Question 1

What is an FFT spectrum analyzer and how does it work?

Accepted Answer

An FFT (Fast Fourier Transform) spectrum analyzer converts a time-domain audio signal into its frequency components. It takes a block of audio samples (the FFT size), applies a window function to reduce spectral leakage, and computes the discrete Fourier transform. The result shows the magnitude (in dB) at each frequency bin, revealing which frequencies are present and how loud they are. The frequency resolution equals the sample rate divided by the FFT size: at 48 kHz with 4096 points, each bin represents 11.7 Hz. Higher FFT sizes give finer resolution but slower update rates. SonaVyx implements the FFT in Rust compiled to WebAssembly for real-time performance, computing thousands of FFTs per second directly in your browser without sending any audio data to a server.

Question 2

What is a window function and why is it necessary?

Accepted Answer

The FFT assumes the input signal is periodic within the analysis block. In reality, the captured audio block rarely starts and ends at the same amplitude, creating an artificial discontinuity. This discontinuity spreads energy across all frequency bins, an effect called spectral leakage. A window function tapers the signal smoothly to zero at the edges, reducing leakage at the cost of slightly broader frequency peaks. Common windows: Hann (general purpose, good frequency resolution), Blackman-Harris (excellent sidelobe rejection, wider main lobe), Rectangular (no window, narrowest main lobe but worst leakage), and Flat-Top (accurate amplitude measurement). SonaVyx defaults to Blackman-Harris for measurement because it provides the best sidelobe rejection (-92 dB), meaning frequency components far from a strong tone are accurately displayed without leakage artifacts.

Question 3

What is the difference between linear and logarithmic frequency scales?

Accepted Answer

A linear frequency scale spaces frequencies evenly: 100 Hz and 200 Hz are the same distance apart as 10000 Hz and 10100 Hz. A logarithmic scale spaces frequencies by ratio: each octave (doubling of frequency) occupies the same width. Human hearing perceives frequency on a roughly logarithmic basis — the interval between 100 Hz and 200 Hz (one octave) sounds the same as 1000 Hz to 2000 Hz (one octave). Audio spectrum analyzers almost always use a logarithmic frequency axis because it shows the perceptually relevant frequency range (20 Hz to 20 kHz) evenly distributed across the display. SonaVyx uses a logarithmic frequency axis with standard audio frequency labels (31.5, 63, 125, 250, 500, 1k, 2k, 4k, 8k, 16k Hz).

Question 4

What is spectral leakage and how do I minimize it?

Accepted Answer

Spectral leakage occurs when a frequency component in the signal does not align exactly with an FFT bin center. The energy from that component spreads (leaks) into adjacent bins, appearing as a broad hump rather than a sharp peak. Leakage is worst with no window (rectangular) and for frequencies between bin centers. To minimize leakage: use a window function with good sidelobe rejection (Blackman-Harris, Flat-Top). Increase the FFT size so bins are closer together in frequency, reducing the maximum distance from a component to the nearest bin center. Use zero-padding to interpolate between bins for sharper peak display. For SonaVyx, the default Blackman-Harris window reduces sidelobes to -92 dB, meaning a 100 dB tone leaks less than 8 dB into distant bins. This is sufficient for most audio measurement applications.

Question 5

What is the frequency resolution of an FFT analyzer?

Accepted Answer

Frequency resolution (also called bin width) equals the sample rate divided by the FFT size: delta_f = fs / N. At 48 kHz sample rate: 1024-point FFT gives 46.9 Hz resolution, 4096 gives 11.7 Hz, 8192 gives 5.9 Hz, 16384 gives 2.9 Hz, and 32768 gives 1.5 Hz. Higher resolution is needed to resolve closely spaced frequencies (such as room modes below 200 Hz or the 50/60 Hz hum harmonics). The trade-off is time resolution: higher FFT size means longer blocks, slower updates, and less ability to track rapidly changing signals. The uncertainty principle sets a fundamental limit: delta_f × delta_t ≈ 1, meaning you cannot have both fine frequency and fine time resolution simultaneously. SonaVyx lets you adjust FFT size from 1024 to 32768 to find the right balance for your application.

Question 6

What is the Nyquist frequency and why does it limit analysis?

Accepted Answer

The Nyquist frequency is half the sample rate and represents the highest frequency that can be accurately captured. At 48 kHz sample rate, the Nyquist frequency is 24 kHz. Frequencies above Nyquist are aliased — they appear as false lower frequencies in the analysis. The human hearing range extends to approximately 20 kHz, so 48 kHz sample rate captures the entire audible spectrum with margin. The FFT output extends from 0 Hz to the Nyquist frequency, with each bin covering sample_rate/FFT_size Hz. In practice, the useful range is limited at the high end by microphone response (phone mics typically roll off above 10-15 kHz) and at the low end by the microphone noise floor and FFT resolution. SonaVyx displays the spectrum from 20 Hz to 20 kHz, covering the full audible range.

Question 7

What is the difference between power spectrum and magnitude spectrum?

Accepted Answer

The magnitude spectrum shows the absolute value of each FFT bin: |X(f)|, typically displayed in dB as 20 × log10(|X(f)|). The power spectrum shows the squared magnitude: |X(f)|², displayed as 10 × log10(|X(f)|²). In dB, they produce the same result (20 × log10(|X|) = 10 × log10(|X|²)). The distinction matters when computing cross-spectra, auto-spectra, and transfer functions, which use power-related quantities. For display purposes, magnitude and power spectra look identical in dB. Power Spectral Density (PSD) normalizes by the frequency bin width to give power per Hz, which allows comparing spectra measured with different FFT sizes. SonaVyx displays the magnitude spectrum in dB by default, which is the standard for audio measurement.

Question 8

Can I use a spectrum analyzer to identify specific sounds or noise sources?

Accepted Answer

Yes, spectrum analysis is one of the best tools for identifying noise sources because different sources have characteristic spectral signatures. Electric hum appears as sharp peaks at 50 or 60 Hz (mains frequency) and its harmonics (100/120, 150/180, 200/240 Hz). Fan noise shows as broadband energy with a broad peak at the blade-pass frequency. Transformer hum is predominantly at 100/120 Hz (twice mains). Motor noise creates peaks at the rotation frequency and harmonics. Feedback appears as a single narrow peak that grows over time. Rattle and buzz create harmonic series at the resonant frequency. By identifying the spectral signature, you can locate and address the specific source. SonaVyx peak detection highlights prominent tones and the problem detector classifies common noise signatures automatically.

FFT Spectrum Analyzer Basics: How Frequency Analysis Works

Frequently Asked Questions