Why does compression seem to improve the fidelity of recording?

Question

I understand the basic idea of compression: increase low amplitude, decrease high amplitude to some 'normalized' middle range.

I especially understand how the effect enhances electric guitar to turn the normally fast decay into a slower decay and good sustain.

But when compression is applied to a vocal signal - one where a notes are held and sung in the mid-range of dynamics - what is happening to the signal? If it's already sustained an mid-dynamic range, it would seem the compression doesn't have anything to change.

Yet when I have recorded myself singing (in the privacy of my home to save others the misery hearing me!) the uncompressed sound was 'thin' while the compressed sound was fuller. It's hard to describe the difference. Certainly just turning up the level on the uncompressed track did not sound like the compressed one. Even singing with a full, resonant voice, the uncompressed track had no 'presence', poor fidelity to what could be heard in the room. Compression made the track sound much more like what I could hear in the room.

I also noticed the same thing when recording acoustic guitar with a microphone. Even close miked, with playback level up the sound was thin and dead. Adding compression didn't really change the sustain of the acoustic guitar, but the sound brightened up and had presence. It sounded like what I heard in the room when playing.

Somehow compression made the miked tracks sound much more like the room sound, but I don't know why. It's like the dry recording just lost lots audio detail and compression magically restored it. That seems different that just compressing amplitudes to some ideal middle range.

In case it helps my mic is something like a Shure PGA48. I don't remember the exact brand and model, but it's shaped like that and the price was entry level.

The compressor was a plugin that came with PreSonus Studio One. I don't have the gear set up now (I didn't have space to keep it set up and so packed it away.) I'm not absolutely certain, but the interface image below is what I remember. I don't remember how I set it, but changing the 'knobs' changed the graph line, and I didn't know how to read the graph, so I probably left it at presets which may not be what is pictured. I wish I could be more specific, but I don't remember exactly what I did. More importantly to me, I mostly remember that just adding the effect did the job. I didn't need to mess about with the settings.

Answer 1

Most of the reason a compressor makes a given signal sound better is indeed simply that it allows you to turn up the gain – much more than would be possible without compression, because then it would either run you into headroom limitations or drown out some other instruments in the mix because the peaks are too loud.

But yes, compression also has a bunch of other results. One you have noticed is that it makes mic recordings “roomier”. The reason is that whenever there is no direct sound, the compressor opens up, amplifying any remaining reverberations. Whereas when there is signal, the compressor pushes it down. So the overall dry/reverb ratio is decreased, which is what you would also get with more reverb. However, compressor-boosted reverb has the advantage that it doesn't mingle as much with the dry signal, and therefore can in practice come out as less “muddy”. Of course you may or may not want this; for a silky smoothening reverb, compression isn't really useful.

Another consequence of such “level-ratio changing” is what you probably mean by “sound fuller”. What makes a voice sound full are the actual tonal parts, vocals and voiced consonants. But except for classically trained singers, those are actually not the loudest bits of a vocal track. If instead hard unvoiced consonants, particularly plosives and s-sounds, stick out as the loudest, then this makes the whole track appear thinner. A compressor can reduce the level of those sounds while boosting the soft tonal components. It can make sense to use a high-pass filter in a sidechain to make the compressor focus on consonants even more. A compressor with very high sidechain highpass is called de-esser.
For guitar this also works similarly: the compressor masks away some of the thin-sounding transients, leaving in relation more of the actual notes.

Then there is the fact that real-world compressors, especially analogue ones based on tubes and/or optoelectronics, introduce quite a lot of distortion – but a smooth kind distortion, which is not very overtly notable yet works out to be pretty useful musically. This is something you also get without a designated compressor, by simply running the mic through a tube preamp. The resulting distortion is itself a kind of compression: it clips away some of the higher-peak components and allows boosting the softer elements. Furthermore, distortion forces the wave shape during voiced consonant to more follow the coherent voice component – that too helps with “full sound”. And finally, the distortion introduces some extra harmonic overtones. This first of all makes the signal more present or “transparent”, something which is often desired in vocals. Engineers often put a treble boost on vocals, but boosting high frequencies has the obvious downside of making it sound thinner too. Distortion from a tube or compressor helps making treble boost less necessary, so in the end you'll have fuller-sounding vocals in the mix.

Answer 2

They don't always improve the quality or fidelity of audio. With the right parameters (or in many contexts, wrong parameters) compressors can:

• Completely change the dynamics of a performance, aka a good chunk of the performer's vision (and often the composer's and producer's and everyone else's but you vision too).
• Induce tremolo-like artifacts.
• Induce distortion.
• Kill the attack.
• Alter the timbre, not always in an expected way.

Maybe some distortion is fine in rock, but in a classical music context it might be considered "bad fidelity" or "bad quality". Context here includes your subjective personal tastes, and the specific traditions or conventions each genre and listening setting has and expects. Same for the other stuff. The right parameters in the right context can make it sound more professional, or with "more fidelity", but wrong parameters can make things sound awful just as easily.

If you are not using presets, or a compressor designed for laymen, it's not easy to make a compressor add to the performance. What's actually easy is to make the effect unnoticeable, or affect the audio in a negative way performance-wise and context-wise.

If everything you throw into a compressor comes out sounding with more "fidelity", one of two things is happening:

• Pure luck, coincidence that the initial parameters happened to shape that specific audio track into something you liked better than the dry version.
• You are using a compressor or preset designed from the start for your usecase.

what is happening to the signal? If it's already sustained an mid-dynamic range, it would seem the compression doesn't have anything to change

If the threshold isn't reached, nothing is happening to the signal. Not all compressors behave this linearly, so what's exactly happening depends on the compressor. If the signal is dancing around the threshold, and specially the voice where compression ratios are traditionally low (assuming a good singer, and some styles do highly compress vocals even if the singer is awesome), what you are hearing is just a little bit of compression, that can turn good dynamic control from the singer into awesome-out-of-this-world dynamic control, but this requires a mixing engineer that knows the tools.

Yet when I have recorded myself singing (in the privacy of my home to save others the misery hearing me!) the uncompressed sound was 'thin' while the compressed sound was fuller. It's hard to describe the difference.

What compressor did you use? Some compressors do not act linearly over frequency, so they compress some frequencies more than others (without the control of a multi-band compressor). Or perhaps you used a compressor with multi-band functionality. Even more linear compressors can sound different on different frequencies because of psychoacoustics.

The point is, maybe you equalized the audio without knowing, perhaps making the lows more noticeable (equalization or psychoacoustics can do this), making the sound less "thin", but it's hard to know and analyze without knowing:

• The compressor you used
• The settings
• It's behavior
• The actual details of the audio you used (sustained stuff, transient stuff, etc)
• Exactly what you mean by "thin"

That's on the frequency side of things. On the amplitude side of things, unless you are a trained musician, your dynamic control is not on point. A compressor can mask that, making the voice more controlled.

Why does it sound better / more professional?

Coincidence, probably. Did you use a preset? Maybe the preset was tailored for guitar and voices. Did you not use a preset? Then maybe it was your knowledge on compressors! If you didn't use a preset and don't know much about compressors, then it most likely the initial patch of your compressor is not initialized to 0, but to something tailored for voice and guitar (which are very common things in music in general).

But compressors can also make things sound awful. It's not just a box that you can connect stuff to and make things sound better.

They tend to make things sound better because that's what they are designed to do. Performers with no control over dynamics sound amateur. Performers with high control over dynamics sound professional (assuming they have control over everything else). The compressor controls dynamics, so can take some care of that for you.

It can also alter our perception of tone (again, even in single-band compressors, but maybe you did use a multi-band compressor), making things sound "brighter", "darker", "more mellow", "thinner", etc.

Answer 3

There are actually two types of compression:

1. Normal "in line" compression, where the signal goes into the compressor and out again. This has the effect of making the loud bits of the signal quieter - but extra gain is then usually applied to bring up the overall level. This is basically "squashing the peaks".

2. Parallel compression. This is where the original signal is compressed as above (often quite heavily) and then mixed in at a lower level with the original. This was standard practice on vocals on Motown recordings. The overall effect is to make the quieter parts louder. This is often great on vocals as it tends to make them sound more "present" without actually sounding as if they are compressed - because the peaks are pretty much unaffected.

Unless it is done badly, compression should not introduce appreciable distortion or alter the overall frequency content significantly.

It could well be that the compressor configuration you are using is actually making the room reverb in the original louder. Without seeing how it is configured, it's not possible to say for sure.

Answer 4

The other answers seem to be addressing ways to alter (and ideally enhance) the original audio waveform. Here's an alternative.

Way back in the vinyl days, the RIAA curve was developed as a way to reduce background noise related to the physics of turntables and contact with the vinyl. Quoting bits from Wikipedia,

The purposes of the equalization are to permit greater recording times (by decreasing the mean width of each groove), to improve sound quality, and to reduce the groove damage that would otherwise arise during playback. RIAA equalization is a form of pre-emphasis on recording and de-emphasis on playback. A recording is made with the low frequencies reduced and the high frequencies boosted, and on playback the opposite occurs. The net result is a flat frequency response, but with attenuation of high frequency noise such as hiss and clicks that arise from the recording medium. Reducing the low frequencies also limits the excursions the cutter needs to make when cutting a groove. Groove width is thus reduced, allowing more grooves to fit into a given surface area, permitting longer recording times. This also reduces physical stresses on the stylus which might otherwise cause distortion or groove damage during playback.

A potential drawback of the system is that rumble from the playback turntable's drive mechanism is amplified by the low frequency boost that occurs on playback. Players must therefore be designed to limit rumble, more so than if RIAA equalization did not occur.

RIAA playback equalization is not a simple low-pass filter. It defines transition points in three places: 75 μs, 318 μs and 3180 μs, which correspond to 2122 Hz, 500 Hz and 50 Hz (rounded values)

Prior to this standard, each record company had a slightly different compression scheme, which of course raised hell because a record player only had one playback filter curve.

Fast-forward a couple decades, and Dolby comes out with curves to handle tapes (primarily hiss) and to match acoustics in movie theatres.

Well after that comes the start of digital companding techniques.

The point is that all those early methods were primarily aimed at eliminating noise for analog recording and playback systems. These were essentially "lossless" except where the original SNR was poor enough that filtering couldn't separate noise from signal.

For that matter, the knobs on your stereo preamp let you tweak the compression to your personal taste.