Auto-routing algorithm for field-programmable photonic gate arrays

Abstract

Programmable multipurpose photonic integrated circuits require software routines to make use of their flexible operation as desired. In this work, we propose and demonstrate the use of a modified tree-search algorithm to automatically determine the optimum optical path in a field-programmable photonic gate array (FPPGA), based on end-user specifications, circuit architecture and imperfections in the realized FPPGA arising, for example, from fabrication variations. In such a scenario, the proposed algorithm only requires the hardware topology and the location of the connections of the FPPGA defining the optical path to be programmed. The routine is able to optimize the path over multiple and competing objectives like the overall length, accumulated loss and power consumption. In addition, should any region of the circuit suffer from any potential damage that may affect the device performance, this algorithm is also able to provide basic self-healing and fault-tolerance capabilities by supplying alternative paths through the photonic arrangement.

Programmable photonic circuits

“…Recirculating architectures (which are generally overdimensioned) require new layers of programming. Two broad classes of such programming techniques are "black-box" optimization methods [13], [94], [95] (for example machine learning), and synthesis techniques. Optimization methods can operate directly on the hardware and do not necessarily require calibration.…”

Multipurpose self-configuration of programmable photonic circuits

“…The information involving the architecture itself, includes the interconnection scheme, and the physical features that define the TBUs, such as their basic unit length (BUL) and basic unit delay 25 . The information regarding the full control of every TBU includes the computation of the nonideal passive offsets of the phase of each TBU, the calibration curve (tuning response) of each phase shifter, estimated insertion loss, power consumption 26,27 , and the tuning crosstalk matrix that characterizes the undesired coupled effects of neighboring phase actuators. These requirements can be obtained by means of the periodic application of selfcharacterization routines based on iterative maximization and minimization methods 28 , and regression-based approximations (see implementation and detailed analysis in Supplementary Note 2).…”

“…(1), and the accumulated penalties are then considered by the optimizer. To perform the auto-routing task, a shortest path tree algorithm with restrictions can be employed 26,27 . More information can be found in Supplementary Note 3.…”

Programmable Photonics: An Opportunity for an Accessible Large-Volume PIC Ecosystem

“…routing, distribution networks, or wavelength filters. Graph-based routing algorithms, such as those developed for FPGA programming, can be of great help here [76], [77]. In a waveguide mesh, not only the efficient use of actively used couplers is important, but also the control of the unused couplers, as the effect of small parasitic light flows mush be minimized [78], [79].…”