John A. C. Bingham

A practical discrete multitone transceiver loading algorithm for data transmission over spectrally shaped channels

In this paper, we present a finite-granularity, loading algorithm for a discrete multitone (DMT) modulation system. The proposed algorithm offers significant implementational advantages over the well-known water-pouring method and the earlier Hughes-Hartogs algorithm, while typically suffering only negligible performance degradation relative to the optimal solution. We also present simulation results of this loading algorithm applied to the newly proposed asymmetric digital subscriber lines (ADSL) service. >

Equalizer training algorithms for multicarrier modulation systems

In a multicarrier data transmission system the length of a symbol is often limited by the maximum permissible delay of data from input to output. If the length of the impulse response of the channel is not negligible compared to the permitted symbol length, the impulse response can be shortened by passing the received signal through a time-domain equalizer before modulation. The best performance for a given computational complexity can then be achieved by appending to each block of samples of the transmit signal a cyclic prefix that has the same length as the shortened impulse response. Several algorithms for designing the time-domain equalizer, which use adaptation in the frequency domain and windowing in the time domain in order to minimize the mean squared error of the equalized response, are described.<<ETX>>

Discrete multitone echo cancelation

Multicarrier transmission methods have long been known to optimize the performance of data transceivers on bandlimited communication channels. One form of multicarrier transmission, known as discrete multitone modulation (DMT), is particularly attractive for its ability to be implemented using efficient digital signal processing techniques. Given a basic DMT system, it is possible to increase the aggregate data rate with full-duplex transmission using echo cancelation. However, DMT echo cancelation at first appears difficult because of the computational complexity required in a straightforward implementation to cancel the cross-echoes produced by each carrier into every carrier. This paper presents high-speed echo cancelation techniques for full-duplex data transmission using DMT systems. The techniques estimate the echo with a method of fast convolution that combines a complex multiply per tone with a short convolution in the time domain. In addition, the frequency-domain update of the echo parameters consists of one complex multiply per tap. As a result, these techniques can achieve much lower complexity than that required by traditional single-carrier technique, such as the least mean square (LMS) algorithm. This approach has been implemented for asymmetric digital subscriber line (ADSL) applications.

A Multicarrier El‐HDSL Transceiver System with Coded Modulation

In this paper, we present a trellis-coded multicarrier transceiver system, known as Discrete Multitone modulation, for the application of transmitting digital data at the El rate of 2.048 Mbit/s over existing copper telephone lines. The service of El-HDSL (El-rate High-speed Digital Subscriber Lines) is currently being studied by ETSI for future deployment in Europe. Because of severe crosstalk and additive white Gaussian noise (AWGN) encountered in El-HDSL, the application of a powerful trellis code is highly desirable to ensure reliable system performance, with a required bit-error-rate (BER) of no more than 10−7, at the desired El data rate. Detailed descriptions of the DMT transmitter and receiver structures are given, and projected performance results, both with and without the addition of a trellis code, are presented. Because of the different transmission environments encountered in the various countries in Europe, we have performed our computer simulation on both British and German (1) copper twisted-pairs. We find that with the addition of Weis powerful 4-D 16-state trellis code, we can achieve the El rate with a positive margin over all loops tested using either one or two sets of twisted-pairs only, depending on the particular transmission scenario.

A data-driven multitone echo canceller

A data-driven echo canceller for full-duplex data transmission with multitone modulation is presented. This echo canceller is shown not to be impaired by the eigenvalue-spread problems that are inherent in signal-driven echo cancellers, and to have numerical performance and cancellation range that equals, and in most cases exceeds, that of data-driven echo cancellers used in data transmission with baseband or quadrature amplitude modulation. The method makes use of frequency-domain updating, but with time-domain implementation of the canceller. It introduces no delay into the received signal path.<<ETX>>

Synchronized DMT for multipoint-to-point communications on HFC networks

This paper presents a method for synchronizing a large number of discrete-multitone-based modems for multipoint-to-point communications on hybrid fiber-coax (HFC) networks. Each newly-installed remote unit is first loop-timed to the central modem master clock. Ranging is then performed to determine the sample delay required to align at the central unit receiver the symbols transmitted by the remote with those transmitted by remote units already operating. Upon completing the synchronization procedure, the remote unit transmits a signal to train the central unit receiver so that the available upstream bandwidth can be used as efficiently as possible during steady-state operation.

High-speed full-duplex echo cancellation for discrete multitone modulation

Methods for echo cancellation that are easily implemented in a discrete multitone modulation (DMT) system using off-the-shelf digital signal processors are presented. Echo emulation is performed with a combination of fast frequency and time-domain techniques, while adaptive updating is performed on a vector of complex one-tap filters. Simulation results in finite precision show that the required 70 dB of linear echo rejection can easily be achieved on digital subscriber lines (DSLs).<<ETX>>

Erasure tagging in a slow Rayleigh-fading environment

Our proposed digital audio broadcast system uses a Reed-Solomon code with hard decoding. If a system can identify codeword symbols that have a high-likelihood of errors and tag them as erasures, the probability of error can be decreased significantly. We introduce a method for tagging erasures in a slow Rayleigh-fading environment. This method combines 4 PSK with multitone modulation using differential detection. To decrease probability of bit error, an erasure-tagging scheme augments a Reed-Solomon code. In this scheme, the 4 PSK constellation, which would normally have 4 decision regions, is divided into 8 decision regions; 4 of these decision regions are considered bad. The scheme takes advantage of the slowly-varying nature of the channel by tagging an erasure when a majority of received tone symbols in a given block of multicarrier symbols fall into bad regions. This method deduces when there is a large SNR loss due to fading and can be used when it is difficult to measure the signal SNR directly.

Improved Methods of Accelerating the Convergence of Adaptive Equalizers for Partial-Response Signals

In order to accelerate the convergence of adaptive equalizers for partial-response signals it is necessary to transform the input signal, and use the transformed values in one of two ways-to calculate both the equalizer output and the tap updates, or to calculate only the tap updates. These methods are called double-sided and single-sided orthogonalizalion, respectively; this paper confirms previous suggestions that the latter is preferable. In an early implementation of the double-sided orthogonalization (DSO) the number of multiplications was proportional to the square of the number of taps (N2); this was later reduced to 2N multiplications; the only implementation of single-sided orthogonalization (SSO) that has been described used of the order of N2multiplications. This paper describes an implementation of SSO that needs only 2N multiplications,and gives explicit formulas for the two partial-response systems of most interest-I and IV. Two examples of the improved convergence are given, and the application of this method to complex equalizers is discussed.