You are on page 1of 2

http://www.spsc.tugraz.

at/courses/adaptive

Adaptive Systems - Problem Class 6 Channel Equalization


Signal Processing and Speech Communication Laboratory Graz University of Technology Inffeldgasse 16c/II http://www.spsc.tugraz.at May 23, 2007

Problem 6.1 - Channel Equalization: Least-Squares Solution


F (z ) d[n] e[n] C (z )

x[n] u[n] H (z )

+
DD

y [n]

u [n ]

Figure 1: Communication channel (H (z )), equalizer (C (z )), and decision device (DD). The symbols to be transmitted are u[n] {1, 1}. The decision device returns the sign of its input signal (zero is regarded as a positive number). The channel is known as H (z ) = 1/2 + 1z 1 + 2/3z 2 . The equalizer is an FIR lter of order 2, i.e., it consists of 3 coefcients. (a) Assume that we do not use any equalization of the channel H (z ), i.e., C (z ) = 1. What is the worst possible inter symbol interference (ISI), i.e., the worst possible input signal combination? Can the decision device always decide correctly or not? (b) The overall transfer function of the channel H (z ) and the equalizer C (z ) is given by (z ) = H (z )C (z ). F Find the matrix product formulation of the overall polynomial coefcient vector f. (c) Derive the optimal equalizer coefcients c by minimizing the squared norm, i.e., minc J (c) = ||f f ||2 , where f is the coefcient vector of the transfer function F (z ).

http://www.spsc.tugraz.at/courses/adaptive

(d) Calculate the optimal equalizer coefcients c in the least-squares sense for the given channel H (z ) and for F (z ) = z 2 . What is the noise gain of the equalizer for a white-noise process with 2 variance w = 0.08?

Problem 6.2 - Channel Equalization: Minimum Mean Squared Error Solution


F (z ) d[n] e[n] C (z )

u[n]

H (z )

x[n] + w[n]

+
DD

y [n]

u [n ]

Figure 2: Communication channel (H (z )) with noise w[n], equalizer (C (z )), and decision device (DD).

The symbols to be transmitted are u[n] {1, 1}. The decision device returns the sign of its input signal (zero is regarded as a positive number). The channel is known as H (z ) = 1/2 + 1z 1 + 2/3z 2 . The equalizer is an FIR lter of order 2, i.e., it consists of 3 coefcients. The additive noise signal w[n] is independent of the input signal u[n]. (a) Find the optimal equalizer coefcients c by minimizing the mean squared error (MSE), i.e., minc J (c) = E {e[n]2 }. (b) Assume that the input process and the noise process are white-noise processes. What are the optimal equalizer coefcients c now? What is the relation to the least-squares solution? (c) Calculate the minimum MSE solution for the given channel H (z ) and for F (z ) = z 2 . The 2 input process u[n] is a white-noise process with variance u = 1 and the noise w[n] is a white-noise 2 process with variance w = 0.08. What is the noise gain of the equalizer?

You might also like