Spectrum analyzer basics · Guide

How to read a spectrum analyzer

A spectrum analyzer graph looks busy at first, but it only answers two questions: which frequencies are there, and how strong is each one? Once you know how the two axes work and what a peak, a harmonic and the noise floor look like, you can read almost any spectrum in a few seconds. Switch between the three sounds below as you read.

Try it

Pick a sound and watch the graph. A pure tone makes one peak, a voice makes a row of peaks, and noise makes a wide carpet.

Pick a sound to start

Strongest —Note —Level —
Open the live analyzer

The two axes

Every spectrum graph has the same layout. Across the bottom is frequency; up the side is level. A point on the line tells you "this much signal at this frequency". Read the axis labels first, every time, because the same sound can look very different on two analyzers with different settings.

Frequency in Hz, log vs linear

Frequency is measured in hertz (Hz), the number of cycles per second. 1 kHz is 1,000 Hz. Low frequencies are on the left and high ones on the right. For sound, the axis usually runs from 20 Hz to 20 kHz, the range a young person can hear. Bass sits roughly below 250 Hz, the voice and most instruments from about 100 Hz to 4 kHz, and the sparkle of cymbals and the letter s above 5 kHz.

On a log scale, each doubling of frequency gets the same width: 20 to 40 Hz takes as much room as 5,000 to 10,000 Hz. This matches how we hear, because each doubling is one octave, the same musical step. Most audio analyzers use it, and the grid lines bunch up toward each 10× mark (100, 1k, 10k).

On a linear scale, each hertz gets the same width, so the bass is squeezed into a thin strip on the left. That sounds worse, but it is better for some jobs: evenly spaced harmonics, such as hum at 60, 120 and 180 Hz, line up like a comb. Radio analyzers almost always use linear frequency.

Level in dB, dBFS and dBm

Level is shown in decibels (dB), a scale that compares two amounts. Every 6 dB up roughly doubles the signal's strength (its voltage), and every 20 dB is 10 times. A peak 40 dB above another is 100 times stronger. That is why a quiet harmonic and a loud note can fit on the same graph.

The label after "dB" tells you what the level is compared with. dBFS (decibels relative to full scale) is used in digital audio: 0 dBFS is the loudest sample the device can record, so every real reading is negative, for example −20 dBFS. dBm is used for radio and compares power with one milliwatt: 0 dBm is 1 mW, and a weak radio station might be −80 dBm. dB SPL is sound pressure in the air, which needs a calibrated microphone.

Reading peaks

A peak is a point where the line rises well above its neighbours. It means a lot of energy at one frequency: a tone. Most analyzers can place a marker on the highest peak and print its frequency and level, like the "Strongest" readout in the demo above. Peaks are where the useful information is.

Fundamental and harmonics

Play a note on almost any instrument and you get more than one peak. The lowest strong peak is the fundamental, the pitch you hear. Above it sit the harmonics, at exactly 2, 3, 4 and more times the fundamental. A note of 125 Hz has harmonics at 250, 375, 500 Hz and so on. Their pattern of heights is the instrument's tone colour, which is why a flute (few weak harmonics) and a trumpet (many strong ones) sound different on the same note.

Press "Voice-like" in the demo: it makes a buzz at 125 Hz, shaped like the vowel "ah". You will see a row of peaks 125 Hz apart, with the tallest ones around 600 to 900 Hz. Those humps in the row are called formants: they come from the shape of the mouth, and they are how we tell one vowel from another.

Reading a note

Musical notes have fixed frequencies. A4, the A above middle C, is 440 Hz; one octave up, A5, is 880 Hz; one down, A3, is 220 Hz. Each of the 12 semitones in between is about 6% higher than the one before. Many analyzers turn the frequency of the strongest peak into the nearest note and show how far off it is in cents (hundredths of a semitone). The pure tone in the demo reads 440.0 Hz, A4 ±0 ¢.

Be careful with low notes: sometimes a harmonic is taller than the fundamental, so the analyzer names the note an octave or more too high. Check that the peaks are evenly spaced and find the lowest one in the row.

Peak width

A perfectly steady tone should be a thin line, but on a real analyzer it has some width. Part of that width comes from the analyzer itself: its resolution, set by the FFT size on an audio analyzer or the RBW on a radio one. With an FFT size of 8,192 at 48 kHz, the lines are about 5.9 Hz apart, so two tones closer than about 12 Hz blur into one peak. A bigger FFT makes peaks thinner but the graph slower to react.

The rest of the width is real. A note with vibrato, a drifting oscillator or a radio signal carrying data spreads over a wider band than a pure tone. A wide hump instead of a sharp spike tells you the signal is changing or carrying information.

The noise floor

Below the peaks, the line never goes perfectly flat or silent. It wanders along an uneven carpet called the noise floor. It is made by background sound in the room, the hiss of the microphone and its amplifier, and the limits of the converter. In a quiet room with a phone microphone, the floor often sits around −90 to −100 dBFS in each FFT line.

The noise floor decides what you can see. A signal below it is hidden, even though it is there. The distance between a peak and the floor is the signal-to-noise ratio: a hum peak at −40 dBFS over a floor at −95 dBFS stands 55 dB clear. If the floor itself rises when you play music, the music has broadband content (drums, breath, noise), not just tones.

The floor also moves with settings. A bigger FFT spreads the same noise over more, narrower lines, so the floor drops while steady tones stay at the same height. This is a useful trick for finding a weak hum hidden in noise.

Five patterns you will see

Most spectrums you meet are one of these five shapes, or a mix of them. Learn to spot them and you can name a sound from its graph alone. All five use a log frequency axis from 20 Hz to 20 kHz.

201k20k
Pure tone. One thin peak, here at 1 kHz, and nothing else. A whistle, a test tone or a tuning fork.
201k20k
Music. A dense forest of peaks from several notes and their harmonics, strongest from 100 Hz to 1 kHz and falling toward the treble.
201k20k
Voice. Evenly spaced harmonics of the voice pitch, with humps (formants) that change with each vowel. Little energy above 6 kHz.
201k20k
Hum. A peak at 50 or 60 Hz plus a few multiples of it, and nothing above. Mains power leaking into the audio.
201k20k
Noise. A rough carpet with no steady peaks. Pink noise, rain or a fan falls slowly toward the right; white noise or hiss rises.

Real sounds mix these shapes. A recording of a singer in a room with a fridge running shows the voice pattern, a small hum peak at 50 or 60 Hz and a raised floor from the room. Read the graph one layer at a time: find the tallest peaks first, then check whether they line up as harmonics, then look at the floor.

Time matters too. A drum hit is a short burst of noise that flashes up and fades, while a hum peak never moves. If you need to see how a sound changes over time, a spectrogram shows the same information as a scrolling picture.

Practice exercises

Open the live analyzer, or use "My mic" above, and try these three. Each takes under a minute.

  1. Whistle. Whistle one steady note. You should see a single tall peak, usually between 1 and 3 kHz. Slide the note up and watch the peak move right. Read its frequency and note name.
  2. Hum. Hum a low "mmm". Now you get a row of peaks: your voice pitch (about 100 to 250 Hz) and its harmonics at equal steps. Change the pitch and watch every peak move together.
  3. Two tones. Play two notes at once on a keyboard or a phone tuner app, for example A4 (440 Hz) and E5 (659 Hz). Find both fundamentals, then their harmonics. Notice that some harmonics of the two notes land close together.

When these feel easy, read about settings explained to learn how FFT size, averaging and range change the picture, or go back to what a spectrum analyzer is. Unfamiliar words are in the glossary. Then open the live analyzer and try it on real music. This page is part of spectrum analyzer basics.

FAQ

What does the x-axis show?

Frequency in hertz, low on the left and high on the right. It is usually on a log scale, so each octave gets equal space.

What does the y-axis show?

Level in decibels: higher means stronger. 0 dBFS is the loudest the device can record.

Why use a log frequency scale?

Our ears hear in octaves: 100 to 200 Hz sounds as big a step as 1,000 to 2,000 Hz.

What is the noise floor?

The low, uneven line made by background noise and by the equipment itself. A signal must rise above it to be seen.

Why are there many peaks for one note?

A note contains its fundamental plus harmonics at 2, 3 and 4 times its frequency, and more.

Written and tested by the SpectrumAnalyzer.app team · Last checked

How we tested: Checked with the three built-in sounds: the pure tone reads 440.0 Hz (A4 ±0 ¢) at −12.0 dBFS, the voice-like sound shows harmonics every 125 Hz with its strongest peak at 750 Hz, next to the 730 Hz vowel formant, and the noise shows a falling carpet with no steady peak.