Capacity of iteratively estimated channels using LMMSE estimators

Abstract

This paper explores how iterative channel estimation, symbol detection and decoding at the receiver affects the achievable capacity compared with a non-iterative, purely pilotsbased scheme. First, a bound is put on the linear minimum mean-square error (LMMSE) estimator of doubly selective radio channels. Subsequently, a bound on the capacity using an LMMSE estimator is found. These bounds take into account the uncertainty in symbol detection on channel estimation, and incorporate the effect of channel estimation error on channel capacity. The paper describes how bounds can be calculated for typical multipath and fast fading radio channels. The interaction between the symbol detector and the decoder is depicted by exploiting an extrinsic information transfer (EXIT) chart, where a bound on the detector curve is found. With optimal LMMSE pilot-based channel estimation, the bounds demonstrate that iterative channel estimation has little advantage except at fading rates greater than 1% of the symbol rate. To make the results more concrete, finite order modulations are also investigated and compared with the case of Gaussian distributed symbols, showing that the latter can be used to approximate the performance of the former in a more convenient way.

Improved scheme for the problem of anti‐fading of DSRC systems in IEEE 802.11p environments

“…Since a matrix inversion is needed every time the transmitted signal in X changes, many scholars have proposed their improved methods, such as replacing the term ( XX H ) −1 in (8) with its expectation E ( XX H ) −1 to become LMMSE, a good case in point is the SVD algorithm proposed by Edfors and Sandell in [19]. Although the computational burden is reduced, their estimation performance is deteriorated owing to the non‐equivalent approximate transformation [25].…”

Adaptive Turbo Equalization for Nonlinearity Compensation in WDM Systems

“…between channel estimator/equalizer and FEC decoder operation, which may not hold for some schemes. To the best of our knowledge, for schemes such as the one proposed in this work, this is currently an open research problem on its own and the interested reader can find a useful discussion in [37].…”