All-fiber content-addressable memory (CAM) for in-sensor computing
The in-sensor computing based on fiber can transmit near-light-speed information, skip photoelectric conversion, shorten data storage and reading delay, and quickly respond to massive data for intelligent optical sensor networks. Content Addressable Memory (CAM) is capable of screening the useful data sensed in intelligent photonic perception networks and filtering out redundant data. Herein, we propose an all-optical CAM based on fiber and Ge<sub>2</sub>Sb<sub>2</sub>Te<sub>5</sub> (GST), the device structure is simple and easy to manufacture. We first use a core of side-etched seven-core fiber (SCF) to achieve 22-level storage. Next, CAM is realized by Mach-Zehnder (MZ) interferometer composed of two fiber cores of SCF. Finally, we achieve a 3-bit CAM by using a single SCF. The maximum average power density for an individual CAM cell is 23.64 nW/μm<sup>2</sup> and requires no continuous power input to maintain the switching states, which is also naturally advantageous given the low-power density and the long intervals between entry updates in modern optical packet switching systems.
Authors
- Siying Cheng (ORCID: https://orcid.org/0000-0002-9406-6873)
- Yuqing Li (ORCID: https://orcid.org/0009-0003-4029-9051)
- Yu Zhang (ORCID: https://orcid.org/0000-0001-8362-5871)
- Xiang Li (ORCID: https://orcid.org/0000-0002-5214-217X)
- Zhihai Liu (ORCID: https://orcid.org/0000-0001-5622-2372)
- Wei Jin (ORCID: https://orcid.org/0000-0003-0016-3160)
Institutions
- Harbin Engineering University (CN)
Publication Details
- Published
- 2025-05-22
- DOI
- https://doi.org/10.1117/12.3061632
- Primary Topic
- Advanced Memory and Neural Computing
- Type
- article
- Field-Weighted Citation Impact
- 0.00