Signal-to-Noise Ratio (SNR)

SNR is a universal quality metric that applies at every stage of the audio chain: microphone self-noise, preamp noise floor, digital converter quantization noise, signal processing headroom, speaker system noise, and the acoustic environment. The weakest link determines the overall system SNR. In acoustic measurement, SNR directly determines which parameters can be reliably extracted. The impulse-to-noise ratio (INR) — essentially the SNR of an impulse response measurement — must exceed 35 dB for valid T20 and 45 dB for valid T30. Low INR causes the decay curve to flatten prematurely as it approaches the noise floor, producing artificially long reverberation times. For speech intelligibility (STI), the relationship between SNR and intelligibility follows a sigmoidal curve. Below 0 dB SNR, intelligibility drops rapidly toward zero. Between 0 and 15 dB SNR, it rises steeply. Above 15 dB SNR, intelligibility plateaus near 1.0 (limited only by reverberation). This relationship is captured in the modulation transfer function: m_noise = 1/(1 + 10^(-SNR/10)). In digital audio, SNR is determined by the bit depth: theoretical SNR = 6.02 × N + 1.76 dB, where N is the number of bits. A 16-bit system provides 98 dB SNR; 24-bit provides 146 dB (though analog electronics limit practical 24-bit systems to about 115-120 dB). Dithering is used to decorrelate quantization error from the signal, trading correlated distortion for random noise. Practical SNR improvement techniques: increase signal level (closer microphone placement, higher source output), reduce noise (HVAC off during measurements, shielded cables), use averaging (each doubling of averages improves SNR by 3 dB), and use longer excitation signals (sweep duration in IR measurements). SonaVyx displays the noise floor and computes SNR metrics for measurement quality validation.