Microphone Spectrum Analyzer: Free Mic Quality Test
Click Start Microphone on this microphone spectrum analyzer to see your live spectrum and Dominant Frequency. Nothing you do in the test mic recording quality ever leaves your device — the whole test runs locally in your browser tab.
Your microphone quality and spectrum check starts the moment you press Start: this free microphone spectrum analyzer turns live audio from your microphone into a real-time FFT graph of every frequency between 20 Hz and 20 kHz. You see which frequencies it captures strongly, where it fades, and how much bass, mid and treble loudness it actually captures — the same view sound engineers use for mixing and mastering, without installing anything.
How to Run a Mic Test on the Frequency Spectrum
A mic test here takes about ten seconds and runs entirely in your browser through the Web Audio API. Nothing is uploaded or stored: your audio is analysed on your own device and discarded straight away, so no recording ever leaves your computer. This online frequency analyser needs only a working input and permission to use it. Nothing you do in the measure audio round trip delay ever leaves your device — the whole test runs locally in your browser tab.
- Click Start and allow microphone access when the browser asks.
- Choose the right input — a built-in mic, a USB headset or an audio interface — from the device list.
- Speak, sing or play a steady sound and watch the frequencies fill the frequency spectrum.
- Stay silent for five seconds to see the hiss your setup produces on its own.
- Switch to the spectrogram to watch how your frequencies change over time, and hover any peak on the spectrum display to read its exact value.
To check playback instead of your microphone — a music app or a video in another tab, or a reference track for mixing — use the browser's tab-capture option and tick "Share tab audio"; where that isn't supported, upload an audio file instead.
Microphone Access, Gain and FFT Size Settings
FFT size sets how finely the frequency axis is sliced: larger sizes separate close tones but react more slowly. Gain lifts a quiet signal so it fills the display, without changing the capsule itself. Smoothing averages successive frames so the trace stops flickering; turn it down when you are hunting short clicks. Leave the frequency axis in its default log view unless you are checking electronics, because a linear scale crams everything below 1 kHz into a sliver on the left and makes audio work much harder to read.
Reading the Logarithmic Scale and Peak Frequency
The horizontal axis runs from 20 Hz (bass, left) to 20 kHz (treble, right) on a logarithmic scale, so every octave takes the same width — the way you hear pitch and the standard for mixing. The vertical axis shows amplitude in dBFS, decibels relative to digital full scale: zero is the loudest value the converter can store, and everything below it is negative. Bar height is loudness. The tallest bar is the peak frequency; hover over it to read its value in Hz, then look for other tall peak frequencies at whole-number multiples of it.
What the Sound Spectrum Analyzer Measures
A microphone records pressure changes over time — a waveform in the time domain. The sound spectrum analyzer runs a Fast Fourier Transform (FFT) on short slices of that audio signal, using the Web Audio API, to move it into the frequency domain, where each bar shows how much energy sits at one frequency. The real-time result covers every frequency between 20 Hz and 20 kHz and tells you what your microphone actually hears: which frequencies it captures strongly, which it misses, and what unwanted sound rides along with every word. Nothing you do in the mic level meter ever leaves your device — the whole test runs locally in your browser tab.
FFT Bins, Frequency Resolution and Sample Rate
An FFT splits the audio band from 0 Hz up to half the sample rate into equal slices called bins. The width of each bin is your frequency resolution, and the length of audio each frame needs is the window:
$$\Delta f = \frac{f_s}{N} \qquad T = \frac{N}{f_s}$$
where \(f_s\) is the sampling frequency in Hz and \(N\) is the number of FFT points. Take a typical USB headset running at 44,100 Hz with an FFT size of 2,048: each bin is 44,100 ÷ 2,048 = 21.53 Hz wide and each frame covers 46.4 ms. Good spectrum analysis is always that compromise between detail and speed:
| FFT points | Bin width at 48 kHz | Window 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 |
Small sizes suit speech and fast transients; large ones separate two tones a few hertz apart. That is why every FFT analyser lets you change it.
RMS Level, Crest Factor and Clipping
The meters beside the graph summarise the signal over time. Peak is the single loudest sample; RMS level is the average power and tracks perceived loudness more closely; the crest factor is the gap between them in dB. Say a spoken line reads a peak of −3.2 dBFS and an RMS level of −17.6 dBFS: that 14.4 dB gap is typical of natural, uncompressed dialogue. Values of 6 dB or less point to heavy compression or a limiter, while a very dynamic performance can run well above 20 dB. If the peak hold touches zero, the input is clipping — lower the input level until peaks stay several dB below full scale, because clipped audio carries distortion you cannot remove later. The VU meters and the oscilloscope also reveal DC offset, a waveform pushed off its centre line.
Spectral Centroid, Dominant Frequency and Note Detection
The dominant frequency is the loudest bin, and note detection maps it to the nearest note (A4 = 440 Hz) with an offset in cents. That helps with tuning instruments and, even more, with naming a noise: a 60 Hz tone reads as B1, 49 cents flat. Because a harmonic can be louder than the fundamental, better tools search for the fundamental frequency behind the overtones rather than trusting the tallest bar, so you get the right note in the right octave. The spectral centroid is the amplitude-weighted average frequency — one number for brightness. If it falls sharply when you swap mics on the same singer, the second microphone has a darker timbre.
Judging Microphone Quality and Spectrum Bands
To judge a microphone, record the same spoken line at the same distance and read the spectrum band by band. A good vocal model shows a smooth curve through the lower and middle bands, a gentle lift in the presence region for clarity, and no spikes at fixed frequencies once you stop talking. The table maps each band of the audio spectrum to what it adds and what a problem looks like.
| Band | Range | What it adds | What a problem looks like |
|---|---|---|---|
| Sub bass | 20 Hz–60 Hz | Weight you feel more than hear | Rumble from desks, stands or traffic |
| Bass | 60–250 Hz | Warmth of vocals, body of a kick drum | Boom and muddiness up close |
| Low mid | 250–500 Hz | Fullness of speech | Boxiness from a small, bare room |
| Mid | 500 Hz–2 kHz | Fundamental of most vocals and instruments | Nasal or honky tone |
| Upper mid | 2–4 kHz | Consonants, where hearing is most sensitive | Harshness on cheap capsules |
| Presence | 4–6 kHz | Articulation and clarity | Dull, muffled words if this band drops away |
| Brilliance | 6–12 kHz | Sparkle on cymbals and hi-hat | Spitty "s" sounds |
| Air | 12 kHz–20 kHz | Openness and room sound | Self-noise sitting flat across the top of the audio band |
Noise Floor and Background Noise
Mute yourself and the trace that remains is the noise floor: the capsule's self-noise, preamp hiss and the room's background noise combined. A healthy setup leaves a low, flat line. A broad hump in the bass points to traffic or ventilation, and a raised, even layer across the treble points to a noisy preamp or an input level set too high. Compare that floor with your speaking level in the online frequency analyser — the further your voice sits above it, the cleaner the audio. If the floor rises as you raise the input level, the problem lives early in the signal chain, not in the room.
Mains Hum at 50 Hz and 60 Hz
Mains hum shows up in the frequency analyser as a narrow spike that stays put whatever you say — 60 Hz in North America, 50 Hz in Europe and much of Asia — with each harmonic stacked at whole multiples above it. Electrical hum almost always comes from the cable path, not the capsule: a ground loop between computer and interface, an unbalanced lead running past a power supply, or poor shielding. Fix it with balanced connections, a single shared power strip or a ground loop isolator, then rerun the test and confirm the spike is gone.
Proximity Effect and Sibilance
Two capsule effects show up clearly among the frequencies. Proximity effect is the bass boost directional mics add as you move closer: record a phrase at 30 cm and again at 5 cm, and the bass energy below 200 Hz climbs while the spectral centroid falls. Sibilance appears on the spectrogram as bright bursts between 5 and 9 kHz on every "s" and "t". If either is excessive, change position or angle before reaching for mixing EQ — a mic placed slightly off-axis often tames both.
Chasing Audiobook Hum with the Online Frequency Checker
I'm about to record a narration sample for an audiobook platform that follows ACX's published specs: RMS between −23 and −18 dB, peaks no higher than −3 dB, and a noise floor no louder than −60 dB RMS. The chain is a second-hand dynamic mic into a two-channel interface at 48 kHz, and the decision is simple: record tonight, or fix something first?
I set the FFT size to 2,048 points, read a test paragraph, then sit silent for ten seconds of room tone. The readings:
- Speech RMS −21.7 dBFS, peak −4.6 dBFS — both inside the window.
- Room tone RMS −64.2 dBFS, a gap of −21.7 − (−64.2) = 42.5 dB, so the −60 dB floor rule passes.
- A narrow spike the peak readout labels 70.3 Hz, at −61.3 dBFS, with smaller spikes above it.
That 70.3 Hz label nearly fooled me. At 48,000 ÷ 2,048 each bin is 23.44 Hz wide, and 60 Hz sits 2.56 bins up, so its energy lands in the 70.3 Hz bin and reads about 0.8 dB low. Raising the FFT size to 4,096 moves the label to 58.6 Hz, with neighbours at 117 and 176 Hz: 60 Hz mains hum and its harmonics, plain as day on the frequency spectrum. The RMS figure alone would have passed this take, yet anyone on headphones would hear that tone in every pause.
The laptop and interface are on different wall sockets, so I move both onto one power strip and rerun with nothing else changed:
| Reading | Before | After one power strip |
|---|---|---|
| 60 Hz spike | −61.3 dBFS | −79.5 dBFS |
| Room tone RMS | −64.2 dBFS | −71.6 dBFS |
| Speech-to-room-tone gap | 42.5 dB | 49.9 dB |
One cable change bought 7.4 dB of quieter floor and pushed the hum down 18.2 dB. My judgement: on this kind of check, the number to watch is the hum spike against your speech, not the overall floor, because a pure tone carries little energy but stands out in silence. I'll record tonight, and rerun the test after any cabling change, since a new USB hub can bring the loop straight back.
Microphone Frequency Response and the Hearing Range
Human hearing spans roughly 20 Hz to 20 kHz, and the hearing range shrinks in the top octave with age. A microphone's frequency response describes how evenly it captures those frequencies. Many vocal models deliberately trim the deep bass and the top of the treble, below 80–100 Hz and above 15–16 kHz; that rolloff is fine for talking but shows clearly in a frequency analysis of music. To compare two mics, save the first as a reference trace, swap mics without moving anything, and read the gap between the two curves in the online frequency checker. A difference of a few dB is a change of character you can correct with EQ; a gap of 10 dB or more in one band is a real limit. Your own mic colours every reading, so for speaker or room work use a flat, calibrated measurement microphone.
Pink Noise, Sweeps and Room Modes
The frequency analyser can also judge the room and speakers around you. Play pink noise — equal energy per octave — from a tone generator through a speaker, put the capsule at the listening position, and peaks and nulls in the bass reveal room modes. White noise has equal energy per hertz and looks tilted upward on the display; a sine wave tests one frequency; a log sweep from 20 Hz to 20 kHz exposes resonances and cancellations, a quick check of room acoustics. The lowest axial mode is the speed of sound divided by twice the room length: a 5.2 m room gives 343 ÷ 10.4 ≈ 33 Hz. Keep levels low, because sweeps at high volume can damage tweeters.
Octave Bands and Z Weighting
Octave bands group the spectrum's frequencies into proportional slices: 20 Hz to 20 kHz is almost exactly ten octaves, so a full-octave view gives about ten bars and a 1/3 octave view about thirty-one. The weighting setting changes how they are summed: Z is flat, A mimics how quiet sounds are heard and trims bass, and C weighting stays flatter in the bass for loud or impact sounds. None of this turns a browser into a calibrated SPL meter — treat readings as relative, not absolute.
Spectrogram and Waveform Views in the Online Frequency Analyser
The bar graph shows one instant; the spectrogram shows time. It scrolls left to right with frequencies running bottom to top and colour showing amplitude, so a steady tone draws a horizontal line, a room resonance leaves a streak that lingers after you stop, and a sung phrase appears as stacked harmonics that bend with your pitch. The waveform oscilloscope shows the raw real-time audio trace instead, which is the quickest way to spot clipped flat tops, while the VU meters track level. When you find something worth keeping, freeze the display, save a screenshot, or export the spectrum as CSV to open in a spreadsheet — this online frequency analyser keeps all of it on your device.
Mid/Side Stereo and Mono Audio Checks
With a stereo pair, the mid/side view splits the signal into what both channels share and what differs between them. A strong side signal means a wide image; a near-empty side means your recording will hold up in mono. For a single USB model, the side channel should be almost silent — if it isn't, one channel may be faulty or out of phase.
Leq, L10 and L90 Noise Sessions
For longer checks, a session logs levels over time. Leq is the energy-average level for the whole run, L10 the level exceeded 10% of the time (the loud moments) and L90 the level exceeded 90% of the time — effectively your room's background floor. A minute suits steady noise such as an air conditioner; give a busy room several minutes so the background reading settles, then export the log. A wide gap between the two tells you the noise comes and goes, which matters when you pick a time to record.
Who Uses a Real-Time Spectrum Analyser
Anyone who needs to see audio rather than guess at it. A real-time spectrum analyser in the browser works as a quick frequency checker for:
- Streaming and podcast setups — confirm your headset or USB mic is clean, free of hum and not overloading before you go live.
- Music production — check vocal and instrument recordings, low-end balance and EQ decisions while mixing.
- Microphone testing — compare two mics, or one mic at different distances, against a saved reference.
- Speaker and room checks — find room resonances and speaker rolloff with a test signal.
- Broadcast and voice work — catch sub-20 Hz rumble, hiss and harsh esses before delivery.
- Learning audio — sing a note, check your tuning, and watch the fundamental and harmonics appear on the spectrogram.
Sound engineers still use dedicated hardware for certified measurements, but for everyday audio analysis this free frequency analyser and online frequency checker cover every microphone quality and spectrum check that matters.
Frequently Asked Questions
- What does a microphone spectrum analyzer show?
- It splits the sound your mic picks up into its frequencies with a Fast Fourier Transform and shows how loud each one is, from 20 Hz bass on the left to 20 kHz treble on the right. Taller bars mean more energy at that frequency.
- Is my audio recorded or uploaded?
- No. The analysis runs in your browser through the Web Audio API, and the audio is processed on your device and discarded. Nothing is sent to a server.
- Which FFT size should I choose?
- Smaller sizes react faster and suit speech and transients; larger sizes give finer frequency resolution for separating close tones such as hum and a bass note. Resolution is the sample rate divided by the FFT size, so 4,096 points at 48 kHz gives about 11.7 Hz per bin.
- How do I find 50 Hz or 60 Hz mains hum?
- Stay silent and look for a narrow spike that does not move at 50 Hz (Europe, most of Asia) or 60 Hz (North America), often with weaker spikes at whole multiples. Balanced cables, a shared power strip or a ground loop isolator usually remove it.
- What is the difference between A, C and Z weighting?
- Z is flat and treats every frequency equally. A follows how the ear hears quiet sounds and reads bass and extreme treble lower. C stays much flatter in the bass and suits loud or impact noise. None of them turns a browser into a calibrated sound level meter.
- What do Leq, L10 and L90 mean?
- Leq is the energy-average level over a session, L10 the level exceeded 10% of the time (the loud moments) and L90 the level exceeded 90% of the time, which is effectively your background noise floor. A large gap between L10 and L90 means the noise comes and goes.
- What do peak, RMS and crest factor tell me about my mic?
- Peak is the loudest single sample and shows how close you are to clipping. RMS is the average power and tracks perceived loudness. Crest factor is peak minus RMS: a high value means natural, dynamic audio, a low value means heavy compression.