Audio Compression
Setting-up the Compressor for the right reasons
Introduction
There are many excellent articles which explain what an audio compressor does and how they work. This article serves to explain the reasons for using compressors and hopefully help you to decide on how you want to control the dynamics of the audio loudness.
Classical Music has a very wide dynamic range; where the quiet parts are very quiet and the loud parts are very loud. A compressor basically makes the loud parts quieter which reduces the dynamic range; thereby smoothing out the difference between soft and loud sound. You have to judge if this makes it better or worse.
I should point out that older recordings of music did not use compressors. Also, many recordings of classical music do not use compression.
For modern recorded music, you can use audio Compressors at any of these 3 stages:
at Recording Input,
on the Audio Track / Channel (or Bus), and
at Mix-Down
The first 2 are compression of individual Tracks; while Mix-Down is a completely separate exercise altogether.
Why Compress?
Do you need to compress at all? For your final Mix-Down, you would likely want to use a Compressor. Note that Alan Parsons [producer, musician and author] doesn't use compression at all and instead uses volume balancing. For Tracks, you may not need to use compression at all.
Tracks which use built-in or plug-in Instruments are already dynamically perfect [in terms of loudness]. If you're listening to a Track using a built-in piano but the loudness volumes are unpleasantly inconsistent, it likely needs to be fixed at the sequencer. If you can fix "bad playing" at the sequencer, do so (instead of using a compressor). MIDI Sound sources (like synths) are also mostly dynamically perfect and will not need to be compressed.
You will want to use compression when there are loudness problems. Typically, problems would arise when
(a) multiple different sounds are ganged-up, eg Drums, or
(b) there is any Sub-bass content, eg Bass.
While Drums and Bass are important elements in modern music, their importance is not the reason we would choose to use compression. For example, if the Drum Track was just using a cow-bell, it would likely not need any compression. For Drums, it's the multitude of sounds in play which makes compression desirable; plus the existence of Sub-bass in the Kick drum. For Bass, apart from Sub-Bass, a lot of modern sound design might use 3 or more different sound sources mixed together.
It's really the capture of sounds from the real world which might need compression. Any audio track recorded using a microphone will likely need a compressor; eg Vocals. Also, Electric Guitars might be recorded using a microphone or sometimes using an effects unit's output (direct).
Once you've recorded the audio with compression, you lose some of the dynamics; and you can't really undo that. As such, the preferred method is to record the signal dry/ unprocessed, then to add the compressor on the Track channel (or Bus).
So what are we trying to do with microphone/ real world recordings? At the very least, we're trying to make the recording respond like what we hear in the real world. What we listen to standing next to a vocalist compared to what the microphone picks up are not the same. This is where a compressor can help.
Also, different genres of music have different expectations of specific sound; which you can conform to by using compressors (or not).
So many Knobs
Quick recap. - An audio recording has quiet parts and loud parts; this difference is the dynamic range. A compressor basically makes the loud parts quieter which reduces the dynamic range; thereby smoothing out the difference between soft and loud sound. But the whole thing is now quieter which means you can crank up the volume, called make-up Gain, to make the loud parts as loud as before. Of course, now the quiet parts are louder too.
Ratio of compression
What do you want to do? How much do you want to compress? The ratio of compression is only applied to the audio signal louder than the Threshold set... but we can deal with Threshold after we look at Ratio.
For recording audio tracks, compression Ratios are generally thought of as follows:-
2:1 = Soft, suitable for Vocals
4:1 = Medium, for Guitar
10:1 = Hard , for Limiting
This "traditional" use of Ratios is derived from the ubiquitous SSL Master Bus Compressor (MBC) which only had Ratios of 2, 4 and 10. Although modern compressors allow for much wider range and more precise choice of compression Ratios, the traditional settings are good starting points.
For mix-down, your choice of Ratio depends on how wildly dynamic the loudness of the tracks are in general. Mostly, a ratio of 2:1 would suffice. You might select 4:1 Ratio if the mix-down is using a lot of "dry"/ unprocessed audio tracks.
If it's hard to hear the effect of the Ratio, try 2:1, then 4:1, then 8:1, then 16:1 to make it more obvious.
Threshold level
Why not just compress the loud parts? Using Threshold, we can leave the quiet parts alone because the Compressor only activates when the loudness level is higher than the Threshold.
It's worth spending a bit of time setting up the Threshold. Everything below is what you want to keep intact and everything above is what you want to compress. Many compressors have visual aids like a bar-graph or flashing LEDs to show when the compressor is activated.
Attack time
You can keep the beginning of a sound uncompressed using Attack time. When the loudness has crossed the Threshold, the Attack time dampens the full activation of the compressor.
Attack time is useful for instruments like Drums and Guitar where the front part of the sound has a lot of timbral content you want to maintain.
What is slow Attack time? Looking at the SSL MBC, a slow Attack is traditionally considered to be about 30ms.
If it's difficult to hear the Attack, try 80ms, then 40ms, then 20ms, then 10ms, then 5ms to make it more obvious.
Release time
When the loudness is above the Threshold, the compressor comes on; when the loudness goes back below the Threshold, the compressor switches off. The Release time will dampen the switching off to become more gradual. The intent is to not have any jarring loudness differences.
What is fast Release time? Looking at the SSL MBC, a fast Release time is traditionally about 100ms.
If the Release time is too long, this will delay the compressor from activating properly at the next Threshold crossing. So, how much time is there between the Threshold crossing for "off" and the next "on"? This depends on the Tempo of the song and if the instrument is playing 1/16ths or 1/8ths of a bar.
Final Note
As mentioned at the beginning, there are many excellent articles on Audio Compression and how it works. Compression is just a corrective tool to smoothen the dynamics to conform to expectations. You will still have to make some qualitative judgement about the dynamics of the audio and how to treat them. I hope this somehow helps in that decision-making.