Machine Learning for Self-Coherent Detection Short-Reach Optical Communications

Round 1

Reviewer 1 Report

Considering that self-coherent detection systems, in which the optical carrier and the signal are transmitted together, are attracting great interest in the field of optical communications for short- to medium-range fast optical transmission links, this comprehensive review of recent advances in self-coherent detection systems and extensive applications of machine learning in self-coherent detection systems may be relevant in this field. I recommend publication of this article if some clarifications and improvements are made:

1) The introduction is missing some important references. For example:

-      -M. P. Yankov, F. Da Ros, U. C. de Moura, A. Carena and D. Zibar, "Flexible Raman Amplifier Optimization Based on Machine Learning-Aided Physical Stimulated Raman Scattering Model," in Journal of Lightwave Technology, vol. 41, no. 2, pp. 508-514, 15 Jan.15, 2023, doi: 10.1109/JLT.2022.3218137.

-        -Milorad Cvijetic, Coherent and Nonlinear Lightwave Communications, Artech House Publishers (January 1, 1996), ISBN-13 ‏ : ‎ 978-0890065907

-        -B. Batagelj, V. Janyani and S. Tomazic, ''Research challenges in optical communications towards 2020 and beyond '', Informacije MIDEM, vol. 44 (2014),no. 3, pp. 177-184 http://www.midem-drustvo.si/Journal%20papers/MIDEM_44(2014)3p177.pdf

-        -E. M. Ip and J. M. Kahn, "Fiber Impairment Compensation Using Coherent Detection and Digital Signal Processing," in Journal of Lightwave Technology, vol. 28, no. 4, pp. 502-519, Feb.15, 2010, doi: 10.1109/JLT.2009.2028245.

-        -Uzunidis, D.; Logothetis, M.; Stavdas, A.; Hillerkuss, D.; Tomkos, I. Fifty Years of Fixed Optical Networks Evolution: A Survey of Architectural and Technological Developments in a Layered Approach. Telecom 2022, 3, 619-674. https://doi.org/10.3390/telecom3040035

-        -Seb J. Savory, Giancarlo Gavioli, Robert I. Killey, and Polina Bayvel, "Electronic compensation of chromatic dispersion using a digital coherent receiver," Opt. Express 15, 2120-2126 (2007), DOI: 10.1364/oe.15.002120

-        -R. W. Gerchberg, “A practical algorithm for the determination of phase from image and diffraction plane pictures,” Optik, vol. 35, pp. 237–246, 1972

-        -J. Winzer and R. . -J. Essiambre, "Advanced Optical Modulation Formats," in Proceedings of the IEEE, vol. 94, no. 5, pp. 952-985, May 2006, doi: 10.1109/JPROC.2006.873438.

-        -K. Kikuchi, "Quantum Theory of Noise in Stokes Vector Receivers and Application to Bit Error Rate Analysis," in Journal of Lightwave Technology, vol. 38, no. 12, pp. 3164-3172, 15 June15, 2020, doi: 10.1109/JLT.2020.2967420.

-        Since this is a comprehensive review paper, the authors must also include these important references in their manuscript.

2) I assume the equations presented are not new. Authors should use references if the equations do not show their own results from your own derivation.

3) The authors use many acronyms in the paper. All acronyms (standard acronyms and author-defined acronyms) should be defined at the first mention in the abstract and in the body of the article. Authors should pay attention to this. For example, the abbreviation ANN (artificial neural network) is never defined in the article. I also recommend an additional abbreviations section defining the abbreviations used in the article.

4) In Figure 5a, which shows the transfer function of the MZM, it is not clear what the “modulation index” is. The author should explain this in the text of the paper or delete the “modulation index” from Figure 5a. The text of the paper text is that the modulation index is the peak value of electrical output. This definition seems to me incorrect or at least imprecise.

5) The author should make sure that the X and Y axes of all graphs are always labelled and the units are clearly indicated. Now the Y-axis labels are missing from the graphs in Figure 1(b), Figure 1(c), Figure 2(a), Figure 5(a), and Figure 6(a).

Author Response

Comment 1:

1) The introduction is missing some important references. For example:

-      -M. P. Yankov, F. Da Ros, U. C. de Moura, A. Carena and D. Zibar, "Flexible Raman Amplifier Optimization Based on Machine Learning-Aided Physical Stimulated Raman Scattering Model," in Journal of Lightwave Technology, vol. 41, no. 2, pp. 508-514, 15 Jan.15, 2023, doi: 10.1109/JLT.2022.3218137.

-        -Milorad Cvijetic, Coherent and Nonlinear Lightwave Communications, Artech House Publishers (January 1, 1996), ISBN-13 ‏ : ‎ 978-0890065907

-        -B. Batagelj, V. Janyani and S. Tomazic, ''Research challenges in optical communications towards 2020 and beyond '', Informacije MIDEM, vol. 44 (2014),no. 3, pp. 177-184 http://www.midem-drustvo.si/Journal%20papers/MIDEM_44(2014)3p177.pdf

-        -E. M. Ip and J. M. Kahn, "Fiber Impairment Compensation Using Coherent Detection and Digital Signal Processing," in Journal of Lightwave Technology, vol. 28, no. 4, pp. 502-519, Feb.15, 2010, doi: 10.1109/JLT.2009.2028245.

-        -Uzunidis, D.; Logothetis, M.; Stavdas, A.; Hillerkuss, D.; Tomkos, I. Fifty Years of Fixed Optical Networks Evolution: A Survey of Architectural and Technological Developments in a Layered Approach. Telecom 2022, 3, 619-674. https://doi.org/10.3390/telecom3040035

-        -Seb J. Savory, Giancarlo Gavioli, Robert I. Killey, and Polina Bayvel, "Electronic compensation of chromatic dispersion using a digital coherent receiver," Opt. Express 15, 2120-2126 (2007), DOI: 10.1364/oe.15.002120

-        -R. W. Gerchberg, “A practical algorithm for the determination of phase from image and diffraction plane pictures,” Optik, vol. 35, pp. 237–246, 1972

-        -J. Winzer and R. . -J. Essiambre, "Advanced Optical Modulation Formats," in Proceedings of the IEEE, vol. 94, no. 5, pp. 952-985, May 2006, doi: 10.1109/JPROC.2006.873438.

-        -K. Kikuchi, "Quantum Theory of Noise in Stokes Vector Receivers and Application to Bit Error Rate Analysis," in Journal of Lightwave Technology, vol. 38, no. 12, pp. 3164-3172, 15 June15, 2020, doi: 10.1109/JLT.2020.2967420.

-        Since this is a comprehensive review paper, the authors must also include these important references in their manuscript.

Response: Thank you for your valuable feedback. We appreciate your suggestion to include important references in our comprehensive review paper. We have carefully reviewed the references you've pointed out and ensured their appropriate incorporation into the manuscript.

We add the references [2-8, 19, 20, 26] in the Reference.

 

Comment 2:

2) I assume the equations presented are not new. Authors should use references if the equations do not show their own results from your own derivation.

Response: Thank you for your feedback. We understand the importance of proper referencing for equations. We have ensured that appropriate references are provided for equations that are not a result of our own derivation. Your input helps us maintain the integrity and accuracy of our manuscript.

  We add Refs. [21], [92], [1] for Equations. (1)-(5), respectively.

 

 

Comment 3:

3) The authors use many acronyms in the paper. All acronyms (standard acronyms and author-defined acronyms) should be defined at the first mention in the abstract and in the body of the article. Authors should pay attention to this. For example, the abbreviation ANN (artificial neural network) is never defined in the article. I also recommend an additional abbreviations section defining the abbreviations used in the article.

Response: Thank you for the feedback. We recognize the need for clear acronym definitions. We have ensured that all acronyms, including both standard and author-defined ones, are defined at their first mention in both the abstract and the body of the article. (The ANN is defined in page6, line 225) Additionally, we have added an abbreviations section to provide a comprehensive list of defined abbreviations according to your advice in the APPENDIX.

 

 

Comment 4:

4) In Figure 5a, which shows the transfer function of the MZM, it is not clear what the “modulation index” is. The author should explain this in the text of the paper or delete the “modulation index” from Figure 5a. The text of the paper text is that the modulation index is the peak value of electrical output. This definition seems to me incorrect or at least imprecise.

Response: Thank you for your feedback. We understand your concern about the definition of "modulation index" in Figure 5a. We have revised “modulation index” to “the peak-to-peak voltage” to avoid possible confusion. Your comment helps improve the accuracy and clarity of our content.

 

 

 

Comment 5:

5) The author should make sure that the X and Y axes of all graphs are always labelled and the units are clearly indicated. Now the Y-axis labels are missing from the graphs in Figure 1(b), Figure 1(c), Figure 2(a), Figure 5(a), and Figure 6(a).

Response: Thank you for pointing out this issue. We appreciate your feedback regarding the labeling of the X and Y axes in the graphs. We have ensured that all graphs, including those in Figure 1(b), Figure 1(c), Figure 2(a), Figure 5(a), and Figure 6(a), have clear and appropriately labeled axes, along with the inclusion of units. Your attention to detail is valuable, and we will take the necessary steps to enhance the clarity and comprehensibility of our figures.

 

Reviewer 2 Report

Comments to the Author,

 

The authors present recent advances in machine learning (ML) for self-coherent detection systems designed for high-speed optical short- to medium-reach transmission links and discuss different neural network architectures and future prospects in ML for self-coherent detection systems. The authors clearly explain the principle of using ML and verify the feasibility of the scheme through simulations and experiments. I think there are some questions need to be clarified before it can be accepted. Below are some detailed comments.

1.     In this paper, is there a corresponding increase for the cost of using machine learning in a direct detection system?

2.     Can the transmission distance limitation problem due to power fading caused by dispersion be solved by machine learning for SCD systems?

3.     Can the noise problem presented in SCD systems be solved by machine learning?

4.     In Figure 3. (a), there is an error for the S₂ component BPD in the Stokes-vector receiver, and it needs to be modified.

5.     Please check carefully for minor grammatical errors in the manuscript.

The English in this article is well expressed, with only a few grammatical errors.

Author Response

Comment 1:

In this paper, is there a corresponding increase for the cost of using machine learning in a direct detection system.

Response: In our paper, we acknowledge that integrating machine learning into direct detection systems can potentially lead to an increase in costs due to the specialized hardware requirements for efficient computation. However, the impact of this cost increase depends on factors such as advancements in hardware technology and the benefits gained from enhanced performance. We appreciate the reviewer's consideration of this aspect and have addressed it in our discussions.

We add the illustration of the cost of machine learning “Integrating machine learning into direct detection systems may raise costs due to specialized hardware needs for efficient computation. The actual impact varies with technology advancements and the benefits gained from enhanced performance.” in the Conclusion.

 

 

Comment 2:

Can the transmission distance limitation problem due to power fading caused by dispersion be solved by machine learning for SCD systems?

Response: In the traditional IM-DD system, ML can help to mitigating the chromatic-dispersion-induced power fading effect due to its strong equalization ability. In the SCD system, the optical field could be recovered in the receiver DSP, where there is no power fading. SCD systems enable to compensate for the chromatic dispersion in the electrical domain and do not suffer from power fading.

We add the illustration of the compensation of chromatic dispersion “SCD systems recover the optical field in receiver DSP, allowing compensation for chromatic dispersion similar to coherent detection. The power fading effect induced by the traditional IMDD channel will no longer be a problem in SCD systems” in the Introduction.

 

 

 

Comment 3:

Can the noise problem presented in SCD systems be solved by machine learning?

Response: Thank you for your question. Machine learning can serve as a noise whitening filter to enhance bit error rate performance in the case of colored noise. However, for randomly distributed Gaussian noise, machine learning cannot improve SNR.

 

 

 

 

Comment 4:

In Figure 3. (a), there is an error for the S2 component BPD in the Stokes-vector receiver, and it needs to be modified?

Response: Thank you for bringing this to our attention. We apologize for the errors in Figure 3(a) related to the S2 component BPD in the Stokes-vector receiver. We will thoroughly review and verify all the figures, and if necessary, make the required modifications to ensure accuracy.

 

 

 

Comment 5:

Please check carefully for minor grammatical errors in the manuscript.

Response: Thank you for your feedback. We appreciate your suggestion to review the manuscript for minor grammatical errors. We are committed to delivering a polished and error-free manuscript, and we have carefully gone through the entire manuscript to ensure the grammatical accuracy.

Reviewer 3 Report

machine learning algorithms have demonstrated remarkable performance in various types of optical communication applications, including channel equalization, constellation optimization, and optical performance monitoring. ML can also find its place in SCD systems in these scenarios. In this paper, the authors provide a comprehensive review of the recent progress in SCD systems designed for high-speed optical short- to medium-reach transmission links, and discuss the diverse applications and the future perspectives of ML for these SCD systems.

The authors show lots of researches referring to literatures, but it is not as good as expected because the authors did not harmoniously synthesize these literatures.

What is the trend of the ML in SCD systems? Are there any challenges for ML in SCD applications? This is a very important and interest problems for readers. The authors should demonstrate the trend and the challenges.

the english is good.

Author Response

Comment 1:

machine learning algorithms have demonstrated remarkable performance in various types of optical communication applications, including channel equalization, constellation optimization, and optical performance monitoring. ML can also find its place in SCD systems in these scenarios. In this paper, the authors provide a comprehensive review of the recent progress in SCD systems designed for high-speed optical short- to medium-reach transmission links, and discuss the diverse applications and the future perspectives of ML for these SCD systems.

The authors show lots of researches referring to literatures, but it is not as good as expected because the authors did not harmoniously synthesize these literatures.

What is the trend of the ML in SCD systems? Are there any challenges for ML in SCD applications? This is a very important and interest problems for readers. The authors should demonstrate the trend and the challenges.

Response: Thank you for your valuable comments. We will discuss the trend of machine learning in SCD systems and the associated challenges. Your point is well taken, and we aim to provide readers with valuable insights into this important aspect of our research.

We have added the illustration of the trend and associated challenges of ML in SCD system “In the context of SCD systems, machine learning techniques are increasingly favored for tasks such as optical field recovery or phase retrieval, tasks that traditional nonlinear equalization algorithms struggle to achieve.” in the Conclusion.

We have added the illustration of the challenges of ML in SCD applications “Regarding challenges linked to applying ML in SCD systems, these involve concerns about computational complexity and hardware requirements, especially for ASIC chips. Consequently, it is essential to focus future endeavors on exploring and resolving the intricacies of ML algorithms to facilitate their practical implementation.” in the Conclusion.

Round 2

Reviewer 3 Report

the authors have made some improvements according to my comments, I agree to recommend it to be accepted.