NEXT-limited capacity

Written on 6:49 AM by ooe

First, the source of noise in the channel is assumed to be generated by NEXT from adjacent and like systems (hence the term self-NEXT, and throughout this chapter, NEXT limited capacity will imply self-NEXT-limited capacity). The term ‘like system’ is used to indicate that the noise sources have the same spectral occupancy or power spectral density (PSD) as the system being interfered with. This is most likely to cause worst-case interference because all the coupled signals will be in-band for the receiver of the interfered system. Also, the worst case will occur when 49 other disturbers in a 50 pair unit are active at one time (Large cables are constructed from groups of smaller units, typically of 50 pairs. The UK term is unit, but the American term binder group is more common in world literature). Within the unit or group the individual pairs are usually ordered and twisted in a particular fashion. When a wire-pair suffers interference from other wire-pairs in the same unit it is the nearest 6–8 interferers which cause most of the noise power or crosstalk. However, it cannot be predicted which ones these will be unless a priori knowledge of the physical location and spectrum of all the possible interferers in the cable is available. Hence, a worst-case assumption is usually made in which it is assumed that all 49 possible disturbers are active and that 1% worst-case coupling is achieved. This is rather pessimistic but yields a planning rule that is conservative and offers margin to cater for the effects of other (less tangible) noise sources.

Second, it must be assumed that the receiver is perfect in all other aspects, namely:

• no non-linear behaviour in the channel, transmitter or receiver;

• no internal sources of noise (in particular no ADC quantisation noise);

• equalisation cancels all inter-symbol interference (ISI) perfectly;

• echo cancellation removes own-system echoes perfectly.
  
  
  

  The 3-D plot shows the surface of maximum capacity (C in Mbit/s) with variation in cable length (l in km) and bandwidth in use from DC to (f_max in MHz). Clearly it can be seen that as length increases the capacity decreases. At long cable lengths (~2–5 km), increasing the frequency of transmission has little or no effect on capacity. Only over relatively short lengths is it advantageous to use higher frequencies. Intuitively, this is to be expected from the band-limited nature of metallic access channels. This is illustrated more clearly by the 2-D contour plot shown on the right of the figure. Contours of fixed capacity are plotted for the following capacities – 10, 8, 6, 4, 2 and 1 Mbit/s. The 10 Mbit/s contour is shown at the top left of the plot, whereas the 1 Mbit/s contour is on the right. The 1 Mbit/s contour shows quite clearly that there is no advantage to using frequencies higher than about 200 kHz on long loops (This assumes a NEXT-limited environment. ADSL avoids this limit by reserving the higher frequencies (up to 1.1 MHz) for one direction only). However, the potential for high-speed transmission over short distances, using higher frequencies, would seem to exist.