Spectrum Analyzer

From MadTracker Wiki

This page is almost literal copy of the the Spectrum Analyzer question thread. Sunbuster wrote it, with some additional information from Qbical.

Basically, a spectrum analyzer gives a visualisation of sound (or any waveform signal for that matter) on three axes: Amplitude (volume if you like), Frequency and Time. This way you can see how a sound changes in both amplitude and frequency components over time. Often the 3rd dimension is represented by a color intensity instead of a 3rd axes. So if you have the frequency on the Y-axis and time on the X-axis of a graph, then the amplitude of the sound at a particular frequency and time can be represented by a color. A brighter color usually means a higher amplitude. Consider the following spectrums:

Figure 1: Orkidea - Unity

Figure 2: Orkidea - Unity (zoomed)

Figure 3: Rammstein - Mutter

Figure 1 shows a spectral view of Orkidea's track "Unity" as it's shown in Adobe Audition 1.5. Figure 2 shows the same thing, but zoomed in on the lower frequency ranges. What you can notice directly from fig. 1 is that the sound cuts almost completely above 16kHz. Why this happens has been raised into question in the discussion forum. One suggestion is that this is because the track was mastered for vinyl. Another suggestion is that the file upon which this spectrum analysis is based on at some point has been converted to an mp3. Which of the claims is correct is left unanswered for the time being, since the original author of the spectrum graph could not verify either claim.

Note the almost straight line around 1.5kHz in figure 1 (and at least one distinct line below that also) in the beginning of the track. That is produced by an organ playing a single note. So, by doing a spectrum analysis on an instrument you can see what frequencies are most important for that instrument. Thus you can use this to aid you in deciding what frequencies of an instrument to filter out in order to give some other instrument more space. Note that I said aid. The final decision of course has to be made using your ears.

What you also can see is that the spectrum is pretty bright yellow, almost white at the bottom of the graph (this is zoomed in in fig. 2). This indicates that the track is pretty heavy on the bass. This intensity ends almost completely before 500Hz. Most dance/trance/techno tracks will have something in this direction since bass is important for these genres, whereas rock/pop [See figure 3] will usually have a wider spread on the bright, meaning a more even balance in amplitude for different frequency components.

Figure 4: Dire Straits - Down to the waterline

Figure 5: Dire Straits - Down to the waterline (zoomed)

In figure 4 I have a graph of the Dire straits classic "Down to the waterline". In figure 5 I zoomed in on a couple of interesting features. It shows the guitar solo in the intro. Note how you can see the various harmonics of the guitar (the yellow dots/lines) and how they fade out. At the bottom you can see the bass coming in a couple of times. Again, this could be used by you to decide what areas of an instrument to filter out to give another instrument more space.

Figure 6: Dire straits - Sultans of swing

Figure 7: Dire straits - Sultans of swing (zoomed)

Figures 6 and 7 show the same procedure from "Sultans of swing". Compare figure 3 with figure 6 and you can see how the Dire Straits track looks more chopped up compared to Rammsteins track. Sultans of Swing is very clean in the sound. It doesn't have more than the bassline, kick, snare, hihats and the occasional ride and crash, the clean guitar and the vocals. All sounds have a fairly quick decay rate. Rammstein's Mutter on the other hand has a lot of distorted guitars that don't decay anything between notes, they create an even wall of sound. The drumming contains a lot of open hihats and crashes. Thus, based on spectrum analysis you could probably do a pretty educated guess on wether a track will sound clean or dirty.

Figure 7 shows the quitar solo in the end of Sultans of swing. Again you can see the harmonics and slides of the guitar.

Spectrum analysis thus can be used to figure out what frequencies are dominating in a sound and how those frequencies decay or don't decay over time.

If you want to go completely overboard on the subject of spectrum analysis then here's a 120 page discussion on the matter: http://cp.literature.agilent.com/litweb/pdf/5952-0292.pdf

It contains all the juicy technical and mathematical details.

Don't rely on spectrum analyzers too much. It does give an impression of the sound, but it won't tell you if it sounds good or bad. The only tool that can do that is the combination of your ears and your brains.

Spectrum Analyzers can also be used for another purpose, namely re-synthesis. For that purpose you can take an instrument (let's say a church bell) and load the sample into a spectrum analyzer. You can then find out wich frequencies (often called partials) are playing in the churchbell and what there decay time is and so on. After that you can reprogram this in an additive synth (that's a synth with a lot of sine oscilators) and you end up with a good recreation of a church bell. A neat tool for that purpose is Spear. It provides a great way to see the partials of a certain sound. You can furthermore shift these partials around and extend them. It works especially well on voices.