ECE 614 – Principles of Digital Communications Homework 4 Assigned on: 02/23/2016 Due by: 03/10/2016 4.1 Suppose a binary symbol a 1 is transmitted across three separate channels to a receiver. The receiver observations are r1 , r2 , r3 , as sketched below: n1 a 1 E n2 n3 r1 r2 r3 N The noise n1 , n 2 , n3 , are i.i.d. N 0, 0 . 2 (a) Sketch the block diagram of the ML detector. Simplify it as much as possible. (b) Find the MAP decision when r1 0.1 , r2 0.2 , r3 0.3 , and a priori pmf of the transmitted symbol is Pra 1 0.4 . Assume E 4 and N 0 0.2 . 4.2 Consider a one-dimensional model for PAM in AWGN, where the received sample is: r an where a is the transmitted symbol, and where n ~ N 0,1 is independent of a. Instead of the usual pulse-amplitude modulation scheme, in which a is chosen from an alphabet of possible amplitudes, consider a noise-variance-modulation scheme, in which the transmitted symbol is a Gaussian random variable, and information is conveyed by modulating its variance 2 1,2,3,4. Assume all four variances are equally likely. (a) Find the ML decision regions. (b) Find a numerical value of the probability of error of the ML detector. 4.3 Instead of an additive-Gaussian-noise channel, consider a multiplicative-Gaussian-noise channel, where the received sample r is related to the transmitted symbol a A by: r an N Assume N 0, 0 is independent of a. Consider a binary alphabet A 1,2 , and assume 2 that Pra 1 2 Pra 2. Find a numerical value for r such that the ML decision is different from the MAP decision. Assume that N 0 1 . 4.4 Consider the scalar channel r a n , where the symbol alphabet is a 1 , and where the noise n is independent of a . If the noise is Gaussian, then the ML detector reduces to a simple quantizer: aˆ ML sgn r . The Guassian pdf is said to be symmetric because it satisfies f n f n . Give an example of a symmetric noise pdf, satisfying f n f n , for which the ML detector does not reduce to aˆ ML sgn r . A precise answer is not needed, a rough sketch of the pdf is sufficient.