Spectrum Analyzer

Spectrum analyzers are indispensable tools for audio engineers, acousticians, and noise control professionals. They reveal what the ear hears as a single sound as a collection of frequency components — each with its own magnitude. This decomposition enables identification of tonal problems, feedback frequencies, HVAC noise bands, and system bandwidth limitations. There are two primary types of audio spectrum analyzers: FFT-based (constant bandwidth) and filter-bank/octave-band (constant percentage bandwidth). FFT analyzers divide the spectrum into equally-spaced frequency bins (e.g., every 5.86 Hz for a 8192-point FFT at 48 kHz). Octave-band analyzers use filters with bandwidth proportional to center frequency (constant Q), which better matches human frequency perception. Resolution in an FFT analyzer is determined by the FFT size and sample rate: Δf = fs/N. A longer FFT window gives finer frequency resolution but slower update rate and poorer time resolution. For real-time applications, 4096-8192 point FFTs at 48 kHz provide a good balance (5.86-11.72 Hz resolution, 85-170 ms windows). For detailed low-frequency analysis, 16384-32768 points may be needed. Windowing is essential to prevent spectral leakage. The rectangular window (no window) gives the narrowest main lobe but worst sidelobes. The Hann window is the standard compromise. The Blackman-Harris window provides excellent sidelobe rejection (-92 dB) at the cost of wider main lobe, making it preferred for identifying low-level tonal components near strong signals. Modern spectrum analyzers like SonaVyx add features beyond simple magnitude display: spectrograph (waterfall) views showing time evolution, peak hold, averaging modes (linear, exponential), octave-band overlay on FFT data, and frequency-weighted display (A, C, Z).