A Smarter Way to Align Light: Celebrating Our Latest Patent in PM Fiber Optics Innovation

Neptec OS proudly unveils a patented innovation that uses AI-driven imaging to help automate the alignment of polarization-maintaining (PM) fibers, paving the way for faster, smarter photonic manufacturing.

(View the patent here)

The user has selected the tolerance of ±3° for alignment relative to the industry standard of  ± 5 °. The software provides real-time feedback on PM fiber alignment, displaying the angle with a binary indicator (OK/OFF) to guide users on manual rotational adjustments. Green(OK) indicates that you are within the selected tolerance and can stop adjusting, the red(OFF) means to keep adjusting until reaching green(OK)

Patent Overview

When precision meets imagination, new ways of doing hard things become possible. That’s exactly what happened when Miao Qi, one of our talented engineers, developed an automated software using deep learning that takes the complexity out of aligning polarization-maintaining (PM) fibers, a process traditionally done by hand. Her patented software enables visual alignment of the stress rods to the slow and fast axes, eliminating the need for costly active alignment equipment typically used to measure PER with a laser source, polarizer, and power meter. 

This process is applicable to both single fiber and more complex multi-fiber alignment applications. Especially when comparing multi-fiber alignment using PER as the guiding parameter, each fiber needs to be measured independently. However, with Miao’s software, you can align any number of fibers at the same time, saving time and effort.

From Manual Precision to Intelligent Automation Software

Aligning PM fibers has always required equipment investment and patience. Micron-level accuracy is critical because even a small misalignment of the fiber’s slow or fast axis can impact optical performance. Neptec’s Miao saw an opportunity to make this process faster, more reliable, and scalable.

Her newly patented software powered by neural networks automatically identify fiber end faces and predicts the precise rotation and direction needed to align them to a reference key.

Multiple PM fibers are analyzed simultaneously in real time. The bottom number indicates the rotation angle needed to align the slow axis.

Neural Networks That See and Think

The innovation lies in how the software “sees” and “decides.” For each fiber captured in an image, the algorithm predicts the rotation angle to ensure perfect alignment within. Even when lighting or image clarity challenges traditional methods, this intelligent approach can adapt and guide the system toward correct alignment in real time.

This automated solution isn’t just a concept; it’s built for integration. Its outputs can easily connect to existing mechanical setups, providing automated feedback for fiber alignment tasks. That means manufacturers and research labs can achieve consistency without sacrificing speed or accuracy (within +/-1° angular tolerance).

Images illustrate automated software detecting the rotation angle of PM fiber alignment, reliably performing even when end-face visibility is imperfect. The robust algorithm ensures consistent results across various working environments.

Pioneering What’s Next in Photonic Assembly

Innovation like this reinforces our commitment to advancing fiber optic assembly technology through both creativity and practicality. By blending optical engineering with machine learning, Neptec continues to set new standards for how photonic systems can think, respond, and refine operations autonomously.

We’re proud of Miao Qi for transforming a complex challenge into an elegant, patented, solution one that illuminates the future of automated fiber alignment by simplifying the process to the point where multi-fiber PM cables have become almost ordinary.