Spode is an opensource differentiable simulator specialized for programmable photonics. It almost has no learning cost.
It is currently under active development.
Spode provides the most comprehensive functions for research, design of programmable photonics.
- Spode is a frequency-domain simulator specialized for programmable photonics.
- It supports derivative calculation of any node response to any parameter.
- The built-in generator enables users easily produce triangular, square, and hexagonal mesh.
- The built-in visualization function produces high-quality figures satisfying academic purposes.
- A few functions are available for analyze the imperfections (e.g., random variation) in programmable photonics.
Spode is written in Python 3, with dependency on Numpy and Scipy. It should be installed successfully with pip:
pip install spode
from spode.util import generate
from spode.core import Circuit
import numpy as np
# generator instance for a 2 by 2 square mesh
# initialize all TBUs in the circuit
init_dict = {'theta': 0.0, 'phi': 0.0, 'l': 250e-6}
circuit_element = generate('square_1', [2, 2], init_dict)
# define the circuit instance and run the simulation
circuit = Circuit(
circuit_element=circuit_element,
mode_info={'neff':2.35}, # effective index
omega=np.linspace(192.5,193.5,1000) * 2 * np.pi, # [192.5Thz, 193.5Thz]
srce_node={'n_0#2_br': 1.0},
prob_node=['n_2#0_br'],
deri_node=['n_2#0_br'],
deri_vari=['tbum_2#1_2#0_v::theta'])
response, grads = circuit.solve(require_grads=True)
# Shapes by pseudo code:
# response.shape = (len(prob_node), len(omega), 2)
# grads.shape = (len(deri_node), len(deri_vari), len(omega), 2)Lesson 1: a tunable basic unit. We show how to use Spode to define a tunable basic unit (TBU), the building block of programmable photonics, and verify the simulation result by comparing with Lumerical Interconnect.
Lesson 2: a 2 by 2 square mesh. We show two ways to define a 2 by2 square mesh (i.e., manually and using built-in generator), and verify the simulation result by comparing with Lumerical Interconnect.
Lesson 3: Automatic circuit generators. We illustrate a few built-in circuit generators, which could be used in a one-line manner to generate triangular, square, hexagonal mesh. We also introduce a systematic way to name the TBUs, ports presented in the circuit.
Lesson 4: Built-in visualization methods. We first illustrate the built-in visualization functions for triangular, square, hexagonal mesh. Then we explain how to visualize a customized topology by taking advantage of our provided functions.
If you find any bugs, or want a new feature, please open an issue, or contact me at zhengqi@mit.edu.
Please cite the following papers if you find the package is helpful in your research.
@article{gao2023automatic,
title={Automatic synthesis of light-processing functions for programmable photonics: theory and realization},
author={Gao, Zhengqi and Chen, Xiangfeng and Zhang, Zhengxing and Chakraborty, Uttara and Bogaerts, Wim and Boning, Duane S},
journal={Photonics Research},
volume={11},
number={4},
pages={643--658},
year={2023},
publisher={Optica Publishing Group}
}
@article{gao2024provable,
title={Provable Routing Analysis of Programmable Photonic Circuits},
author={Gao, Zhengqi and Chen, Xiangfeng and Zhang, Zhengxing and Lai, Chih-Yu and Chakraborty, Uttara and Bogaerts, Wim and Boning, Duane S},
journal={Journal of Lightwave Technology},
year={2024},
publisher={IEEE}
}
@article{gao2024gradient,
title={Gradient-Based Power Efficient Functional Synthesis for Programmable Photonic Circuits},
author={Gao, Zhengqi and Chen, Xiangfeng and Zhang, Zhengxing and Chakraborty, Uttara and Bogaerts, Wim and Boning, Duane S},
journal={Journal of Lightwave Technology},
year={2024},
publisher={IEEE}
}