Reverse Engineering EDM’s Mix Profile

This article touches on something that is very easy to do at home, and yet I’ve managed to procrastinate for maybe half a year before finally getting around to blogging about it!

While my own music does not usually adhere to mix norms — since I come from an experimental and electroacoustic music background where more rules are usually broken than followed : ) — proverbially one is supposed to at least know the rules before allowing oneself to break them. This article outlines the ‘rules’ of spectral intensity structures in contemporary EDM mixing practices.

I don’t think ‘spectral intensity structures’ is a term yet (Google says it isn’t) but it should be. Google thinks this term relates to earthquakes at the time of this writing. A spectral intensity structure (SIS?) is illustrated in the banner image at the top of this article — it basically summarizes the main slope features of mixing practices as discoverable in a spectrum analyzer, in this case focused on EDM.

First, apologies for the hand-drawn scribble (actually, a trackball mouse drawing with the scribble tool in Google Sheets), but it adequately conveys the same information as the screenshot below, from a moment in the song Reload:

A moment of Reload (Sebastian Ingrosso, Tommy Trash, John Martin)

Or, how about a still from Generate (Eric Prydz) instead?

A moment of Generate (Eric Prydz), also exhibiting the spectral intensity structure abstracted in the banner image.

Or how about Titanium?

Yep, basically same mix profile as Generate and Reload and who know how many other songs. Titanium by David Guetta.

Since the banner image is way at the top of this article, let’s repeat it below:

EDM’s Spectral Intensity Structure (despite what Google might say, this is not an earthquake-related term!)

These spectrograms are stills from screen-recorded videos of Voxengo’s SPAN plugin while playing various songs through them. At the bottom of this article, you can see the videos to get a fuller sense of how they evolve, but note that they are silent, to avoid song copyright issues. You’ll have to refer to the openings of the respective songs (all videos start at the song’s start) if you want to match the videos to their absent audio.

If doing this kind of analysis for yourself, make sure you set SPAN’s slope (or that of any similar spectrum analyzer) to 3 dB per octave instead of its default 4.5, for reasons I discuss in this article. In a nutshell, if you run pink noise though a spectrum analyzer set to 3dB per octave slope, it will read as a flat line, which is easier on the eyes and the thinking one does with the help of one’s eyes.

I will ignore the level information in this discussion, since I captured these songs from online streaming platforms and the songs’ streamed levels won’t equate to their master’s levels. With respect to dynamics, all of these tracks have an average dynamics range in the usual 8–10 LU (loudness units) readings on analyzers like the Youlean Loudness meter.

The slope structure in general is:

Clear the Mud: the ultra low end needs to attenuate dramatically before reaching the theoretically lowest level of human hearing, 20 Hz, so that the mud in the mix is definitively cleared away.

The kick or bass drum is like Kong in the Godzilla movie franchise (or even Godzilla, actually — spoiler alert, since there are actually two apex predators…)— The Kick’s Peaks represent the apex predator frequency bands of a mix, the most dominant at least from a signal energy perspective.

Most EDM mixes I examined seem terrified that you may miss a fraction of a second of a kick beat, and so via The Kick’s Valley they gouge a huge spectral channel in the frequencies immediately after the kick, so that bass sounds don’t prevent you from hearing the entire dynamic envelope of each kick sound, like, down to millisecond accuracy, thanks to these bass gouges.

The next important feature is the Melodic and/or Vocal Plateau. After the kick peaks, this is the second most prominent (in terms of spectral energy) area that is sorta flat line-ish (hence, a plateau…) between ~600 Hz and 3 kHz. This range varies because reverb and delay effects may be doing some high and low frequency shenanigans not to mention EQ precision shaping and the differences between male and female voices. This is the ‘sacred mix space’ of vocals or, if the track is an instrumental or during the instrumental part of a song, whatever instrument is carrying the main melody is (in pseudo-musicological terms) the functional stand-in for the vocals and is assigned to this plateau.

Then there is the High End Awning, which kind of looks like an awning (to me at least 🙂 as it is lower than the roofline of the Vocal/Melodic Plateau. The High End Awning gently slopes gradually towards the 20 kHz Deep Dive. I live in Vancouver (Canada) where it rains a lot — I am familiar with awnings and they do slope down like this, so that rain slides off them. Sometimes, though, there is some high end sizzle on a hi hat or noise uplifter that produces a bit of high frequency peakiness, which is a bit un-awning-like (as can be seen with the Generate screenshot above).

Symmetry is a powerful force in the universe and in Star Wars movies, so we need a high end equivalent to Clear the Mud which we can call Clear the (Digital) Aliasing or alternately, The 20 kHz Deep Dive, which sounds bette . The idea here is to get those nasty too-high frequencies out of our ears along with any unwanted digital artifacts or even to make mixes sound more analog (like digitized old records) since the 20 kHz Deep Dive doesn’t seem to care much about the much higher sampling rates that are available today!

A practical method to help you quickly achieve this mix profile is to use a subtle EQ (usually an analog emulation) on the pre-mix (i.e. all track elements sans the kick drum) to dampen the highs and carve out some bass space adjacent to the kick drum:

The exact settings shown will be highly dependent on the spectral content of your instruments. But in broad strokes, it helps with getting this EDM mix profile into your mix.

Fans of bass sounds might be saddened by the prevalence of the deep gouge in frequencies represented by the Kick’s Valley, but keep in mind that this slope feature is merely a major trend and not a law enacted by legislators — you can of course make bass sounds as loud as you want! A great example of an EDM track that omits the Kick’s Valley is Ellie Goulding’s Lights, which has so much bass it basically ramps up and merges with the kick’s zone:

If you really like bass, you can skip the Kick’s Valley.

You might be wondering, OMG, where’s Ellie in all that bass heavy mix? Actually, Ellie is very easy to see. Ellie is always hanging out very strongly at the center frequency of human hearing, or 1 kHz. Above and below, you can see Ellie totally dominating 1000 Hz. Actually, in this mix, Ellie is the apex predator, dominating even over the Kick’s Peaks:

Ellie Goulding, the 1 kHz Apex Predator of her mix’s spectral energy : )

Mixes of course evolve over time, and screenshots can’t do justice to the evolution of a mix unless you include more screenshots! Two other main features of these slopes over time have to do with how many voices are present in the mix. When there are the greatest number of voices present (here, I mean voices to include instrumental and human voices), that will usually indicate The Drop. During The Drop, the entire mix profile (except for the Kick’s Peak) tends toward flat lining (in a good, not bad way!). At a slope of 3dB/octave in a spectrogram, the drop looks more like the profile of pink noise:

A drop moment — with lots going on (vocals, instruments, effects), the spectral profile starts to resemble Pink Noise, i.e. equal energy per octave.
The spectral profile of pink noise in SPAN (at 3 dB / octave).

Despite what Voxengo’s default preset it, I recommend setting your spectrogram slopes to 3 dB/octave because flat lines and pink noise are easy to grasp conceptual entities. The mix profiles in visualizations set to 4.5 dB/octave look very strange in comparison.

During a song’s opening, and during the quieter interludes, you will get what I call the Tilda 45 Down (which sounds like a football play!), but just means that the frequencies slope at an approximately (hence, the tilda ~) 45-degree angle downwards. Below you can see the Tilda 45 Down with many and with just a few voices (remember, ‘voice’ here means any musical element, and doesn’t have to be the singer):

A Tilda 45 Down with a few voices.
A Tilda 45 Down with many voices.

This slope mainly reflects natural orders of overtones, with the fundamental frequencies having the most energy and diminishing towards the higher partials, as captured below in some overtone graphs of acoustic instruments:

For your own reference, below you can see silent spectrograms of the above-mentioned top-streamed EDM songs playing from their start times, in case you’d like to follow along asynchronously with the videos which are ‘legally’ (i.e. copyright saintly) silent.

Below are two tracks of mine that incorporate the techniques discussed above:

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>