The photonic neural processing unit (PNPU) demonstrates ultrahigh inference speed with low energy consumption, and it has become a promising hardware artificial intelligence (AI) accelerator. However, the nonidealities of the photonic device and the peripheral circuit make the practical application much more complex. Rather than optimizing the photonic device, the architecture, and the algorithm individually, a joint device-architecture-algorithm codesign method is proposed to improve the accuracy, efficiency and robustness of the PNPU. First, a full-flow simulator for the PNPU is developed from the back end simulator to the high-level training framework; Second, the full system architecture and the complete photonic chip design enable the simulator to closely model the real system; Third, the nonidealities of the photonic chip are evaluated for the PNPU design. The average test accuracy exceeds 98%, and the computing power exceeds 100TOPS.
A novel square-assisted ring-core fiber characterized by a square region of low refractive index in the core was proposed. This fiber structure allows for spatial mode modulation. When the subscript m of the LPmn modes supported by the fiber is an even number, the spatial modes of the LPmn modes are degenerately separated. This fiber supports 15 modes over the entire C-band. The effective refractive index difference (Δneff) between spatial modes at 1550nm is greater than 1.18×10-4. At the same time, the Δneff between all adjacent modes is in the range of (1.13~13.52)×10-4. The polarization separation level of each mode is below 7.12×10-6, which is two orders of magnitude lower than the level of degenerate separation of spatial modes. Numerical simulations show that the dispersion values range from -16 ps/nm/km to 15 ps/nm/km for 15 modes over the entire C-band. This fiber has a good tolerance for process error, the optical fiber processing requirements can be met by using the existing optical fiber preparation technology.
Polarization beam splitter is an important part of integrated optical system to overcome the strong polarization dependence of silicon nanodevices, and has broad application prospects in optical fiber communication and polarization imaging. In this paper, a polarization beam splitter of metalens based on the fiber end face is designed. Through the planar light field control ability of metalens and the coordinated regulation of the geometric phase and propagation phase, the device adjusts the medium duty cycle in the meta-unit and the rotation angle of the nanoantenna to realize the polarization beam splitting and focusing simultaneously. In order to simplify the manufacturing process and achieve high focusing efficiency even further, Si with high transmittance in the near infrared band is selected as the nanoantenna, and SiO2 is used as the substrate. Since the substrate used the same material with fiber cladding, compared with the traditional polarization beam splitter, the structure can directly etch the nanofins periodically on the fiber end face, which is convenient for optical system integration. Theoretical design and numerical simulation results show that any polarized light in fiber cladding can realize polarization focusing through fiber end face.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.