The pSim electroic-photonic co-circuit simulation tool verifies multimode, bi-directional and multi-channel photonic circuits for photonic and electronic signal processing and analysis in the time or frequency domain. Its design environment is like that of traditional EDA electronic circuit simulation tools, facilitating the design of optoelectronic cooperative links using multiple devices; its hierarchical design environment and intuitive visual interface allow designers to easily interconnect devices and simulate and analyze the overall system to verify the system's response in both the time and frequency domains. pSim users can take full advantage of the GUI as well as the Python 3 programming language to build links, set up simulations, and perform post-processing. Designers can also import a rich library of third-party intelligent codes for machine learning-assisted reverse design, optimization design, statistical analysis, yield analysis, etc., and support convenient and flexible secondary development.
This case is based on micro-loop cascade structure with automatic parameter design by genetic algorithm, aiming to design dense WDM links with wave peaks located at 1542nm/1544nm/1546nm/1548nm. the model of micro-loop consists of directional coupler and waveguide, and the specific parameter settings of directional coupler and waveguide can be referred to the pSim User's Manual. The following will describe the design idea and algorithm implementation respectively.
Design Planning
The design planning is shown in Figure 2. First, we need to initialize the link model and create the basic link structure, here we choose the cascade micro-loop structure, the specific implementation code is as follows.
After the basic link creation, we enter the part of automatic design of the link parameters. We need to have clear design goals and give them in the form of numerical or mathematical models, so that the algorithm can clearly optimize the direction to make further optimizations. In addition, we need to set some of the initial values, mainly including the parameters related to the simulation environment and the parameters of the structural devices, which need to be in a reasonable range. After completing the above steps, we must choose the appropriate optimization algorithm to realize the automatic design of the device parameters.
Algorithm Implementation
The genetic algorithm in this case is chosen as the automatic design parameter of the optimization method. The genetic algorithm in this case is implemented by directly calling the library, and the following describes the fitness function solution and the installation and calling of the library respectively.
The fitness function is the value function of the genetic algorithm to determine whether the output condition is satisfied, and here we choose the objective function as the fitness function. The objective of this case is to design dense WDM links with wave peaks located at 1542nm/1544nm/1546nm/1548nm, respectively, and the specific implementation code is as follows.
To invoke the genetic algorithm, you first need to ensure that the sko library is installed, and after the installation is complete, you can invoke it to achieve automatic parameter design.
func is the fitness function, n_dim is the parameter dimension, size_pop is the population size, max_iter is the maximum number of iterations, lb is the parameter starting boundary, ub is the parameter maximum boundary, and precision is the design precision.
The final simulation results show that the parameters of the final design meet our target requirements and that we can arbitrarily select channels for the extraction of signals in the corresponding wavelength range by means of this algorithm with automatic parameter setting. The above results also show the feasibility of inverse design for integrated WDM links. In addition, the algorithm can be ported to any optoelectronic integrated link. In conclusion, the combination of the algorithm and pSim makes the integrated circuit design much easier and more efficient.