Professional Analog Mastering.
Professional Analog Mastering.
220K+ subscribers
Compression is a powerful tool in mixing or mastering, and a deep understanding of it can unlock the full potential of the plugins you use, and greatly improve your projects.
To better understand compression, let’s first briefly cover:
It probably seems strange to start with this, but we need to know some unique aspects of what our compressors are actually measuring and affecting.
dBSPL is based on atmospheric pressure - our threshold for hearing is 0.00002 Pascals. Meanwhile, the threshold for pain is about 20 Pascals, so that’s a huge jump in the air pressure needed to make that transition.
But our hearing becomes even more interesting when we consider how a conversation is just 0.2 pascals, busy traffic is 0.5 pascals, and even a jackhammer from only 1 meter away is just 2 pascals. Yet, we can hear all the way down to 0.00002, up to well beyond 20. This clearly shows that our perception of loudness is not directly proportional to the air pressure needed to create that loudness.
To help with this issue, we started using dBSPL, where 0dBSPL is our threshold for hearing, 120dBSPL is the threshold for pain, and everything else in between is given a value that represents the linear way in which we interpret loudness.
I bring this up because the way we measure audio in a digital system, or dBFS, is similar to Pascal’s—it’s not a linear relationship.
You may think, okay, the signal is peaking at -40dB, so if I want to double the perceived loudness, I should amplify it to -20dB or something along those lines. But, each amplification of 6dBFS doubles the amplitude, and
If silence is multiple 0s, and as loud as a digital system allows is multiple 1s, increasing the signal 6dBFS doubles the number of 1s, and decreasing 6dB doubles the number of 0s. It’s weird, it doesn’t make much sense causally speaking, but that’s just how it works.
Now, I detailed all of this just to give you a solid argument for why 6dB is about as much as you can attenuate a signal in a digital system before that attenuation becomes very noticeable to everyday listeners.
When you use a compressor to attenuate by 6dB, you may think you’ve only attenuated a small portion of the overall signal, but actually, you’ve cut its perceived loudness, and its amplitude in half.
Of course, this change is less noticeable with makeup gain, but I’ll cover some of the nuances of that in a moment.
For now, let’s compress a signal by 6dB and notice how this seemingly small amount of attenuation causes a drastic change in the perceived volume.
Watch the video to learn more >
Compression is almost synonymous with peak attenuation - we think of it as a way to reduce peaks. This is part of what they do, but not why we use them.
When we attenuate peaks, we create more distance from the noise floor to 0dB. With this additional space, we can amplify the remaining signal, in turn, increasing all of the quieter parts of the signal.
What was being masked or covered up by louder signals a moment ago is now loud enough to hear. Depending on what’s in this lower amplitude range, this post-compression makeup gain may make the signal sound super upfront - or if the range is comprised of mainly reverb reflections, it may cause it to sound further away.
Similarly, if it’s unwanted noise or hum, then that’s what we get more of. This is why problems captured during recording can’t all be fixed with mixing.
The same goes for mastering. What’s wrong with the recording will be emphasized when mixing, and what’s wrong with the mix will be emphasized when mastering, at least if that problem can’t be fixed with some creative method.
So, compression isn’t about controlling the dynamic range, it’s about controlling the dynamic range so that we can adjust what’s there.
Let’s listen to this track that’s been compressed and then amplified after compression, and notice how much more upfront the signal becomes.
Watch the video to learn more >
Usually, I show you how one thing affects something else - for example, how compressing affects our perception of loudness. But in this instance, I want to quickly share in what ways frequency-specific compression does and doesn’t change our perception of loudness.
If we look at a loudness meter, we’ll notice how specific frequencies register as having a higher LUFS or perceived loudness. For example, -10dBFS of 30Hz. is not percieved as being as loud as -10dBFS of 3.5kHz.
That’s because LUFS meters take into account how we perceive different frequencies as being louder or quieter than others, even when they have identical amplitudes.
Seemingly conversely, if I attenuate 30Hz. by 1dB, the LUFS is reduced by about 1 unit. Similarly, if I attenuate 3.5kHz by 1dB, the LUFS is also reduced by about 1 unit.
This indicates that attenuation, or the reduction of dB, has the same effect on loudness regardless of the frequency being attenuated. However, this doesn’t apply to complex signals.
Say I’m mastering a track, and I attenuate 3.5kHz by 6dB. Yes, I’ve reduced the range by about half its amplitude, as we discussed earlier, but overall, I’ve attenuated the primary contributing factor to the track’s perceived loudness.
If I attenuate 30Hz by 6dB, again, I’ve reduced its perceived loudness by half—but in the grand scheme of the overall track, I’ve made a much less significant change to the track’s perceived loudness.
Because 3.5kHz contributes to the track’s loudness much more than 30Hz, compressing one frequency range does not have the same perceived effect as compressing another.
To show this, let’s use a multiband compressor. With it, I’ll set up two bands, one centered on 3.5kHz and one centered on frequencies below 60Hz, and then I’ll alternate between enabling the 2.
Notice how attenuation to 3.5kHz has a significant impact on the track’s perceived loudness, while attenuation to 60Hz and below has a much less significant effect, even when the amount of attenuation is identical.
Watch the video to learn more >
Harmonic distortion is more often associated with saturation—we think a compressor compresses, and a saturator saturates. But when triggered, compressors introduce significant harmonic distortion, and the way in which they do so varies greatly between compressor types.
For the sake of simplicity, let’s use this Pro C 2 and adjust the algorithms to showcase different compressors.
Using Plugin Doctor, I’ll input a -10dB 100Hz sine wave. Notice the THD measurement at the top right. This shows the level of added distortion compared to the original sine wave.
When no compression occurs, the distortion is imperceivable. As soon as the threshold is lowered below -10dB, even a miniscule amount, we instantly notice harmonic distortion.
The level of harmonic distortion increases in correlation with the amount of compression before leveling off. But it’s not just the amount of attenuation that affects harmonics.
If I adjust the knee shape from hard to soft, you’ll notice the THD is reduced, even though more attenuation occurs. As I alter the compressor type, the THD also changes. Additionally, the quicker the attack and release, the more aggressive the distortion—this is because the compressor is cutting into and releasing the waveform faster than 1 full oscillation.
Notice that when I increase the frequency of the sine wave, the THD is lowered - because the oscillation is quicker, less of the waveform is being cut into relative to the number of overall oscillations.
With enough lookahead, I can remove harmonic distortion entirely. When enabled, the compressor receives the information ahead of time and adjusts the amplitude without cutting into the waveform.
Although not exactly like it, compression is a form of wave-shaping. To attenuate peaks, it inevitably alters the waveform, resulting in both harmonic and intermodulation distortion that varies with the amount compressed, the knee, the ratio, the attack and release, the compressor type, the frequency range that’s compressed, and whether lookahead is introduced.
So a compressor doesn’t just alter the perceived loudness through attenuation or the perception of quieter details through post-compression makeup gain; it also alters a signal’s timbre through the introduction of subtle to not-so-subtle harmonic distortion, and alterations to a waveform’s ADSR.
Let’s listen to a track compressed with identical amounts of attenuation - first with a very clean compressor that introduces nearly no harmonic distortion, then a very colorful compressor that introduces significant harmonic distortion.
Notice how much the timbre varies between the 2 examples.