When I simulate with Gaussian inputs, I get rates slightly below that even for large n. Is that due to finite blocklength? Or must I also shape the codewords to have exactly power P, not just in expectation?
I’m working through Cover & Thomas, Chapter 10. For a real AWGN channel: ( Y = X + Z ), ( Z \sim \mathcal{N}(0, N) ), average power constraint ( \frac{1}{n}\sum x_i^2 \leq P ). Shannon says capacity ( C = \frac{1}{2}\log_2\left(1 + \frac{P}{N}\right) ) bits per use.