Machine Learning Assisted Inverse Design for Ultrafine, Dynamic and Arbitrary Gain Spectrum Shaping of Raman Amplification

Round 1

Reviewer 1 Report

the authors presents a neural networks to for Raman Amplifier. the results are interestinting and the manuscripts is well writen. However before publish i suggest a some change.

• some equation needs references
• this works is only theoretical. Dificulties to implement  the experimental parts.
• comparation with other. Fast? precision? the author needs include a comparasion.

I suggest a minor revision before publish.

Author Response

Thanks for the suggestions.

Here are our responses for your questions:

 

Question1:

Actually, we have reference for Raman transmission equation, other equations are all about some equations in mathematics. And they are used as loss functions. BCE was proposed by Reuven Y.Rubinstein in 1999. And we add the reference[26] Rubinstein, R. Y. (1999). The simulated entropy method for combinatorial and continuous optimization. Methodology and Computing in Applied Probability, 2, 127–190. for it.

 

Question2:

This is a useful suggestion. Because of lack of time, this work is still a theoretical program so far. But we have the plan to take experiments to validate the simulation results. And we will actually measure this work by experiments and estimate. I believe it can well work.

 

Question3:

In the first part of this paper, we reckon that this is a novel scheme in this area. And before, we had similar simulations in 4-mode Raman system in reference[21]. Compared with that system, we think the scheme in this paper will attain more ultrafine and dynamic spectra in accordance with more situations, meanwhile it can be faster.

Reviewer 2 Report

The paper is very well written, structured and logical. I believe it deserves acknowledgment and be published in photonics Journal. 

I would only pinpoint one typo mistake on page 7: 

"The gain flatness at of each configuration 10 dB on-off gain is used to measure the accuracy of optimization." should be 

The gain flatness of each configuration at 10 dB on-off gain is used to measure the accuracy of optimization.

I wish authors all the best and further success in research. 

Author Response

Thanks for your advise. It's absolutely a typo mistake at the point. And it would be corrected in revised version.

 

Best wishes.

Reviewer 3 Report

The authors have investigated a distributed raman amplification scheme with broadband ASE source as the pump instead of discrete pump lasers. The broadband pump is optimized by machine learning based inverse design and shaped by programmable wave shaper, in order to realize the spectrum shaping of Raman amplification with dynamic gain. The manuscript is interesting and useful for the design of optical fiber communication systems, and is acceptable to be published in Photonics, provided the following issue can be addressed

 

1. Some abbreviations should be defined at their first appearance
2. Specify the parameters used in the transmission setup and summarize them in a table
3. Add some discussion regarding the potential improvement when the proposed scheme is used
4. Add some discussions regarding the inter-channel Raman scattering (IRS) effect in the new DRA system.
5. Improve the language a bit, e.g. spectrums -> spectra

Author Response

Thanks for the suggestions.

 

Here are our responses for your questions:

 

Question1:

We have checked all the abbreviations and add definitions where they first appear. e.g amplified spontaneous emission(ASE)

 

Question2:

We have listed these parameters in table1:

 

Revision:[Part: Broadband pump based DRA and inverse design]:

The parameters used are shown in Table 1:

 

Question3:

We have discussed this in revised version.

 

Revision:[Part: Result and discussion]

Last paragraph:

Considering the distortion of some device is time-variant, we can take advantage of waveshaper’s programmability to compensate the distortion in real-time, as Fig. 10 shows. Optical spectrum

analyzer (OSA) or other device is used to monitor the spectrum of output signal. The

adaptive compensation module will calculate the pump according to current spectrum and

target, and then control the pump module to update the gain of DRA. This configuration

can adaptively control the spectrum of output signal, avoiding dynamic change induced

by other devices. For instance, we can regard this system as an integral module. When

the signal fluctuates or changes occur during signal transmission, we can use this system

to compensate it. When it happens, the compensation waveform is obtained through

calculation, and then the system works to keep the signal stable or make the signal reach

the target waveform.

 

Question4:

 

IRS can be expressed as the transmission of optical power between signals. Because the signal optical channel interval is small and the Raman gain is weak, the IRS effect is also very weak. When we simulate and calculate the gain spectrum of signal light, IRS is also included in the calculation range.

 

Question5:

 

We have replace all “spectrums” with “ spectra”. e.g :

 

Revison[Part: Broadband pump based DRA and inverse design]

Paragraph1:

Once the NN has been well trained, it can be recognized as the mapping between gain spectrum and pump spectrum. Different target profiles of gain spectra are input to the trained NN for attaining corresponding pump spectrum.

 

Reviewer 4 Report

The manuscript presents a numerical method to optimize the wavelength response of Raman fiber amplifiers pumped by broad band light sources, the targeted response is achieve width high precision (including gain flattening) by tailoring the spectral shape of the pump . The work is original, the basic ideas are clearly presented and the results can be useful for researchers in the field.

Before publication, the authors must make clear in the abstract and in the introduction the theoretical character of their work in order to guide  future readers.

Author Response

Thanks for your suggestion and we also think it’s essential to simplify abstract and make the introduction clear. We make some changes in revised version.

 

e.g:

Revision:[Part: Abstract]

Distributed Raman amplifier (DRA) has been widely studied in recent decades because

of its low noise figure and flexible gain. In this paper, we present a novel scheme of DRA with

broadband amplified spontaneous emission(ASE) source as pump instead of discrete pump lasers.

The broadband pump is optimized by machine learning based inverse design and shaped by programmable waveshaper, so as to realize the ultrafine, dynamic and arbitrary gain spectrum shaping of Raman amplification. For the target of flat gain spectrum, the maximum gain flatness of 0.1086 dB is realized based on the simulation results. For the target of arbitrary gain spectrum, we demonstrate four gain profiles with maximum root mean square error (RMSE) of 0.074 dB. To further measure the performance of arbitrary gain spectrum optimization, the probability density functions (PDF) of RMSE and Errormax are presented. Meanwhile, the numeral relationship between the bands of broadband pump and signal is also explored. Furthermore, this work has great application potential to compensate the gain distortion or dynamic change caused by other devices in communication systems.