Proceedings Article | 20 June 2021
KEYWORDS: Light emitting diodes, LED displays, Optical design, Spatial resolution, Ray tracing, Cameras, 3D displays, Prototyping, Product engineering, Optical injection molds
LED technologies expectedly will dominate future displays by outperforming current mainstream technologies, while LED walls have already become the key solution in large-scale displaying. LED walls are not just display walls anymore; they offer a creative modular platform for displaying in any size, aspect, shape, and on any surface. At the time of this paper, there is no real 3D solution for such an attractive technology. In the recent years, rapid miniaturization of LEDs, including microLEDs, and LED control ICs has paved the way for novel LED wall displays with extremely high LED density. In this line, Holografika has developed a unique glasses-free 3D LED wall technology and this article describes the implementation of the 3D LED display prototype, focusing on the challenges in designing the optical elements, and also describes the software components that drive it. To match the high expectations for image quality, the optical system was designed with great care, balancing the conflicting 2D- and 3D-related parameters, like pitch size, angular resolution, FOV, considering practical use cases with typical viewing distances, covering specific LED layouts, up to the theoretical limits with regard to the chip sizes; novel optics for wide angle and high angular resolution beyond what usual lenticular systems can provide; preparing for volume production through designing plastic injection molded optics and tools; custom designed record high-density LED backplane, including PCB, seamlessly tileable LED panels where known slanted structures were not acceptable. The software system was designed to be run on GPUs and to be flexible enough to handle various configurations of display geometry and electronics. Not only are arbitrary layouts supported for the arrangement of the pixels and color channels of the 3D pixel under the sheet array, but also the geometries of the individual cabinets of the LED wall are separately addressable. The software calculates the correct ray field for the light field geometry and the color channel and pixel position shuffling table for the individual cabinets. The rendering consists of two distinct phases. The first phase is rendering the 3D scene for all camera rays. The system can render using various algorithms, such as real-time ray tracing; rasterization with orthogonal projection in the horizontal and perspective projection in the vertical axis, with individual images corresponding to directions selected by the lens array; rasterization of an array of horizontally asymmetric perspective cameras coupled with a ray-selection GPU render pass. The second phase enables shuffling the pixels and color channels to the correct byte order for the LED controller ICs. This allows the system to address individual cabinets. It also allows the usage of different ICs in subsequent generations of the display, without the need to deploy costly FPGAs with large area and power consumption requirements to handle the correct shuffling. Furthermore, the system enables the rendering hardware to scale across multiple PCs and GPUs. This is necessary to allow the system to render the high amount of pixels required to drive the LEDs.
