Pulse or digital communications – Receivers – Interference or noise reduction
Reexamination Certificate
2000-02-14
2003-12-02
Vo, Don N. (Department: 2631)
Pulse or digital communications
Receivers
Interference or noise reduction
C375S233000, C375S341000, C375S346000, C708S323000, C714S795000
Reexamination Certificate
active
06658071
ABSTRACT:
BACKGROUND OF THE INVENTION
The present invention relates generally to cellular and wireless communication and, more particularly, to an equalizer hypothesizing modulation symbols received over multipath fading channels.
Signals in a wireless communication system are subject to a number of phenomena that degrade signal quality. Each signal is reflected from many different man-made and natural objects. The receiver thus receives a number of signals delayed by one or more signal periods, called “multipath”, as each reflected signal is received. If the period of delay is more than the time required to transmit one symbol, producing intersymbol interference (ISI), then a receiver decoder may not be able to decode the symbol. This can cause poor sound or data quality to the user. Many different algorithms are used at the receiver to attempt to compensate for such effects. These techniques are discussed, for example, by G. L. Stüber in
Principles of Mobile Communication
, Chapter 6, Kluwer Academic Publishers, 1996. One such algorithm is the maximum likelihood sequence estimation (MLSE) or “Viterbi” algorithm.
Digital cellular and personal communication systems based on IS-136 and GSM (global system for mobile communications) require equalization to handle ISI arising from time dispersion. With the advent of the high date rates and the high level modulation in EDGE (enhanced data rates for global evolution) systems, the extent of ISI has increased considerably. To cover the same delay spread as today's GSM, EDGE receivers are specified to resolve a 5-tap channel. Furthermore, these systems are convolutional coded. To enable soft-decision decoding, it is, therefore, highly desirable for the equalizers to produce soft information at the bit level.
For M-ary modulation, a traditional MLSE equalizer requires M
L
at each trellis stage in order to properly model an intersymbol interference (ISI) channel driven by L+1 M-ary phase-shift keying (MPSK) symbols. Because of the high computational and storage requirements this implies, the delayed decision feedback sequence estimation (DDFSE) equalizer uses a kind of hybrid of MLSE and DFE (decision feedback estimation) in which only the K+1 most recent symbols are hypothesized as part of the state model, while the remaining L−K symbols required to compute each path metrics are determines from the surviving paths of each state. Therefore, a trellis of only M
K
states is needed in this case and the L−K “DFE” symbols constitute side information of the states. DDFSE equalizers are discussed, for example, by A. Duel-Hallen and C. Heegard in “Delayed Decision Feedback Sequence Estimation,”
IEEE Transactions on Communications
, vol. 37, no. 5, pp. 428-436, May 1989. One convenient way of describing the DDFSE operation is to say that the energy in the most recent K+1 “MLSE” symbols is used to equalize the data, while the energy in the remaining “DFE” symbols is simply canceled. The performance of DDFSE equalizers can be improved by pre-filtering the received signals with minimum-phase filters to maximize the energy of the leading ISI taps. Such filtering techniques are discussed, for example, in application Ser. No. 09/378,314 entitled “Method and Apparatus for Computing Prefilter Coefficients” of K. C. Zangi et al. and filed Aug 20, 1999.
In a convolutional coded system, a Log-MAP (maximum a posteriori) equalizer is preferred because of its ability to generate high-quality soft information for the Viterbi decoder. Traditional MAP algorithms, as described in L. R. Bahl, J. Cocke, F. Jelinek and J. Raviv (BCJR) in “Optimal Decoding of Linear Codes for Minimizing Symbol Error Rate,”
IEEE Transactions on Information Theory
, vol. 20, pp.284-287, March 1974, are not very suited to practical implementation because these algorithms require many multiplication and logarithm operations. The Log-MAP algorithms circumvent this complexity problem by operating the signals in the logarithmic domain, as discussed, for example, by P. Robertson, E. Villebrun and P. Hoehner, in “A Comparison of Optimal and Sub-Optimal MAP decoding algorithms operating in the log domain,”
Proceedings of IEEE International Communications Conference
'95, pp. 1009-1013, June 1995. However, both approaches are designed to generate optimal soft information for trellis with static state spaces, i.e., the definitions of the states do not change over time. Moreover, these definitions are known a priori and are usually given exogenously. For example, the trellis of a convolutional code falls into this category: the states are defined by the contents of the encoding shift-registers and the complete trellis can be constructed even before the codewords are transmitted. For multipath equalization, the traditional MAP algorithms thus require the construction of the complete state space as in the case for MLSE equalizers. For many practical applications, this proves to be excessively complex. For example, a MAP equalizer for 8PSK with 5 channel taps (EDGE specifications) requires 4096 states.
Since the Log-MAP equalizers suffer from the same complexity problems as the MLSE equalizers, it would be of great practical value if the principle of DDFSE could be combined with the Log-MAP algorithms to reduce complexity. An approach to this problem is discussed in the following.
Several problems arise when this MLSE/DFE hybrid method is applied to a forward-backward Log-MAP equalizer. More specifically, the forward DDFSE recursion would use an anti-causal prefilter to create a minimum-phase channel response. A DDFSE run backwards through the trellis, on the other hand, would need a causal prefilter to crease a maximum-phase channel, because in this case the MLSE symbols appear at the other end of the true state, as illustrated in FIG.
1
. This results in two major issues:
1. The forward-backward Log-MAP algorithm breaks down. The two passes are now using two sets of data which are correlated with each other (because of the two prefilters with overlapping support). Because of this correlation, the Markovian property of the data sequence, which means that the future data symbols are independent of the past data symbols given the current state, no longer holds. This Markovian property plays an essential role in the derivation of the forward-backward algorithm, without which the Log-MAP algorithm is invalid.
2. The state space from the two recursions do not match. A complete state space is used in a Log-MAP equalizer, which guarantees the forward and backward recursions work on the same state space. On the other hand, for a naive delayed decision feedback Log-MAP equalizer, only the “MLSE” symbols can be matched up. The “DFE” symbols setup by the forward and backward recursions generally do not agree with each other. The overlapping supports of the two prefilters makes the situation even more confusing.
The definitions of the states in a DDFSE trellis are constructed endogenously by the equalization recursion itself and, therefore, become unknown a priori. The traditional MAP algorithms are thus not directly applicable in these situations. In order to extract soft information, ad hoc post-processing methods such as the soft-output Viterbi algorithms (SOVA) are usually used instead. The SOVA algorithm is discussed, for example, by J. Jagenauer and P. Hoehner in “A Viterbi Algorithm with Soft-Decision Outputs and its Applications,”
Proceedings of IEEE Global Telecommunications Conference
'89, pp. 1680-1686, November 1989. These algorithms, however, are sub-optimal and their shortcomings become more apparent in iterative processing settings.
Accordingly, there is a need for an equalizer and method that generates optimal soft information at the bit level for MPSK and MDPSK modulations over multipath fading channels.
SUMMARY OF THE INVENTION
The present invention meets this need by providing a delayed decision feedback Log-MAP equalizer that generates optimal soft information at the bit level for MPSK and MDPSK modulations over multipath fading channels.
Broadly, ther
Coats & Bennett P.L.L.C.
Ericsson Inc.
Ghulamali Qutbuddin
Vo Don N.
LandOfFree
Delayed decision feedback log-map equalizer does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Delayed decision feedback log-map equalizer, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Delayed decision feedback log-map equalizer will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3094708