Browse Prior Art Database

Reducing the Acquisition Time in an Automatic Equalizer

IP.com Disclosure Number: IPCOM000084269D
Original Publication Date: 1975-Oct-01
Included in the Prior Art Database: 2005-Mar-02
Document File: 5 page(s) / 22K

Publishing Venue

IBM

Related People

Nussbaumer, H: AUTHOR

Abstract

Described is a process which permits a drastic reduction in the acquisition time of automatic equalizers. This process is based upon the use of training sequences of a length equal to that of the equalizer. The number of tap iterations per fixed internal of time is increased drastically without an increase of peak computing power requirements, by taking full advantage of the repetitive nature of the training sequence and dividing the initialization period into three phases: precomputation, tap iterations, postcomputation.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 39% of the total text.

Page 1 of 5

Reducing the Acquisition Time in an Automatic Equalizer

Described is a process which permits a drastic reduction in the acquisition time of automatic equalizers. This process is based upon the use of training sequences of a length equal to that of the equalizer. The number of tap iterations per fixed internal of time is increased drastically without an increase of peak computing power requirements, by taking full advantage of the repetitive nature of the training sequence and dividing the initialization period into three phases: precomputation, tap iterations, postcomputation.

A special initialization sequence of a length equal to that of the equalizer is used and the precomputing of correlation coefficients is performed in a first step: the equalizer will then be organized in such a way that the computing power required for tap adjustments during initialization is drastically reduced. This permits a considerable speedup of the initial tap adjustment by recirculating the initialization sequence in the receiver and, therefore, contributes in solving the important problem of reducing the turnaround time of high-speed modems.

In a conventional baseband equalizer, the output sample y(i) is obtained by convolving the demodulated output samples x(i) ... x(i-k) with the equalizer coefficients a(k) so that:

(Image Omitted)

An equalized signal error dy(i) is derived by comparing the equalized signal with a locally generated reference u(i).

The tap coefficients a(k) are adjusted by increments da(k) that are extracted by correlating the equalized signal error with the input samples (2) da(k) = x(i-k) dy(i).

The signal is extracted from noise by resorting to the fact that the received signal samples are statistically independent, so that the long-term averages x(i-j)
. x(i-k) are zero except for j = k.

Such an approach for separating the signal from noise requires a large number of operations. It is, therefore, essential to reduce the computational workload, in order to speed up the modem acquisition phase.

In an n-tap equalizer, during initialization, the transmitter sends a respective n sample pseudo-random patter . In a typical example n is chosen equal to 8.

Under these conditions, the successive equalized signal samples y(1) ... Y(8) are related to the unequalized signal samples x(1) ... x(8) by:
(3) Y(1) = X(1) . A(1) with:

(Image Omitted)

1

Page 2 of 5

The taps are adjusted by tap increments that are extracted by correlating equalized signal errors with unequalized signal samples.

Assuming the averaging is performed over 8 signal samples, the tap increments are given by: (4) dA(1) = x/T/(1) . Y(1) - X/T/(1) U(1) X/T/(1) being the transpose of X(1) and U(1) being the locally generated reference;

(Image Omitted)

In a conventional equalizer, computing the tap increments da(i) requires computing the y(1) (equations 3) and computing da(1) (equations 4) In the case of the 8-tap equalizer, this corresponds to 128 multiplications and 112 addit...