Noise Frequency Analyzer
This noise frequency analyzer helps you identify mystery noises by their frequency fingerprint. Capture a quiet baseline, then record the problem noise to see exactly what changed in the baseline noise comparison. The tool matches against 30+ known noise signatures, classifies noise color (white/pink/brown), and provides actionable noise troubleshooting advice — all processed locally in your browser.
Noise Frequency Analyzer Tool
| Noise Source | Freq Range | Status |
|---|
| Time | Source | Freq | Confidence |
|---|
How to Use the Noise Frequency Analyzer
-
Start Listening
Click "Start Listening" and grant microphone permission. Select your preferred microphone from the dropdown if needed. The live spectrum begins immediately.
-
Capture Baseline
In a quiet moment (when the problem noise is absent), click "Capture Baseline". The tool records 3 seconds of ambient sound as your reference point.
-
Capture the Noise
When the problem noise is present, click "Capture Noise". The tool records 3 seconds and computes the difference spectrum, highlighting exactly which frequencies the noise adds.
-
Review the Diagnosis
The tool matches detected frequencies against 30+ noise signatures and shows the most likely source with its icon, description, and a specific troubleshooting fix. The noise color classification tells you the spectral character.
-
Export or Copy Results
Use "Copy Noise Report" for a quick summary or "Export CSV" to download the full detection history for documentation or further analysis.
Understanding Your Results
Noise Signature Matching
The tool maintains a database of 30+ common noise sources, each defined by a characteristic frequency range. When your audio contains energy in one of these ranges above the threshold, the tool helps you identify the noise frequency and provides a confidence percentage based on how closely your signal matches the known pattern. Sources range from narrow-band tonal interference (electrical hum, CRT whine) to broadband mechanical noise (HVAC rumble, appliance motors).
Baseline vs Noise Comparison
The difference spectrum subtracts your baseline recording from the noise recording, showing only the frequencies that changed. This eliminates ambient sounds (air conditioning, computer fans, etc.) and isolates the actual problem noise, making noise signature matching far more accurate. The same spectral subtraction technique is used in professional acoustic measurement software for characterizing environmental noise and identifying room modes in studio monitoring setups. If you want to track overall decibel levels over a session rather than pin down a specific frequency, the noise level logger is a complementary tool.
Noise Color Classification
Noise is classified by its spectral slope — how energy is distributed across frequencies. This noise color identification mirrors the terminology used in psychoacoustics, audio engineering, and electronic circuit noise characterization:
- White noise has equal energy at all frequencies (flat slope, 0 dB/octave)
- Pink noise drops 3 dB per octave — equal energy per octave, common in nature and used as an audio reference signal
- Brown noise drops 6 dB per octave — deep, rumbling character like thunder or ocean waves
- Blue noise increases 3 dB per octave — hissy, high-frequency emphasis, sometimes called azure noise
Frequency Fingerprint
The fingerprint visualization shows the unique spectral shape of the detected noise as a compact barcode-style display. Each noise source has a characteristic fingerprint pattern that helps with visual identification even when multiple sources are present. Tonal noise sources like mains hum produce sharp vertical bands at their fundamental and harmonic frequencies, while broadband sources show a more diffuse pattern across the spectrum.
Confidence Level
The match confidence indicates how well the detected frequency pattern matches a known noise signature. Higher confidence means a stronger, cleaner match. Multiple simultaneous sources may reduce individual confidence scores. For narrow-band tonal noise (electrical hum, buzzing transformers), confidence is typically high. For broadband sources like HVAC systems or background traffic, the spectral overlap with other noise types naturally yields lower confidence scores. To improve accuracy, use the peak frequency detector alongside this tool to confirm the dominant frequency before cross-referencing signatures.
Technical Background
Noise identification combines spectral analysis with pattern matching. The tool analyzes incoming audio in real time using the Web Audio API's FFT implementation and compares the spectral characteristics against known noise profiles.
Baseline Subtraction
The baseline capture stores the average power spectrum of your environment over 3 seconds. When you capture the noise, the tool computes the spectral difference in dB, revealing only the frequencies that increased above the ambient floor. This is the same principle used in professional noise analysis — similar to how noise-canceling headphones identify and remove unwanted sound. The approach is also analogous to octave-band analysis used in room acoustics, where a known reference state is subtracted to reveal the contribution of a specific source. You can also upload a pre-recorded WAV or MP3 file to run the same analysis on an existing recording, making it possible to investigate intermittent noise you captured earlier with a dedicated frequency detector session.
Signature Matching Algorithm
Each noise signature in the database defines a frequency range and optional harmonic structure. The matcher calculates energy concentration within each range and checks for expected harmonic relationships (e.g., 60 Hz + 120 Hz + 180 Hz for mains hum). Confidence is weighted by how much energy falls within the expected range versus outside it.
Spectral Slope Estimation
Noise color is determined by fitting a linear regression to the log-frequency vs. dB power spectrum. The slope of this line classifies the noise: 0 dB/octave = white, -3 dB/octave = pink, -6 dB/octave = brown. The regression uses octave-spaced sample points for an unbiased estimate across the audible range.
Frequently Asked Questions
What is the difference between baseline and noise capture?
The baseline records the ambient sound when the problem noise is absent — your normal background. The noise capture records the sound when the problem is present. The tool then subtracts the baseline from the noise to isolate exactly which frequencies the problem noise adds, eliminating false matches from normal ambient sounds.
Do I need to capture a baseline to use the tool?
No — the tool works in real-time mode without a baseline. However, baseline capture significantly improves accuracy by removing ambient noise from the analysis. Without a baseline, the tool matches against the raw spectrum, which may include false matches from normal background sounds.
What are noise colors (white, pink, brown)?
Noise colors describe the spectral distribution of random noise. White noise has equal energy at every frequency (like TV static). Pink noise has equal energy per octave, sounding more balanced and natural (like rainfall). Brown noise has more energy at low frequencies, creating a deep rumble (like thunder or ocean waves).
Why does the tool detect 60 Hz or 50 Hz hum?
A persistent 60 Hz (North America) or 50 Hz (Europe/Asia) tone is electrical mains hum caused by electromagnetic interference from power lines, transformers, or cables. The harmonics at 120/100 Hz are also common. Use shielded cables, a ground loop isolator, or move away from power sources.
How many noise signatures does the database contain?
The database contains 30+ noise signatures covering electrical hum (50/60 Hz and harmonics), ground loops, fluorescent lighting, HVAC systems, refrigerators, computer fans, dimmer switches, transformers, CRT whine, USB power supply noise, water pipe vibration, appliance motors, and more. Each signature includes the frequency range, description, and a specific troubleshooting fix.
Can this tool identify multiple noise sources at once?
Yes. The tool scans for all matching signatures simultaneously and lists every match with its confidence level. For example, it can identify both a 60 Hz mains hum and a 120 Hz ground loop buzz in the same audio signal, showing each match in the results.
Is my audio data private?
Absolutely. All frequency analysis and noise matching runs 100% in your browser using the Web Audio API. No audio data is ever recorded, transmitted, or stored on any server. The tool works completely offline once loaded.
What microphone should I use for best results?
Any microphone works, but results improve with a flat-response measurement microphone or a good-quality condenser microphone. Built-in laptop microphones tend to have poor low-frequency response, which may miss sub-bass noise sources. An external USB microphone placed close to the noise source gives the best results.
Can I analyze a noise from an audio file instead of a live microphone?
Yes. Use the drag-and-drop file upload area to load a WAV, MP3, OGG, or FLAC recording directly. The tool runs the same FFT-based signature matching and noise color classification on the file's audio as it would on a live microphone signal. This is useful for investigating intermittent noises you recorded earlier, or for analyzing a sample captured on a dedicated recorder with a better microphone than your device has built in.
Why does my HVAC or air conditioning unit show a low confidence match?
HVAC systems produce broadband noise spread across many frequencies rather than a narrow tonal peak, so their spectral pattern overlaps with other mechanical and ambient noise sources. This makes confidence scores lower than for tonal sources like 60 Hz mains hum. To get a better reading, capture your baseline with the HVAC running normally, then capture the noise when an additional sound appears. The difference spectrum will isolate only what changed, improving match accuracy significantly.
What does changing the FFT size do?
A larger FFT size gives finer frequency resolution at the cost of slightly slower temporal response. The default 8,192-point FFT resolves frequencies to roughly 5 Hz bins at a 44.1 kHz sample rate, which is sufficient to separate 50 Hz and 60 Hz mains hum. Increasing to 16,384 bins helps distinguish closely spaced harmonics (e.g., 60 Hz vs 63 Hz mechanical vibration). Use the smaller 4,096 setting only if the display feels sluggish on an older device.
Related Tools
Frequency Detector
Full real-time frequency detection with gauge, pitch accuracy, and multi-peak separation.
Try it →Peak Frequency Detector
Find the single loudest frequency in any sound with top-5 peaks and noise diagnosis.
Try it →Sub-Bass Detector
Specialized detection of ultra-low frequencies below 60 Hz with enhanced resolution.
Try it →