Approximation for SQRT( I2 + Q2 ) with Gaussian Noise Present
When a complex signal is represented in terms of in-phase (I) and quadrature-phase (Q) components, a commonly used approximation for the signal root-mean-square (RMS) value is
Assuming a large number of OFDM subcarriers in the intended application, I and Q appear to both be Gaussian thereby making r a Rayleigh-distributed random variable where
We are interested in the ratio
From symmetry, it suffices to consider a limited range for θ as π/4 . Letting
where γ = 0.375. It is worthwhile evaluating just how good this choice for γ really is. We seek then to minimize the mean-square error given by
To read the article in its entirety in PDF format, please follow this link.