OCDMA systems can support multiple classes of service by differentiating code
parameters, power level and diversity order. In this paper, we analyze BER
performance of a multi-class 1D/2D OCDMA system and propose a new approximation
method that can be used to generate accurate estimation of system BER using a
simple mathematical form. The proposed approximation provides insight into
proper system level analysis, system level design and sensitivity of system
performance to the factors such as code parameters, power level and diversity
order. Considering code design, code cardinality and system performance
constraints, two design problems are defined and their optimal solutions are
provided. We then propose an adaptive OCDMA-PON that adaptively shares unused
resources of inactive users among active ones to improve upstream system
performance. Using the approximated BER expression and defined design problems,
two adaptive code allocation algorithms for the adaptive OCDMA-PON are
presented and their performances are evaluated by simulation. Simulation
results show that the adaptive code allocation algorithms can increase average
transmission rate or decrease average optical power consumption of ONUs for
dynamic traffic patterns. According to the simulation results, for an adaptive
OCDMA-PON with BER value of 1e-7 and user activity probability of 0.5,
transmission rate (optical power consumption) can be increased (decreased) by a
factor of 2.25 (0.27) compared to fixed code assignment.Comment: 11 pages, 11 figure