it to Dolby AC3 for DVD Architect, something awful happens: All the speech sounds dead, and gets attenuated, as if it's coming from the next room. The music sounds just fine, though, nice and crisp. I've made three separate mixes, dialing down the music track more and more, but it doesn't seem to make any difference.

I have a feeling it must have something to do with a Dolby feature called "Dialogue Normalization." In the default template it's set to -27 dB. I've read everything I can find on the Web about it, and I'm still in the dark. It doesn't behave like the usual sort of normalization. I've read posted in this forum that you should set the Dialogue Normalization to the very bottom of the scale (-31 dB), which I've tried. The result is still a drastically shrunken waveform, when compared to the waveforms I see on the Vegas timeline. The Vegas manual says next to nothing about this, and I'm frustrated. If anyone is getting a decent render to AC3, I'd like to know how you do it.

--
Michael

Comments

farss wrote on 5/18/2007, 4:25 PM
Preprocessing tab, change Line Mode and RF Mode profile to None.

Bob.
johnmeyer wrote on 5/18/2007, 4:37 PM
This came up the other day in this thread:

http://www.sonycreativesoftware.com/forums/ShowMessage.asp?ForumID=4&MessageID=526141

These are the instructions I gave there:

Change the dialog normalization to -31

I think this is the same thing Bob recommends, and that has been recommended many times before. However, in that last thread, and in other threads, people claim that this doesn't make the sound equivalent. So ... I decided to run a test:

I put some audio into Vegas. I then put a static title on the video line. I rendered that to an MPEG-2 file. I then rendered the audio to AC-3, first using the Default template, and then using the settings recommended above. Since Vegas will not accept a raw AC-3 file, but can be tricked into reading and editing AC-3 by simply authoring the AC-3 into a VOB using DVDA, I did that and brought the resulting VOB files into the original project in Vegas. I then looked at the waveforms. They seemed identical, although the AC-3 file had shifted by 5 milliseconds. I don't know how to eliminate that shift.

Hope that helps!
TGS wrote on 5/18/2007, 8:13 PM
Which way did it shift? Ahead or behind? Maybe we can copy the sound track when it's a wave and shift it 5 milliseconds to compensate and mute the original before rendering to ac3..
farss wrote on 5/18/2007, 9:57 PM
Sorry John,
just to be clear, are you saying that the encode using the default and the one using the recommendations looked the same?

I haven't tried bringing the two files into Vegas but from memory they looked very different in DVDA.

Also we might need to factor in how the decoder is working say in Vegas compared to a STB DVD player. AC3 is fairly complex dynamic encoding and decoding system.

Bob.
Jay Gladwell wrote on 5/19/2007, 5:04 AM

Michael, I'm sorry I don't have an answer for you. Too, I'm sorry you're having such a problem with the Vegas and DVDA.

I'm nowhere near technically inclined as John and Bob are. I don't know how to measure the differences in two audio tracks in milliseconds. (Can the human ear discern a 5-millisecond difference in audio? I don't know.)

All I can say is that I am, indeed, "getting a decent render to AC3." The only issue I've ever had with it was that the default settings gave an audio track that required the volume to be turned up considerably when the DVD was played. That issue was resolved by using the recommended settings given in the posts above.

The only question I would have at this point is: What is the quality of the raw voice audio? What, if anything, are you doing to the voice tracks that you're not doing to the music tracks?


farss wrote on 5/19/2007, 5:21 AM
Measuring things in Vegas is very easy. Put the two tracks one above the other and look for a sharp spike. If the spike on the two tracks doesn't line up you draw a region between the peaks on track one and two and read off the time. Set the ruler to time.
If you want to compare two audio tracks invert one and render out. The resulting audio file is the difference between them.

Jay Gladwell wrote on 5/19/2007, 5:33 AM

The resulting audio file is the difference between them.

I know how to compare tracks. What I said was "I don't know how to measure the differences in two audio tracks in milliseconds."

Vegas measures in milliseconds?


farss wrote on 5/19/2007, 8:03 AM
Vegas measures in milliseconds?

Sure, change the ruler to time. Actually you can go finer if you change it to samples but then you'd need to do a little arithmetic.
Internally Vegas tracks time much finer than even that, down to about the microsecond.

Bob.
johnmeyer wrote on 5/19/2007, 9:49 AM
just to be clear, are you saying that the encode using the default and the one using the recommendations looked the same?

Ah, sorry I was unclear. The two AC-3 files looked different from each other (thank goodness, or the recommendations would be useless). The AC-3 that was made with the recommendations you gave, and which I reiterated, looks identical in volume levels to the original audio. The AC-3 that was made with the default settings has volume that is MUCH lower than the original.

As for the phase shift between the AC-3 and the original, the AC-3 file lags the original (i.e., on the timeline, the peak on the AC-3 file occurs later in time than the same peak in the original).

Velcro Face wrote on 5/19/2007, 11:09 AM
Thank you for the advice on settings and the link to another thread on the subject. Seems Dolby digital AC3 is a lot more complicated than I first thought. In the software I used before coming over to Vegas, selecting Dolby AC3 was a simple matter of checking a box, and it rendered out at the level of the mix. In Vegas there are four tabbed dialogues that let you tweak various settings. Obviously, I'm now entering the realm of a truly professional system.

I read the other thread, which referred to other threads, which led me to forums outside the Sony site, all of which were very helpful to my understanding of what Dolby digital is all about. In the end, I feel it's actually better to keep some compression. Instead of setting the Line Mode and RF Mode profiles to "none," I put them at "light film."

Getting the Dialogue Normalization right seems to be the key to making a decent render. If it's off, you get that "pumping" quality and artifacts that make the dialogue sound muffled. I followed the instructions to use the "Scan Levels" button in the Normalization dialogue of Sound Forge to determine the RMS level, which I used for the Dialogue Normalization setting. The resulting render sounds very good, even though it's at a lower volume than the mix.

In the past I wondered why my DVD's always sounded louder than commercial DVD's. Now the levels match. Again, thank you for your generous help.

--
Michael