Oversampling SNR Improvement

Formula

Description

Oversampling an ADC by a factor of N and then averaging (decimating) improves the effective signal-to-noise ratio by spreading the quantization noise across a wider bandwidth. When the data is low-pass filtered and downsampled, only the noise within the signal bandwidth remains, effectively reducing the noise floor. Each 4x oversampling improves SNR by 6 dB, equivalent to gaining one bit of resolution. This technique is used in sigma-delta ADCs, digital audio, and precision measurement systems to achieve resolution beyond the native ADC capability.

Variables

• SNR_eff — Effective SNR after oversampling and decimation (dB)
• SNR — Original single-sample SNR (dB)
• N — Oversampling ratio

Practical Notes

This assumes the noise is white (uniformly distributed across the Nyquist bandwidth). If the noise is not white (e.g., 1/f noise), oversampling is less effective at improving SNR. For ADC applications, the input signal must contain dithering noise of at least 1 LSB amplitude for oversampling to work correctly. Without dither, quantization errors are correlated and do not average out properly.