Dual-drop microring weight banks with linear configuration

Traditional electronic computing methods are hitting their limits as data processing demands continue to escalate. Optical neural networks (ONNs) present a promising solution to overcome this challenge. The microring resonator (MRR) weight bank (WB), which provides distinct weights for various light wavelengths in optical neural networks (ONNs), is advantageous due to its compact size, scalability, ease of tuning, and natural compatibility with wavelength division multiplexing (WDM). Despite these benefits, MRRs are highly sensitive and experience uneven transmission power with wavelength shifts, leading to inefficient and imprecise tuning. In this work, we successfully demonstrate a dual-drop microring weight bank (DDMRR-WB) with linear configuration, maintaining weight precision over 4.3 bits.