A channel noise compensation method based on photon number resolvable detector

Abstract

The present application relates to the field of coherent optical communication technology, and discloses a channel noise compensation method based on a photon number distinguishable detector. The method includes the following steps: a transmitter generates a signal light pulse with a known coding format and an unknown coding state; a receiver prepares local oscillator light pulses with the same coding format and different coding states for signal light pulses in different time partitions that have been attenuated after transmission through a long-distance noisy channel; a shift operation is performed on the signal light pulses in the same coding state in different time partitions by interference multiple times; a photon number distinguishable detector is used to detect the light pulses after interference to obtain the photon number distribution result of the interference light pulse; and deep learning is combined to guide the preparation of the coding state of the next local oscillator light by matching the photon number distribution result each time with the coding state of the corresponding prepared local oscillator light. The use of this method will reduce the detection bit error rate of the coherent optical communication system for weak intensity and unknown coding state signal light after transmission through a long-distance noisy channel.

Description

Technical Field

[0001] The present application relates to the field of coherent optical communication technology, and more specifically, to a method and device for compensating channel noise based on a photon number resolvable detector. Background Art

[0002] As transmission distances gradually increase, the coherent light received by the detector as a signal (hereinafter referred to as signal light) continues to attenuate. When it attenuates to photon-level intensity, coherent optical communication systems using traditional coherent detection technologies (homodyne and heterodyne detection) will be limited by the sensitivity of commonly used detectors (such as balanced detectors) and shot noise, making it difficult for the detection bit error rate of the entire coherent optical communication system to fall below the standard quantum limit (SQL) under the corresponding coding format.

[0003] The quantum-enhanced detection receiver based on single-photon detectors prepares a local oscillator light pulse with the same frequency, spatial mode, polarization direction, coding format, and known coding phase as the weak signal light pulse (weak means that the average number of photons in a single pulse is less than 10, which can be detected and distinguished by a photon number-resolvable detector) as a displacement operator, and completes an interference-based displacement operation with the weak signal light to be measured: when the prepared local oscillator light pulse is completely consistent with the coding state of the signal light pulse, the weak signal light can be moved from the coherent state to the photon number state of 0 (the single-photon detector fails to detect the photon); if they are inconsistent, the single-photon detector can detect that there are still photons remaining after the displacement operation. Therefore, based on the local oscillator light pulse of known coding state prepared at this time and the detection result (whether there are photons remaining), it is possible to determine whether the coding state of the signal light detected and received in a single measurement is consistent with the coding state of the local oscillator light pulse. Subsequently, adaptive feedback can be used to complete the preparation of local oscillator light pulses of different coding states under the same coding format, and the coding state judgment of the signal light of the same coding state can be completed through multiple measurements, achieving a bit error rate lower than that of the coherent optical communication system, and even lower than the standard quantum limit.

[0004] However, quantum-enhanced detection receivers based on single-photon detectors can only form a judgment on the coding state of weak signal light by detecting the presence or absence of photons after the shift operation. The noise accumulated in the actual channel under long-distance transmission will affect the detection results (especially the result of no photons remaining when the coding state of the prepared local oscillator light is consistent with that of the weak signal light). A photon-number resolving detector (PNRD) that can distinguish the number of photons of detected photon-level weak signals can achieve "photon number resolving" based on the detection result of "the presence or absence of photons". It can obtain more information from the detection results and apply it to multiple detections of subsequent adaptive feedback, thereby achieving a lower bit error rate judgment of the coding state of weak signal light under the actual channel noise background.

[0005] In addition to traditional estimation algorithms that can model noise in the transmission channel and adaptively reduce the impact on the effective determination of the coding state of weak signal light through compensation-prepared coded state local oscillator light, with the increase in data volume and computing power, deep learning has also been applied to noise compensation. The patent named "A channel estimation method based on deep learning and data pilot assistance" and the application number CN202110195357.X has also applied deep learning to communication systems to achieve compensation for phase noise. The data set generated by phase noise is trained using the stochastic gradient descent method, and the weighted function optimization model is calculated to compensate for and improve the performance deterioration of the communication system caused by noise.

[0006] However, in the field of quantum-enhanced coherent optical communication systems, deep learning algorithms have not yet been applied to compensate for the intensity / phase noise in the channel noise, and cannot effectively reduce the impact of channel noise on the detection of weak signal light coding states, nor can they effectively reduce the bit error rate of the coherent optical communication system when detecting weak signal light. Summary of the Invention

[0007] In order to solve the above problems, the present invention provides a method and device for compensating channel noise based on a photon number resolvable detector, which will reduce the impact of channel noise on the detection of weak signal light and reduce the bit error rate of detecting weak signal light.

[0008] To achieve the above object, according to a first aspect of the present invention, a method for compensating channel noise based on a photon number resolvable detector is provided, the method comprising:

[0009] The transmitting end generates a signal light pulse with a known coding format and an unknown coding state, and divides the signal light pulse into time partitions within a time period corresponding to each pulse width of the signal light pulse;

[0010] The receiving end prepares local oscillator optical pulses with the same coding format but different coding states for the signal optical pulses in different time partitions after being attenuated by long-distance noisy channels.

[0011] Using the local oscillator light pulse as a shift operator, within a period corresponding to a pulse width, the same coded state signal light pulse is shifted multiple times in different time partitions by interference, outputting an interference light pulse.

[0012] The interfering light pulse is detected by a photon number resolvable detector to obtain the photon number distribution result of the interfering light pulse;

[0013] By matching the photon number distribution results each time with the corresponding coded state of the prepared local oscillator light, combined with deep learning network feedback, the offset of the coded state of the local oscillator light relative to the coded state of the signal light under the influence of channel noise is calculated, and the next coded state of the local oscillator light is prepared by using intensity and phase modulation to compensate for the noise based on the standard coding format.

[0014] Furthermore, by matching the photon number distribution results each time with the corresponding LO light coding state, combined with the deep learning network feedback to calculate the offset of the LO light coding state relative to the signal light coding state under the influence of channel noise, and guiding the next LO light to use intensity and phase modulation to compensate for the noise in the standard coding format, including the photon number distribution results. and the corresponding prepared LO light encoding state (|β k >) Form a data set The dataset A channel noise compensation model is constructed using the data as input to the deep learning network. The channel noise compensation model is used to calculate the impact of channel noise on the coded state of the signal light. Based on the impact of the channel noise on the coded state of the signal light, intensity and phase modulation compensation for the coded state of the local oscillator light are calculated to form feedback instructions. Based on the feedback instructions, compensation is performed on the next coded state preparation of the local oscillator light until the average value of the photon number distribution result is 0.

[0015] Furthermore, the channel noise compensation method based on the photon number resolvable detector also includes that a pulse width of the signal light pulse is T, a time partition of the signal light pulse is M, and the pulse width of the prepared local oscillator light pulse is less than T / M.

[0016] Furthermore, the above-mentioned channel noise compensation method based on the photon number resolvable detector also includes the photon number distribution result of the interference light pulse conforming to the Gaussian distribution. (μ, σ 2 ), μ represents the average value of the photon number distribution, σ 2 Represents the fluctuation of the photon number distribution around the average value.

[0017] Furthermore, the coding state preparation of the next local oscillator light is compensated according to the feedback instruction until the average value of the photon number distribution result is 0, including when μ=0, determining that the coding state of the signal light pulse is consistent with the coding state of the local oscillator light pulse; when μ≠0, according to the photon number distribution result of the interference light pulse (μ, σ 2 ), prepare new coded state local oscillator light pulses under the same coding format, and combine μ and σ 2 The numerical value of and the channel noise compensation model are used to calculate the intensity and phase noise caused to the signal light pulse during channel transmission. On the basis of preparing a new standard coded state local oscillator light under the coding format, the modulation compensation of intensity and phase is completed.

[0018] According to a second aspect of the present invention, a channel noise compensation device based on a photon number resolvable detector is also provided, which includes a signal light generation module for generating a signal light pulse with a known coding format and an unknown coding state; a local oscillator light generation module for generating a local oscillator light pulse with the same coding format as the signal light pulse and an adjustable coding state; an interference module for interfering the signal light pulse and the local oscillator light pulse to form a displacement operation of the local oscillator light pulse on the signal light pulse, and output an interference light pulse; a photon number resolvable detection module for detecting the interference light pulse and obtaining a photon number distribution result of the interference light pulse; and a data acquisition and adaptive feedback module for calculating, based on the photon number distribution result, the offset of the coding state of the local oscillator light relative to the coding state of the signal light under the influence of channel noise, and providing feedback guidance for preparing a new coding state of the local oscillator light pulse with the same coding format as the signal light pulse and after noise compensation.

[0019] Furthermore, the signal light generation module includes a narrow linewidth continuous light fiber laser, an acousto-optic modulator, an electro-optic phase modulator, an adjustable optical attenuator, and a long-distance transmission optical fiber connected in sequence; the local oscillator light generation module includes a narrow linewidth continuous light fiber laser, an acousto-optic modulator, an electro-optic phase modulator, an electro-optic intensity modulator, and a compensation adjustment module connected in sequence. The compensation adjustment module includes an optical fiber polarization controller and an optical fiber delay line, which are used to adjust the local oscillator light pulse and the signal light pulse to maintain the same polarization and consistent pulse leading edge.

[0020] Furthermore, the interference module includes a fiber coupler, and the photon number resolvable detection module includes a photon number resolvable detector.

[0021] Furthermore, the data acquisition and adaptive feedback module includes an FPGA, a host computer communicating with the FPGA, and a digital-to-analog conversion module; the FPGA is used to synchronously complete the recording of the prepared coding state of the local oscillator light and the matching photon number distribution results, form a data set, and upload it to the host computer; the host computer is used to input the data set into the channel noise compensation model to calculate the impact of channel noise, and output the output result as a feedback instruction to the FPGA; the digital-to-analog conversion module is used to convert the feedback instruction into an electrical signal, complete the modulation of the electro-optical phase modulator and the electro-optical intensity modulator in the local oscillator light generation module, and realize compensation for the modulation of the coding state of the local oscillator light / channel noise impact next time.

[0022] Furthermore, the training method of the channel noise compensation model includes step I, taking the photon number distribution matrix and the intensity and phase data of the local oscillator light in the coded state corresponding to the detected photon number distribution as sample data; step II, establishing a channel noise compensation model based on the simplified U-Net network structure of deep learning, using the simplified U-Net network structure as the main structure to sequentially extract shallow features using 32×3×3 convolution layers, sequentially extract middle-level features using 64×3×3 convolution layers, and sequentially extract high-level features using 128×3×3 convolution layers; inputting the high-level features extracted by the convolutional neural network into the simplified U-Net decoding network, and obtaining the second decoding features and the first decoding features in turn after 64×2×2 and 32×2×2 upsampling convolution layers, and performing a splicing operation on the second decoding features and the middle-level features during the decoding process. The first decoding network output feature is obtained, and the first decoding feature and the shallow feature are spliced ​​to obtain the second decoding network output feature; the second decoding network output feature is flattened to obtain one-dimensional feature data and input into the fully connected layer, where the size of the fully connected layer is (32, 2), and the output result of the fully connected layer is the intensity and phase data of the coded state local oscillation light; step III, initialize the channel noise compensation algorithm model parameters, optimizer, and learning rate; step IV, input the sample data into the channel noise compensation model for training, and determine the loss function; step V, backpropagate the error of the channel noise compensation model, adjust the parameters of the model, and optimize the channel noise compensation model; return to step I, iteratively train the channel noise compensation model until the loss function converges, and obtain the final channel noise compensation model after training is completed.

[0023] In general, the above technical solutions conceived by the present invention can achieve the following beneficial effects compared with the prior art:

[0024] (1) The present invention provides a channel noise compensation method based on a photon number resolvable detector. In the process of transmission, signal light of unknown coding state is interfered by complex channel noise, and the signal strength becomes weak after long-distance transmission. At the receiving end, a local oscillator light with the same coding format and known coding state is prepared to perform interference displacement operation on the signal light with weak intensity and coding state affected by channel noise. Then, the photon number resolvable detector is used to detect the photon number distribution of the light beam after the interference between the signal light and the local oscillator light. The detected photon number distribution results and the matching coding state of the local oscillator light are collected to establish a data set. Combined with the deep learning method, the complex noise in the channel (mainly considering phase noise and intensity noise) is compensated for by adaptive feedback into the preparation of the new coding state of the local oscillator light, so that the coding state of the local oscillator light can better adapt to the weak signal light under the influence of channel noise, thereby achieving better and more accurate detection results of the photon number resolvable detector. This method can not only effectively reduce the impact of channel noise on the detection of weak signal light coding states, but also provide more accurate judgment of weak signal light coding states based on the detection results of photon number-resolved detectors, thereby achieving a lower bit error rate in coherent optical communication systems under quantum-enhanced detection, and can surpass the standard quantum limit.

[0025] (2) The present invention provides a channel noise compensation method based on a photon number resolvable detector. Under the condition of weak incident signal light intensity, the compensation algorithm can be based on the displacement operation of interference, the detection of the photon number resolvable detector, and the simplified U-Net neural network structure in deep learning. While being closer to actual long-distance channel transmission and introducing more noise analysis, it can achieve intelligent compensation for channel noise and ensure that the detection bit error rate of the coherent optical communication system is lower than the standard quantum limit. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0027] Figure 1 A flow chart of a method for compensating channel noise based on a photon number resolvable detector provided in an embodiment of the present application;

[0028] Figure 2 A schematic diagram of the structure of a device for compensating channel noise based on a photon number resolvable detector provided in an embodiment of the present application;

[0029] Figure 3 A schematic diagram of the internal structure of each module of a device for compensating channel noise of a photon number-resolvable detector provided in an embodiment of the present application;

[0030] Figure 4 The standard QPSK format phase-coded modulation provided in the embodiment of the present application and the corresponding phase space representation diagram of the QPSK coding state under the influence of intensity noise and phase noise after channel transmission;

[0031] Figure 5 Phase space characterization of the local oscillator light displacement operation (interference) on the signal light provided in the embodiment of the present application and the photon number distribution diagram of the light pulse after interference based on PNRD measurement;

[0032] Figure 6 Schematic diagram of the QPSK-coded signal light pulse in the time domain, the signal light pulse after channel transmission, and the local oscillator light preparation / compensation detection test phase corresponding to the signal light pulse partition provided in an embodiment of the present application;

[0033] Figure 7 The coding state prepared for detecting the QPSK coded signal provided in the embodiment of the present application |β k >Photon number distribution of matching interference light pulse detection results The records will form a data set Schematic diagram of a convolutional neural network that uses a matrix to compensate for channel noise. DETAILED DESCRIPTION

[0034] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely for the purpose of explaining the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.

[0035] The terms "first," "second," "third," and the like in the specification and claims of this application and the accompanying drawings are used to distinguish between different objects, not to describe a particular order. Furthermore, the terms "including," "having," and any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus comprising a series of steps or elements is not limited to the listed steps or elements, but may optionally include steps or elements not listed, or may optionally include other steps or elements inherent to the process, method, product, or apparatus.

[0036] like Figure 1 As shown, a method for compensating channel noise based on a photon number resolvable detector is provided, comprising the following steps:

[0037] In step 101 , a transmitting end generates a signal light pulse with a known coding format and an unknown coding state, and time-partitions the signal light pulse within a time period corresponding to each pulse width of the signal light pulse.

[0038] The signal light pulse is a pulse form of coherent light serving as a signal. The transmitting end may include a narrow linewidth continuous light fiber laser, an acousto-optic modulator, an electro-optic phase modulator, and a variable optical attenuator connected in sequence.

[0039] In step 102, the receiving end prepares local oscillator optical pulses having the same coding format but different coding states for the signal optical pulses in different time partitions that have been attenuated after being transmitted through a long-distance noisy channel.

[0040] The local oscillator (LO) optical pulse and the signal optical pulse have the same coding format but different coding states, and both the LO and signal optical pulses are coherent. The receiving end may include a narrow-linewidth continuous light fiber laser, an acousto-optic modulator, an electro-optic phase modulator, an electro-optic intensity modulator, and a compensation adjustment module, connected in sequence.

[0041] In step 103 , the local oscillator light pulse is used as a shift operator to perform shift operations on the signal light pulse of the same coding state in different time partitions by interference multiple times within a period corresponding to a pulse width, and output an interference light pulse.

[0042] Step 104 : Detect the interference light pulses using a photon number resolvable detector to obtain a photon number distribution result of the interference light pulses.

[0043] Step 105 , by matching the photon number distribution result each time with the corresponding coded state of the prepared local oscillator light, combined with the deep learning network feedback, the offset of the coded state of the local oscillator light relative to the coded state of the signal light under the influence of channel noise is calculated, and the next coded state of the local oscillator light is prepared by using intensity and phase modulation to compensate for the noise based on the standard coding format.

[0044] The above-mentioned channel noise compensation method based on the photon number resolvable detector relies on a deep learning algorithm and combines the influence of the channel noise (mainly including the phase and intensity noise in the transmission process) reflected by the photon number distribution results of the interference light pulse detected by the photon number resolvable detector on the coding state of the signal light. The adaptive parameter adjustment fits the noise model and realizes low bit error rate detection and discrimination of weak signal light by preparing and compensating the coding state of the local oscillator light.

[0045] In one embodiment, Figure 2 As shown, a compensation device for channel noise of a photon number resolvable detector includes:

[0046] A weak coherent signal light generation module (also referred to as a signal light generation module) is used to generate a signal light pulse, wherein the coding format of the signal light pulse is known (taking the QPSK coding format as an example) and the coding state is unknown;

[0047] A local oscillator light generation module, used to generate a local oscillator light pulse having the same coding format as the signal light pulse;

[0048] An interference module is used to interfere with the signal light pulse and the local oscillator light pulse, thereby performing a displacement operation of the local oscillator light pulse on the signal light pulse and outputting an interfered light beam (also called an interference light pulse);

[0049] The photon number resolvable detection module is used to detect the interfering light pulse and obtain the detection result of its photon number distribution (referred to as the photon number distribution result);

[0050] The data acquisition and adaptive feedback module is used to collect the photon number distribution results of the interference light pulse and the corresponding local oscillation light coding state, and upload them to the host computer. The host computer uses a deep learning-based neural network to analyze the data and, based on the analysis results, controls the local oscillation light generation module to re-encode the local oscillation light pulse and compensate for channel transmission noise.

[0051] Specifically, within the time period of a signal light pulse, the host computer continuously modulates the coding state of the prepared local oscillator light, interferes with the signal light, and forms a corresponding interference light pulse. The interference light pulse is then combined to form a change in the coding state of the next prepared local oscillator light, forming a multi-round adaptive feedback cycle.

[0052] The channel noise compensation device based on a photon number-resolvable detector provided in this embodiment can, when the incident signal light pulse is of weak intensity and unknown coding state, be based on the displacement operation of the interference module, the photon number-resolvable detection module, and the deep learning algorithm. While being closer to actual long-distance channel transmission and introducing more noise analysis, it can achieve intelligent compensation for channel noise and ensure a quantum-enhanced detection bit error rate below the standard quantum limit.

[0053] In one embodiment, Figure 3As shown, in a channel noise compensation device based on a photon number resolvable detector, the signal light generation module includes a narrow-linewidth continuous light fiber laser 1, an acousto-optic modulator 2, an electro-optic phase modulator 3, an adjustable optical attenuator 4, and a long section of optical fiber 12. The components and equipment are all connected via a flange. The narrow-linewidth continuous light fiber laser 1 is used to generate continuous linearly polarized light, the acousto-optic modulator 2 is used to modulate the continuous linearly polarized light into optical pulses with a pulse width T (the pulse width T is generally tens to hundreds of μs, divided into M equal regions in the time domain), the electro-optic phase modulator 3 is used for phase encoding under the modulation format (taking the QPSK encoding format as an example), the adjustable optical attenuator 4 is used to adjust the intensity of the encoded signal light pulse (also called the coded state signal light pulse), and the long section of optical fiber 12 is used to transmit the coded state signal light pulse. After the coded state signal light pulse is transmitted over a long distance, a weak signal light pulse is obtained.

[0054] Figure 4 This is the representation of the QPSK coding format in the phase space. The horizontal and vertical coordinates are X and P components respectively, and the four quadrants represent the coherent states |α0>, |α1>, |α2>, and |α3> under the QPSK coding format. Figure 4 (a) represents the standard QPSK coding format of the signal light pulse, (b) represents the influence of intensity noise on the standard QPSK coding state after channel transmission, and (c) represents the influence of phase noise on the standard QPSK coding state after channel transmission. As the signal light pulse is transmitted through the optical fiber 12 in the channel, in addition to the attenuation caused by the optical fiber, it is also affected by intensity noise and phase noise. The evolution of the corresponding QPSK coding state is shown as follows: Figure 4 Phase space representation in (b) and (c).

[0055] The local oscillator light generation module includes a narrow linewidth continuous light fiber laser 1, an acousto-optic modulator 2, an electro-optic phase modulator 3, an electro-optic intensity modulator 5, and a compensation adjustment module 6. The narrow linewidth continuous light fiber laser 1 is used to generate continuous linearly polarized light, and the acousto-optic modulator 2 is used to modulate the continuous linearly polarized light into a light pulse with a pulse width of T0, where the pulse width satisfies The electro-optical phase modulator 3 and the electro-optical intensity modulator 5 are used to complete the encoding and intensity adjustment of the optical pulse. The encoding format is consistent with the signal optical pulse (also the QPSK encoding format. The specific correspondence between the local oscillator optical pulse generated by the modulation and the signal optical pulse in the time domain is as follows: Figure 6 (c)(d) shows), the compensation adjustment module 6 includes a fiber polarization controller and a fiber delay line, which are used to adjust the polarization state and pulse leading edge of the local oscillator light pulse to be consistent with the signal light pulse, ensuring high interference in the interference module.

[0056] The interference module includes a fiber coupler 7, which has two input ends. One input end is used to receive the signal light pulse whose intensity becomes weak and is affected by channel noise after being transmitted through the long section of optical fiber 12. The other input end is used to receive the local oscillator light pulse. The output end has one output end, which is used to output the interference light pulse. The coupling ratio of the fiber coupler 7 can be 99:1. The fiber coupler 7 is used to use the local oscillator light pulse to convert the weak signal light pulse from a coherent state to a photon number state through the displacement operation of interference. The displacement operation of interference is represented by the phase space of QPSK encoding as follows Figure 5 As shown in (a)(b).

[0057] The photon number resolvable detection module includes a photon number resolvable detector (PNRD) 8, which is used to measure the interference light pulse and output a level signal corresponding to the photon number distribution through its own electrical amplifier module. Figure 5 The displacement operation (interference) process shown in (a) and (b) shows the different coding states corresponding to the interference light pulses. The photon number distribution results obtained by the photon number resolvable detector 8 are as follows: Figure 5 As shown in (c): When the shift operation is performed to the zero state of the photon number, the photon number distribution is a Gaussian distribution with an average value of 0 (μ = 0); when the shift operation is not performed to the zero state of the photon number, the photon number distribution is related to the intensity of the interference light pulse and appears as a Gaussian distribution with different average values. (μ, σ 2 ), μ reflects the average value of the photon number distribution when PNRD receives a certain light intensity light pulse (also the photon number value corresponding to the maximum peak of the detected photon distribution), σ 2 It reflects the fluctuation of the photon number distribution around the average value when PNRD receives a light pulse of a certain light intensity (the detection performance of PNRD will vary when facing light pulses of different light intensities).

[0058] The data acquisition and adaptive feedback module includes a field programmable gate array 9 (FPGA), a host computer 10 communicating with the FPGA 9, and a digital-to-analog conversion module 11. The FPGA 9 is used to collect the photon number distribution results detected by the photon number resolvable detector 8. And the photon number distribution results The coded state of the local oscillator light |β k >Forming the dataset The data set is uploaded to the host computer 10, and the host computer 10 As the input data of the simplified U-Net neural network based on deep learning, a channel noise compensation model is constructed; through the channel noise compensation model, the compensation corresponding to the selection of the coding state preparation of the next local oscillator light / channel noise influence is calculated and judged, and a feedback instruction is formed, which is output to FPGA9.

[0059] The digital-to-analog conversion module 11 is used to convert the feedback instruction into a level signal of modulation phase and intensity, which is loaded into the electro-optical phase modulator 3 and the electro-optical intensity modulator 5 respectively to complete the compensation of the modulation / channel noise influence of the coding state of the next local oscillator light.

[0060] In this embodiment, the pulse width T of the prepared signal light pulse is regarded as a period. In each period, the preparation of local oscillator light pulses with the same coding format but different coding states is repeated M times, and the influence of channel noise is compensated to complete the interference of M local oscillator light pulses on the signal light pulse. The standard coding format of the weak signal light pulse (taking QPSK as an example) is represented by the phase space as follows Figure 6 As shown in (a), after being transmitted over a long distance through an optical fiber and affected by channel noise (intensity noise, phase noise), the coding state of the signal light pulse changes as follows: Figure 6 (b) shown.

[0061] The relationship between the coding state of the signal light pulse (taking the QPSK coding format as an example) and the preparation of the coding state of the local oscillator light and noise compensation in the time domain is as follows: Figure 6 (c) (d) As shown. The FPGA9 collects the photon number distribution results of each interfering light pulse detected by the photon number resolution detector 8, as shown in FIG. Figure 6 As shown in (c)(d), in the first stage (Phase I), the host computer 10 uses a deep learning neural network to complete the selection of the coding state preparation of the next local oscillator light in the cycle, and completes all possible standard coding state tests under the QPSK standard coding format one by one through polling. The local oscillator light pulse is modulated to complete multiple interference measurements, and the most likely coding state is determined by the algorithm based on the photon number distribution result of the interference light pulse; then, in the second stage (Phase II), based on the most likely coding state of the local oscillator light and the photon number distribution result of the interference light pulse, the host computer 10 uses a deep learning neural network to calculate the impact of intensity / phase noise in channel transmission on the coding state of the signal light, and feeds it back to the FPGA 9, and with the help of the digital-to-analog conversion module 11, intensity / phase modulation compensation based on the standard coding format is completed.

[0062] This embodiment provides a channel noise compensation device based on a photon number resolvable detector, which has the following beneficial effects:

[0063] In the photon number-resolvable detection module, a photon number-resolvable detector (PNRD) is used to replace the single photon detector (SPD) or superconducting nanowire single photon detector (SN-SPD) commonly used in quantum-enhanced detection receivers. After generating coded local oscillator light with the same coding format and converting the signal light from a coherent state to a photon number state through a shift operation (interference), traditional SPDs and SN-SPDs can only distinguish between the "0" state (absence of photons) and non-"0" states (presence of photons) in the detection and reception system. Therefore, in coding formats such as QPSK with more than two coding states, the presence of non-"0" states makes it difficult to effectively distinguish between the inconsistent coded states of the LO light and the signal light when a non-"0" state is detected. However, the photon number-resolvable detector (PNRD) can provide a more detailed distinction between the photon number distribution of the non-"0" state interference light detection results, effectively reducing the error rate of detection results. In addition, when the signal light has difficulty maintaining a stable standard coding state under the influence of channel noise, it is difficult for the "0" state to appear when interfering with the prepared standard coding state local oscillator light pulse. At this time, PNRD can make a good distinction between all non-"0" state detection results, providing sufficient judgment basis for the preparation of subsequent coding state local oscillator light pulses.

[0064] The following further describes a method for compensating for weak signal transmission channel noise based on a photon number resolvable detector. The method includes the following steps:

[0065] In step 1, the continuous light generated by the narrow-linewidth continuous light fiber laser 1 is output through the acousto-optic modulator 2 and the electro-optic phase modulator 3 to prepare a coded signal light pulse of a certain coding format (taking the QPSK coding format as an example). After passing through the adjustable optical attenuator 4 and the long section of optical fiber 12, a weak QPSK-coded signal light pulse is generated after long-distance optical fiber transmission. The devices and equipment are all connected through flanges.

[0066] In step 2, another narrow-linewidth continuous-light fiber laser (1) is connected via a flange to an acousto-optic modulator (2), an electro-optic phase modulator (3), and an electro-optic intensity modulator (5). This generates coded-state local oscillator (LO) pulses with the same encoding format as the signal light. A compensation adjustment module (6), consisting of a fiber delay line and a fiber polarization controller, controls the pulse leading edge and polarization state to maintain consistency with the signal light pulse after transmission through the long-distance fiber channel. The level signal corresponding to the coded state generated by first modulating the LO light through the electro-optic phase modulator (3) is output and recorded by the FPGA.

[0067] The narrow-linewidth continuous light fiber laser 1 generating the local oscillator light in step 2 and the narrow-linewidth continuous light fiber laser 1 generating the signal light in step 1 are set by the operating software to maintain the output of linearly polarized light with the same frequency (same wavelength), same polarization, and same spatial mode.

[0068] In step 3, the weak coded signal light pulse, which has been transmitted through a long-distance optical fiber channel and affected by channel noise, is connected to the two input ends of the optical fiber coupler 7 at the same time as the prepared local oscillator light pulse with the same coding format. Through the interference displacement operation, the coded signal light pulse is converted into a photon digital state output.

[0069] Step 4: Input the interfered light pulse into the photon number resolvable detector (PNRD) 8 to complete the photon number distribution. (μ, σ 2 ) measurement.

[0070] Step 5: The detected photon number distribution result and the corresponding coding state of the local oscillator light are collected and recorded by FPGA9 and uploaded to the host computer 10. The feedback instruction is output through the digital-to-analog conversion module in a polling manner (taking QPSK as an example, the four coding states are prepared and tested one by one), and the electro-optical phase modulator 3 is adjusted to complete the preparation of the coding state of the next local oscillator light. Through multiple interferences between the local oscillator light of each coding state under the coding format and the signal light of the same coding state, the host computer 10 combines the photon number distribution obtained by PNRD 8 to screen and obtain the photon number distribution. (μ, σ 2 ) is the coded state of the local oscillator light corresponding to the detection result closest to μ = 0.

[0071] Step 6: Keep the coded state of the local oscillator light selected in step 5, and interfere with the signal light again. The photon number distribution detection result of the light pulse after interference and the coded state of the local oscillator light form a data set. The data is uploaded from FPGA 9 to host computer 10, and the influence of channel noise on the coding state of the signal light is calculated by combining the channel noise compensation network. The coding state of the local oscillator light |β is obtained through the output result calculated by the channel noise compensation model. k > intensity and phase data, determines the selection of the next LO light coding state preparation / channel noise compensation, and generates feedback instructions for output to FPGA 9. Finally, the channel noise compensation network is combined to determine the phase / intensity compensation of the LO light coding state, which is output as feedback instructions to FPGA 9. The digital-to-analog conversion module 11 modulates the phase, and the electro-optical intensity modulator 5 compensates for the next LO light preparation until the detected photon number distribution satisfies μ = 0.

[0072] The electrical signal output after conversion by the digital-to-analog conversion module 11 matches the parameters of the RF input end of the electro-optical phase modulator 3. The duration of the single feedback process in steps 5 and 6 (the feedback process includes PNRD measurement, FPGA data acquisition and upload, host computer processing and calculation, and feedback instructions sent back to the FPGA, and the digital-to-analog conversion module controls the electro-optical phase modulator and the electro-optical intensity modulator to complete the new coding state preparation / noise compensation to generate a new coding state local oscillator light output) is equal to T / M. The pulse width of the local oscillator light is less than T / M, where M is the number of time domain partitions of the coding signal light pulse, and its value is generally greater than the number of codable states of the selected coding format. For example, for the QPSK coding format, M>4 is satisfied. When executing step 5, the first 4 partitions of the M partitions in the time domain are used to poll and traverse to clarify the most likely coding state under the QPSK coding format, corresponding to Figure 6 (d) Phase I; the remaining (M-4) partitions execute step 6, and based on the setting of the most likely coding state of the local oscillator light, the influence of the channel noise on the signal light is calculated through the channel noise compensation algorithm to form a noise compensation feedback for the coding state of the local oscillator light, corresponding to Figure 6 (d) Phase II.

[0073] In step 7, the pulse width T of the prepared coded-state signal light is considered a cycle. For each coded-state signal light pulse, steps 1 to 6 are repeated within the cycle. After completing multiple measurement feedbacks within each cycle, the coded state of the signal light is compared with the coded state of the finally selected local oscillator light to calculate the system bit error rate of the entire detection and intelligent compensation device.

[0074] In one embodiment, in step 5, the host computer 10 extracts the photon number distribution obtained by polling. (μ, σ 2 ) as a dataset to train the embedded channel noise compensation model based on a simplified U-Net network structure based on deep learning. The dataset can be represented by a two-dimensional matrix, where the column vectors represent the distribution of detected photons, and the row vectors represent the local oscillator light in different QPSK coding states.

[0075] refer to Figure 7 The process of training a channel noise compensation model (also called a channel noise compensation algorithm or a channel noise compensation network) based on a simplified U-Net network structure based on deep learning includes the following steps:

[0076] I. Determine the sample set for neural network training by the operations of steps 1-5, which includes the photon number distribution matrix and the distribution of the number of detected photons The corresponding LO light encoding state |β k >intensity and phase data.

[0077] II. Based on the simplified U-Net network structure in deep learning, a channel noise compensation algorithm is established. The specific simplified U-Net channel noise compensation network is as follows: the simplified U-Net encoding network is the main structure, using 32×3×3 convolution layers to extract shallow features F1, 64×3×3 convolution layers to extract middle features F2, and 128×3×3 convolution layers to extract high-level features F3; the high-level features F3 extracted by the convolutional neural network are input into the simplified U-Net decoding network, and after 64×2×2 and 32×2×2 upsampling convolution layers, the decoding features F2 are obtained in turn. ′ (second decoding feature) and F1 ′ (First decoding feature), F2 is used in the decoding process ′ The concatenation operation with F2 obtains the first decoding network output feature F4, F1 ′ The output feature F5 of the second decoding network is finally obtained by concatenating it with F1. Finally, the output feature F5 of the simplified U-Net decoding network is flattened to obtain one-dimensional feature data and input into the fully connected layer, where the size of the fully connected layer is (32, 2). The final output of the fully connected layer is the encoding state of the local oscillator light |β k >intensity and phase data.

[0078] III. Initialize the training process, including initializing the channel noise compensation algorithm model parameters, the optimizer, and the learning rate. Preferably, the LeakyRelu activation function is used in all the above operations.

[0079] IV. Sample data (including photon number distribution matrix and the distribution of detected photons The corresponding LO light encoding state |β k >intensity and phase data) are input into the channel noise compensation model for training, and the loss function is determined, specifically, the mean square error loss function is determined as the overall loss function of the channel noise compensation model.

[0080] V. Backpropagate the error of the channel noise compensation model, adjust the model parameters, and optimize the channel noise compensation model. Return to step I and iteratively train the channel noise compensation model until the overall loss function converges. After training is complete, the final channel noise compensation model is obtained.

[0081] This embodiment provides a method for compensating channel noise based on a photon number resolvable detector, which has the following beneficial effects:

[0082] In the actual context of long-distance optical fiber signal transmission, the intensity and phase noise introduced in the channel transmission will have a disturbing effect on the signal light of the standard coded state, resulting in deviations in both intensity and phase between the signal light in the phase space and the local oscillator light of the standard coded state using the same coding format. It is difficult to form high-sensitivity and low-bit-error-rate detection of weak signal light through the combination of "shift operation (interference) to photon number state + SPD / SN-SPD" in the traditional quantum enhanced detection receiver. On the basis of PNRD, the present invention introduces a deep learning neural network algorithm and establishes a channel noise compensation algorithm based on a simplified U-Net network structure. It can identify and extract the influence of intensity / phase noise in channel transmission from the detection results, and rely on the feedback module to form intelligent and real-time channel noise compensation. On the basis of keeping the detection efficiency, interference degree and other parameters consistent with the reported experiments, the overall system can achieve a lower bit error rate for weak coded state signal light under the influence of channel intensity / phase noise. Therefore, the "PNRD+intelligent noise compensation algorithm" provided by the present invention can effectively assist coherent optical communication systems in detecting weak intensity, unknown coded state signal light under the influence of long-distance transmission and channel noise, thereby achieving a lower bit error rate and overall performance improvement.

[0083] It should be noted that for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that this application is not limited by the order of the actions described, because according to this application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by this application.

[0084] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0085] The above description is merely an exemplary embodiment of the present disclosure and is not intended to limit the scope of the present disclosure. That is, any equivalent changes and modifications made in accordance with the teachings of the present disclosure are still within the scope of the present disclosure. After considering the specification and practicing the disclosure herein, those skilled in the art will easily think of the implementation scheme of the present disclosure. This application is intended to cover any variations, uses or adaptations of the present disclosure, which follow the general principles of the present disclosure and include common knowledge or customary technical means in the art that are not recorded in the present disclosure. The description and examples are to be regarded as exemplary only, and the scope and spirit of the present disclosure are defined by the claims.

[0086] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0087] It will be easily understood by those skilled in the art that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for compensating channel noise based on a photon number resolvable detector, characterized in that: include: The transmitting end generates a signal light pulse with a known coding format and an unknown coding state, and divides the signal light pulse into time partitions within a time period corresponding to each pulse width of the signal light pulse; The receiving end prepares local oscillator optical pulses with the same coding format but different coding states for the signal optical pulses in different time partitions after attenuation through a long-distance noisy channel. Using the local oscillator light pulse as a shift operator, within a period corresponding to a pulse width, the same coded state signal light pulse is shifted multiple times in different time partitions by interference, outputting an interference light pulse. Detecting the interference light pulses using a photon number resolvable detector to obtain a photon number distribution result of the interference light pulses; By matching the photon number distribution results each time with the corresponding coded state of the prepared local oscillator light, combined with deep learning network feedback, the offset of the coded state of the local oscillator light relative to the coded state of the signal light under the influence of channel noise is calculated, and the next coded state of the local oscillator light is prepared by using intensity and phase modulation to compensate for the noise based on the standard coding format.

2. The method according to claim 1, wherein The method matches the photon number distribution result of each time with the corresponding coded state of the prepared local oscillator light, calculates the offset of the coded state of the local oscillator light relative to the coded state of the signal light under the influence of channel noise in combination with deep learning network feedback, and guides the preparation of the coded state of the local oscillator light next time using intensity and phase modulation to compensate for the noise in the standard coding format, including: The photon number distribution results and the corresponding coded state of the prepared local oscillator light Forming the dataset { }; The dataset { }Use it as input data for the deep learning network to build a channel noise compensation model; The influence of the channel noise on the coding state of the signal light is calculated by using a channel noise compensation model, and intensity and phase modulation compensation of the coding state of the local oscillator light is calculated based on the influence of the channel noise on the coding state of the signal light to form a feedback instruction; The coding state preparation of the next local oscillator light is compensated according to the feedback instruction until the average value of the photon number distribution result is 0.

3. The method according to claim 1, wherein The method further comprises: The pulse width of the signal light pulse is T, the time partition of the signal light pulse is M, and the pulse width of the prepared local oscillator light pulse is less than T / M.

4. The method according to claim 2, wherein The method further comprises: The photon number distribution of the interference light pulse conforms to the Gaussian distribution , represents the average value of the photon number distribution results, Represents the fluctuation of the photon number distribution around the average value.

5. The method according to claim 4, wherein The compensating the next coded state preparation of the local oscillator light according to the feedback instruction until the average value of the photon number distribution result is 0 includes: when , determining that the coding state of the signal light pulse is consistent with the coding state of the local oscillator light pulse; when When the photon number distribution of the interference light pulse is , prepare new coded state local oscillator light pulses under the same coding format, and combine and The numerical value of and the channel noise compensation model are used to calculate the intensity and phase noise caused to the signal light pulse during channel transmission. On the basis of preparing a new standard coded state local oscillator light under the coding format, the modulation compensation of intensity and phase is completed.

6. A device for compensating channel noise based on a photon number resolvable detector, characterized in that: include: A signal light generating module, used to generate a signal light pulse with a known coding format and an unknown coding state; A local oscillator light generation module is used to generate a local oscillator light pulse with the same coding format as the signal light pulse and an adjustable coding state; an interference module, configured to interfere the signal light pulse and the local oscillator light pulse, thereby performing a displacement operation of the local oscillator light pulse on the signal light pulse and outputting an interference light pulse; a photon number resolvable detection module, configured to detect the interference light pulses and obtain a photon number distribution result of the interference light pulses; The data acquisition and adaptive feedback module is used to collect the photon number distribution results of the interference light pulse and the corresponding coding state of the local oscillator light. A deep learning-based neural network is used to calculate the offset of the coding state of the local oscillator light relative to the coding state of the signal light under the influence of channel noise based on the photon number distribution results and the corresponding coding state of the local oscillator light. Feedback is used to guide the preparation of a new coding state of the local oscillator light pulse with the same coding format as the signal light pulse and after taking into account noise compensation.

7. The device according to claim 6, characterized in that The signal light generation module includes a narrow linewidth continuous light fiber laser, an acousto-optic modulator, an electro-optic phase modulator, an adjustable optical attenuator and a long-distance transmission optical fiber connected in sequence; The local oscillator light generation module includes a narrow linewidth continuous light fiber laser, an acousto-optic modulator, an electro-optic phase modulator, an electro-optic intensity modulator, and a compensation adjustment module connected in sequence. The compensation adjustment module includes a fiber polarization controller and a fiber delay line, which are used to adjust the local oscillator light pulse and the signal light pulse to maintain the same polarization and consistent pulse leading edge.

8. The device according to claim 6, wherein The interference module includes a fiber coupler, and the photon number resolvable detection module includes a photon number resolvable detector.

9. The device according to claim 6, wherein The data acquisition and adaptive feedback module includes an FPGA, a host computer communicating with the FPGA, and a digital-to-analog conversion module; The FPGA is used to synchronously complete the recording of the coded state of the prepared local oscillation light and the matching photon number distribution result, form a data set, and upload it to the host computer; The host computer is used to input the data set into the channel noise compensation model to calculate the influence of channel noise, and output the output result as a feedback instruction to the FPGA; The digital-to-analog conversion module is used to convert the feedback instruction into an electrical signal, complete the modulation of the electro-optical phase modulator and the electro-optical intensity modulator in the local oscillator light generation module, and realize compensation for the modulation / channel noise influence of the coding state of the next local oscillator light.

10. The device according to claim 9, wherein The training method of the channel noise compensation model includes: Step I, taking the photon number distribution matrix and the intensity and phase data of the local oscillator light in the coded state corresponding to the detected photon number distribution as sample data; Step II: Establish a channel noise compensation model based on a simplified U-Net network structure of deep learning. The simplified U-Net network structure is a main structure that uses 32×3×3 convolution layers to sequentially extract shallow features, 64×3×3 convolution layers to sequentially extract middle features, and 128×3×3 convolution layers to sequentially extract high-level features. The high-level features extracted by the convolutional neural network are input into the simplified U-Net decoding network. After 64×2×2 and 32×2×2 upsampling convolution layers, the second decoding features and the first decoding features are obtained in sequence. During the decoding process, the second decoding features and the middle-level features are concatenated to obtain the first decoding network output features, and the first decoding features and the shallow features are concatenated to obtain the second decoding network output features. The second decoding network output features are flattened to obtain one-dimensional feature data and input into a fully connected layer, where the size of the fully connected layer is (32, 2). The output result of the fully connected layer is the intensity and phase data of the coded state local oscillator light. Step III, initialize the channel noise compensation algorithm model parameters, optimizer, and learning rate; Step IV, inputting the sample data into the channel noise compensation model for training, and determining a loss function; In step V, the error of the channel noise compensation model is back-propagated, the parameters of the model are adjusted, and the channel noise compensation model is optimized; returning to step I, the channel noise compensation model is iteratively trained until the loss function converges, and the final channel noise compensation model is obtained after the training is completed.

Images