Arbitrary spectral shaping and digital synthesis of an M-ary data time series.

Digital filtering as a means of precise spectral control has long played an important role in advanced data communication systems, where it offers benefits ranging from crosstalk interference suppression to bandwidth-efficient Nyquist channels. In this work the authors describe a novel approach to the arbitrary spectral shaping and digital synthesis of transmitted M-ary data signals. The technique is based on suitably truncating the weighted impulse functions that make up the data stream, defining a binary address from a sliding but finite sequence of encoded multilevel symbols that span the truncation interval, and using that address to call from digital memory at a rate equal to or greater than the Nyquist sampling frequency, a spectrally unique output that corresponds to an a priori superposition of symbol-weighted, shaped pulses. Unlike other digital methods, notably binary transversal filters, this memory-based method facilitates the arbitrary shaping of high-level (M > 2) data signals, affords precise pulse definition, and eliminates or significantly reduces the need for discrete summing networks. The paper presents a detailed discussion of the conceptual basis and analytic model, augmented with functional circuit descriptions, the latter including simple realizations that would otherwise require prodigious and sometimes impractical memory storage. Also included are two experimental examples that illustrate design approaches and performance attainments for bandlimited, raised-cosine Nyquist channels.