Professional Analog Mastering.
Professional Analog Mastering.
226K+ subscribers
A very skilled listener can tell when something has been turned down by 1dB - some engineers report they hear smaller changes.
But for your casual listener, a change of 3dB is typically the threshold. When I say a change in 3dB, I’m talking about isolated signals.
For example, if I have a stereo mix and lower it by 3dB, this is the threshold for your everyday music listener.
However, we seem to be more sensitive to relative levels.
For example, if we have an acoustic guitar and a vocal, even a casual listener will be able to hear a 2dB drop in either the vocal or the acoustic.
If one signal remains the same level, and another changes, it’s easier to notice the change.
This explains one of the biggest challenges in mixing - if the vocal has multiple takes, and one is say 2dB higher in amplitude than the other, even a casual listener will notice, let alone an experienced engineer.
All this to say, be sure to keep gain staging in mind and try to capture consistent performances. If you have multiple takes, automation of effects, or different tracks for different effects, do your best to ensure the level is consistent between each stage of processing.
This makes controlling the relative amplitude between multiple signals much easier.
Let’s take a listen to a stereo mix. I’ll alter the amplitude by 3dB, then 2dB, then 1 dB. Let me know in the comments if you think the volume has been increased or decreased for each change.
Watch the video to learn more >
Unlike simple amplitude changes, there’s a lot more to consider when determining the audible threshold for EQ filters.
The amplitude, center frequency, and Q value all play a role in whether a change can be detected.For example, if I boost 20 kHz by 1dB, it’s a lot less audible than if I boost 1 kHz by 1 dB.
We’re also more sensitive to amplification than attenuation. For example, attenuating 1kHz by 1dB is less noticeable than amplifying 1kHz by 1dB, if using identical Q values.
Although it’s a difficult thing to determine and depends on the context, I’d recommend making no changes greater than 3dB when mixing and 1.5dB when mastering - if you’re goal is to retain a natural sound.
That is, excluding HP filters which of course make much greater changes than this.
If you’re not concerned with that or you’re making creative changes, then it’s genuinely whatever works in that context.
Also, the narrower the filter, the less natural the sound. For example, if I amplify 2kHz with a 1 octave bandwidth by 3dB, it’ll sound more natural than a 1/3rd octave filter.
In an A/B comparison, it may be easier to hear the 1 octave change, since we’re amplifying the overall signal by a greater amount. However, if the filter is applied for the entire track, the 1/3rd octave filter sticks out more.
The same could be said for attenuation. A sharper filter may be harder to hear while switching between on and off, by it will sound unnatural when compared to a change to a larger bandwidth.
To show this, let’s listen to a mix with 1 Oct filters and a mix with 1/3rd octave filters. Although the filters are more precise, notice how the 1/3rd filters are more audible when there’s no on/off cycle.
Watch the video to learn more >
Like EQ, it depends on a lot of factors. The general rule for an individual instrument is no more than 6dB of compression, for an instrument bus no more than 3dB, and for a stereo mix, no more than 1.5dB.
However, different genres have different listener expectations.
Before we get into that, let’s discuss what makes compression audible.
RMS compression, which detects and processes the average loudness of a signal, is the least audible. Peaks will get through, but it allows for general dynamic control that’s harder to notice.
Peak compression is much more audible due to its accuracy and its impact on transients.
Generally speaking, a longer attack and moderate release will result in the least audible compression. An attack longer than 40ms will let the transient through without altering its timbre too aggressively, and a release around 100 - 200ms will avoid distortion, without greatly impacting the ADSR of peaks after the peak that triggered the compressor.
Also, peak compression has a good deal of interplay between distortion and changing the ADSR.
For example, a 20ms attack will cause a good deal of distortion, the same could be said for a 40ms release; however, the slight delay of the attack and quick behavior of the release retain the original dynamics more than most settings.
Alternatively, if I use lookahead, a longer release, and a soft knee to compress gradually, I’ve greatly reduced the amount of distortion. However, the accuracy of the detection means the transients are completely altered.
The gradual compression changes the ADSR of lower amplitude aspects of the signal, and the longer release impacts potentially unrelated signals that occur after the initial crossing of the threshold.
As I mentioned earlier, the amount of compression that can occur before it’s noticed by the listener varies greatly between genres and the listeners of that genre.
If I compress a guitar by 6dB in a rock track, that’s not an aggressive change. If I compress a saxophone by 6dB in a jazz track, that probably won’t go over well with listeners.
So ultimately, it comes down to how much distortion can be tolerated in the genre, how greatly ADSR changes can be tolerated, and if the genre’s listeners expect or don’t expect a compressed sound.
Let’s listen to 2 compressor settings. The first will minimize distortion but alter the ADSR. The second will retain the ADSR but result in more distortion.
Watch the video to learn more >
This is definitely the most difficult to figure out. Unlike faders, EQs, or compressors, most saturation plugins don’t show the amount of processing applied.
Additionally, some saturation plugins introduce changes to the frequency response aside from harmonics. For example, each algorithm or mode of Saturn 2 introduces a unique frequency response.
But if we really want to assign a value to distortion, we could measure the THD - this is the percentage of the harmonics as it relates to the original signal.
The one plugin I know of that offers this is Softube’s Harmonics plugin, but honestly, I’ve found this to be super inaccurate.
Alternatively, you’d have the do a null test and find the ratio of the original to the saturated signal, while trying to compensate for any EQ changes the plugin.
The best plugin I found to test this is with Logic’s stock exciter, which lets me introduce either 2nd or 3rd order harmonics with color 1 and color 2, respectively, and then set a THD percentage.
Using separate tracks, I could gain match a heavily distorted sine wave and the original sine wave. Interestingly, the differences are minor once the gain is matched.
Additionally, odd-order harmonics seem to change the timbre more than even-order harmonics. Despite the 3rd-order harmonic being lower in amplitude, once gain-matched to the original signal, the timbre is different.
My guess is that since the even-order harmonic is a perfect octave, it’s easier to group the harmonic with the original signal. Meanwhile, the 3rd order harmonic is a different note from the original, making it stand out more.
So for saturation or distortion, you’ll need to use your ears. Additionally, even order harmonics will blend more seamlessly with the original signal than odd.
Let’s take a listen to these 3 signals, and let me know if either the even or odd harmonics sound more distinct.