What are your best AC-3 settings?

craftech wrote on 10/25/2003, 8:28 AM

I shoot musicals and stage productions for the most part. I have produced some excellent looking DVDs with Vegas + DVDA, but I normally use the default settings for AC-3 when I render in Vegas. That gives me a digital 2 channel stereo track in the DVD. It sounds OK. My primary reason for choosing AC-3 is that it takes up less space thereby allowing me to fit more video on the disc.

Farss mentioned in another thread that he changes the dialogue normalization from -27dB to -21dB when he renders because the audio in the DVD isn't loud enough otherwise. Assuming that is the only change he makes, what are the rest of you doing to "punch up" or enhance the audio (using the normal tools that come with Vegas)?

I haven't tried 5.1 yet because I don't think it will enhance the sound particularly with the vocalists. I want to be careful not to "change" their voices too much or add unwanted background noise.

Thanks,
John

PS: I don't have Sound Forge.

Comments

craftech wrote on 10/25/2003, 2:57 PM

C'mon. The subject can't be THAT boring guys.

Most sincerely,
John

Elizabeth Lowrey wrote on 10/25/2003, 8:23 PM

Right off the top, I'm perplexed by what you said another poster did to *louden* the audio. From my reading of the excellent guide to AC-3 encoding put out by Dolby Labs, the only thing you can introduce by adjustment of the "dialog normalization" parameter is *attenuation* of the audio signal, not gain.

The reference dialog norm level with AC-3 is -31dB (or dBFS, to be exact). The aim with setting dialog norm is to have all "dialog", or average loudness of the program material, emerge from the encoder at -31dB so that there is loudness standardization among different programs and program types (e.g., movies with music passages and loud, transient sound effects, pure speech, pure music, etc.) The amount of attenuation applied to the signal by the encoder is determined in relation to this reference level.

So if you set a dialog norm level of "-31", you are telling the encoder that your dialog and the "average" loudness of your material is right where it should be, at -31 dB. Therefore no attenuation is applied. If you set the dialog norm parameter to "-27", you are telling the encoder that the average level or dialog level in your material is at -27 dB, or 4 dB higher than the desired reference level, thus the encoder will apply a straight-line attenuation to the material of 4 dB. If you set the dialog norm level to "-21", you will introduce 10 dB of attenuation to the material, and so on.

Expressing this mathematically:

G (gain applied) = (31 + DN (dialog norm setting)) x -1

Only integer values between -1 and -31 are valid for the DN level, thus it is possible only to apply 0 or negative gain (i.e. attenuation) to the audio signal. You cannot ever raise it beyond what it is at the input stage.

I verified this recently. I've only used Vegas (and AC-3) since March and have only produced one project that wound up on DVD. It was a dance recital, so the whole event was essentially a musical soundtrack.

My initial audio render, done entirely with the default Vegas AC-3 settings and before I'd taken the time to read the aforementioned AC-3 encoding guide, proved very unsatisfactory, not only because of the compression introduced by the settings on the "preprocessing" tab but because of the overall lowered signal level.

Since my project was essentially music only, I had no real concern for the very philosophy or reasons why AC-3 dynamic concepts were invented/implemented in the first place. I wasn't looking to standardize the loudness of my DVD with Hollywood movies or other mass-produced DVD or televised content. I wasn't looking to balance the dynamic range of dialogue with music or loud, transient sound effects. I simply wanted my audio to sound exactly like it did when I finished editing (or as close thereto as possible) while utilizing the data compression that you get with AC-3. I wanted my dynamic range to be entirely preserved and respected. However backward it may seem, my attitude was to treat this like a music CD and not a DVD in terms of dynamic range and signal level.

So I set dialog norm to "-31" and set "line mode profile" under dynamic range control (DRC) on the preprocessing tab to "none". This gave me full dynamic range, equal loudness encoding that was indistinguishable to me from the orignal PCM audio file.

I'm sure if I were producing more traditional video content I'd want to make more traditional use of at least the dialog norm setting. But honestly, I'm a tweaker and control freak who does a lot in editing to adjust the dynamic range, EQ, and relative loudness of the differing source materials to my liking at that stage. I'm not anxious to then have another layer of compression put on in the AC-3 encoding stage where I have much less control of it. I want to hear in the editor what my finished product is going to sound like.

Then I suppose if Paramount Pictures was paying me to master their movies on DVD, I might have to change my work preferences a bit.:-)

craftech wrote on 10/25/2003, 10:41 PM

EClaire,
What you are saying makes sense. I searched for the post that Farss made regarding changing the default -27dB to -21dB to increase the loudness and noticed that in response to my asking him why he made the adjustment, he responded...."the audio ends up at a very low level otherwise. I normally apply mild compression (2:1) on most VV audio so that the ac3 encoder doesn't need that much headroom".

I guess he is applying compression in Vegas first, then changing the dialog normalization from -27dB to -21dB.

John

riredale wrote on 10/25/2003, 11:18 PM

EClaire: nice explanation.

farss wrote on 10/26/2003, 1:11 AM

You guys are right on the ball. I'd seen the recommendation to change the dial norm to -21 dB elsewhere and followed it and it did give a boost to the levels.
However since then I'e researched a bit more and am now leaving it at the default -27dB and change the line level to 'none'. This works much better.

I have read a lot of the material and had some input from SPOT on this as well. I think Sony are a bit remiss in not giving more guidelines here, ac3 is a complex beast and to unleash it without a lot of hand holding is inviting problems.

BTW, the dialogue normalisation value is supposed to be set to a value know as the Longterm Average Equalization . I know of nothing that will derive that value for a soundtrack although SF will give you the RMS value which is close. Seems a lot of expense to buy SF just to find that out.

My original point was that I simply want to be able to encode using ac3 and get it out of the DVD player at much the same level as if I'd use PCM. I fully understand the reasons for the complexities of the ac3 encoder but I don't need that in my situation and I imagine most VV users don't either.

I don't need to alter dynamic range outside of VV, if the ac3 encoder is going to change that the only way to check is to burn a DVD and play it back in a STB.