A coherent optical communication system and a LUT-based soft decision output sequence detection method

Abstract

The application discloses a coherent optical communication system and a sequence detection method based on LUT soft decision output, wherein at the sending end of the coherent optical communication system, binary bit stream original data is first mapped into QPSK symbols, the original QPSK signal is filtered through a narrow-band digital filter to obtain a narrow-band filtered QPSK signal, then the pre-filtered QPSK signal is Nyquist sampled to obtain a digital signal to be sent, after the sending end DSP is completed, the digital signal to be sent needs to be converted into an analog signal through a digital-to-analog converter, the analog signal is optically modulated through an optical modulator to obtain an optical domain sending signal, the optical domain sending signal is transmitted through an optical fiber to reach a receiving end, coherent optical reception is realized through a coherent receiver, a digital signal is obtained through an analog-to-digital converter, a series of receiving end DSP pretreatments are carried out, finally, a BCJR algorithm based on LUT is used to perform sequence detection on the pretreated signal to recover original data. The system can realize long-distance coherent optical communication with high transmission rate and high spectral efficiency.

Description

Technical Field

[0001] This application relates to the field of optical communication technology, specifically to a coherent optical communication system based on narrowband filtering at the transmitting end and a soft-decision output sequence detection method based on LUT. Background Technology

[0002] In narrowband filtered coherent optical communication systems, the severe ISI introduced by narrowband filtering can be effectively eliminated by sequence detection algorithms, which are effective methods for equalizing inter-symbol interference (ISI). Existing methods for eliminating ISI mainly include Maximum-a-posterior (MAP), Maximum Likelihood Sequence Estimation (MLSE), and Bahl-Cocke-Jelinek-Raviv (BCJR).

[0003] Existing research has shown that the MLSE algorithm significantly outperforms the MAP algorithm. Traditional MLSE algorithms first require channel estimation at the receiver to obtain channel response coefficients, which are then used as weights to calculate branch metrics in the trellis graph. Finally, the Viterbi algorithm is used to reconstruct the original data. For example... Figure 1The diagram illustrates a current flexible bandwidth compression sequence detection algorithm that combines narrowband filtering at the transmitter with a multi-symbol distortion lookup table (LUT)-based MLSE process at the receiver. This scheme controls the degree of signal bandwidth compression by adjusting the bandwidth of the narrowband digital filter and utilizes a LUT to record the multi-symbol correlation features introduced by the narrowband filtering, guiding the MLSE process at the receiver to recover the original signal. The scheme employs a Viterbi-based MLSE algorithm, which calculates branch metrics under different paths and uses an accumulation-comparison-selection process to find the path with the smallest error as the decision result, finally decoding and outputting the original signal. However, this scheme only provides hard information and cannot obtain the probabilistic information of the decision symbols, thus failing to obtain the soft information that can be used with the decoder, and therefore cannot achieve better performance. Furthermore, for narrowband filtering optical communication systems, the compression of the signal spectrum in the frequency domain leads to pulse broadening in the time domain, causing time-domain pulse overlap and introducing correlation between symbols. According to Shannon's channel tolerance theory, the performance limit of a communication system can only be reached when the transmitted and detected signals approach infinite duration. However, simply combining independent individual symbols is meaningless. Only when the combined symbols exhibit correlation can the system performance approach the channel tolerance. Therefore, to better balance the ISI introduced by narrowband filtering and more accurately recover the original signal, it is necessary to effectively acquire and record the correlation characteristics between symbols.

[0004] Currently, there is a MLSE scheme that combines narrowband filtering at the transmitter with a look-up table (LUT) at the receiver. This scheme uses the LUT to record the correlation features between symbols introduced by the narrowband filtering and guides the sequence detection process. However, this scheme cannot obtain soft information that can be used with the decoder. Therefore, the current LUT-based MLSE scheme is not optimal for narrowband-filtered coherent optical communication systems.

[0005] Based on the LUT-MLSE scheme, a simplified scheme is available that uses a double binary integer method at the transmitter and a two-symbol MLSE algorithm at the receiver, as follows: Figure 2 As shown, this scheme introduces ISI by superimposing the previous symbol with the current symbol in the time domain, while simultaneously compressing the spectrum in the frequency domain. Since each symbol is only correlated with the previous symbol, this scheme can only use MLSE with a maximum tap length of two to recover the original signal. Furthermore, the system bandwidth of this scheme is limited by the system's baud rate, thus lacking flexible bandwidth compression capabilities and unable to improve system performance by increasing the sequence detection length. This scheme has limitations: firstly, the degree of signal bandwidth compression is limited by the system baud rate; secondly, the MLSE algorithm can only use a maximum of two tap lengths.

[0006] Soft information combined with forward error correction coding (FEC) achieves superior performance compared to hard information combined with FEC. Therefore, in recent years, sequence detection algorithms that can output soft information of decision symbols, namely the BCJR algorithm, have been further developed and studied in narrowband filtered optical communication systems.

[0007] Currently, there is a scheme that combines post-filtering and BCJR soft-decision output at the receiving end, such as... Figure 3 As shown, this scheme requires adding a post-filtering structure in the receiver DSP to introduce controllable ISI, followed by a BCJR processing procedure. However, this scheme does not consider, and cannot record, a series of symbol correlation issues between the transmitter, channel, and receiver.

[0008] The above description of the technical background is provided solely for the purpose of clearly and completely explaining the technical solutions of this application and facilitating understanding by those skilled in the art. It should not be assumed that these solutions are known to those skilled in the art simply because they have been described in the background section of this application. Summary of the Invention

[0009] To overcome the aforementioned shortcomings, this application discloses a coherent optical communication system based on narrowband filtering at the transmitting end and a soft-decision output sequence detection method based on LUT. This method can effectively record the relevant features of the pre-filtered signal and output the probability information of the decision symbol in a soft-decision manner for sequence detection.

[0010] To achieve the above objectives, this application adopts the following technical solution:

[0011] A coherent optical communication system, characterized in that it comprises:

[0012] Transmitter DSP module, digital-to-analog converter and optical modulator;

[0013] The transmitting DSP module maps the received raw information into a QPSK signal according to the modulation format. The QPSK signal is filtered by a narrowband digital filter and then transmitted to the digital-to-analog converter. After conversion by the digital-to-analog converter, it is transmitted to the optical modulator, where it is modulated into a transmit signal. The transmit signal is then transmitted to the receiving end via optical fiber.

[0014] The receiving end includes: an integrated coherent receiver, an analog-to-digital converter, and a receiving end DSP module.

[0015] The integrated coherent receiver receives the transmitted signal and transmits the received and converted signal to the analog-to-digital converter (ADC). The ADC converts the signal into a digital signal, which is then transmitted to the receiving DSP module.

[0016] The receiving DSP module receives the digital signal converted by the analog-to-digital converter and transmits it to the preprocessing DSP module. The received signal from the preprocessing DSP is then transmitted to the LUT-based BCJR processing module, which outputs the original data after sequence detection.

[0017] In one embodiment, the narrowband digital filter compresses the QPSK signal to less than 3dB bandwidth.

[0018] In one embodiment, the narrowband digital filter is a Gaussian narrowband digital filter, which compresses the spectrum of the original signal.

[0019] In one embodiment, the BCJR processing module outputs probability information of the decision symbol.

[0020] In one embodiment, the coherent optical communication system includes:

[0021] The received signal obtained after processing by the preprocessing unit separates the training sequence from the effective data, and transmits the training sequence to the LUT training and generation module. The training and generation module generates an N-symbol distortion signal LUT and guides the subsequent BCJR sequence detection and processing module.

[0022] This application provides a LUT-based soft-decision output sequence detection method based on the above-mentioned coherent optical communication system, including:

[0023] After receiving the transmitted signal, the receiving end performs coherent optical reception through a coherent receiver and obtains the received digital signal through an analog-to-digital converter. The received signal is transmitted to the DSP processing module of the receiving end, and processed by the preprocessing unit to separate the real and virtual parts of the preprocessed signal and separate the training sequence and the effective data sequence. The training sequence is input to the LUT training generation module to generate an N-symbol distortion signal LUT. The generated N-symbol distortion signal LUT is transmitted to the BCJR processing module.

[0024] In one embodiment, in the LUT-based soft-decision output sequence detection method, the BCJR processing module,

[0025] Select the calculation of the received signal and the first row of each line in the LUT. Branch transition probabilities between columns and based on the following formula:

[0026] Obtain the transition probabilities of different branches

[0027] Among them, u k It is the symbol sent at time k, y k It is the symbol received at time k, E s It is the symbolic energy, N0 is the noise power spectral density, P(u k ) is the time when u is sent. k The probability of a symbol.

[0028] In one embodiment, the LUT-based soft-decision output sequence detection method further includes, after obtaining the branch transition probability:

[0029] Calculate the forward metric separately and backward metric

[0030] Then, based on the log-likelihood ratio, the soft value information of the decision symbol is finally obtained.

[0031] In one embodiment, in the LUT-based soft decision output sequence detection method, the LUT training and generation module generates an N-symbol distortion signal LUT based on the QPSK signal filtered by a narrowband filter.

[0032] In one embodiment, in the LUT-based soft-decision output sequence detection method, the LUT training and generation module generates an N-symbol distortion signal LUT based on the sequence after sequence detection decision or error correction.

[0033] Beneficial effects

[0034] Compared to existing technologies, the embodiments of this application can achieve flexible compression of the signal spectrum; compared to the current LUT-MLSE technology, it can output the probability information of the decision symbols and use it as soft information required by the decoder, thereby improving the bit error rate performance. Compared to the traditional BCJR algorithm, it effectively utilizes the correlation between symbols and guides the BCJR algorithm, thereby improving spectral efficiency and performance. Compared to existing schemes combining post-filtering and BCJR, it can use LUTs to record a series of symbol correlations at the transmitter, channel, and receiver of the optical communication system, guiding BCJR to achieve improved performance without the need for a post-filtering structure. Attached Figure Description

[0035] Figure 1 This is a system block diagram of an existing MLSE scheme that combines a first-order Gaussian narrowband filter at the transmitter with an N-symbol distortion signal lookup table at the receiver.

[0036] Figure 2 This is a system block diagram of an existing scheme combining dual binary filtering and a two-tap MLSE at the receiver.

[0037] Figure 3This is a system block diagram of the existing combined post-filter structure and the BCJR approach.

[0038] Figure 4 This is a system block diagram of a coherent optical communication system according to an embodiment of this application.

[0039] Figure 5 This is a schematic diagram illustrating the process of applying the N-symbol distortion signal LUT to the BCJR algorithm in an embodiment of this application.

[0040] Figure 6 This is a schematic diagram illustrating how the receiving end of this application corresponds the original transmitted sequence and the received training sequence to each other.

[0041] Figure 7 This is a schematic diagram of a method for generating an N-symbol distortion signal LUT at the transmitting end according to an embodiment of this application.

[0042] Figure 8 This is a schematic diagram illustrating how, according to an embodiment of this application, the N-symbol distortion signal feature LUT is generated by feeding back the sequence after sequence detection decision or error correction at the receiving end. Detailed Implementation

[0043] The above-described solution will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of this application. The implementation conditions used in the embodiments may be further adjusted according to the conditions of specific manufacturers, and the implementation conditions not specified are generally those in routine experiments.

[0044] This application discloses a coherent optical communication system and a soft-decision output sequence detection method. This method, designed for optical communication systems with pre-filtering at the transmitting end, effectively records the correlation features of the pre-filtered signal and outputs the probability information of decision symbols using a soft-decision approach. The method utilizes a lookup table to record the multi-symbol correlation features of the pre-filtered signal to guide the sequence detection process at the receiving end, and employs a soft-decision approach to output the probability information of the decision symbols. Combined with a decoder, it achieves improved bit error rate performance under the same optical signal-to-noise ratio.

[0045] The following description, with reference to the accompanying drawings, illustrates the coherent optical communication system and soft-decision output sequence detection method proposed in this application.

[0046] A block diagram of a coherent optical communication system is shown below. Figure 4 As shown, this coherent optical communication system is used for long-distance coherent optical communication and includes:

[0047] The transmitting side includes: a transmitting DSP, a digital-to-analog converter, an optical modulator, and a laser; the receiving side includes optical fiber.

[0048] The receiving signal side includes: an analog-to-digital converter, a coherent receiver, a laser, and a receiver DSP.

[0049] When the system is running:

[0050] On the transmitting side, the original data (binary bitstream) is first mapped to QPSK symbols. The original QPSK signal is then passed through a narrowband digital filter (with a 3dB bandwidth less than the signal baud rate) to obtain a narrowband filtered QPSK signal. Next, the pre-filtered QPSK (Quadrature Phase Shift Keying) signal is subjected to Nyquist sampling to obtain the digital signal to be transmitted. After completion by the DSP at the transmitting end, the digital signal to be transmitted needs to be converted to an analog signal via a digital-to-analog converter. This analog signal is then optically modulated by an optical modulator to obtain the optical domain transmission signal. This transmission signal is then transmitted via optical fiber.

[0051] from Figure 4 As can be seen, the original QPSK signal has 4 constellation points. After narrowband filtering, the high-frequency components of the signal in the frequency domain are filtered out, while in the time domain, the original signal pulses are broadened. This causes the symbols to overlap, introducing deterministic ISI. It is precisely because of the ISI introduced by narrowband filtering that the original 4 constellation points are expanded to 36 constellation points, thus introducing correlation between symbols.

[0052] After the transmitted signal reaches the receiving end (receiving signal side) via optical fiber, it is received coherently by a coherent receiver and then converted into a QPSK signal (digital signal) by an analog-to-digital converter. Following this, a series of receiver-side DSP processing steps are performed, including dispersion compensation, clock recovery, polarization demultiplexing, polarization mode dispersion compensation, frequency offset compensation, and phase offset compensation. After receiver-side DSP preprocessing, the real and virtual components of the received signal are separated, and the training sequence and valid data sequence are also separated. The training sequence is input into a LUT (Look-up Table) training generator to generate an N-symbol distortion LUT.

[0053] like Figure 5 The diagram shows the process of applying the generated N-symbol distortion signal LUT to the calculation of branch transition probabilities in the BCJR algorithm.

[0054] Because the current sampled symbol is subject to ISI from neighboring symbols, when calculating the branch transition probability under different inputs,

[0055] Select the calculation of the received signal and the first row of each line in the LUT. The branch transition probability between columns yields the transition probability of different branches.

[0056] As shown in the formula below:

[0057]

[0058] Among them, u k It is the symbol sent at time k, y k It is the symbol received at time k, E s It is the symbolic energy, N0 is the noise power spectral density, P(u k) It is sent at time k. k The probability of a symbol.

[0059] After obtaining the branch transition probabilities, calculate the forward metric respectively. and backward metric Then, based on the log-likelihood ratio, the soft value information (decoding / soft information) of the decision symbol is finally obtained.

[0060] For example, with a QPSK modulation format, modulation order M=4, and sequence detection length N=3, after DSP preprocessing at the receiver, the real and imaginary parts of the received signal are separated, and the training sequence and valid data are also separated. For a training sequence with known permutations, three consecutive symbols are grouped together. Because the real and imaginary parts are separated, the real or imaginary part of the symbol will contain either "-1" or "1". Therefore, there are a total of 2... 3 There are several types. For a training sequence separated from the received signal, since the values ​​of its original transmitted sequence are known, the two sequences can be correlated. For example... Figure 6 As shown, by mapping the original transmitted sequence and the received training sequence to each other, when the original transmitted sequence is {-1, -1, -1}, the sampled values ​​of sequence 2 in the register are sequentially recorded in the 3-symbol distortion signal LUT. Each time the register value of sequence 1 is {-1, -1, -1}, the sampled values ​​of sequence 2 in the register are recorded and accumulated. After all symbols have passed through the register, the average of each row of the LUT is calculated based on the number of accumulations, and finally, a 3-symbol distortion LUT based on the average sampled values ​​can be obtained.

[0061] Table 1 provides an example of a lookup table for a 3-symbol distortion signal.

[0062]

[0063] Table 1

[0064] Where ε is the pre-filtered symbol amplitude reduction factor, and Δ is the ISI of the current symbol from the adjacent symbols.

[0065] After obtaining the 3-symbol distorted signal LUT, the LUT is applied to the calculation of branch transition probabilities in the traditional BCJR algorithm. Since the current sampled symbol is affected by ISI from adjacent symbols, when calculating the branch transition probabilities under different inputs, we choose to calculate the Euclidean distance between the received signal and the second column of each row in the LUT, and finally obtain different branch transition probabilities.

[0066] In one embodiment, when generating a LUT using the training sequence at the receiving end, the LUT can be directly generated at the transmitting end using the narrowband filtered signal. For example... Figure 7 The method shown for generating an N-symbol distortion signal LUT at the transmitting end, compared to the method of generating a LUT at the receiving end using a training sequence, does not require a training sequence and has higher robustness. That is, under different channel environments and for a fixed bandwidth constraint, only one LUT needs to be generated. However, the LUT generated by this method cannot record the damage to the signal caused by ISI introduced by narrowband filtering and other channel effects, so its performance is slightly worse, and its performance will further degrade when strong symbol correlation is introduced by channel and receiving end impairment effects.

[0067] In one embodiment, the method of generating a LUT using separate training sequences at the receiving end can directly use the sequence after sequence detection decision or error correction to generate an N-symbol distortion signal feature LUT, such as... Figure 8 As shown, this method can record the impact of narrowband filtering at the transmitting end, as well as other integrated channel responses and nonlinear impairments, on the characteristics of the original transmitted signal, and more effectively achieve universal equalization processing for various signal impairments. However, compared with the method of directly generating the LUT at the transmitting end, this method requires adding a LUT training and generation mechanism at the receiving end, which increases the system complexity.

[0068] The above embodiments are only for illustrating the technical concept and features of this application, and are intended to enable those skilled in the art to understand the content of this application and implement it accordingly. They should not be used to limit the scope of protection of this application. All equivalent changes or modifications made in accordance with the spirit and essence of this application should be included within the scope of protection of this application.

Claims

1. A coherent optical communication system, characterized in that, include: Transmitter DSP module, digital-to-analog converter and optical modulator; The transmitting DSP module maps the received raw information into a QPSK signal according to the modulation format. The QPSK signal is filtered by a narrowband digital filter and then transmitted to the digital-to-analog converter. After conversion by the digital-to-analog converter, it is transmitted to the optical modulator, where it is modulated into a transmit signal. The transmit signal is then transmitted to the receiving end via optical fiber. The receiving end includes: an integrated coherent receiver, an analog-to-digital converter, and a receiving end DSP module. The integrated coherent receiver receives the transmitted signal and transmits the received and converted signal to the analog-to-digital converter (ADC). The ADC converts the signal into a digital signal, which is then transmitted to the receiving DSP module. The receiving DSP module receives the signal converted into a digital signal by the analog-to-digital converter and transmits it to the preprocessing DSP module. The received signal from the preprocessing DSP is then transmitted to the LUT-based BCJR processing module, which performs sequence detection and outputs the original data. The BCJR processing module selects and calculates the received signal and the first [number]th ... Branch transition probabilities between columns and based on the following formula: The transition probabilities of different branches are obtained. , After obtaining the branch transition probability, the following steps are also included: calculating the forward metric separately. and backward metric , Then, based on the log-likelihood ratio, the soft value information of the decision sign is finally obtained, where, It is the symbol sent at time k. It is the symbol received at time k. It is symbolic energy. It is the noise power spectral density. It is sent at time k. The probability of a symbol.

2. The coherent optical communication system as described in claim 1, characterized in that, Narrowband digital filters compress QPSK signals to less than 3dB bandwidth.

3. The coherent optical communication system as described in claim 1, characterized in that, The narrowband digital filter is a Gaussian narrowband digital filter, which compresses the spectrum of the original signal.

4. The coherent optical communication system as described in claim 1, characterized in that, The BCJR processing module outputs the probability information of the decision symbol.

5. The coherent optical communication system as described in claim 1, characterized in that, The received signal obtained after preprocessing by the DSP separates the training sequence from the valid data, and transmits the training sequence to the LUT training and generation module. The training and generation module generates an N-symbol distortion signal LUT and guides the subsequent BCJR sequence detection and processing module.

6. A method for detecting soft-decision output sequences based on a LUT in a coherent optical communication system as described in any one of claims 1-5, characterized in that, After receiving the transmitted signal, the receiving end performs coherent optical reception through a coherent receiver and obtains the received digital signal through an analog-to-digital converter. The received signal is transmitted to the DSP processing module of the receiving end. After the preprocessing DSP processing, the real and virtual parts of the preprocessed signal are separated, and the training sequence and the effective data sequence are separated. The training sequence is input to the LUT training generation module to generate an N-symbol distortion signal LUT. The generated N-symbol distortion signal LUT is transmitted to the BCJR processing module.

7. The LUT-based soft-decision output sequence detection method as described in claim 6, characterized in that, The BCJR processing module, Select the calculation of the received signal and the first row of each line in the LUT. Branch transition probabilities between columns and based on the following formula: , Obtain the transition probabilities of different branches , in, It is the symbol sent at time k. It is the symbol received at time k. It is symbolic energy. It is the noise power spectral density. It is sent at time k. The probability of a symbol.

8. The LUT-based soft-decision output sequence detection method as described in claim 7, characterized in that, After obtaining the branch transition probability, the following is also included: Calculate the forward metric separately and backward metric , Then, based on the log-likelihood ratio, the soft value information of the decision symbol is finally obtained.

9. The LUT-based soft-decision output sequence detection method as described in claim 7, characterized in that, The LUT training and generation module generates an N-symbol distortion signal LUT based on the QPSK signal filtered by a narrowband filter.

10. The LUT-based soft-decision output sequence detection method as described in claim 7, characterized in that, The LUT training and generation module generates an N-symbol distortion signal LUT based on the sequence after sequence detection decision or error correction.

Images