Determination of optical performance of distributed Raman amplifiers using deep neural networks
Systems and methods include receiving inputs comprising a Raman gain target for a Raman amplifier, fiber parameters, and fiber characteristics, each for an optical span; analyzing the inputs with a deep neural network (DNN) having a plurality of layers; providing a plurality of outputs from the DNN; and utilizing the plurality of outputs for a link budget calculation on the optical span. The Raman amplifier includes a plurality of pumps each at a fixed wavelength.
1 . A non-transitory computer-readable medium comprising instructions that, when executed, cause one or more processors to perform steps of:
receiving inputs comprising a Raman gain target for a Raman amplifier, fiber parameters including any of fiber attenuation coefficient, fiber span loss, and fiber loss margin, and fiber characteristics including any of fiber effective area, zero chromatic dispersion wavelength, fiber attenuation at Raman pump frequencies, and a peak Raman gain coefficient value, each for an optical span;
analyzing the inputs with a deep neural network (DNN) having a plurality of layers;
providing a plurality of outputs from the DNN, the outputs including any of a calculated Raman gain spectrum, an effective noise figure, an incremental multi-path interference (ΔMPI), and an Optical Service Channel (OSC) gain; and
utilizing the plurality of outputs for a link budget calculation including determination of amplifier settings, launch powers, channel occupancy, spacing, and performance margin for the optical span.
2 . The non-transitory computer-readable medium of claim 1 , wherein the Raman amplifier includes a plurality of pumps each at a fixed wavelength.
3 . The non-transitory computer-readable medium of claim 1 , wherein the Raman amplifier includes a plurality of pumps each at a fixed wavelength, and wherein the Raman amplifier supports amplification across both the C-band and the L-band.
4 . The non-transitory computer-readable medium of claim 1 , wherein the Raman gain target specifies a gain over a spectrum of interest, and wherein the plurality of outputs include a calculated Raman gain representing an expected gain achieved over the spectrum of interest.
5 . The non-transitory computer-readable medium of claim 1 , wherein the Raman amplifier includes a plurality of pumps each at a fixed wavelength, and wherein the Raman amplifier is configured to set a pump power of each of the plurality of pumps based on the Raman gain target.
6 . The non-transitory computer-readable medium of claim 1 , wherein the Raman amplifier operates in a plurality of operating modes and switching between two of the plurality of operating modes causes a discontinuity in performance of the Raman amplifier.
7 . The non-transitory computer-readable medium of claim 6 , wherein the DNN includes one or more classification layers to identify an operating mode of the plurality of operating modes, followed by one or more regression layers to receive outputs from the one or more classification layers and provide the plurality of outputs of the DNN.
8 . The non-transitory computer-readable medium of claim 6 , wherein the Raman amplifier includes a plurality of pumps each at a fixed wavelength, and wherein the Raman amplifier supports amplification of channels across both the C-band and the L-band, with the plurality of operating modes configured to minimize undesirable interference between the plurality of pumps and the channels in either of the C-band and the L-band.
9 . The non-transitory computer-readable medium of claim 1 , wherein the steps further include
performing a simulation based on non-linear effects, hardware specifications of the Raman amplifier, and software control of the Raman amplifier to obtain a set of training data; and
training the DNN with the set of training data.
10 . A method comprising steps of:
receiving inputs comprising a Raman gain target for a Raman amplifier, fiber parameters including any of fiber attenuation coefficient, fiber span loss, and fiber loss margin, and fiber characteristics including any of fiber effective area, zero chromatic dispersion wavelength, fiber attenuation at Raman pump frequencies, and a peak Raman gain coefficient value, each for an optical span;
analyzing the inputs with a deep neural network (DNN) having a plurality of layers;
providing a plurality of outputs from the DNN, the outputs including any of a calculated Raman gain spectrum, an effective noise figure, an incremental multi-path interference (ΔMPI), and an Optical Service Channel (OSC) gain; and
utilizing the plurality of outputs for a link budget calculation including determination of amplifier settings, launch powers, channel occupancy, spacing, and performance margin for the optical span.
11 . The method of claim 10 , wherein the Raman amplifier includes a plurality of pumps each at a fixed wavelength.
12 . The method of claim 10 , wherein the Raman amplifier includes a plurality of pumps each at a fixed wavelength, and wherein the Raman amplifier supports amplification across both the C-band and the L-band.
13 . The method of claim 10 , wherein the Raman gain target specifies a gain over a spectrum of interest, and wherein the plurality of outputs include a calculated Raman gain representing an expected the gain achieved over the spectrum of interest.
14 . The method of claim 10 , wherein the Raman amplifier includes a plurality of pumps each at a fixed wavelength, and wherein the Raman amplifier is configured to set a pump power of each of the plurality of pumps based on the Raman gain target.
15 . The method of claim 10 , wherein the Raman amplifier operates in a plurality of operating modes and switching between two of the plurality of operating modes causes a discontinuity in performance of the Raman amplifier.
16 . The method of claim 15 , wherein DNN includes one or more classification layers to identify an operating mode of the plurality of operating modes, followed by one or more regression layers to receive outputs from the one or more classification layers and provide the plurality of outputs of the DNN.
17 . The method of claim 15 , wherein the Raman amplifier includes a plurality of pumps each at a fixed wavelength, and wherein the Raman amplifier supports amplification of channels across both the C-band and the L-band, with the plurality of operating modes configured to minimize undesirable interference between the plurality of pumps and the channels in either of the C-band and the L-band.
18 . The method of claim 10 , wherein the steps further include
performing a simulation based on non-linear effects, hardware specifications of the Raman amplifier, and software control of the Raman amplifier to obtain a set of training data; and
training the DNN with the set of training data.