Short-Reach MCF-Based Systems Employing KK Receivers and Feedforward Neural Networks for ICXT Mitigation
Round 1
Reviewer 1 Report
This is a well written paper that discuss very interesting topic. It will be better if the author can briefly clarify the following points or suggestions:
- For the link lengths considered in this study, Chromatic dispersion will have major impact? how does it was eliminated or compensated?
- What about other non-linear effects of optical fiber?
- I think its better if the author add more main results of the study to the abstract like BER etc.
Author Response
Please, see the attachment (Reviewer 1 responses) .
Author Response File: 
Author Response.pdf
Reviewer 2 Report
In this paper, the authors analyze the short-reach MCF-based networks employing KK receivers and K-means, KNN and FNN for ICXT mitigation, here are some comments.
1) The training overhead should be added and commented.
2) The algorithm complexity should be calculated.
3) Does the algorithm fit the online learning requirements? Since the ICXT is a time-varying problem.
4) The quality of Fig. 8 is not good, the authors should consider to separate the lines of wo ML, K-means, KNN.
Author Response
Please, see the attachment (Reviewer 2 answers) .
Author Response File: Author Response.pdf
Reviewer 3 Report
This paper focuses on the research of using ML to reduce the ICXT in the transmission system of specific networks (240 Gb/s short-reach networks employing MCF and KK receivers). The topic is meaningful. The proposed system model is reasonable and using FNN techniques improve the performance.
However, in my opionion, the manuscript still requires some further clarifications, as follows.
1、 The main claimed innovation of this paper is to apply FNN technology in the system to reduce ICXT. What is the current mainstream method except FNN? What is the difference between them？Please elaborate.
2、 When discussing the topic of networks , we usually concern the related problems caused by switching between more than 3 nodes. It seems that this paper is not in this scope. The focus of this paper seems to be the transmission system. If so, the network is only an application scenario of the system. Please decide whether to modify the title and other contents as appropriate.
3、 This paper investigates three ML methods -- FNN, K-means clustering and KNN to estimate the BER. Can you explain more about the state of that art of applying ML methods in this problem? From my side, the context and related work are not clear enough.
Author Response
Please, see the attachment (Reviewer 3 answers) .
Author Response File: Author Response.pdf
Reviewer 4 Report
The author demonstrated three kinds of machine learning (ML) techniques in short-reach optical communication systems. But there are several key points that the author does not make clear. Here are the detailed comments.
1. In the introduction part, first sentence, the author claim 100 Tb/s per single core fiber, it is unbelievable.
2. The author claims that the system is 64 Gbaud SSB-16QAM and using KK receiver. The signal is Nyquist shaping (roll off=0.5). Why the author set almost 30 GHz guard band? The KK receiver is no need for such large guard band, and the O/E and E/O bandwidth should be large than 90 GHz.
3. Why the system rate is 240 Gb/s? The signal is modulation on single polarization, and detected by single PD. What is the significance of Figure 3?
- The author employing multicore fiber (MCF) in this system, but 2 cores are too few. And 0 dBm launch power is too low. Maybe the ICXT is very small. The author should give a figure to prove the ICXT’s impairment is important.
- The author should give a figure about BER vs. OSNRs to show the system performance.
Author Response
Please, see the attachment (Reviewer 4 answers) .
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
The paper can be accepted.
Author Response
We thank you for the review
Reviewer 3 Report
The authors answered my questions perfectly.
Author Response
We thank you for the review
Reviewer 4 Report
The author claimed that 35 km transmission by using KK and MCF no need for optical amplifier. I do NOT think so! Although the author focus on ICXT impairment mitigation. The author do NOT use optical amplifier in simulation. The author must give a final result to show the system performance, such as BER vs. ROPs.
Author Response
The authors followed the reviewer recommendation and evaluated the mean BER as a function of the received optical power. This information was added to the manuscript. Please, see figure 10, the last paragraph of the section 3, the abstract and conclusion.
Author Response File: Author Response.pdf
