Skip to content

Fourier shows the ingredients

Any sound can be described as a sum of sine waves. See how harmonics build a waveform, how to hear them, and what an analyzer leaves out.

Updated · 7 min read Lesson 5 of 14 in Audio Science

In short

  1. The Fourier transform describes any sound as sine waves, each with its own frequency, level and phase.
  2. A pitched note is a fundamental plus harmonics at whole-number multiples, and their balance is most of its tone.
  3. An analyzer shows the level of each ingredient and hides the phase, so use it to locate what you hear, not to judge it.
In this lesson (8)

Mixing a moving line

Open almost any home studio session and there is a on the master, a moving line full of peaks and dips. The producer sees a bump in the low mids and pulls it down. They see the highs sloping away and boost them. They are fixing a graph.

The graph is real. It comes from one of the most useful ideas in audio, the , and it shows something your ears also hear: the ingredients of a sound. Knowing what those ingredients are, and what the graph leaves out, turns the analyzer into a tool instead of a boss.

Why it matters

A peak is not a problem by itself

Fourier's idea is that any sound can be described as a sum of sine waves, each with its own frequency, level and . A plucked bass note is a sine at the pitch you play, the , plus quieter sines at two, three and four times that frequency, the . The balance of those harmonics is most of what makes a bass sound like a bass and not like a flute playing the same note.

Figure 1. Each row adds sines at whole-number multiples of the fundamental, the nth one at 1/n of its strength. By the 10th harmonic the smooth wave is close to a sawtooth. Fourier analysis runs this in reverse.

That is why an analyzer helps when you hear a problem you cannot place, such as a build-up in the low mids or a ringing tone in a vocal. It is also why it cannot tell you whether a sound is good. A vocal with a strong peak in the low mids may look lumpy, but that peak can be the chest tone that gives the singer authority. Cut it because it looks wrong and you remove the part you liked.

Listen

Find a frequency by ear

A narrow boost you can sweep across a synth loop or a real mix. Move it slowly, stop where it sounds worst, then check the number.

Loading demo…
100%

Start with your speakers or headphones low.

The science

From time to frequency

A waveform shows pressure over time. The Fourier transform rewrites the same signal as level and phase over frequency:

X(f)=∫−∞∞x(t) e−i2πft dtX(f) = \int_{-\infty}^{\infty} x(t)\, e^{-i 2 \pi f t}\, dt

Read it as a test. For each frequency ff, the transform multiplies the signal x(t)x(t) by a sine and a cosine at that frequency and adds up the products over time. If the signal contains that frequency, the products line up and the total is large. If it does not, they cancel. The result X(f)X(f) is a complex number: its size is how much of that frequency is present, and its angle is the phase.

For a note with a steady pitch, the ingredients sit at whole-number multiples of the fundamental. A sawtooth contains every harmonic, the nth at 1/n1/n of the fundamental's amplitude: the 2nd is 6 dB down, the 3rd 9.5 dB down, the 4th 12 dB down. A square wave keeps only the odd harmonics. A sine has one ingredient, itself.

Figure 2. A sawtooth at 110 Hz as an analyzer sees it: a line at every multiple of 110 Hz, the nth one 1/n as strong. The 3rd harmonic, at 330 Hz, is an E, which is why a narrow boost there sounds like a new note.

An analyzer shows only the size of each ingredient and throws the phase away. Two signals with the same harmonics at the same levels draw the same lines even when their waveforms look nothing alike. Phase starts to matter as soon as two signals are added, which is why phase problems in a mix rarely show up on a spectrum display.

Figure 3. Both waves hold the same three harmonics at the same levels. Only their phases differ. An analyzer that shows level per frequency draws the same three lines for both, because it throws the phase away.

In a DAW the analyzer uses the discrete version of this idea, computed with the fast Fourier transform () on short blocks of samples. How long those blocks are and how they are shaped decides what the display can resolve. The lesson on reading a spectrum analyzer covers those settings.

Try it

Build a sawtooth, then take it apart

0 of 7 steps done

A single note is a stack of sines you can pick out by ear once you know where they are. The analyzer draws the same stack. Your ear decides which parts of it matter.

Common mistake

EQing until the line looks flat

A flat line on an analyzer looks like a balanced mix, so you until you get one. Music carries more energy per hertz in the low end than in the highs, and each higher octave spreads its energy over more hertz. On an analyzer with no tilt, a balanced mix therefore slopes downward, roughly like pink noise, which has equal energy in every octave and falls 3 dB per octave on such a display. Many analyzers can add a tilt so that slope reads flat. Check which setting yours uses before you decide a mix has too much bass or too little top.

The second mistake is looking for phase problems on the analyzer. Because it hides phase, a kick and bass that partly cancel simply look a little quieter. Find those with your ears, the switch and a mono check.

Takeaway

Use it as a flashlight

Keep the analyzer closed while you balance. Open it when you hear something you cannot place, when you need to see sub-bass your monitors cannot reproduce, or to check what a boost is really doing. Make the EQ move by ear, then look to confirm the frequency. Over time your guesses land closer to the number on the screen, and you need the screen less.

3 questions

Check yourself

  1. You play A2 (110 Hz) on a sawtooth synth. Where is its 3rd harmonic, and how loud is it?
  2. Two signals show identical lines on a spectrum analyzer. What can still be different?
  3. Why does a balanced mix slope downward on an analyzer with no tilt?

Sources

Show all 2

Next lesson · 6 of 14

How to read a spectrum analyzerFFT size, window and averaging decide what an analyzer can show. Learn to set it for the question you are asking, and keep your ears in charge.

Previous: Why aliasing is a ghost frequency problem

Portrait of Virzy Guns

Written by Virzy Guns

Songwriter and producer, founder of Virzy Guns Production. Now building HealingWave.