Audio Spectrum Analyzer

Reading a Frequency Spectrum

Look for broad patterns before trusting a single peak. A sine near 440 Hz should create a compact cluster; a signal containing 440 and 880 Hz can produce two clusters. A periodic instrument may show harmonics at multiples of its fundamental frequency. Distortion can add harmonics too, so a taller upper peak does not prove that the source played a higher note.

The audio spectrum analyzer draws sparse low-frequency bins as separate points and combines dense bins by their strongest value in each screen column. The chart starts at 20 Hz and ends at the lower of 20 kHz or the runtime Nyquist frequency. Nyquist is half the sample rate; there are no measured bins beyond that boundary.

Frequency, Magnitude, and FFT Bins

Bin spacing equals sample rate divided by FFT size. At 48000 samples per second, FFT 4096 gives about 11.7 Hz per bin, while FFT 16384 gives about 2.9 Hz. A 100 Hz signal therefore need not land exactly on a bin center. Windowing spreads its energy into neighboring bins. The displayed peak is rounded and should be interpreted within this resolution.

Digital magnitude is a logarithmic representation of the windowed FFT output. It is related to a full-scale digital reference, but the value for one spectral bin is not a calibrated broadband dBFS meter reading. It is never physical dB SPL. Changing the minimum and maximum dB controls only changes the visible vertical range; it does not amplify the input.

Why a Logarithmic Frequency Scale Helps

On the default scale, each doubling of frequency takes the same horizontal distance. The interval from 100 to 200 Hz is as wide as 1000 to 2000 Hz. This makes bass, midrange, and treble easier to compare in one view. Linear scale instead gives equal space to equal hertz intervals, which is useful when examining evenly spaced sidebands. Neither option changes the FFT resolution.

Choosing FFT Size and Smoothing

A larger FFT separates nearby tones more clearly but examines a longer span of audio. At 48 kHz, 16384 samples span about 341 milliseconds, so a short event can be spread across that window. Try 2048 for quick changes, 4096 for general viewing, and 8192 or 16384 for sustained low notes or neighboring tones. Size alone cannot make a microphone accurate.

Smoothing blends recent magnitudes, reducing flicker at the cost of response speed. Set it near zero to follow moving peaks, or raise it for a steadier view of sustained sound. The audio spectrum analyzer offers peak hold as a separate visual aid: it remembers the highest observed value in each bin during this session. Clear Peak Hold removes that trace; changing FFT size or input also clears it. It is not a timestamped recording.

Microphone Mode vs Audio File Mode

Microphone mode observes your input device without monitoring it through the speakers. The browser is asked to disable echo cancellation, noise suppression, and automatic gain control. Hardware, drivers, or the browser may ignore those requests. Input gain, microphone direction, distance, and room reflections all influence the spectrum you see.

Audio File mode decodes a mono or stereo recording locally and follows its current playback position. Listening volume sits after analysis, so turning it down does not lower the displayed spectrum. Stereo is averaged for analysis, which can cancel opposite-phase content. The decoded rate is the AudioContext rate and may differ from the original file rate because decoding can resample. Seek to compare passages and Pause to inspect one frame.

Peak Frequency vs Fundamental Pitch

The peak readout in this audio spectrum analyzer is the strongest displayed spectral bin above your chosen chart floor. A harmonic can be stronger than the fundamental, and two instruments can overlap. There is no musical-note detector in this view. Silence or a signal below the chart floor produces no peak reading instead of a made-up frequency.

For sustained single-note pitch estimates, open the Instrument Tuner. It uses a separate pitch-detection workflow.

Common Uses for an Audio Spectrum Analyzer

Compare a sine with a harmonic-rich waveform, check whether a recording contains low-frequency rumble, or observe how a sweep moves through the spectrum. White noise gives a broadly distributed, fluctuating trace rather than one steady line. A short modulation-generator WAV can reveal carrier and sideband regions when FFT spacing is fine enough. Keep input, gain, FFT size, and smoothing consistent when comparing examples.

Create an AM or FM example with the Modulation Generator, export it, and open that WAV here for direct digital inspection.

Speaker, Room, and Microphone Limitations

A microphone recording of a loudspeaker combines the speaker response, room reflections, input response, placement, and system processing. A dip could be cancellation at one listening position; a peak could come from the room. This audio spectrum analyzer does not separate those influences or measure a calibrated frequency response. A proper acoustic measurement also requires a known stimulus, suitable reference equipment, and a defined procedure.

Why This Is Not a Calibrated SPL Meter

Sound-pressure level needs a known relation between acoustic pressure and digital samples. This page has no microphone sensitivity calibration, acoustic reference, or certified weighting filters. A digital peak of −30 dB cannot tell you the sound pressure at your ears. Do not use these readings for workplace exposure compliance, medical hearing assessment, or equipment certification.

Privacy and Local Processing

The audio spectrum analyzer does not upload microphone samples or files. Microphone permission is requested only after your action, and access stops when you stop it, switch inputs, hide the page, or leave. Removing a file releases its decoded samples; switching input modes also clears the loaded file. Browser permission settings are separate and may remember your choice. File size, decoded memory, and codec support limit what the browser can open.