The performance of adaptive equalization for digital communication systems corrupted by interference

Abstract

PAGE Form Approvedpu,,l1c rep,,"in burde 1Wr itt coliection of intormaetion is estimated to averag18 I hour, per response indluding the time for re,,iewing insructionis, searching existing data souces. 94-08836San Diego, CA 92152 -5001 SPONSOR~INGJMONTORING AGENCY NAMIE(S) AND ADORESS(ES)800 North Quincy StreetC 12a. DISTRIBUTION/AVAILABIU1TY STATEMENT 12.DISTRIBUTION CODEApproved for public release; distribution is unlimited. ABSTRACT (UALXhnur 200 words)This paper analyzes the effects of interference on the steady-state performance of several popular adaptive equalization algorithms. It is shown that adaptive equalizers based on the linear equalizer structure have a built-in capability to reject narrowband interference, however performance deteriorates as the bandwidth of the interference increases. The exidstence of a time-varying misadjutment component in the adaptive equalizer weight vector is shown to affect the interference cancellation properties. In addition, simultaneous multipath propagation and interference is investigated. These effects on the perfor-

Bi-scale LMS equalization for improved performance

“…and produces its output (3) (2) (1) better predictor of MSE performance for the entire range of mean-square stable step-sizes for LMS [7].…”

“…Although equalizers are usually employed to mitigate the channel effect on x n , we assume in this paper that the channel is ideal as we are interested in the equalizer behavior when mitigating narrowband interference. All signals are at complex baseband and sampled at the symbol rate of X n • The transversal adaptive equalizer forms an M-tap input vector In various applications involving narrowband processes, such as noise canceling, prediction, and equalization in interference-dominated environments, transversal least mean square (LMS) adaptive filters will often perform better than expected from the optimal filters of the same structure [1][2][3][4]. These non-linear or non-Wiener effects are due to the capability of adaptive filters to produce dynamic, timevarying behavior in their weights [3][4][5].…”

Nonlinear effects in LMS adaptive equalizers

“…However, it has recently been reported that an LMSimplemented adaptive equalizer operating with a temporally correlated interferer can produce better probability-of-error performance than the corresponding Wiener filter [12]. Subsequent simulations have revealed the unexpected result that with the proper choice of the step-size parameter, the nonlinear nature of the LMS algorithm can be exploited to generate MSE that is less than the Wiener MSE.…”

A Performance Bound for a Cascade LMS Predictor

“…The textbook [1] analyzes the LMS algorithm performance based on "independence assumptions" approach and constat that an LMS adaptive filter operates random fluctuations of its weighting coefficients around the optimal "Wiener solution" in a stationary environment at its steady-state value. This approach has been shown to be a successful model of analyzing the LMS algorithm in many situation, e,g., such assumptions can be justified for a true vector signal like that emerging from a sensor array, but not for the important cases of a tapped-delay line TDL, noise cancelling to a sinusoidal interference with a strong deterministic coherence between successive input vectors [2,3]. Some papers have reported cases where the performance of the LMS filter exceeds that of the finite Wiener filter [2,3].…”

“…This approach has been shown to be a successful model of analyzing the LMS algorithm in many situation, e,g., such assumptions can be justified for a true vector signal like that emerging from a sensor array, but not for the important cases of a tapped-delay line TDL, noise cancelling to a sinusoidal interference with a strong deterministic coherence between successive input vectors [2,3]. Some papers have reported cases where the performance of the LMS filter exceeds that of the finite Wiener filter [2,3]. Later, Quirk et al derived a performance bound for the LMS estimator without using "independence assumptions" and simply relying on the wide-sense stationarity of the signals, and showed some examples in which the LMS estimator outperforms the finite Wiener filter [4].…”