What a Spectrum Actually Shows (and How to Read One)

The bars are not decoration. Once you know what the axes mean, a spectrum tells you where a mix lives, what an instrument is made of, and which problems are real.

Published 2026-09-10 · Analysis · Spectrum · FFT

A waveform shows you a sound over time: loud here, quiet there. A spectrum shows you the same sound cut a different way — at this instant, how much energy sits at every frequency. Both are the same audio. They answer different questions, and most of the confusion about visualizers comes from asking the wrong one.

Left to right is pitch, up is loudness

Read a spectrum left to right as low to high: the far left is sub-bass you feel more than hear, the middle is where voices and most instruments sit, the far right is cymbals, air and sibilance. The height of a band is how much energy is there right now. A kick drum is a tall spike on the left that collapses fast. A hi-hat is a wide, low patch on the right. A sung vowel is a picket fence — a tall fundamental with evenly spaced harmonics marching up behind it.

The number of bands matters less than their spacing. BLOOB draws 96 bands on a logarithmic axis, which means each octave gets roughly the same width. On a linear axis — the default output of a raw FFT — the octave from 40 Hz to 80 Hz would be a hair's width while 10 kHz to 20 kHz sprawls across half the screen, and every bass instrument you own would live in two pixels. Logarithmic spacing matches how hearing works, so what you see lines up with what you hear.

What peak holds are for

The live bars move faster than an eye can track. The thin markers that hang above them and drift down are peak holds, and they are the part you should actually read: they preserve the loudest value each band hit in the last moment, so you can see a transient that came and went before you focused on it. If a peak hold parks itself far above the live level, something is spiking. If peak holds sit flat and low across the top end, the recording is dull or the source is band-limited.

A rough map worth memorising

• 20–60 Hz. Sub. An 808 lives here. Small speakers cannot reproduce it, which is why it disappears on a phone but ruins a club system.
• 60–200 Hz. Kick body, bass guitar, the low end of a male voice. Most of a mix's total energy is in this band.
• 200–500 Hz. Warmth, or mud. When a mix feels blanketed, it is usually a pile-up here.
• 500 Hz–2 kHz. The core of nearly everything: vocal body, snare crack, guitars. Crowded by definition.
• 2–6 kHz. Presence and intelligibility — where a voice cuts through and where harshness starts.
• 6–20 kHz. Air, cymbal shimmer, sibilance. Also where lossy compression quietly removes things.

How to actually read one, in order

1. Open the spectrum analyzer and load a track you know well, or press M to use the microphone.
2. Choose Spectrum Bars or Mirror EQ — the literal scenes. The artistic ones use the same data, but they trade precision for looks.
3. Watch the whole picture for one chorus before you judge anything. A single frame of a spectrum is nearly meaningless; the shape over a few seconds is the information.
4. Compare with a reference: play a commercial track in the same genre and look at the balance between the left third and the right third. That comparison is worth more than any absolute number.
5. Now look at one instrument at a time. Solo a part, or play it live, and note where its fundamental sits and how far its harmonics reach.
6. Check the extremes last. Rumble below 30 Hz that you did not put there, or a hard cliff at 16 kHz on something that should be lossless, are both worth investigating.

What a spectrum cannot tell you

It has no opinion about whether something sounds good. It cannot show phase, so two takes that cancel each other can look healthy. It cannot separate instruments that share a range — a bass and a kick overlap on screen exactly as they overlap in your speakers. It says nothing about stereo image or direction. And it flatters loudness: making everything louder makes every bar taller without improving anything at all.

It is also worth knowing that a display like this is a moving average by necessity. Energy is measured over a short window, so very fast events are smeared a little; the peak holds exist precisely to give those events somewhere to show up.

Why a visualizer needs more than a spectrum

If a visual is driven by spectrum alone, it wobbles pleasantly and never hits anything. That is why BLOOB runs separate onset detection on the low, mid and high bands — a kick, a snare and a hat are events, not just energy — and predicts the beat grid slightly ahead so a movement lands on the beat rather than a frame after it. The spectrum decides what the picture is made of; the onsets decide when it moves.

Once you can read the bars, the artistic scenes stop looking arbitrary. Signal Ridges is a spectrum turned into terrain. Resonance Plate is a spectrum reorganising a Chladni pattern. Deep Sonar is the beat grid drawn as a sweep. They are all the same measurements, wearing different clothes.