JLT24 – phiPU

Digital signal processing (DSP) has become a cornerstone of optical signal transmission. Today, a plethora of tasks inherent to a wide range of optical communication systems are being transferred to the digital domain to perform signal shaping, modulation and recovery. At the same time, optical-layer subsystems are kept as lean as possible, even though they are recognized for their ample bandwidth and high degree of energy efficiency. Towards exploring these gains, we propose and experimentally demonstrate the concept of a multi-functional photonic processing node that is tasked to decode partial-response formats, to demodulate multi-level signals, to generate high-frequency radio signal carriers, to cancel crosstalk in full-duplex transmission schemes involving non-orthogonal modulation formats, and to perform neuromorphic signal processing. As we will demonstrate, all these tasks can be accomplished with a rather small set of photonic toolbox elements dedicated to an efficient electro-optic frequency translation and spectral signal manipulation. We prove that the majority of the aforementioned functions can be executed with superior performance when compared to DSP-based methods and with great potential for energy- and footprint-efficiency.