Audio compression in FCE with AUPeakLimiter and AUDynamicsProcessor

Question

I recently edited a video with poorly recorded audio, and learn some useful things about FCE that I want to share. Hope it will help someone else. I also posted this info, with pictures, here: http://members.dslextreme.com/users/craig.lawson/fce/audio_compression.html
Although the video was shot in a studio, the audio levels ranged from nearly 0 dB to down below 36 dB. Some of the talent were experienced and projected their voices, others were inexperienced, young kids, who did not. The audio system had intermittent ground-loop hum. And one of the XLR cables gave occasional bursts of static, perhaps when someone trod on it. And then there was the rare full-scale pop. A real mess.
The destination was broadcast (community) TV. The station's audio guidelines are, "at audible levels without distortion". Not much of a guideline, so I used the guidelines in the FCE manual and set as my goal -12 dB to -24 dB (I know, the recommendation is a 6 dB range, not 12 dB, but 6 dB sounded bad; maybe they use a compressor). In other words, I want to decrease my dynamic range by 24 dB (from 36 dB to 12 dB), with a maximum at 12 dB. How to achieve this in FCE?
Several people have written about Compressor/Limiter (and it's in the manual), but I was not satisfied with the result. Instead, I used AUPeakLimiter and AUDynamicsProcessor and it worked well. These are the settings I used and why.
*Component 1: Audio Filters > Apple > AUPeakLimiter*
A Peak Limiter prevents audio signals from clipping (in FCE, that means exceeding 0 dB). It does this by turning down the volume at the loud parts, and leaving it alone otherwise. The middle of this page has useful graphs which explain peak limiting: http://www.bluehaze.com.au/unix/level.html
Settings
*Attack: 0.01, units: seconds.* This tells AUPeakLimiter to respond to over-limit peaks within 10 msec. I chose the maximum setting for a more gradual volume changes (not sure if that is best - it may cause sudden peaks to go over limit).
*Release: 0.04, units: seconds.* When the peak has passed, return volume to normal within 40 msec. I chose the maximum setting for a more gradual volume change.
*Pre-gain: 20, units: dB.* Increase the input by 20 dB before doing the peak limiting. This makes the quiet sounds louder, and pushes the loud sounds will above the clipping limit (0 dB). But this is PRE-gain, and after applying this gain the peak limiter brings the signals above 0 dB back down. End result: my quiet sounds are louder, and my very loud sounds and perhaps only a little louder. And no clipping. This setting controls the lower value of the filter chain's dynamic range.
*Limiting amount.* FCE shows this as a setting, but in Garage Band it is a read-out. In Garage Band, it displays the amount of limiting currently being applied. I think that for FCE to display this as a setting is a program bug. Also, adjusting it doesn't do anything.
Result at this stage: all audio is loud, including the noise floor. In other words, it's like a compressor with a hard knee and low headroom.
*Component 2: Audio Filters > Apple > AUDynamicsProcessor*
A Dynamics Processor is a Compressor and Expander in one unit. The compressor decreases the dynamic range of the loud sounds, and the Expander increases the dynamic range of the quiet sounds, making them quieter. More on compression here: http://en.wikipedia.org/wiki/Dynamicrangecompression
*Compression threshold: -40, units: dB.* With AUDynamicsProcessor, the compression does not begin at the threshold, but instead begins about halfway to 0 dB. With this setting, the knee is around -20 dB. The Compressor appears to have a "soft knee" (see Wikipedia), but the softness is not adjustable.
*Headroom: 4, units: dB.* The headroom setting adjusts the amount of compression. Lower headroom values result in higher compression. Think of it as the amount of allowed dynamic range above the knee. This setting controls how much peaks may exceed the filter chain's dynamic range: set headroom low if you really don't want the output to exceed your target dynamic range, and set it higher if occasional peaks above your limit is OK.
*Expansion threshold: -60, units: dB.* Sounds below this threshold are expanded. Unlike the compression threshold setting, this sets the position of the knee directly. This setting, combined with expansion ratio, reduce background noise during quiet segments.
*Expansion ratio: 2.* How much expansion to apply to sounds below the threshold. Example: with threshold at -60 dB and ratio at 2, sounds levels at -70 dB (10 dB below the threshold), and pushed down to -80 dB (20 dB below the threshold).
*Attack time: 0.01, units: seconds.* Same as for AUPeakLimiter.
*Release time: 0.100, units: seconds.* Same as for AUPeakLimiter.
*Master gain: -12, units: dB.* This applies post-gain. This setting sets the upper value of the filter chain's dynamic range.
*Comp amount.* FCE shows this as a setting, but in Garage Band it is a read-out. In Garage Band, it displays the amount of compression currently being applied. I think that for FCE to display this as a setting is a program bug. Also, adjusting it doesn't do anything.
Summary
This filter chain effectively compressed my dynamic range, and pushed down my noise floor during quiet segments. Unfortunately, when the talent was talking quietly, the noise became apparent, but to address that requires an adaptive noise reduction filter (e.g. Sound Soap).
The dynamic range mapping created by this filter chain is not so great. Because pre-gain of AUPeakLimiter is so high, it makes the compression ratio infinite above the knee. Then AUDynamicsProcessor lowers the knee and compresses again. A better mapping would use something less than an infinite ratio, and that can be accomplished by putting AUDynamicsProcessor first in the chain. But the resulting filter chain would have more than 2 components, and FCE won't play it in real time (on my system).
If you want the effect I've described, the settings above will get you started. To tune your settings:
1. Disable AUDynamicsProcessor, turn your speaker volume way down, and then play with the AUPeakLimiter's pre-gain until your audio falls within your chosen minimum range on the audio meter. Pay attention to the quiet sounds, and AUPeakLimiter will take care of the loud ones.
2. When your lower range looks good, turn on AUDynamicsProcessor, set master gain, and then play with the compression threshold and headroom. Pay attention to the loud sounds.
3. Finally adjust the expander settings to quiet the noise floor.
Once you have filter settings you like, apply them to your top-most sequence.
Message was edited by: c555

Answer 1

Thanks Sasha for the reply, The Audio Rate on my files are 48KHz and Higher, when I mention the 16 bit I was talking about the Audio Format. The files I have are 16 bit integer where my sequence says it 32 bit floating point Audio Format (not really sure what that means). Would you know why the audio would be perfect on my laptop and not on the pc that is connected to the TV? Could it be because of the Audio Format? because I don't see what else it could be... :/
Message was edited by: Doodfsanfkldsjabfsda