Audio Mastering Help Needed!

[javertim]

Hello all.

I need help in mastering a new audio of mine that has really extreme dynamics. I want to preserve the nuances of the areas of high volume, so I don't want to do any Hard-Limiting (which sounds horrible anyway). Instead, I want to create some sort of filter (gate, maybe?) that will boost the lower volume areas only. For example, I want everything in the audio to be at least -18 dB, but I don't want any areas louder tahn -18 to be tampered with or changed.

I'm really having a difficult time with this. I've done a lot of trading recently and get the same reponse every time: Crystal-clear, beautiful sound, but the volume is just too low in some places and too high in others. Therefore, I want to boost the quiet stuff. Limiting the loud stuff just sounds too "capped" and the ceiling gets really low.

So, does anyone have any suggestions? I've been playing around with Adobe Audition's Dynamics Processing features (under "Amplitude" effect), and I think that the key is somewhere in the "traditional" tab. I've been trying lots of things, but I always seem to raise both the quiet *and* the loud stuff. I really need help.

Thanks again! :-)

Tim

[dex Otaku]

Good dyanmics processing can be difficult to achieve. Keeping things sounding natural, especially when you're amplifying the quiet parts, is probably among the most difficult to attempt.

As I don't use Audition, I can't give you advice on how to use its plugins. In generalised terms though, what you want is sometimes called reverse-compression and is basically the exact opposite of a gate.

If you were to look at compression as on a graph [sound Forge's dynamics method uses this] what you'd be wanting to do is set a threshold, but rather than compressing everything above it [lowering, or attenuating, the levels above the threshold], you want to compress everything -below- it by amplifying it.

As I said, in my experience at least it can be very difficult to do this without things sounding very unnatural. Also, the more dynamic the signal is [the bigger the difference between the loud parts you want to leave alone and the quiet parts you want to amplify] the more obvious the compression will be.

Hopefully there are some Audition users who can contribute to this for you.

[ozpeter]

I'd suggest asking on Adobe's own Audition forum or on the AudioMasters forum where there are many users with many ideas, and also FAQs covering some aspects of this. And it might be deeply boring for the others here.

This link might make a start. Remember the more you squash the higher levels, the more you can raise the lower levels into the 'space' you have made at the top of the waveform.

[Peel]

I'd suggest asking on Adobe's own Audition forum or on the AudioMasters forum where there are many users with many ideas, and also FAQs covering some aspects of this.  And it might be deeply boring for the others here.

This link might make a start. Remember the more you squash the higher levels, the more you can raise the lower levels into the 'space' you have made at the top of the waveform.

←

Use a compressor with long attack & decay values. Also try a low threshold (around the level of the quiet parts) & a soft ratio (probably under 2:1, but play around). If the loud parts come in suddenly, you might have to reduce the attack time, or put a limiter after the compressor just to catch the highest peaks. And of course you'll be adding some out-gain to bring the overall level up.

Or, draw an amplitude envelope to manually ride the volume.

[dex Otaku]

Here's a comparison between normal compression and what I was talking about before, using Sound Forge's graphic dynamic tool. These are exaggerated for the sake of getting the point across.

Normal compression:

[attachmentid=134]

What I referred to as "reverse-compression":

[attachmentid=135]

In the 2nd diagram what's happening is that everything -below- the threshold is being compressed "upward" at a ratio of 2:1. The loud parts are not compressed at all.

Note that this greatly increases the noise floor of your recording [in the example above, the absolute lowest level is now 48dB louder than it originally was]; while I have used this technique before, its success really depends on how noisy the source is.

Another example, in which I try to reduce the noise-floor, but which really has to be tuned to match the recording exactly, has both an upper and lower threshold:

[attachmentid=136]

The upper threshold sets the level below which sound will be amplified; the lower sets the level at which they will fall back to "normal".

With fine-tuning this method can work very well - but it has to be set specifically for the recording, or the results will be very obvious.

Most people just end up using compression at the top of the dynamic range with gain over the whole recording to bring the peak levels up to or near 0dBfs.

I find that the best thing to do is simply play around with your compression plugins to see what works best in the end.

Perhaps it's ironic that with current MD and HiMD recording equipment, it's possible to make recordings whose dynamic range greatly exceeds that of what we actually want to listen to.

[javertim]

Thanks for all the help, guys! :-) I've managed to find a good setting that work well for what I need (I am actually using a 3:1 ratio). I just have to fool around with the attack time because some of the suddenly loud parts are clipping.

[ozpeter]

I know you said that "hard limiting sounds horrible" but you might want to try using it after the compressor if using a rack of effects in Audition's multitrack view. The idea of hard limiting is just to catch the occasional extreme peak - perhaps the ones that slip through the compression net. It does not change the sound at all on audio below its cutoff point - I've tested that scientifically and proved that only the peaks are modified (which is the whole idea). Really, if you can hear it working, you're not setting it up correctly - you must be hacking into the main signal rather than just the out-of-the-ordinary peaks.

[javertim]

Thank you, Ozpeter. I just ran a hard-limiter at 3dB to lessen the applause and clapping (which wasn't that extreme, but was technically the loudest parts on the recording) and then normalized the whole thing at 99%. This way, the quiet stuff is raised even more, although, of course, the loud stuff is, too.

Once again, thanks for all the help, guys! :-)

Edited March 23, 2005 by javertim