OFDM is an attractive modulation scheme for communicating over Wide-Band frequency selective channels. In the context of the smart grid, OFDM is to be employed over various links, both wireless and wireline: LAN / WAN / FAN / HAN. While OFDM-based communications has been developed and tailored for each of the above links, it would have to go through some changes and adjustments in order to allow optimal operation of a unified network. OFDMA employs OFDM modulation to support multiple access. Multiple users (MU) are served by assigning users with different set of frequencies and time resources. The responsibility for properly providing this allocation process is imposed on a central element in the network (e.g. base-station, or an access point). Such centralized allocation, however, is not always a viable scenario. In this work we study a new approach to OFDMA drawing its main idea from IR-UWB systems (ultra-wideband impulse radio). In IR-UWB systems, each user transmits its information using a narrow pulse located at random points along the frame (frame). While putting no burden on the multiple-access mechanism, the problem is how to effectively handle anticipated conflicts, collisions, between users. In classical IR-UWB systems this problem is tackled by employing repeated transmissions of the same data, which is a spectrally inefficient approach. In the current study we employ the time and frequency domains in a random manner, random access OFDMA, while providing a more effective treatment for collisions. A large number of potential users are expected to simultaneously occupy the network with relatively short messages. Correspondingly, we employ short, high-rate error correction codes to handle both the channel impairments and multiuser interference resulting from direct collisions.. A major activity in this research is the development of an effective MU detector based on a graphical representation of the complete systems, i.e. using a variant of a bipartite graph, which describes both, the OFDMA network connections and the coding. Using such a graph should facilitate efficient iterative detection/decoding algorithms.