Best strategy for eliminating master audio clipping?

NickHope wrote on 12/20/2004, 2:01 AM

Hi, I'm mixing the audio for an underwater documentary DVD. It will have AC3 audio (stereo) and will be authored and burnt by me.

I've got 3 audio tracks containing WAV's:

1. Voice-over track, normalize switch 'on' (peak level -0.1dB), track compressor applied (input gain 0, output gain +3dB, threshold -12.0, amount 3.0, attack 10ms, release 250ms, auto gain compensation off, smooth saturation off).

2. Music track. Normalize switch 'on' (peak level -0.1dB). The music track ducking down to -7.9dB under the voice-over sections (c/o Excalibur plugin).

3. Ambient reef noise, coming and going but peaking at something like -10dB. Not normalized.

Because of the event normalisation and the additive effect of the tracks, there is periodic of clipping in the master bus.

What's my best strategy for eliminating the master clipping and a good result?

1. Reduce the normalize peak setting to something like -1.5dB instead of -0.1dB?
2. Reduce the output gain in the voice-over compressor from 3dB?
3. Just reduce the volume of the individual tracks?
4. Just reduce the master volume? Or is this too late because the clipping has already occurred (i.e. distortion will remain)?
5. Wave hammer on the individual tracks (in place of compressor on the voice-over track).
6. Wave hammer on the master bus? Again, is this too late in the chain re distortion?

Any guidelines you guys can give me to reduce my options a bit would be much appreciated.

Also, is there a way for me to preview what the maximum clipping will be, without leaving the whole 2 hours to play and then coming back to look at the figure in the red box?

thanks!

Comments

farss wrote on 12/20/2004, 2:13 AM

Lot's of questions and no one shoe fits all answer.
To answer one, SF can detect clipping and place markers where they happen.
However wht worries me most is you seem to have everything compressed which is going to kind of kill any dynamics.

Firstly I'd knock the output of the VO track compressor back to 0dB, You need to leave room for everything else to fit. Then I'd bring me music track back a few dB.
Normal rule of thimb is to normalise last, all normalising is doing is winding the gain up as far as it can be without clipping, it has no affect on dynamics. Totally different to compression.
Alos I'd be looking at the Eq of everything, the ambient reef noise and music in particular. If you;ve got a whole lot going on in the same part of the spectrum no matter what you do things can get muddy.
And last and far from least, monitor on GOOD speakers. I've been caught out so many times by this. My next big purchase is going to be decent mid field monitors! I cannot believe how much difference what you listen on makes to a mix or more specifically where things sit in a mix, truly caught me off guard that one.
Bob.

NickHope wrote on 12/20/2004, 2:51 AM

Thanks for the quick reply Bob! Here's some feedback...

>> "However wht worries me most is you seem to have everything compressed which is going to kind of kill any dynamics."<<

Actually so far I only have the voice-over compressed, and from what I've read, this seems to be best practice. It sounds better too. I haven't applied any other compression/wave hammer yet, just aired it as an option.

>>"Firstly I'd knock the output of the VO track compressor back to 0dB, You need to leave room for everything else to fit. Then I'd bring me music track back a few dB."<<

My calculations tell me that the ambient reef noise track should only be adding well under 1dB to the master mix, because it's so much quieter, so let's disregard that for a moment. If both the other tracks are peaking at 0dB, then even if they peaked at exactly the same moment (which they never actually do because of the ducking) the clipping would only be 3dB, right? So in theory I shouldn't have to knock these tracks down more than 3dB each. And just 2dB should suffice because of the ducking, shouldn't it?

>>"Normal rule of thimb is to normalise last, all normalising is doing is winding the gain up as far as it can be without clipping, it has no affect on dynamics. Totally different to compression."<<

How would I actually normalise last, other than doing it manually? I have a bunch of voice-over events with slightly different levels, because they were recorded over a few different sessions. And I have a bunch of music tracks from different composers with greatly differing volume levels. Surely I need to get these evened out before I do anything else???

Former user wrote on 12/20/2004, 5:11 AM

"Any guidelines you guys can give me to reduce my options a bit would be much appreciated."

Do NOT normalize at all. Use light-non colored compression to get some firmness to the sound and make sure your master bus is peaking at -3db and no more. What's with the 0db? This isn't analog. -3db on the master bus should be fine.

Also - remember what people systems are going to do to your over compressed audio when they attempt to play it on their systems...depending on the system - your audio will sound lifeless and dead...even tiring to listen to if the crap is compressed out of it.

VP

farss wrote on 12/20/2004, 5:57 AM

Fair enough, yes speech normally needs a medium amount of compression, just so it can punch through although good VO guys don't need much if any. That doesn't mean you cannot vary the level after you've compressed it though, just to avoid loosing all dynamics. You''ll notice Attenborough does this a lot and it add a bit of interest, I'd say that'd work well with some underwater stuff too.
With normalisation, you can render everything out to a new track with a bit of headroom and then normalise that track
The great thing about digital audio is you don't have to keep everything hitting the stops because you have a noise floor to worry about. From my limited experience you certainly don't want to go down too far but if you work with 24 bit audio you've got a lot more room to move I think.
I've really found to get a good mix you need a few breaks away from it, if you've got the time, stick a copy onto a DVD and watch it next day in the living room, not where you edit, shifting your mindset makes you see and hear stuff you just plain don't hear when you're editing. Better still, make a CD of the audio track, yip no vision. See if that by itself tells the story.
I know as video guys this seems backwards but I tend to cut the audio and then find the images.
Now this is going to be a bit different in your case, you've probably got these great underwater shots and then the image is the thing. But you still need a story that hangs the images together.

Getting back to what you were asking last. If you use compression then you need to get your levels to match between clips otherwise the compression will be different. Probably in that case normalising or manually adjusting the levels makes a lot of sense.

Bob.

Spot|DSE wrote on 12/20/2004, 6:01 AM

make sure your master bus is peaking at -3db and no more. What's with the 0db? This isn't analog. -3db on the master bus should be fine.

Could you explain that a little more, Vocalpoint?

NickHope wrote on 12/20/2004, 6:39 AM

>>"stick a copy onto a DVD and watch it next day in the living room, not where you edit"<<

I edit in my living room! :) :)

Former user wrote on 12/20/2004, 7:47 AM

"make sure your master bus is peaking at -3db and no more. What's with the 0db? This isn't analog. -3db on the master bus should be fine."

Spot - All I was getting at here is headroom. Pinning the master bus at 0db (and then piling on the Normalize and a ton of compression) makes for a very tiring audio experience.

Hey - I used to be one of these that subscribed to the ole trap: "Slap a brickwall limiter on the master bus and pound the RMS levels up until it sounds like every other hypercompressed POS mix out there today".

After doing some research and reading the very good Bob Katz book - Mastering Audio...I wised up a lot regarding my master bus usage and pretty much everything else in the chain when mixing VO, music, spots...whatever.

Bottom line - if your master bus is clipping - you have major problems - and you will have truly nasty distortion in the final product. Either your individual elements are up so high that they have taken all available headroom or you have piled too much crap (compression, limiting) into the mix (or BOTH).

Remember - there is no such thing as "going over" in the digital realm...0.0db is the line in the sand - this is not like driving analog tape past 0db and enjoying some nice warm tape distortion. A "clip" on your master bus - cannot be represented as a warm compressed element in the digital realm - it will be represented in playback as a very nasty, grainy, truly unlistenable distortion. Just take a listen to any current rock CD that has had the complete living s**t mixed out of it by hyper compression to get the vibe I am talking about.

I really took note of this back in 2002 when Rush released Vapor Trails. This CD was the culmination of all that was wrong with the record business and their obsession with the "loudness wars". Now - Rush is one of my all time favorite bands...but this CD is just dog shit due to the hypercompressed mix and the constant overbearing volume onslaught on one's ears. As a result - this CD is impossible to listen to for any length of time and while Bubble's project is not a Rush album...if he insists on slamming the master bus with everything he has listed - the effect will generally be the same...a boring tiring over compressed mix that will have viewers of this DVD reaching for the volume knob to turn it down.

Leave a little breathing room - peaks of -3.0db on the master bus is just about right. As a listener/viewer - if I want this DVD louder - I will turn it up myself.

VP

NickHope wrote on 12/20/2004, 8:09 AM

Well, it's definitely not a Rush album, that's for sure.

Thanks for your advice Vocalpoint. I'm getting the picture, and chilling out about getting everything peaking perfectly at -0.000001 dB ;-)

I've got a question for you... If you're recommending that I don't normalize at all, how do you recommend getting my mixed bag of voiceover and music events to sound equally loud? Just with volume envelopes? I can't do it just with compression, can I? If I CAN do it just with compression, could someone possibly spare the time to tell me how?

thanks!

Former user wrote on 12/20/2004, 8:58 AM

Bubble,

Compression is one component to firm things up...again - in just the right doses. Personally - of much more importance is EQ. This is the critical elementin every mix and for my stuff - much more important than compression when it comes to ensuring everything cuts through. Don't just slam a compressor into the mster bus and assume it's going to sound good. Also loudness - does not equal goodness.

A real good solid mix should need very little plugin help to deliver the point. I try and sculp almost every individual element with EQ but WITHOUT compression first. I know it's sound boring etc etc...but if I can get to a point where it's starting to sound pretty good while remaining completely bare plugin wise...then I am confident that I can blend in selective compression/limiting and watch the mix really start to firm up with even more positive improvement. (all the while - keeping those peaks in check!)

Plugins are very handy but can truly kill any project with over-zealousness and really need to be applied carefully and sparely.

BTW - what are you using for monitors and do you have any other plugins at your disposal or are you just using the Sony supplied ones?

VP

NickHope wrote on 12/20/2004, 9:36 AM

Thanks again Vocalpoint. I hear what you're saying. The compressor has definitely helped my voiceover track but I'll only use a compressor anywhere else or on the master bus if I can see a good reason for it.

I'm having to listen to the output through a cheap AIWA hi-fi at the moment, with it's EQ off. I'm also comparing using a pair of Sony MDR-D77 headphones. I know this is way from ideal. The voiceover was recorded on a Fostex M521 dynamic mic. I'm hoping this project will help fund a nice pair (or set) of Edirol, M-Audio or Yamaha monitors and a nice condenser mic. But for now, I've got to make do with what I have. Phuket's not exacly littered with quality audio gear shops anyway, and you can't buy anything online here unless it's from abroad, and then Thai customs start their fun and games. Grrr.

My soundcard is decent. It's a Terratec DMX 6-fire 24/96.

I've only got the Sony plugins that came with Vegas and Sound Forge.

I know I need to get my nose in some audio text books, but in the meantime I need to get this done.

farss wrote on 12/20/2004, 2:19 PM

Now that's sad :)
I think most of what I could say has been pretty well covered except for one thing. Don't confuse loudness with level!
If your final mix peaks at -10 or 0dBFS to the listener it's the same, they just turn the volume up or down, same effect.
Take a good VO track, add wavehammer voice preset. Now it's LOUD. Even if you turn the volume down 20dB it still sounds LOUD. That's the art of audio. Take a thin female voice. As far as I know there's just no way you can make it sound loud. There simply isn't the energy there to start with. One way to see what's happening is SFs spectrum analysis, I'm using that tool more and more. Something that fills the audible spectrum sounds much louder than something that just fits in a couple of octaves. That's one reason fuzz boxes make guitars sound louder, partly it's the compression but equally it's the added harmonics, the sound is filling more of the spectrum.
The converse is also true. If you've got audio that fills the spectrum trying to add something over the top of it is tricky. Lower that part of the spectrum that what your adding uses and things sit much nicer.
Bob.