To fix masking in your mix or master, pay particular attention to your low frequencies - which both have immense overlap in terms of the musical scale, and can greatly affect higher frequencies. The best tool to reduce masking is an EQ, but saturation and compression also play a role.
Before we start talking about how to fix masking issues, let’s get a better understanding of what it is.
When we discuss masking in producing, mixing, or mastering, we’re usually discussing spectral masking, in which lower frequencies tend to mask or cover up higher ones. For example, if I played a 500Hz sine wave while simultaneously playing a mix, that wave would to some extent cover the mix.
In other words, we’d perceive less of the original mix than if that mix was playing by itself. Let’s listen to these 2 signals, first only the mix, then with the sine wave added to see if we perceive less of the mix.
In the last chapter, I briefly showed a graph that shows how 250Hz covers up other frequencies, but let’s look at it in greater detail.
The red bar represents a sine wave, which if we observe the x-axis, we’ll notice is at 250Hz - while the y-axis shows the amplitude of the sine wave. Combining the y-axis with the white curved lines will show how much masking occurs to various frequencies.
It’s kind of convoluted, but I’ll explain it a bit better with some examples. Let’s say the sine wave is 40dB, then at around 500Hz, the sine wave would mask about 30dB of that signal.
If for example, the sine wave is 70dB, then 2kHz will be masked by it by about 30dB.
Similar graphs exist for a 500Hz sine wave, a 2kHz sine wave, and so on, but 250Hz and slightly below is by far the most destructive or impactful when it comes to spectral masking.
So let’s listen to a mix again, and play 250Hz to see if it masks our mix more than the 500Hz sine wave used in our previous demonstration.
In the past 2 chapters we’ve been discussing only spectral masking, but this one covers both spectral and a psychoacoustic phenomenon often called the “Cocktail Party Effect.” In short, our brain can only focus on so many stimuli at once, so the more instrumentation we have, the less we perceive.
What’s interesting is that if we reduce the amount of instrumentation, we reduce this effect and spectral masking due to there being fewer competing frequencies.
Let’s listen to a mix, and enable each stem one at a time to see if we begin to lose detail both from a spectral perspective and in terms of what we can perceive.
Now that we better understand masking, let’s spend the rest of the video considering ways to lessen its effect.
The simplest way to reduce masking is with an EQ. We can create a bell filter over 250Hz, or that area in general, and reduce it by up to 3dB on a full mix, or more on individual instruments. Additionally, we could amplify higher frequencies.
Let’s listen to a mix with a cut in the lower range, and a boost to the higher range and notice how we hear more detail come through.
Saturation can both improve or worsen the effect of masking. For example, if the saturator introduces a second-order harmonic, this is going to amplify lows, which will likely increase masking to frequencies above the harmonic - but if saturate higher frequencies, we can expect the opposite effect.
Let’s listen to saturation on low frequencies, then saturation on high frequencies, and notice how the first causes masking, while the second creates a clearer sound.
How compression affects masking is a complex topic - by reducing dynamics and then amplifying the signal we can reduce masking by bringing quieter parts of the signal upward. Additionally, we can reduce masking if we isolate the compression to the low frequencies, having a similar effect to our EQ earlier.
So let’s listen to multi-band compression with only the lows compressed, and notice how the effect of masking is reduced.
If you’re using any form of parallel processing, it helps to use linear phase settings to avoid temporal masking - AKA phase cancellation. For example, if I set up a parallel signal, and I use an EQ to isolate the highs, timing differences between the 2 signals will cause temporal masking.
By enabling linear phase on the EQ, I create a pre-determined delay amount for the processor that my DAW can now compensate for, in turn eliminating temporal masking.
Let’s listen to this first without linear phase enabled, and then with it enabled.
The reason I wanted to discuss temporal masking in the last chapter, is so I could share this tip with you. Let’s set up a parallel track, and knowing what we know now, use a linear phase EQ - like before I’ll isolate the high frequencies using this EQ.
Then I’ll insert an upward compressor - with which I’ll capture, compress, and amplify quieter parts of the signal - in this case, only the high frequencies since we isolated them.
With this trick, we’re using compression to reduce spectral masking, since the quieter parts of the highs will now be louder, and more capable of competing for space with the more powerful low frequencies.
Let’s listen and notice how we can perceive a lot more detail.
Oversampling increases the frequency range that a signal can occupy - in turn reducing something called aliasing distortion. Aliasing distortion occurs when frequencies exceed the limits of our session’s sampling rate, and bounce back down the frequency spectrum in the form of out-of-key, unpleasant sounding distortion.
Aside from having an unpleasant timbre, this distortion can also cause both spectral and temporal masking to our high frequencies.
Let’s listen to a mix with processing that would cause aliasing, and then this same mix and but with our processor’s oversampling enabled.
In chapter 4 I said that EQ is the easiest way to reduce masking, but the Gullfoss EQ, which automatically adjusts the response of a signal to reduce masking, is even easier. All I need to do is increase the recover function, and maybe some of the tame function.
After measuring the incoming signal, the EQ creates a curve that alters the relationship between frequencies to best suit how our ears perceive complex signals.
Let’s listen and notice how the EQ works surprisingly well, and consider how it accomplishes something that isn’t possible with a typical EQ.