Sparse and Dense: An Achievable Region for Code-Domain NOMA with Mixed Users

Spectrally efficient non-orthogonal multiple access (NOMA) schemes are of paramount importance for 5G and beyond wireless networks, among which code-domain NOMA is a prominent technology. Two extreme paradigms have lead to analytically tractable benchmarks for code-domain NOMA performance. The first relies on independent random dense spreading signatures (fully utilizing the available orthogonal resources) and suits settings with sporadic user activity, where signature coordination is hard to enforce. The second paradigm is regular sparse NOMA (where a small fixed number of resources is allocated to each user, and vice versa). It exhibits superior performance, while facilitating near-optimal multiuser detection using iterative message-passing algorithms. However, it also requires fully coordinated signatures and might be impractical in certain use-cases. In this paper, we investigate for the first time a mixed setting, where one class of users obeys the latter paradigm, while a second class obeys the former. The achievable ergodic class-throughput region is derived, while considering the large-system limit. Particularly, the total achievable sum-rate is characterized by means of a free additive convolution of probability measures, while relying on a recent strong representation theorem. The corresponding low-SNR characterization is further shown to admit closed-form expressions. The analysis provides insightful tools for investigating various use-cases of interest for future networks.