<p>The proposed automatic bias control technique for Mach-Zehnder modulators (MZMs) utilizes Root Mean Square Propagation (RMSProp) and optical feedback to adjust the DC bias voltage based on the gradient of the MZM’s transfer function. This approach maximizes the signal-to-noise ratio (SNR) without the need for RF dither signals, relying instead on optical output power measurements and numerical gradient estimation. Bias drift, primarily caused by environmental variations such as temperature changes and mechanical disturbances, severely affects modulation stability in high-speed optical communication systems. To validate its effectiveness, the proposed method was experimentally evaluated. In experiments conducted over 30&#xa0;h, the proposed method maintained the bias voltage within <InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="11082_2025_8391_Article_IEq1.gif" Format="GIF" Height="13" Rendition="HTML" Resolution="72" Type="Linedraw" Width="18" /> </InlineMediaObject> <EquationSource Format="TEX">\(\:\pm\:\)</EquationSource> </InlineEquation>0.1&#xa0;V of the optimal point for over 97% of the time, and sustained a bit error rate (BER) of approximately 4<InlineEquation ID="IEq2"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="11082_2025_8391_Article_IEq2.gif" Format="GIF" Height="13" Rendition="HTML" Resolution="72" Type="Linedraw" Width="18" /> </InlineMediaObject> <EquationSource Format="TEX">\(\:\times\:\)</EquationSource> </InlineEquation>10<sup>−9</sup>. The system was implemented on a closed-loop optical transmission link using NRZ-OOK modulation with a 1550&#xa0;nm laser source. Compared to Adaptive Moment Estimation (Adam), RMSProp delivered comparable performance with lower computational cost, making it well suited for real-time deployment in optical transmitters.</p>

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

Automatic bias control of Mach-Zehnder modulator using RMSProp-based gradient feedback optimization

  • Dongmin Choi,
  • Serin Kim,
  • Yong-Yuk Won

摘要

The proposed automatic bias control technique for Mach-Zehnder modulators (MZMs) utilizes Root Mean Square Propagation (RMSProp) and optical feedback to adjust the DC bias voltage based on the gradient of the MZM’s transfer function. This approach maximizes the signal-to-noise ratio (SNR) without the need for RF dither signals, relying instead on optical output power measurements and numerical gradient estimation. Bias drift, primarily caused by environmental variations such as temperature changes and mechanical disturbances, severely affects modulation stability in high-speed optical communication systems. To validate its effectiveness, the proposed method was experimentally evaluated. In experiments conducted over 30 h, the proposed method maintained the bias voltage within \(\:\pm\:\) 0.1 V of the optimal point for over 97% of the time, and sustained a bit error rate (BER) of approximately 4 \(\:\times\:\) 10−9. The system was implemented on a closed-loop optical transmission link using NRZ-OOK modulation with a 1550 nm laser source. Compared to Adaptive Moment Estimation (Adam), RMSProp delivered comparable performance with lower computational cost, making it well suited for real-time deployment in optical transmitters.