3 January 2024 in Optics Express
Optoelectronic integrated circuits (OEICs) have enhanced integration and communication capabilities in various applications. With the continued increase in complexity and scale, the need for an accurate and efficient simulation environment compatible with photonics and electronics becomes paramount. This paper introduces a method using the Verilog-A hardware language in the electronic design automation (EDA) platform to create equivalent circuit and compact models for photonic devices, considering their dispersion, polarization, multimode, and bidirectional transmission characteristics. These models can be co-simulated alongside electrical components in the electronic simulator, covering both the time and frequency domains simultaneously. Model parameters can be modified at any stage of the design process. Using the full link of an optoelectronic transceiver as an example, analyses from our Verilog-A model system show a mean absolute percentage error of 1.55% in the time-domain and 0.0318% in the frequency-domain when compared to the commercial co-simulation system (e.g., Virtuoso-INTERCONNECT). This underscores the accuracy and efficiency of our approach in OEICs design. By adopting this method, designers are enabled to conduct both electrical-specific and photonic-specific circuit analyses, as well as perform optoelectronic co-simulation within a unified platform seamlessly.
11 December 2025 in Journal of Lightwave Technology
In next-generation computing interconnects and high-speed communication scenarios, cross-domain co-simulation and co-optimization between photonic and electronic components are urgently required during the initial design stage of electronic-photonic convergence (EPC) integrated chips. Leveraging the maturity of complementary metal oxide semiconductor (CMOS) platforms and electronic design automation (EDA) tools, photonic device modeling using the hardware description language (HDL) enables the integration of photonic simulation into microelectronic platforms. Due to the limitations of HDL in numerical computation and the absence of embedded functions for flexible domain conversion, existing photonic device modeling efforts primarily focus on wavelength-independent devices, while modeling high-order dispersion effects remains challenging. Therefore, this paper proposes a standardized Verilog-A modeling methodology for photonic devices in both the time and frequency domains under the EDA platform. Fourier transform and convolution algorithms, explicitly formulated and implemented in Verilog-A, enable internal frequency-time domain conversion, capture dispersion and delay effects under cascaded-device conditions, and handle loop structures in time-domain links. Consequently, wavelength-dependent photonic devices, such as waveguides and linear fibers, can be accurately modeled using scattering parameters, numerical operations, windowing, and finite impulse response (FIR) filters within the module. The proposed Verilog-A models are validated through three representative EPC system case studies, with simulation results compared to commercial software demonstrating a maximum mean absolute percentage error (MAPE) of 1.4112%. With this methodology, electronic-photonic systems can be co-designed, and both time- and frequency-domain analyses can be effectively co-simulated within a single platform.
20 February 2025 in Applied Optics
In this paper, we present a complete theory for designing an all-optical time differentiator that enables the realization of arbitrary differentiation orders using the microring resonator (MRR). Based on this theory, a detailed quantitative analysis of the correspondence between the differentiator orders and the microring parameters is provided. All combinations of the amplitude transmission coefficient t and single-pass amplitude transmission ξ can be determined for desired orders. Furthermore, the theory is extended to cascaded rings for high-order differentiators, facilitating the design of arbitrary order differentiators guided by this theory. As an example, a 0.54th-order differentiator is demonstrated, with two parameter sets derived from the theory validated through a FDTD simulation, yielding errors of less than 10%. Additionally, the theory is applied to the design of higher-order differentiators, with the 1.44th order and the 2.1th order used as case studies. The errors are 4.05% and 1.54%, respectively. The proposed theory provides good guidance for designing an arbitrary order differentiator, enabling significant new applications.
5 March 2026 in Optics Express
Efficient photonic circuit simulation tools are essential for the design and optimization of photonic integrated circuits (PICs). However, the computational speed of current commercial simulators remains insufficient to support large-scale circuit optimization and statistical verification. In this paper, a multimode scattering matrix computation method based on a multimode subnetwork growth approach (referred to as MSG) for efficiently calculating the frequency-domain response of photonic circuits is proposed. The presented method starts with accurate photonic devices modeling leveraging the multimode scattering matrix. It then utilizes the established multimodal models to precisely characterize component parameters and topology information of the arbitrarily interconnected multiport photonic circuit. By employing an iterative strategy that connects only two subnetworks at each step, the global response is progressively computed, which significantly reduces the complexity of matrix inversion. The sequence of network connections is optimized using a suboptimal ordering algorithm, further minimizing the total computational cost. Benchmark results demonstrate that our approach enables fast multimode simulation, delivering over twice the speed of commercial simulators and strong engineering applicability.
16 September 2026 in Optics Express
Scaling photonic neural networks (PNNs) based on Mach-Zehnder interferometer (MZI) meshes to high-dimensional classification tasks remains challenging. As the transformation dimension increases, the number of interferometric elements grows quadratically, resulting in a rapidly increasing device footprint. Here, a polarization-controlled MZI core is introduced to process optical signals in a joint spatial-polarization basis. With polarization rotators integrated at the two input ports, the local transfer matrix is expanded from 2 × 2 to 4 × 4 without increasing the number of physical waveguide ports. The rotation angles are fixed by device geometry and material properties, whereas phase-shift variables are retained as programmable controls during network training. For four- and six-vowel classification, a six-port physical MZI mesh is used to implement an effective 12 × 12 transformation, yielding accuracies of 98.3% and 95.6%, respectively. For handwritten-digit classification, the MZI core count is reduced by 75.4% relative to the conventional implementation. Accuracies of 97.2%, 97.0%, and 96.1% are achieved at transverse magnetic mode losses of 0.2, 0.5, and 0.8 dB per MZI core, respectively. The proposed method can substantially reduce the number of photonic devices required for large-scale on-chip optical PNNs while maintaining accuracy.