The Study On Blind Equalization For Communication System

In the modern communication system, are the important factors of theinfluence communication quantities because of the intersymbol interference andchannel interference that the distortion and aberrance caused by the limitedbandwidth, need the valid equalization technique of channel to remove. Blindequalization is a new adaptive technology without resorting to a trainingsequence, which only utilizes the prior information of transmitted signals toequalize the channel character, and makes sure that the output sequenceapproximates the transmitted signals as accurate as possible. Therefore, it canimprove channel efficiency and equalize the not ideal character of channelavailably and overcome the intersymbol interference and reduce the error rate andacquire better equalization performance because of not need to send out thetraining sequence. It becomes currently one of the key technologies in the field ofdigital communication because of the superior performance, and also a hotresearch issue in communicating, signal processing and detecting, which is veryimportant for instructing both the theory and the applications in communication,radar, sonar, control engineering, earthquake prospecting, biomedicineengineering and so on.This paper systematically analyse the basic theory and algorithm form ofblind equalization and the basic algorithm and convergence criterion of it. Thebasic theory and property is described and analyse three algorithm in commonused of Bussgang. CM blind equalization algorithm of Bussgang algorithms is theemphasis. This paper introduces and derives the CM algorithm, and analyses theantinomy between the convergence rate and accuracy because of using the fixedstep in traditional CM algorithm. In order to improve the performance of constantmodulus algorithm, we introduce the thought of variable step-size, and make useof the theory of self-adaptive variable step-size into CM algorithm. It is to find aparameter to control the step-size, which can reflect the work state of equalizer.When the equalizer works not stably, we can use this parameter to adjust thestep-size to make the equalizer to stabilization again. This algorithm is to increasethe step-size in the initial time to speed the convergence rate and reduce thestep-size near convergence to get little residual error to improve the convergenceaccuracy. At last, the summary of the dissertation of the tendency direction of theblind equalization for the future researches is prospected.The main achievement of this paper as following:The constant modulus algorithm (CMA) is the most widely used blindequalization algorithm of Bussgang in digital communication systems. In thispaper we introduces and analysis the inherent limitations of traditional CM blindequalization algorithm, which lead to a tradeoff between convergence rate andmisadjustment, and propose a new variable step-size CM blind equalizationalgorithm. The new algorithm makes use of the square of average grads as theparameter to produce the new variable step size.We systematically analyse the blind equalization algorithm, which use theinstantaneous error power as the parameter to adjust the step. This algorithm hasmore advisable advantage than use fixed step. But this algorithm still has biggerstep in the end of convergence because of the input noise that will make weightcoefficient having bigger motion around the best value and lead to biggermaladjustment. The residual error is very small when the algorithm turns intosteady state. In this time, when the system parameter or the input signals changed,this algorithm will not follow the change of the system nicely because of the stepstill is very small. In order to overcome the defects upwards, we improved on thealgorithm upward by using the step adjustment principle. We use the product ofresidual error and input of the equalizer to adjust variable step-size, in which thestep is adjusted by the square of the average of the grads of the weight modulusof the equalizer of the last time. Its principle is: When the algorithm doesn't yetconverge to steady field, the average value of the grads is larger and theconvergence rate is more quickly;When the algorithm enters the steady field, theaverage value of the grads reduce to minimum, and the step is also reduce tominimum. This produces the pimping maladjustment in the best valueneighborhood. This algorithm has better maladjustment performance because ofthe moment grads is to the input noise not sensitive. And because the step of theimproved algorithm is common adjusted by the input signals and error outputs ofthe blind equalizer, so when the input signals changed, it has the betterperformance on the trail.In conclusion, the new algorithm we proposed by research the constantmodulus blind equalization algorithm in this paper solved the antinomy betweenthe convergence rate and accuracy because of using the fixed step in traditionalCM algorithm and improved the performance of the CM algorithm. Thetheoretical analysis and the simulation results proved to show that the newalgorithm has improved performance of the convergence speed and residual errorthan traditional CM algorithm and can be applied in the self-adaptive blindequalization system availably.