In this paper, we analyze the system performance of a stochastic network, which is in the stationary state. Different from previous full-loaded assumptions that base stations (BSs) are always transmitting signals, which overestimates the inter-cell interference (ICI), dynamic traffic is considered in our model and BSs can be silent in some time slots. Packet retransmission is adopted, which brings a coupled relationship between the outage probability and the buffer empty probability. That is, as the outage probability increases, packets that are not successfully received will remain at the head of the buffer and be retransmitted until they are successfully received, which increases the proportion of active BSs and aggravates the ICI in turn. To better analyze this coupled relationship, the ICI is analyzed with the aid of the probability generating function and the Laplace transform. Then a queuing model is established to depict the packet arrival and departure processes. Some important system metrics, including the outage probability, the buffer empty probability, and the stable throughput region, are derived. Numerical results verify our theoretical analysis, and it is revealed that packet retransmission can enhance the system performance when the packet request rate is within the stable throughput region.

With the common property of the sparse structure, multi-carrier low density signature (MC-LDS) and low density parity-check (LDPC) codes can be combined into a joint sparse factor graph (JSFG) while the joint detection and decoding (JDD) algorithm could bring significant gains in terms of the bit error rate (BER). In this paper, considering the coupling character of the soft information interaction in the elaborate JSFG, we carve the distribution of the merged decoding soft information from both sides and theoretically analyze the error performance of the uplink LDPC-coded MC-LDS system. Correspondingly, not only the average BER but also the practical decoding threshold could be effectively derived for any given number of iterations. Furthermore, based on the theoretical analysis, we optimize the degree distribution of the JSFG by leveraging the differential evolution method. Finally, simulation results show that the practical threshold of the JDD receiver with the optimized code approaches the signal to noise ratio of the theoretical capacity.