It can be difficult reading about compression without hearing what it does. Here are a number of audio files to accompany this article, with a detailed explanation of what they demonstrate.
Here is a selection of audio files to accompany this month's Compression Made Easy. You can listen to these as MP3s in the media player in the main article (/sos/sep09/articles/compressionmadeeasy.htm) or you can download from this page
Vox Balance Uncompressed 
voxbalanceuncompressed.mp3
Here's a section of my Mix Rescue remix of Pocketlips' 'Rock Show' in SOS December 2008, but with the lead vocal compression and level automation both bypassed. Given the lyric 'Heading for the rock stop at the front gate because I'm banned from the rock show', you can hear how syllables such as '-ing', 'the front', and 'be-' in particular are getting lost in the mix, which compromises the comprehensibility of the lyrics.
Vox Balance Compression voxbalancecompression.mp3
This is how the same section sounds once my compression has been added in. Not only are those low-level syllables pulled up in the balance, but so are lots of other lower-level details such as consonants, and this improves the intelligibility of the lyrics in context. The tone has also changed as well, of course, the nature of which was dictated by the specific models of compressor used.
Vox 1 to 1 vox1to1.mp3
An small section of the uncompressed vocal from my Mix Rescue of 'Big Dummy Shake in SOS March 2009. The following three files demonstrate how different choices of compressor and preset can make a considerably difference to the subjective compression sound, even for roughly similar levels of gain-reduction.
Vox Block fish voxblockfish.mp3
The demonstration vocal file compressed using the 'Industrial Vocals' preset from Digital Fishphones' freeware Blockfish plug-in. This preset used the plug-in's VCA mode with the Air switch active, and involved lots of saturation and a time setting on the combined attack/release control (which is labelled Response in the graphical interface).
Vox LEA voxlea.mp3
The demonstration vocal file processed using the freeware Tin Brooke Tales TLS3127 Leveling Amplifier plug-in in its compression mode. The processor's Peak Reduction and Gain controls were adjusted to roughly match the amount of compression in the VoxBlockfish file.
Vox LA voxla.mp3
The demonstration vocal file compressed using the freeware Tin Brooke Tales TLS2095 Limiting Amplifier plug-in. Again, the Input and Output controls here were balanced to roughly match the degree of compression in the VoxBlockfish file.
Vox Balance Automation voxbalanceautomation.mp3
To demonstrate how level automation can take vocal intelligibility beyond what's possible with compression, here's how the compressed vocal you can hear in the VoxBalanceCompression example can be brought even further upfront with detailed level automation, something that's pretty much par for the course on most professional productions. Such an effect is well nigh impossible to achieve simply with compression, because the automatic processing just isn't intelligent enough to judge how audible the vocal needs to be within the context of your specific mix.
Snare 1 to 1 snare1to1.mp3
Here is a snare sample for the purposes of demonstration, without any additional compression processing.
Snare 20 to 1 Fast Att Fast Rel snare20to1fastattfastrel.mp3
This is what the Snare1to1 file sounds like when processed at a ratio of 20:1 with fast attack and fast release times. Because the compressor reduces and resets the gain so quickly, the result is that the initial percussive transient is dipped in relation to the sound of the drum as a whole.
Snare 20 to 1 Fast Att Slow Rel snare20to1fastattslowrel.mp3
Using the same fast-attack 20:1-ratio compression setting as in Snare20to1FastAttFastRel, for this example I've lengthened the compressor's release time so that the gain reduction resets negligibly during the drum hit. This means that all the compressor really does is reduce the level of the whole drum hit, rather than significantly affecting the hit's character.
Snare 20 to 1 Slow Att Slow Rel snare20to1slowattslowrel.mp3
If I now keep the release time of my 20:1-ratio compressor the same as in Snare20to1FastAttSlowRel, but increase the attack time to around 30ms, a small section of the initial percussive transient can sneak past the compressor before it has the chance to react. The end result of this is that the drum transient is emphasised in relation to the sound's sustain tail -- the opposite effect than the one I produced with the same compression ratio for the Snare20to1FastAttFastRel file.
Bass Att Rel Distortion bassattreldistortion.mp3
Here is a bass sound that I've compressed fairly firmly at a 4:1 ratio using Stillwell Audio's The Rocket compressor. The attack and release time settings are already quite fast at the beginning of the file (0.3ms and 70ms respectively), but they gradually shorten through the file to end up at 1 microsecond and 1ms respectively. As this occurs, you can first hear the attack of each bass note being dulled, and then distortion creeping in as the compressor begins tracking the individual waveform excursions.
Kick 1 to 1 kick1to1.mp3
This is a kick-drum sample for the purposes of demonstration without any additional compression processing.
Kick 20 to 1 20ms Att Fast Rel kick20to120msattfastrel.mp3
Here is the same kick-drum sound as in Kick1to1, now compressed at a ratio of 20:1 with a 20ms attack time and an even faster release, to demonstrate how the compression can appear to rob the sound of some bass frequencies.