A phase-uniform minimum-distance optimum does not establish an energy-saving wireless system. This preprint examines two-user square-QAM superposition through an exact conditional soft demapper, a coded packet comparison with optimized orthogonal references, and an explicit system-energy boundary. Conditional marginalization reduces equal-constellation demapping from O(M^2 log M) to O(M^(3/2) log M) without max-log approximation. In the recorded CPU implementation, joint 16-QAM demapping is 3.02 times faster, while orthogonal 256-QAM benefits more from the same structural optimization.
A predeclared 36,000-frame screen and 20,000 fresh confirmation frames use LDPC coding, CRC, distributed noisy pilots, and Wiener phase drift. Selected joint and orthogonal16 points at application Eb/N0 of 14 and 8 dB yield 5 and 13 information-pair failures per 10,000 frames. These are distinct operating points, not an exact matched-error SNR comparison. Joint access reduces nominal pair occupancy from 340 to 178 symbols, but no DC energy reduction is measured. For the proposed 64-node workload, a 20% saving from gateway activity alone requires an active-to-sleep power ratio of at least 7.65 even before other costs. This threshold is workload-specific.
Complete Paper 3 deposit: current manuscript PDF and LaTeX; the deployment feasibility package containing Python/C++ code, dependencies, configurations, simulation results, CPU benchmarks, figures, verification records, technical audit, provenance, and RF measurement/pilot protocols; the earlier exploratory research-start archive; and the subsequent engineering feasibility assessment with reproducible energy/cost calculations. PAPER_3_README.md explains reading order and chronology, and PAPER_3_SHA256SUMS.json supplies deposit checksums. Earlier archive findings do not supersede the current paper and must not be pooled with its experiments.
The subsequent assessment examines hardware compatibility, stronger incumbent comparisons, and conditional annual energy and adoption-cost bounds. Its numerical examples are explicitly hypothetical. No physical RF measurements, measured power traces, deployed energy savings, field coverage improvement, historical-priority claim, or demonstrated societal benefit are asserted. This is a software feasibility study and falsifiable physical-experiment specification, not hardware validation. Archived private/unpublished wording describes the files' preparation history. Third-party code retains its original license; no additional reuse license is asserted for the author's new work.