Design method of high-gain flat broadband erbium-doped fiber and Raman hybrid amplifier based on neural network

Abstract

The invention relates to a method for designing a high-gain flat broadband erbium-doped fiber and Raman hybrid amplifier based on a neural network. The method comprises the following steps: setting a target total gain of the hybrid amplifier; selecting the pumping power of the erbium-doped optical fiber amplifier; subtracting the gain of the erbium-doped optical fiber amplifier from the target total gain to obtain the gain of the Raman amplifier; and determining a Raman pumping parameter value according to the gain of the Raman amplifier by using a neural network. According to the invention, Raman pumping parameter values are rapidly determined by using the neural network, and it is ensured that the hybrid amplifier provides a high-accuracy broadband flat gain spectrum.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of fiber amplifier design, in particular to a high-gain flat wideband erbium-doped fiber and Raman hybrid amplifier design method. BACKGROUND

[0002] In long-distance optical fiber transmission networks, in order to increase the transmission distance, optical amplifiers are needed to compensate for the power loss of optical signals during link transmission, so optical amplifiers are key devices to ensure the normal operation of long-distance optical fiber transmission networks. Among various optical amplifiers, erbium-doped fiber amplifiers and optical fiber Raman amplifiers based on multi-wavelength pumping technology are two commonly used amplifiers.

[0003] In the use of erbium-doped fiber and Raman hybrid amplifiers, the pump parameters usually need to be manually adjusted according to the target gain of the amplifier. However, the manual adjustment method is time-consuming and may not guarantee that the hybrid amplifier provides the most ideal wideband flat gain spectrum. With the widespread attention of artificial intelligence technology in academia and industry, neural network technology has developed rapidly, and its application in optical fiber communication systems has also been valued. In the field of optical amplifier design, a trained neural network can also be used to calculate the pump parameter values required for the target gain of a flat wideband erbium-doped fiber and Raman hybrid amplifier. SUMMARY

[0004] (I) Technical problems to be solved

[0005] The present application proposes a high-gain flat wideband erbium-doped fiber and Raman hybrid amplifier design method based on neural networks, which aims to ensure that the erbium-doped fiber and Raman hybrid amplifier outputs an accurate target wideband flat gain spectrum.

[0006] (II) Technical solutions

[0007] To solve the above technical problems, the present application proposes a high-gain flat wideband erbium-doped fiber and Raman hybrid amplifier design method based on neural networks, including the training and application stages of the neural network:

[0008] The specific steps of the neural network training stage include:

[0009] Set the target total gain of the hybrid amplifier;

[0010] Select different erbium-doped fiber amplifier pump powers and adjust the Raman pump parameters to ensure that the hybrid amplifier reaches the target total gain;

[0011] Record the gain provided by the Raman amplifier for each signal light and the corresponding pump parameters as training data;

[0012] building a neural network, and training the neural network using the training data;

[0013] The specific steps of the neural network application stage include:

[0014] The selected erbium-doped fiber amplifier pump power is used to subtract the erbium-doped fiber amplifier gain from the total gain of the hybrid amplifier to obtain the gain required by the Raman amplifier for each signal light;

[0015] The gain required by the Raman amplifier is used as the input of the neural network, and the Raman pump parameter value is derived.

[0016] As a preferred example, the neural network is a fully connected neural network.

[0017] As a preferred example, the number of Raman pumps is three, and the Raman pump parameters include pump power and wavelength.

[0018] As a preferred example, the erbium-doped fiber amplifier pump is a 980 nm or 1480 nm laser.

[0019] (III) Advantages

[0020] The application discloses a high-gain flat wideband erbium-doped fiber and Raman hybrid amplifier design method based on a neural network. Compared with the prior art, the application realizes fast and automatic selection of the pump parameters of the hybrid amplifier by using a neural network, and ensures that the hybrid amplifier provides a high-accuracy flat wideband gain spectrum. BRIEF DESCRIPTION OF DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0022] Figure 1 is a structure diagram of an erbium-doped fiber and Raman hybrid amplifier;

[0023] Figure 2 is a flow chart of an erbium-doped fiber and Raman hybrid amplifier design method based on a neural network;

[0024] Figure 3 is a structure diagram of a neural network model;

[0025] Figure 4 is a probability density function (pdf) result diagram of the root mean square error (RMSE) and the maximum error (Error max ) of the predicted output power of signal light and the target output power.

[0026] Figure 5 is a plot of the signal light predicted output power versus target output power.

[0027] BRIEF DESCRIPTION OF DRAWINGS

[0028] 1: Erbium-doped fiber amplifier

[0029] 2: Raman amplifier

[0030] 3: Erbium-doped fiber

[0031] 4: Wavelength division multiplexer

[0032] 5: Pump of Erbium-doped fiber amplifier

[0033] 6: Single-mode optical fiber

[0034] 7: Wavelength division multiplexer

[0035] 8: First pump of Raman amplifier

[0036] 9: Second pump of Raman amplifier

[0037] 10: Third pump of Raman amplifier DETAILED DESCRIPTION

[0038] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0039] The purpose of the present invention is to provide a design method for a high-gain flat broadband erbium-doped fiber and Raman hybrid amplifier based on a neural network, so as to ensure that the hybrid amplifier outputs a broadband flat gain spectrum.

[0040] In order to make the above-mentioned objects, features and advantages of the present invention more clearly understood, the technical solutions of the present invention are further described in detail below with reference to the accompanying drawings and specific embodiments:

[0041] like Figure 1 As shown, the hybrid amplifier described in this embodiment includes an erbium-doped fiber amplifier part and a Raman amplifier part. The output end of the erbium-doped fiber amplifier is connected to the input end of the Raman amplifier. The erbium-doped fiber amplifier is composed of an erbium-doped fiber, a wavelength division multiplexer, and an erbium-doped fiber amplifier pump. The Raman amplifier is composed of a single-mode fiber, a wavelength division multiplexer, a Raman amplifier pump 1, a Raman amplifier 2, and a Raman amplifier 3.

[0042] like Figure 2 As shown, the neural network-based high-gain flat broadband erbium-doped fiber and Raman hybrid amplifier design method described in this embodiment includes a neural network training stage and a neural network application stage:

[0043] The specific steps in the neural network training phase include:

[0044] Step 1.1: Set the target overall gain of the hybrid amplifier.

[0045] Step 1.2: Select different erbium-doped fiber amplifier pump powers and adjust Raman pump parameters to ensure that the hybrid amplifier achieves the target total gain.

[0046] Step 1.3: Record the gain provided by the Raman amplifier for each signal light channel and the corresponding Raman pump parameters as training data.

[0047] Step 1.4: Build a neural network and train the neural network using the training data.

[0048] The specific steps in the neural network application stage include:

[0049] Step 2.1: Select the pump power of the erbium-doped fiber amplifier, and subtract the erbium-doped fiber amplifier gain from the target total gain of the hybrid amplifier to obtain the gain required to be provided by the Raman amplifier for each signal light.

[0050] Step 2.2: Using the gain required to be provided by the Raman amplifier as an input of the neural network to derive the pump parameter value of the Raman amplifier.

[0051] like Figure 3As shown, the neural network described in this embodiment is composed of an input layer, a hidden layer and an output layer, wherein the input layer contains n neurons, G1...G n Represents the gain value of n different wavelength signal lights; the output layer contains 6 neurons, P p1 …P p3 Represents three pump light power values, λ p1 …λ p3 represents three pump light wavelength values; the number of hidden layers is 3 and each layer contains 200 neurons; the input layer, hidden layer and output layer of the neural network are connected in sequence using the S-type transfer function sigmoid.

[0052] This embodiment also provides a specific example, using the hybrid amplifier to amplify 40 signal lights with a frequency of 192.1THz to 196THz, a frequency interval of 0.1THz, and an input power of -40dBm. The target output gain of the signal lights after amplification by the hybrid amplifier is 0dBm (i.e., 1mW). 500 sets of different pump power combinations that enable the hybrid amplifier to achieve the target gain are selected as training data, and another 100 sets of pump power combinations are selected as verification data to test the accuracy of the method. The verification results are used as follows: Figure 4 The root mean square error (RMSE) and maximum error (Error max ) is measured by the probability density function (pdf), and the results show that the mean and variance of RMSE are 0.233mW and 0.004mW respectively, and the Error max The mean and variance are 0.499mW and 0.007mW respectively.

[0053] like Figure 5 The following graphs compare the predicted signal light output gain and the target gain obtained using the design method. The solid line represents the target signal light output power, and the dashed line represents the predicted signal light output power. The corresponding erbium-doped fiber amplifier pump power is 30mW; the Raman pump powers are 218.5mW, 442.6mW, and 125.7mW, respectively, and the Raman pump wavelengths are 1441.4nm, 1467.3nm, and 1501.6nm. The graphs show that the predicted signal light output power has a small error compared to the target output power.

[0054] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A neural network-based high-gain flat wideband erbium-doped fiber and Raman hybrid amplifier design method, comprising a training and application stage of a neural network, characterized in that: the specific steps of the neural network training stage are: step 1.1, setting the target total gain of the hybrid amplifier; step 1.2, selecting different erbium-doped fiber amplifier pump powers, adjusting the Raman pump parameters to ensure that the hybrid amplifier reaches the target total gain; step 1.3, recording the gain provided by the Raman amplifier for each signal light and the corresponding Raman pump parameters as training data; step 1.4, building a neural network and training the neural network using the training data; the specific steps of the neural network application stage are: step 2.1, selecting the erbium-doped fiber amplifier pump power, and using the target total gain of the hybrid amplifier minus the erbium-doped fiber amplifier gain to obtain the gain required by the Raman amplifier for each signal light; step 2.2, taking the gain required by the Raman amplifier as the input of the neural network, and deriving the Raman amplifier pump parameter value. The neural network is a fully connected neural network. The number of Raman pumps is 3, and the Raman pump parameters include pump power and wavelength. The erbium-doped fiber amplifier pump is a 980 nm or 1480 nm laser. ​ ​ ​ ​ ​ 2. The neural network based high gain flat wideband erbium doped fiber and Raman hybrid amplifier design method of claim 1, wherein ​ 3. The neural network based high gain flat wideband erbium doped fiber and Raman hybrid amplifier design method of claim 1, wherein, ​ 4. The neural network based high gain flat wideband erbium doped fiber and Raman hybrid amplifier design method of claim 1, wherein, ​

Images