Equalization techniques for high order qam signals

Abstract

Equalization techniques counterbalance the distortion and additive white

Gaussian noise (AWGN) introduced by communication channels which are not known

beforehand to reduce the inter-symbol interference (lSI),. Since the channel is unknown a

priori, adaptive algorithms are used for this purpose. Broadly categorized, there are two

methods by which adaptive algorithms approximate the inverse of the channel impulse

response, (i) by using a training sequence (ii) without using a training sequence.

Conventional adaptive algorithms require a properly synchronized (at both transmitter

and receiver) training sequence which makes it possible to adjust the equalizer

coefficients according to the employed algorithm, thereby minimizing the mean square

error. This type of equalization is called Non-Blind equalization. However, there are

many situations which require equalization without the use of a training sequence. One

example of such a situation is multipoint data networks. For this purpose, a Blind

equalizer has to be built into the receiver design. Blind equalizers estimate the

transmitted signal and the channel parameters without using a known training sequence.

Minimization of a class of non-convex cost functions is used as a criterion for adaptation.

By these cost functions, lSI is characterized independently of the data symbol

constellation and of the carrier phase used in the transmission system [1]. In this project,

three blind equalization algorithms (Constant Modulus Algorithm (CMA), Modified

Constant Modulus Algorithm (MCMA) and Variable Step-size Modified Constant

Modulus Algorithm (VSS-MCMA)) were studied mathematically and their performance

is shown for high order QAM (Quadrature Amplitude Modulation) signals by

simulations in this dissertation. These algorithms were then implemented with a decision

feedback equalizer to further improve the symbol error rate (SER) performance. The

following essential findings were made:

1) Blind algorithms are robust with respect to distortions and, by using appropriate

step-size and reference gain parameters we can ensure convergence to optimal

gains.2) Since CMA is phase-blind in nature, it cannot correct the phase errors. Whereas,

simulations show that MCMA corrects the phase error and results in a lower

MSE as well.

3) In VSS-MCMA, by using different step-size parameters we can ensure faster

convergence and lower MSE.

4) CMA with DFE provides better SER than CMA with linear equalizers. SER is

further improved by implementing a decision feedback equalizer (DFE) using

MCMA and VSS-MCMA.