Design of a novel intelligent active magnetic canceling system for mitigating extremely low frequency magnetic...

Electromagnetic interference has a significant impact on semiconductor manufacturing quality and product yield. The image quality of the high precision equipment, such as scanning electron microscopes and electron beam lithography machines, is often degraded by electromagnetic interference. Conventional filtered-x least mean squares (FxLMS) algorithm used for active magnetic field cancellation is subject to hardware-induced processing delays on the order of 5-20 ms, which can degrade the noise suppression performance. This paper presents a novel algorithm-based approach that integrates long short-term memory (LSTM) networks with FxLMS adaptive filtering to mitigate the hardware limitations via predictive compensation. The proposed LSTM-enhanced FxLMS algorithm employs the adaptive sequence length calculation and the dynamic step size adjustment to compensate the hardware-induced system latency. Compared to the conventional FxLMS algorithm, the proposed algorithm achieves an 8.75% improvement for single-frequency input and a 13.58% enhancement for multi-frequency cases. Furthermore, the proposed algorithm maintains stable noise suppression performance under processing delays of up to 40 ms.