Fast optical simulation from a reduced set of impulse responses using SystemC-AMS

In this paper we propose a methodology to simulate the optical filtering system of a camera module with limited access to proprietary data. The target of the simulation is the virtual prototyping of the overall camera module for a fine tuning of the auto-focus mechanism. For the optical system modeling, the methodology is based on the usage of some point spread functions (PSFs). The use of the full set of PSFs is computationally costly and memory space consuming hence compromising the usability of the optical model in the full system virtual prototyping. To improve the model execution time, PSFs interpolation and free-space propagation techniques are used: they allow reducing the sampling space with minimal impact on the accuracy of the model (sharpness error less than 2%). The total speed-up gain with respect to the standard non-optimized model is provided by two contributors. First, the interpolation technique leads to a speed-up linked to the PSFs number reduction. Second, the caching of computationally intense processes enables speed-up scaling with the number of frames.