Method and arrangement for demodulating data symbols

Details Method and arrangement for demodulating data symbols Method and arrangement for demodulating data symbols

1998-09-18

2001-07-31

Pham, Chi (Department: 2631)

Pulse or digital communications

Equalizers

C375S340000, C375S227000

Reexamination Certificate

active

06269116

ABSTRACT:

This application claims priority under 35 U.S.C. §§119 and/or 365 to 9703389-8 filed in Sweden on Sep. 19, 1997; the entire content of which is hereby incorporated by reference.
FIELD OF INVENTION
The present invention relates to a method for demodulating data symbols having been transmitted through a communication channel, particularly a channel suffering from one or more impairments, such as frequency selective fading, inter symbol interference and distortion, which may be temporally constant or time-varying.
The invention also relates to an arrangement for carrying out the method.
DESCRIPTION OF THE PRIOR ART
When digital-data-modulated signals are to be transmitted over a rapidly fading communication channel, such as a radio channel in a mobile radio communications system, a commonly used technique to cope with the changing channel is to include a predetermined data symbol sequence in the transmitted signal at suitably frequent intervals. The known symbol sequence is used in a receiver by a channel equaliser to adapt a demodulator to the channel's characteristics. This procedure is known as training or pre-setting the equaliser.
The equaliser typically used models the channel as a linear FIR filter (FIR=Finite Impulse Response), that is a transversal filter or a tapped delay line having complex multiplication weights applied to the tap outputs.
In
FIG. 1
is shown an example of such an equaliser, more precisely a decision feedback equaliser (DFE), which regularly adapts two linear filters
100
;
130
to the changing communication channel. The first filter
100
is a prefilter, having a first transfer function f, which is defined by a first set of filter coefficients, and the second filter
130
is a feedback filter, with a second transfer function b, which is defined by a second set of filter coefficients. A so called slicer
120
produces hard data decisions {circumflex over (d)}
k
and a metric computation unit
140
calculates, for each hard data decision {circumflex over (d)}
k
, a corresponding soft value s
k
. A summation unit
110
subtracts the feedback filter's
130
output signal {circumflex over (d)}
k
*b from the prefilter's
100
output signal {overscore (d)}*h*f+w′ and delivers a difference signal {overscore (d)}*h*f+w′−{circumflex over (d)}k*b.
The communication channel is assumed to have an impulse response h. Received signal samples, representing sent data symbols {overscore (d)}, are here represented by a sampled vector {overscore (&rgr;)}. Transmission of the data symbols {overscore (d)}, via the communication channel, corresponds to convoluting the information vector {overscore (d)} with the channel's impulse response h. Moreover, during the transmission noise w is added. The received signal samples {overscore (&rgr;)}={overscore (d)}*h+w are sequentially filtered through the prefilter
100
, which is regularly adapted, so that the communication channel {overscore (d)}*h*f+w′ becomes minimum phase, i.e. has its impulse response energy concentrated as much as possible to the initial part (w′ here represents the noise component w filtered through the prefilter
100
, i.e. w′=w*f). The prefilter
100
is also optimised to remove anti-causal ISI (ISI=Inter Symbol Interference), while only moderately amplifying the noise contents w in the signal samples {overscore (d)}*h+w. The feedback filter
130
is regularly adapted to reduce remaining causal ISI between the received data symbols, i.e. to remove the taps after the main tap of h*f. The prefilter
100
is designed so that this main tap h*f also is real (i.e. includes no imaginary component). For each received burst of signal samples {overscore (d)}*h the DFE performs: calculation of prefilter
100
and feedback filter
130
coefficients; prefiltering f; feedback filtering b; generation of hard data decisions {circumflex over (d)}
k
and generation of soft values s
k
. An estimated burst quality is in most cases also weighed in into the soft values s
k
.
FIGS. 2
a
and
2
b
show per se known methods for respective forward- and backward-demodulation of received signal samples in a data burst. The data burst is assumed to comprise a leading tail T
1
of known data symbols, a first set of unknown data symbols &Dgr;
1
, a known training sequence TR, a second set of unknown data symbols &Dgr;
2
and a trailing tail T
2
of known data symbols. Either the data burst is demodulated in the forward direction, whereby primarily F
1
the first set of unknown data symbols &Dgr;
1
is demodulated by using the leading tail T
1
and secondly F
2
the second set of unknown data symbols &Dgr;
2
is demodulated by using the training sequence TR; or the data burst is demodulated in the backward direction, whereby primarily B
1
the second set of unknown data symbols &Dgr;
2
is demodulated by using the trailing tail T
2
and secondly B
2
the first set of unknown data symbols &Dgr;
1
is demodulated by using the training sequence TR.
If forward-demodulation is selected for a particular data burst, the unknown data symbols &Dgr;
1
received initially and the training symbols TR received thereafter are demodulated according to the procedure F
1
; F
2
, as described with reference to
FIG. 2
a
above. Nonetheless, before the unknown data symbols &Dgr;
2
received after the data symbols in the known training sequence TR are demodulated, the feedback filter
130
is reset and its contents is replaced with the corresponding symbols, which instead are read from a memory unit at the receiving party. The analogous is, of course, also true when backward-demodulation B
1
; B
2
is selected.
Returning to
FIG. 1
, after the prefilter
100
each signal sample; is subtracted with a feedback filtered version of a demodulated preceding subset of signal samples, This reduces as much as possible the influence from previously received samples, as well as from later received samples. After that, a hard data decision {circumflex over (d)}
k
is taken by the slicer
120
. The slicer
120
here simply applies a set of symbol decision boundaries to the real part of the current signal value at its input. The demodulated hard data symbol {circumflex over (d)}
k
is then given by the interval, within which the real part of the current signal value falls.
The soft values s
k
are computed in the metric computation unit
140
, from prefiltered signal samples {overscore (d)}*h*f+w′, which are subtracted with feedback-filtered demodulated hard data symbols decisions {circumflex over (d)}
k
*b. Every soft value s
k
is a vector, whose elements are probability functions, that for each of the possible symbols in the symbol alphabet used, reflect the probability of that symbol being sent. The hard data decision {circumflex over (d)}
k
, made by the slicer
120
, naturally implies selection of the most probable symbol sent, which is indicated by the corresponding soft value s
k
. For binary symbols it is sufficient for the soft value vector s
k
to only contain one single element, whose sign indicates a corresponding hard data decision {circumflex over (d)}
k
and whose modulus reflect the certainty of the hard data decision {circumflex over (d)}
k
.
Generally, data symbols that are located at a small Euclidean distance from the demodulated signal are given a higher probability, than data symbols at larger Euclidean distances. Furthermore, an estimated burst quality is normally weighed into each soft value s
k
. An estimated high burst quality gives a higher a soft value s
k
certainty, than a lower estimated burst quality does.
Further detailed descriptions of equalisers in general, and the DFE in particular, can be found in J. G. Proakis “Digital Communications, 3rd Edition”, McGraw-Hill Inc. New York, 1995.
A disclosure of a more efficient demodulator, the so called Decision-Feedback Sequence Estimator (DFSE) is available in A. Duel-Hallen & C. Heegard “Delayed Decision-Feedback Sequence Estimator”, IEEE Transactions on Communications, vo

Affiliated with

Also associated with

Rating

Method and arrangement for demodulating data symbols does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Method and arrangement for demodulating data symbols, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and arrangement for demodulating data symbols will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-2529637