Tradeoff Between Calculation Precision and Information Rate in Eigendecomposition-Based Faster-Than-Nyquist Signaling

This study measured the achievable performance of the recently developed precoded faster-than-Nyquist (FTN) signaling scheme with truncated power allocation from an information-theoretic perspective. More specifically, the effects of the eigenvalue distribution of the FTN-specific inter-symbol interference matrix on the information rate were investigated. To explicitly evaluate the effects of significantly low eigenvalues, we performed 1024-bit eigendecomposition calculations, which are significantly more accurate than calculations in the standard double-precision environment. The results demonstrated a novel performance tradeoff between the information rate versus the calculation precision of a transmitted signal specific to the precoded FTN signaling scheme with truncated power allocation.