Model-Based Identification of POTS Local Loops for DSL Connectivity Prediction
MetadataShow full item record
Accurate prediction of the Digital Subscriber Line (DSL) access rate - over an existing plain old telephone service (POTS) twisted pair local loop - is vital to DSL providers. We have approached the challenge as an identification problem of the local loop structure and its twisted pair parameters, utilizing single-point measurements made at the provider end of the loop. Once the loop model is obtained, its connectivity measure is readily obtained through simulation.
The single-point measurement setup of a twisted-pair local loop is modeled as a linear time-invariant (LTI) system. Its frequency response is simulated based on the VT-TDL twisted-pair lumped circuit model. Study of the time-domain reflectometry (TDR) response, based on transmission line theory, reveals that the TDR response is a linear combination of reflections associated with loop nodes. An iterative modeling method is presented in which one loop node at a time is identified by analyzing embedded reflections in order of (temporal) appearance. This procedure is found not effective when several TDR reflections are heavily overlapping. To resolve the overlapping reflections, two variations of the Method Of Direction Estimation (MODE) algorithm are considered. The first is a hybrid of MODE and a weighted Fourier transform relaxation-based algorithm (MODE-WRELAX algorithm) and the second is the MODE-type algorithm. The MODE-type algorithm, with its signal model showing behavior analogous to that of the dispersive twisted pair behavior, is found more effective than MODE-WRELAX. To combine the iterative modeling procedure and the MODE-type algorithm, a second identification procedure is proposed which processes a limited number of frequency response data. The latter frequency domain procedure is found to be capable of correctly identifying 70 % of the Carrier Serving Area (CSA) and American National Standard Institute (ANSI) test loops.
- Masters Theses