Asymmetric digital subscriber lines (ADSL) transmit high bit rate data in the forward direction to the subscriber, and lower bit rate data in the reverse direction to the central office, both on a single copper telephone loop. This paper provides a comprehensive description of the performance of ADSL's that use currently available technologies with optimized parameters. Passband ADSL's encounter no self-near-end crosstalk (NEXT) since the forward and reverse channels are separated in the frequency domain. The absence of NEXT allows reliable transmission at high bit rates on long loops. The passband ADSL is disturbed by background noise, self-far-end crosstalk (FEXT), and NEXT from other digital transmission systems that share its spectrum, such as Basic Access DSL, HDSL, and T1 lines. This paper determines the performance of DS1 rate passband ADSL's in the presence of each of these impairments. Although baseband ADSL's are disturbed by NEXT from the reverse channel, they are suitable for data rates at 3 Mb/s or above since they experience less high frequency loss than passband ADSL's. The range of reliable baseband ADSL transmission is determined for forward data rates between 1.5 and 9 Mb/s, and reverse data rates up to 1.5 Mb/s. An analysis and comparison of different codes that could be applied to an ADSL is also presented. The performance of forward error correction codes and trellis codes is found for ADSL's in the presence of crosstalk and impulse noise.>

Using an ideal decision feedback equalizer (DFE), the SNR of quadrature amplitude modulation (QAM) and baseband pulse amplitude modulation (PAM) in the presence of self near-end crosstalk is computed for a large sample of loops within a carrier serving area (CSA). When baud-space feedforward filters are used, PAM has 1-2 dB more SNR than QAM, where the type of PAM is the 2B1Q line code. However, when using fractionally spaced feedforward equalizers (FSEs), the SNRs of 2B1Q and QAM are almost equal for loops at the extreme range of a CSA. Four- and 16-state trellis-coded modulation is applied to PAM and QAM. Coded and uncoded PAM and QAM are simulated with parallel decision feedback estimation. Viterbi receivers and coding gains are computed. QAM has up to 1 dB higher coding gains that PAM. However, the higher coding gains of QAM do not compensate for the lower SNR of uncoded QAM, and coded QAM has worse performance than coded PAM in the presence of self near-end crosstalk. The error rates of PAM and QAM with impulse noise are computed using a collection of measured impulse noise events. Results indicate that QAM has a lower error rate than PAM in the presence of impulse noise.>

Self near-end crosstalk (NEXT) is assumed to be the dominant source of impairment, and the subscriber loop configuration is governed by carrier serving area (CSA) design rules. The structured channel signaling (SCS) technique decomposes the physical channel into multiple parallel independent signaling subchannels by exploiting the combined eigenstructure of the channel and the correlation of the (NEXT) interference. Computer performance evaluation studies reveal two distinct patterns. For a given loop configuration, as the block length increases, the coding gain usually increases, and for a fixed block code length, the coding gain degrades as the loop length (including bridged taps) increases. For loops at the extreme range of a CSA, block codes of at least 20 symbols are required to achieve a performance commensurate with that of a decision feedback equalizer (DFE) composed of an optimal nine-tap minimum mean-square error (MMSE) feedforward filter and an ideal feedback canceler.>