Spectrum Analyzer Online: Free Audio Spectrum Analyzer
Overview
Spectrum Analyzer Online. Click Start Microphone and the Spectrum Analyzer turns your live input into a Fast Fourier Transform (FFT) display — adjust the Sensitivity slider and switch between Frequency Bars and Waveform mode to watch your room's dominant frequency and peak level update as you go. For a rough sound pressure level estimate, the live decibel meter offers peak hold plus min, max and average readouts from your microphone.
Ever wondered which pitches are hiding inside a hum, a voice or a guitar note? This spectrum analyzer listens through your microphone and splits the sound into its parts, so you see the magnitude of every pitch across the full frequency range instead of the single wiggly line an oscilloscope draws in the time domain. Your signal never leaves the browser.
How This Spectrum Analyzer Turns Sound into a Frequency Spectrum
Press Start, allow the microphone, and the spectrum analyzer begins reading your input signal many times per second. Each reading becomes a plot of amplitude versus frequency: pitch runs along the x axis in hertz, level runs up the y axis in dBFS (decibels below digital full scale). The result is a live frequency spectrum that shows the spectral composition of whatever the microphone hears. Find your lowest and highest note, in scientific pitch notation, with the online vocal range check and a quiet room.
A hardware analyzer gets there with a local oscillator that mixes the signal down to an intermediate frequency, a narrow filter and a detector. This tool does the same job with digital signal processing (DSP) on blocks of samples.
Time Domain Waveform vs Frequency Domain View
Your microphone delivers a waveform: the signal's voltage changing over time. A chord or a fan's hum looks like a tangle in that view. Moved into the frequency domain, the same tangle separates into clear spectral components, one spike for each tone in the signal, so the frequency components that make up the sound are easy to read off by eye. That is the whole point of a spectrum analyzer: it shows what the time trace hides.
The Fast Fourier Transform Step by Step
The analyzer collects a block of N samples of the signal, runs the fast fourier transform (a fast way to compute the discrete fourier transform, or DFT) and turns the block into N/2 frequency bins. Bin k sits at:
$$f_k = k \times \frac{f_s}{N}$$
Here is a worked example. With a sample rate of 48,000 Hz and a block of 4,096 samples, each bin is \(48{,}000 \div 4{,}096 = 11.72\) Hz wide, each block covers 85.3 ms of sound, and the top bin sits at 24,000 Hz. A desk fan whine that peaks in bin 37 reads as \(37 \times 11.72 = 433.6\) Hz. Unlike a hardware sweep, every bin is measured at once, so the spectrum analyzer refreshes several times a second while you watch the measurements update.
Frequency Analysis Settings: FFT Size and Resolution Bandwidth
The block size you choose in the spectrum analyzer works like the resolution bandwidth on a bench instrument. Bigger blocks give finer frequency resolution; smaller blocks react faster. The trade-offs look like this:
- A narrow RBW (large FFT size) separates closely spaced signals, such as two motors a few hertz apart.
- A narrow RBW also needs a longer block, the browser version of a longer sweep time, so short clicks smear.
- A wide RBW updates quickly and suits a speech or music signal that changes every moment.
- Pick the RBW from the smallest gap you need to see, then stop there.
| Block size (samples) | Bin width at 48,000 Hz | Block length |
|---|---|---|
| 1,024 | 46.88 Hz | 21.3 ms |
| 2,048 | 23.44 Hz | 42.7 ms |
| 4,096 | 11.72 Hz | 85.3 ms |
| 8,192 | 5.86 Hz | 170.7 ms |
| 16,384 | 2.93 Hz | 341.3 ms |
Start and Stop Frequencies and the Span
You rarely need the full 0 to 24,000 Hz view. Zooming in sets the spectrum analyzer's limits:
- Set the start and stop frequencies around the frequency range you care about, for example 20 Hz to 2,000 Hz for hum and engine drone.
- Check the span (stop minus start). A narrow span spreads the bins out so nearby tones stop overlapping.
- Switch to a logarithmic scale when the frequency range runs across several octaves, since each octave then gets equal room and the bandwidth of every peak is easier to judge.
Reference Level, Attenuation and Clipping
On a bench unit, the reference level is the top line of the screen and attenuation keeps a hot input from causing overload. In a browser spectrum analyzer the same ideas apply to microphone gain:
- If the loudest peak touches 0 dBFS the signal is clipping, and fake tones appear. Move the mic back or lower the input gain, which is attenuation by another name.
- Keep the loudest peak around -10 to -20 dBFS so weak and strong signals both stay inside the dynamic range of the converter.
- Treat the reference level as a fixed ruler: once set, compare every reading against it.
Reading Signal Analysis Results: Peaks, Harmonics and Noise Floor
Back to the fan. The tallest spike in the signal is the dominant frequency at 433.6 Hz and -14.2 dBFS. Most tools let you hover or drop markers on a peak to read its exact bin and level, and the power in each bin is shown in decibels, so a 6 dB step means double the amplitude. Read the trace from left to right, starting with the biggest spike in the spectrum analyzer display, and note each level against the bottom of the plot. Check bass response from 80Hz down to 20Hz with the subwoofer kick test online and listen for rattling, distortion or a sudden roll-off.
Harmonic Distortion and Spurious Signals
Smaller spikes at whole-number multiples of the fundamental frequency are harmonics. In the fan example they sit at 867.2 Hz (-32.6 dBFS) and 1,300.8 Hz (-41.0 dBFS), 18.4 dB and 26.8 dB below the fundamental. Combine them and the harmonic distortion works out to:
$$\text{THD} = \sqrt{0.120^2 + 0.046^2} \approx 12.9\%$$
Spikes that are not multiples of anything are spurious signals: a phone charger, a monitor's whine or aliasing. A spectrum analyzer can't tell you their source, but distortion and spurs look different on screen once you know which is which.
Signal-to-Noise Ratio and Interference
The grassy line along the bottom is the noise floor, here -81.5 dBFS. The gap between the fundamental and that floor is the signal-to-noise ratio: \(-14.2 - (-81.5) = 67.3\) dB. On any spectrum analyzer, a floor that jumps when you plug in a laptop charger points to interference rather than sound in the room.
Tracking Down a Home Studio Hum with the Spectrum Analyzer
An audiobook narrator needs to know one thing before re-recording chapter 7: is the low buzz in the quiet passages bad enough to fail Audible's ACX check, which caps the noise floor at -60 dB RMS? The answer here turns out to be yes, by about 2.7 dB.
The interface records at 44,100 Hz, so that is the first setting entered. Mains hum sits at 60 Hz and its overtones, so the start and stop are set to 20 Hz and 500 Hz, and the block size goes to 16,384 samples. That gives 44,100 ÷ 16,384 = 2.69 Hz bins and a 372 ms block, which is fine for a steady buzz.
With the mic open and nobody speaking, three spikes stand up from a -96.2 dBFS floor:
| Component | Before | After moving the interface plug |
|---|---|---|
| 60 Hz | -58.7 dBFS | -79.4 dBFS |
| 120 Hz | -71.9 dBFS | -88.6 dBFS |
| 180 Hz | -63.4 dBFS | -84.1 dBFS |
| Combined hum | -57.3 dBFS | -77.8 dBFS |
The combined figure is a power sum, \(10\log_{10}\sum 10^{L/10}\), checked by hand against the readout. At -57.3 dBFS the hum alone is above the ACX limit, even though no single spike crosses -58 dBFS. That is the trap: a spectrum analyzer shows level per bin, while ACX measures the total, so reading only the tallest peak would have passed a recording that fails.
Two smaller details matter. The 60 Hz spike reads 59.2 Hz, because 60 Hz falls between bin 22 (59.22 Hz) and bin 23 (61.91 Hz); with a 4,096 block the bins are 10.77 Hz wide and the label would be off by nearly 5 Hz. Energy split across two bins also reads up to about 1.4 dB low with a Hann window, so the real hum is slightly worse than shown. And the strong 180 Hz line, louder than 120 Hz, points to a ground loop rather than the room.
The next step is specific: plug the interface into the same power strip as the laptop and measure again. The rerun puts the combined hum at -77.8 dBFS, 17.8 dB inside the limit, and 66.9 dB below the -12.5 dBFS speech peaks. In my view the setting that matters most here is block size; without fine bins, the 120 Hz and 180 Hz lines blur into the floor and the diagnosis is guesswork.
Types of Spectrum Analyzers Compared
Hardware spectrum analyzers come in three broad designs, and this browser spectrum analyzer belongs to the digital family. The table sums up how they differ for RF and audio testing.
| Type | How it measures | Strength | Typical use |
|---|---|---|---|
| Analog spectrum analyzer | Swept tuner, one point at a time | Simple and robust | Basic RF checks, teaching |
| Digital spectrum analyzer | Transforms on sampled signal blocks | Fine RBW, wide dynamic range | Wireless and RF design |
| Real-time spectrum analyzer | Gap-free transforms on every sample | Catches transient events | Monitoring, signal analysis |
Swept, FFT and Real-Time Designs
- Swept spectrum analyzers tune across the band and can miss a tone that comes and goes between passes.
- Digital spectrum analyzers compute many bins at once, which is why they replaced most swept models and why a browser can do the job.
- Real-time spectrum analyzers process every sample with no gaps, so a transient click or signal bursts lasting microseconds still show up.
- That makes real-time spectrum analyzers the choice for spectrum monitoring and interference hunting.
- Many digital spectrum analyzers add demodulation, decoding AM, FM and other modulated signals.
Benchtop, Portable, Handheld and Networked Units
- Benchtop models run on mains power and live in the laboratory, where size and weight don't matter.
- Portable units add a battery for work in the field.
- Handheld spectrum analyzers trade some performance for low weight.
- Networked spectrum analyzers have no screen; you reach them through a web interface and spread several input channels across distant sites.
The browser tool on this page sits closest to a handheld unit: it gives up absolute accuracy for convenience, and your phone or laptop microphone takes the place of the input connector.
Where Spectral Analyzers Earn Their Keep
Bench spectral analyzers are standard kit for engineers in telecommunications and electronics, where signal analysis covers RF transmitters, electromagnetic interference (EMI) scans and troubleshooting circuits in research. A browser spectrum analyzer covers the sound side: audio analysis and testing of speakers, instruments, room hum and machinery. The frequency analysis you see is only as honest as the signal source, though. Phone and laptop microphones are tuned for speech, so their levels are relative, not absolute, without calibration against a known reference. Use the pitch readings with confidence and the dB readings as comparisons.
Frequently Asked Questions
- What is a spectrum analyzer?
- A spectrum analyzer measures how much energy a signal carries at each frequency and plots amplitude against frequency. Where an oscilloscope shows a waveform changing over time, a spectrum analyzer shows the separate tones that make it up.
- How does this spectrum analyzer work?
- It records short blocks of sound from your microphone and runs a fast Fourier transform on each block. Every block becomes a set of frequency bins, and the height of each bin is shown in dBFS, decibels below digital full scale.
- Which FFT size should I choose?
- Use a large FFT size (8192 or more) for steady sounds such as hum or a held note, where you need to separate tones a few hertz apart. Use a small one (512 to 1024) for speech, drums or anything that changes quickly. 2048 is a good middle ground.
- Why does the peak frequency read slightly off?
- Each bin covers a band equal to the sample rate divided by the FFT size, so a tone between two bins is labelled with the nearer one. A 60 Hz hum at 44,100 Hz and 16,384 samples reads 59.2 Hz. Raise the FFT size to narrow the error.
- Is a network analyzer the same as a spectrum analyzer?
- No. A spectrum analyzer passively measures the frequencies already present in a signal. A network analyzer sends its own test signal into a device and measures how the device changes it, such as gain, loss and reflection.
- Can a browser spectrum analyzer check Bluetooth or Wi-Fi?
- No. Your microphone only picks up sound, roughly 20 Hz to 20,000 Hz. Bluetooth and Wi-Fi run at 2.4 GHz and above, which needs a hardware RF spectrum analyzer with an antenna input.
- Should I use the linear or logarithmic scale?
- Pick logarithmic for music and voice, since each octave gets equal space, matching how you hear pitch. Pick linear when you are checking evenly spaced harmonics or high-frequency content.