Confidence intervals for power estimates

Abstract

Confidence intervals for power estimates based on samples of signal-plus-noise from a radar receiver are computed for true cases of interest, a fluctuating (SWII) and a nonfluctuating signal. Exact equations for computing confidence intervals for these two cases are given, along with a normal approximation for use in the steady-signal case. We show that for the fluctuating signal the width of a 95% confidence interval as a function of the number N of samples, is essentially independent of the signal-to-noise ratio (SNR) and, for fixed N and a confidence interval of /spl plusmn/1 dB, the size (in percent) of the confidence interval approaches an asymptotic value less than 100% as the SNR increases. When the signal is constant, the size of a /spl plusmn/1 dB confidence interval approaches 100% as the SNR increases.