Abstract
M.Ing. (Electrical & Electronic Engineering)
This thesis investigates the problems associated with providing end to end
in-service performance monitoring on leased line data networks. These
problems are of both practical and theoretical nature. A solution to this
problem is presented and is based on linear block coding using some
overhead bandwidth available within the channel. The coding scheme relies
on the characteristics of shortened cyclic codes. A method is given to obtain
the undetected error probability for an arbitrary code generation polynomial
with a specified degree of shortening. References to the literature explain
why this type of calculation is not straightforward. This method is used to
compare various candidate codes. A method is discussed which facilitates
rapid isolation of groups of favourable candidate codes for further study.
Software is developed to aid in the code selection process and optimum
codes are selected for leased line application. It is shown that in most cases
improvements in undetected error probability are achievable, when
comparison is made to previously published codes, proposed for error
detection. This is because code generators optimum at their published or
natural coding rate are seldom optimum at the rate which results from
shortening the code.