Fig. 2
Experimental setup of three-layer D2NN with reflective vHOEs. (a) Schematic of the experimental setup, where \(\lambda\) is \(532 \, \text{nm}\). The distance between the SLM and vHOE1 is \(1.12 \times 10^6 \lambda\), the distance between vHOE1 and vHOE2 is \(5.64 \times 10^5 \lambda\), the distance between vHOE2 and vHOE3 is \(7.52 \times 10^5 \lambda\), and the distance between vHOE3 and the complementary metal-oxide-semiconductor (CMOS) is \(1.50 \times 10^6 \lambda\). The vHOE2 and the vHOE3 are shifted by \(1.32 \times 10^4 \lambda\) from vHOE1 in the x-axis direction. The output wavefront on the CMOS is shifted by \(3.52 \times 10^4 \lambda\) from vHOE2 and vHOE3 in the x-axis direction. HWP: half-wave plate, M: mirror, PBS: polarizing beam splitter. (b) Phase distribution of each layer obtained by learning. (c) Actual optical experimental setup. The following equipment was used. CMOS: The Imaging Source DMK 33UX183, SLM: JVCKENWOOD D-ILA pixel pitch \(4.8 \, \upmu \text {m}\), laser: Cobolt \(532 \, \text{nm}\) Samba 05. In vHOE3 shown in the photograph, vHOE is inside the red box and the label above vHOE is information about this setup and this layer. The other vHOE outside the red box is a backup sample that does not affect the experiment.
