Frequency-Domain Signal Processing for Spectrally-Enhanced CP-OFDM Waveforms in 5G New Radio

Orthogonal frequency-division multiplexing (OFDM) has been selected as the basis for fifth-generation new radio (5G NR) waveform developments. However, effective signal processing tools are needed enhancing OFDM spectrum in various advanced transmission scenarios. In earlier work, we have shown that fast-convolution (FC) is a very flexible and efficient tool filtered-OFDM generation receiver-side subband filtering, e.g., mixed-numerology scenarios of 5G NR. FC filtering approximates linear convolution through fast Fourier transform (FFT)-based circular convolutions using partly overlapping blocks. with continuous overlap-and-save overlap-and-add models fixed block-size overlap, FC-processing blocks cannot be aligned all symbols frame. Furthermore, NR numerology does not allow to use lengths shorter than 128 because this would lead non-integer cyclic prefix (CP) lengths. article, present schemes which solve or avoid mentioned limitations. These based on dynamically adjusting overlap periods extrapolating CP samples, make it possible align each symbol, even case variable This reduces complexity latency, mini-slot transmissions and, an example, allows 16-point transforms 12-subcarrier-wide allocation, greatly reducing implementation complexity. On receiver side, proposed scheme makes effectively combine cascaded inverse forward FFT units FC-filtered processing. Transform decomposition used simplify these computations, leading significantly reduced A extensive set numerical results also provided, terms radio-link performance associated