What Is a Spectrum Analyzer?
A spectrum analyzer is a live graph that shows which frequencies are present in a sound and how loud each one is, and you can open one right now with the free Audio Spectrum Analyzer. Frequency runs from left to right, low to high, on a logarithmic scale. Level in decibels runs from bottom to top. The display redraws many times a second, so as a voice, an instrument, or a mix changes, the shape on screen moves with it.
The name covers two different instruments. In electronics and radio, a spectrum analyzer is bench test equipment that measures radio frequencies, up into the gigahertz, to check transmitters and interference. This article is about the audio kind, which looks only at the range people can hear, roughly 20 Hz to 20 kHz, and which producers, podcasters, and home-studio owners use to see what their ears are telling them.
How to read an audio spectrum analyzer
Read it as a graph with two axes: where along the bottom tells you the pitch, and how high the line rises tells you how much of that pitch there is.
The horizontal axis is frequency. Bass sits on the left, treble on the right. If you need a refresher on what those numbers mean, audio frequencies covers the basics, and the audible frequency range explains why 20 Hz and 20 kHz are the edges. The scale is logarithmic, so each octave gets the same width on screen: 20–40 Hz takes up as much room as 10–20 kHz, even though the second spans 500 times more hertz. That matches how we hear pitch, where an octave sounds like the same step anywhere on the keyboard. On a linear scale, everything below 1 kHz would be squashed into a sliver on the left.
The vertical axis is level in decibels. The top is 0 dB, or full scale (the loudest level digital audio can hold; see what a decibel measures), and quieter content sits lower. Vocal Cut's analyzer runs from 0 dB down to -100 dB, with a gridline every 20 dB. These are levels per frequency, so they're most useful compared against each other: this peak is 12 dB above its neighbours.
Once the axes make sense, the shapes start to mean something:
- A bass-heavy mix shows a big hump on the left that tails off steeply toward the right.
- A thin mix has little going on below 100–200 Hz, so the left side of the graph sits low.
- A bright or harsh mix holds its level unusually high on the right, from a few kHz upward.
- A single sung or played note isn't one spike. You'll see a tall peak at the fundamental, the pitch you'd name, and smaller peaks above it at 2×, 3×, 4× that frequency. Those are the harmonics, and their pattern is part of why a violin and a voice on the same note sound different.
Most real music looks like a lumpy slope, higher on the left and lower on the right. That's normal: voices and instruments carry more energy in the lows and mids than in the top octaves.
How it works: the FFT in plain language
A spectrum analyzer takes a short slice of audio and runs a fast Fourier transform (FFT) on it. The FFT works out how much energy the slice contains at each of a set of evenly spaced frequencies, called bins. Draw the bins as a curve and you have one frame of the display. Grab the next slice, repeat, and the curve moves.
The FFT size is the number of samples in each slice, and it sets two things at once:
- Bin width = sample rate ÷ FFT size. That's how far apart, in hertz, the analyzer can tell two frequencies.
- Window length = FFT size ÷ sample rate. That's how much time each reading covers.
At a 48 kHz sample rate, the numbers work out like this:
| FFT size | Bin width | Time per reading |
|---|---|---|
| 2048 | ~23.4 Hz | ~43 ms |
| 4096 | ~11.7 Hz | ~85 ms |
| 8192 | ~5.9 Hz | ~171 ms |
| 16384 | ~2.9 Hz | ~341 ms |
Bigger FFTs give finer frequency detail but react more slowly, because each reading covers a longer stretch of sound. A 50 ms drum hit gets smeared across a 341 ms window. Smaller FFTs catch fast changes but blur neighbouring frequencies together.
The trade-off bites hardest in the bass. The low E on a four-string bass is about 41.2 Hz, and the F a semitone above is about 43.7 Hz, only 2.5 Hz apart. With 23 Hz bins both notes land in the same bin. With 2.9 Hz bins you can tell them apart. At 4 kHz a semitone is more than 200 Hz wide, so even the smallest FFT resolves it.
Vocal Cut's analyzer offers 2048 (labelled fastest), 4096, 8192, and 16384 (finest), and starts at 8192, which suits most music. At a 44.1 kHz sample rate the figures shift slightly, but the trade-off is the same.
Smoothing
Smoothing averages each new reading with the ones before it, which calms the constant jitter of raw FFT output. Higher is steadier and easier to read; lower reacts faster. The tool's slider runs from 0% to 95% and starts at 70%. Turn it up to judge a mix's overall balance, and down to watch a kick drum or a plosive land.
Peak hold
Peak hold leaves a marker at the highest level each frequency has reached since you started, or since you last pressed Reset peaks (switching the view or FFT size also clears them), while the live curve keeps moving underneath. It catches anything brief: a transient, the loudest note in a phrase, the moment feedback starts to build. Play the passage, then see where the peak line ended up.
Line view vs 1/3-octave bars (FFT vs RTA)
The analyzer can draw the same data two ways, and they don't look alike.
The line view is the full-resolution FFT curve. It's best for pinpointing exact frequencies: a hum at 60 Hz, a resonance at 2.8 kHz, the harmonics of a note.
The 1/3-octave view groups the spectrum into 31 bars, centred at the standard third-octave frequencies from 20 Hz to 20 kHz. That's the layout of a 31-band graphic EQ and of a classic real-time analyzer (RTA), and it's easier to read at a glance. (For how graphic bands differ from freely tunable ones, see parametric vs graphic EQ.)
Here's the part that catches people out: "flat" depends on the view. FFT bins are all the same width in hertz, but third-octave bands get wider as you go up, so the 10 kHz band covers far more hertz than the 100 Hz band.
- White noise has equal energy per hertz. On an FFT line it looks flat. On a classic RTA, which adds up all the energy inside each band, it rises about 3 dB per octave, because each band up collects twice as many hertz.
- Pink noise has equal energy per octave. On a classic RTA it reads roughly flat, which is exactly why it's the standard test signal for tuning speakers and rooms. On an FFT line it slopes down about 3 dB per octave.
(For how pink, white, and brown noise differ by ear and in use, see brown noise vs white noise.)
One detail about Vocal Cut's version: its bars show the loudest reading inside each band rather than adding up the band's total energy. That makes them a simplified, EQ-style picture rather than a power-summing RTA, so pink noise will still lean downward on them. When you test a room or speaker with pink noise, judge the bumps and dips against the overall slope rather than expecting a level line.
What is a spectrum analyzer used for?
Mostly, it finds problems your ears have noticed but can't place. Hearing that a mix sounds boxy is easy. Saying it's 340 Hz is hard without a picture.
Spotting buildup in a mix. When several instruments pile energy into the same region, it shows up as a hump. A broad bulge somewhere around 200–500 Hz is the usual cause of a muddy, boxy sound, and the analyzer shows where it peaks.
Seeing masking between kick and bass. Solo the kick and note where its energy sits, then do the same with the bass. If both crowd the same 50–100 Hz region, one of them needs to give way.
Finding a ringing resonance or hum. This is where the "Strongest frequency" readout earns its place: it names the loudest frequency on screen in Hz, the nearest musical note, and its level in dB. Mains hum shows up as a spike at 50 Hz or 60 Hz, depending on where you live, often with smaller spikes at its multiples. A room or drum resonance is a narrow peak that keeps ringing after everything else decays. With the exact frequency in hand, you can cut a narrow notch with the Equalizer instead of guessing. If EQ is new to you, what is EQ (equalization) explains the controls.
Comparing against a reference track. Play a commercial track in the same genre, note its overall shape, then play your mix. Don't match it line for line; look for big differences, like twice the low end or a missing top octave.
Tuning a room. Play pink noise from the Noise Generator through your speakers, from another window or a second device, and set the analyzer's mic at your listening position. Peaks and dips in the bass, usually below a few hundred hertz, are mostly the room (and the speakers themselves): standing waves between parallel walls boost some low notes and cancel others. Move the mic half a metre and watch them shift.
Checking a microphone or a test tone. The analyzer shows whether a mic picks up the lows and highs at all, or adds hiss across the top. For levels and input devices, see how to test your microphone. If you're testing speakers with sine waves, the tone generator guide covers the setup; a pure tone should appear as one clean spike.
Limits: what an analyzer can't tell you
It can't tell you whether something sounds good. Two mixes with nearly identical curves can feel completely different. Treat the curve as a pointer, then let your ears judge the fix.
A microphone reading includes the microphone. Every mic has its own frequency response, and phone and laptop mics tend to roll off the deep lows and colour the highs. Vocal Cut asks the browser for the raw mic signal, with echo cancellation, noise suppression, and automatic gain control switched off, so software processing doesn't reshape what you see. The hardware still colours it, so treat a built-in mic as a guide to where the peaks are, not a calibrated measurement.
It shows one moment, not a timeline. You see what's happening now, not how the spectrum changed over the last 30 seconds. That's a spectrogram's job.
It isn't a loudness meter. The dB values are per-frequency readings, not perceived loudness. For that, see LUFS, the measure streaming platforms use.
Spectrum analyzer vs spectrogram
A spectrum analyzer shows the frequency content of one moment, live. A spectrogram shows a whole recording over time as an image: time along the bottom, frequency up the side, colour for level.
Use the analyzer while you're playing, recording, or adjusting an EQ. Use the Spectrogram to look back through a file and find where a click happens or when a hum starts. What is a spectrogram explains the idea, and how to read a spectrogram walks through the shapes you'll see.
Try it in your browser
The Audio Spectrum Analyzer runs in the browser, with nothing to install.
- Choose Use microphone (your browser will ask permission) or Analyse a file. A file plays through the page's own player, and the display follows it as it plays.
- Pick a Display: Line for exact frequencies, or 1/3-octave bars for an EQ-style overview.
- Set the FFT size: 8192 for general use, 16384 for bass detail, 2048 for fast transients.
- Adjust Smoothing until the curve is steady enough to read.
- Leave Peak hold on to catch the loudest moments; Reset peaks clears them.
- Press Freeze to hold a shape on screen, and Stop when you're done.
The microphone signal is analysed on your device, isn't recorded, and is released when you press Stop. Your audio never leaves your device.
Frequently asked questions
Is a spectrum analyzer the same as an RTA? An RTA, or real-time analyzer, is a type of audio spectrum analyzer that groups the spectrum into fixed bands, usually 31 third-octave bands from 20 Hz to 20 kHz, and adds up the energy in each. An FFT spectrum analyzer shows the full-resolution curve instead. Both show level against frequency in real time. The difference is how finely the frequencies are split and how each band's level is calculated.
What FFT size should I use on a spectrum analyzer? Start with 8192, which gives about 5.9 Hz of frequency detail at 48 kHz and works for most music. Use 16384 when you need to separate close bass notes, since its bins are about 2.9 Hz wide. Use 2048 when you want the display to react quickly to drum hits and other transients, accepting that nearby frequencies will blur together.
Why does pink noise look sloped on my spectrum analyzer? Pink noise has equal energy per octave, but an FFT line view measures energy in equal-width frequency bins. Higher octaves contain more bins, so each bin gets less energy, and the line slopes down about 3 dB per octave. Pink noise reads roughly flat only on an RTA that adds up all the energy within each third-octave band.
Can I use my phone's microphone with a spectrum analyzer? Yes, and it's useful for finding where peaks, hum, and resonances are. A phone or laptop mic has its own frequency response, though, usually with weak deep bass and an uneven top end, and the analyzer shows that colouring along with the sound. Treat the reading as a guide to relative peaks and dips, not as a calibrated measurement.
What's the difference between a spectrum analyzer and an oscilloscope? An oscilloscope, like the waveform view in an audio editor, plots level against time and shows the shape of the signal as it moves. A spectrum analyzer plots level against frequency and shows which pitches make up the sound at that moment. A waveform tells you when something is loud. A spectrum analyzer tells you which frequencies are making it loud.