COMPASS

English

한국어

English

한국어

Appearance

Roadmap ​

This document outlines the planned development direction for COMPASS. Items are organized by priority and estimated timeline. Priorities may shift based on community feedback and research needs.

Near-term (v0.2.0) ​

Web UI ​

  • Browser-based simulation setup and result visualization
  • Real-time pixel stack editor with live cross-section preview
  • Interactive QE spectrum and field distribution plots
  • Job queue for long-running simulations
  • Built on a Python backend (FastAPI) with a modern frontend framework

Auto-optimization ​

  • Gradient-free optimization of pixel parameters (microlens height, BARL thicknesses, DTI width)
  • Objective functions: maximize broadband QE, minimize crosstalk, maximize color separation
  • Bayesian optimization (Optuna) for efficient parameter search
  • Constraint support: total stack height, fabrication design rules, material availability
  • Multi-objective Pareto front for QE vs crosstalk trade-off

Additional FDTD solvers ​

  • fdtdz: JAX-based systolic GPU FDTD for extreme performance on TPU/GPU clusters
  • Meep: MIT Electromagnetic Equation Propagation integration for the most mature open-source FDTD

Enhanced cone illumination ​

  • Ray file import from Zemax OpticStudio (ZRD format)
  • Arbitrary pupil shapes (circular, rectangular, annular)
  • Vignetting and pupil aberration modeling
  • Per-field-point F-number variation

Mid-term (v0.3.0) ​

TCAD integration ​

  • Interface with Sentaurus or similar TCAD tools for electrical simulation
  • Export COMPASS optical generation profiles as input to drift-diffusion solvers
  • Combined optical + electrical QE prediction
  • Carrier diffusion model for approximate electrical crosstalk without full TCAD

Advanced material models ​

  • Drude-Lorentz dispersion for metals (broadband FDTD compatibility)
  • Temperature-dependent optical constants
  • Quantum dot and perovskite color filter materials
  • Organic photodetector materials for flexible sensors
  • Anisotropic materials (birefringent layers)

Large unit cell support ​

  • Efficient handling of 4x4, 6x6, 8x8, and 10x10 unit cells
  • Memory-optimized RCWA for large matrices (Fourier order > 20)
  • Domain decomposition for FDTD on multi-GPU systems
  • Quad Bayer and nonacell (3x3) pattern support

Batch processing and HPC ​

  • Slurm/PBS job submission for cluster environments
  • Distributed wavelength sweep across multiple nodes
  • Checkpoint and restart for long sweeps
  • Progress monitoring dashboard

Long-term (v1.0.0) ​

3D CAD import ​

  • GDSII layout import for real foundry pixel designs
  • STL/OBJ mesh import for complex 3D structures
  • Automatic voxelization for FDTD grid generation
  • Layer-by-layer process flow definition

ISP co-simulation ​

  • Image Signal Processing pipeline integration
  • Simulated raw Bayer image generation from COMPASS QE data
  • Noise model (photon shot noise, read noise, dark current)
  • Demosaic and color correction evaluation
  • End-to-end image quality metrics (SNR, color accuracy, MTF)

Machine learning surrogates ​

  • Neural network surrogate models trained on COMPASS simulation data
  • Real-time QE prediction for interactive design exploration
  • Transfer learning across pixel geometries
  • Physics-informed neural networks for extrapolation beyond training data

Inverse design ​

  • Topology optimization of pixel structures using adjoint methods
  • Freeform microlens and metasurface design
  • Fabrication-constrained optimization (minimum feature size, layer count limits)
  • Multi-wavelength, multi-angle simultaneous optimization

Additional solver backends ​

  • Lumerical FDTD interface (commercial solver for validation)
  • COMSOL Multiphysics FEM interface
  • Custom FMM (Fourier Modal Method) implementation with analytic gradient support

Community contributions ​

We welcome contributions in all areas. High-impact contribution opportunities:

  • New solver adapters (implement SolverBase ABC)
  • Material data (measured n,k spectra for sensor materials)
  • Validation benchmarks (comparison against published data or commercial tools)
  • Documentation and tutorials
  • Performance optimization (GPU kernels, memory reduction)

See Contributing for details on how to get started.