Pulse or digital communications – Equalizers – Automatic
Patent
1997-02-07
1999-06-01
Le, Amanda
Pulse or digital communications
Equalizers
Automatic
375234, 375348, 375349, 3647242, H03H 730
Patent
active
059094664
DESCRIPTION:
BRIEF SUMMARY
The present invention relates to an equalizer circuit (or "equalizer") for receivers in digital communications systems.
In present digital communications systems, receivers include a certain number of functions including demodulation, i.e. transposing the received signal to baseband, equalization, synchronization (bit rate and carrier), decision-making, and channel decoding.
Equalization consists essentially in removing as much intersymbol interference (ISI) as possible, which interference is associated with the fact that the overall transmission system does not satisfy the so-called "Nyquist criterion". This can result from poor strategy in transmission and reception filtering, from poor setting of sampling instants, which is merely a special case, or from a phenomenon of multi-path propagation.
It is recalled that a digital communications system can be considered as comprising a source that transmits, at a rate 1/T and via a channel having a given discrete impulse response, symbols having values taken from a generally complex alphabet of finite size. This transmission rate constitutes the modulation speed and is generally expressed in Baud, T representing the time interval defined by transmitting two successive symbols (symbol time).
Historically, the first circuits used for countering the phenomenon of intersymbol interference were described by Lucky in the following publication: 44, pp. 547-588, April 1965.
Essentially he described adaptive linear transversal filters that are "synchronous" (i.e. using only one sample per symbol time), with adaptation being made necessary by the varying nature of the transmission channel. The coefficients of the transversal filter were updated using an ISI zero-forcing criterion, with that procedure leading to the folded spectrum being equalized, hence the term "equalizer".
That kind of transversal equalizer is shown in FIG. 1 in which there can be seen a front filter 1 of transfer function B(z) together with a decision circuit 2 situated downstream from the front filter.
It was only later that adaptive equalizers appeared that use, as their optimization criterion, minimizing mean square error (MSE). It turned out that with noisy channels, ISI zero forcing could give rise to a significant increase in noise at the output from the equalizer, thus contributing to a large decrease in performance, whereas, in contrast, the criterion of minimizing MSE turned out to be a good compromise enabling a substantial reduction in ISI to be obtained without significant increase in noise.
Equalization was conventionally done in two modes. During the first mode, the circuit was directed by a learning sequence of sufficient length to guarantee convergence, and then in the second mode it became self-adaptive, i.e. it was directed by its own decisions.
Later still, proposals were made, in particular in the following publication: Proceedings of the IEEE 67 (8), August 1979, for non-linear recursive equalizers with decision feedback (decision feedback equalizers or DFEs) in which detected data was fed back into a feedback filter 3 constituting the recursive portion of the equalizer, as shown in FIG. 2.
Under the best circumstances, that approach made it possible to eliminate or at least considerably reduce ISI of causal origin, providing of course there was no decision error. That new structure was expected to give advantageous performance with particularly difficult channels. Unfortunately, those circuits could simultaneously be extremely sensitive to decision errors, such that under severe conditions an error propagation phenomenon could be observed that sometimes went so far as to cause the circuit to diverge. In such situations, it was almost inevitable that the circuit needed to be reinitialized, which meant that a new learning sequence was required and it became essential for the learning sequence to be sent periodically. For any given application, this required, at best, a non-negligible increase in line transmission rate.
From another point of view, although it is true that a decision
REFERENCES:
patent: 5285480 (1994-02-01), Chennakeshu et al.
patent: 5546430 (1996-08-01), Liao et al.
IEEE Transactions on Communications, vol. 35, No. 9, Sep. 1987, New York, pp. 877-887.
IEEE Transactions on Signal Processing, vol. 39, No. 2, Feb. 1991, New York, pp. 522-536.
International Conference on Acoustics, Speech, Signal Processing, Apr. 19-22, 1994, New York, pp. 481-484.
Global Telecommunication Conference, Nov. 29, 1993-Dec. 2, 1993, New York, pp. 87-91.
International Conference on Communications, May 23-26, 1993, New York pp. 1144-1148.
Labat Joel
Laot Christophe
Macchi Odile
France Telecom
Le Amanda
LandOfFree
Adaptive equalizer for digital communications systems does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Adaptive equalizer for digital communications systems, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Adaptive equalizer for digital communications systems will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-960178