摘要： Early attempts in the field of optical computing were futile because they aimed to construct an all-optical equivalent of the digital computer [1]. However, photonics does have potential for performing certain computing functions in the analog domain [1]. These analog photonic hardware accelerators precede an optical-to-electrical conversion and alleviate bandwidth, dynamic range and power consumption bottlenecks on the subsequent electronics [2]. Far from a general purpose computer, photonic hardware accelerators are custom designed to perform specific signal transformations in real time and in analog domain. The computational operations that have already been realized by these analog optical computing primitives include integration [3], data compression [4], range compression [7]. Optical time~bandwidth implementation of temporal differentiation has been demonstrated in two ways: field differentiators realized using a micro-ring resonator [8] and intensity differentiators implemented using phase modulation [9]. Inspired by the physics of Photonic Time Stretch technology [10], and its spin-off in the form of the Phase Stretch Transform [11, 12] we propose a novel computational approach for calculating mathematical derivatives. The proposed all-optical differentiator consists of group delay dispersion primitives and coherent detection as shown in Figure 1. The re-configurability to compute different orders of differentiation is realized simply by controlling the applied group delay dispersion spectrum. The simulation results presented in Figure 2 demonstrate that the computation of any-order of differentiation of the input signal is possible. As shown in Figure 2, the imaginary component of the output gives the m-th order of differentiation while the real component of the output recovers the input signal. The proposed approach has promising applications in signal classification and is also easy to implement enabling realization of a real-time, low power and broad band optical differentiation functions.