A non - linear compensation method for jointly using LUT at the transceiver in an MLSE detection scheme

Abstract

The present invention relates to the field of high-speed optical signal processing technology. More specifically, it relates to a non-linear compensation method for jointly using a Look-Up Table (LUT) at the transmitter and receiver in an MLSE detection scheme. For the first time, the technology of jointly using LUT at the transmitter and receiver (Joint Tx and Rx LUT-MLSE) and using MLSE detection are proposed to effectively mitigate the non-linear distortion of the system and eliminate the inter-symbol interference problem. The present invention can significantly improve the receiver sensitivity of the system and is more suitable for application in intensity modulation direct detection systems with high-speed high-order modulation formats (PAM-8) limited by bandwidth.

Description

Technical Field

[0001] The present invention relates to the technical field of high-speed optical signal processing, and more specifically, to a non-linear compensation method for jointly using a LUT at a transceiver end in an MLSE detection scheme. Background Art

[0002] With the emergence of various new application scenarios such as virtual reality / augmented reality, 5G communication, and the metaverse, the demand for large-capacity transmission inside and between data centers (DCs) has grown rapidly, becoming an urgent problem to be solved. Data center interconnect (DCI) needs to consider power consumption, architecture complexity, and latency. High-speed IM / DD systems have advantages such as low power consumption and simple structure, and thus have been widely studied. The next-generation Ethernet links of 800 GbE or 1.6 TbE have been put on the agenda. To meet the requirements of the next-generation Ethernet links, it is necessary to increase the bit rate per channel in short-distance optical interconnections, which requires transceiver devices with a larger bandwidth to be used for low-order PAM formats. However, the bandwidth development of commercial devices based on CMOS technology is insufficient. Therefore, in order to alleviate the requirements for bandwidth of transceivers to a certain extent, higher-order PAM-8 modulation formats have been considered. However, higher-order modulation formats face two problems in bandwidth-limited IM / DD systems. One is the ISI problem, and the other is the non-linear distortion problem, among which the non-linear distortion problem is more serious. There has been a certain degree of research on how to solve these two problems, and some methods have been proposed:

[0003] (1) Feed-forward equalization - post-filtering - maximum likelihood sequence estimation scheme (FFE-Postfilter-MLSE): This scheme can effectively alleviate the linear ISI problem when the bandwidth limitation is not serious. However, as the bandwidth limitation becomes more severe and higher-order modulation formats such as PAM-8 are adopted, it has no effect on non-linear distortion. The non-linear distortion will significantly affect the performance of MLSE and deteriorate the BER performance.

[0004] (2) Transmitter LUT pre-distortion scheme (Tx LUT): This method uses a training sequence to establish a LUT. The difference between the signal equalized by a feed-forward equalizer (FFE) and the corresponding signal in the training sequence is statistically recorded in this LUT. Then, this LUT is used to pre-distort the signal to be transmitted, thereby overcoming the non-linear problem in the system. However, this method can alleviate the non-linear distortion brought by the system to the signal when the non-linear distortion is not serious. However, when the non-linear problem is serious, the alleviating ability of this method is limited. And this scheme cannot effectively solve the ISI problem faced by high-rate signals in the transmission of bandwidth-limited devices.

[0005] (3) Receiver LUT-MLSE Scheme (Rx LUT-MLSE): The algorithm used to recover the signal at the receiver in this method is the Feed-Forward Equalizer (FFE)-Postfilter-Maximum Likelihood Sequence Estimation (MLSE) scheme. The RxLUT scheme uses the training sequence to establish the LUT, and the output signal of the Postfilter is statistically counted in this LUT. Then this LUT is used in the calculation of the branch metric in the MLSE algorithm to alleviate the ISI and nonlinear problems. However, this method can only alleviate the nonlinear distortion problem to a certain extent. When high-rate high-order modulation format signals are transmitted in a bandwidth-limited system, the ISI and nonlinear problems are very significant. The ability of this method to eliminate nonlinearity is insufficient and will affect the MLSE performance, resulting in the system still facing an unsatisfactory BER performance.

[0006] Due to the relatively serious nonlinear distortion in the transmission of high-rate high-order modulation format signals (PAM-8), simply adopting a scheme such as FFE-Postfilter-MLSE can only partially alleviate the linear ISI problem, and the nonlinear problem will still affect the system performance. Using only the Tx LUT or the Rx LUT-MLSE can only partially alleviate the nonlinear distortion of the system, and the remaining nonlinear distortion will still deteriorate the BER performance of the system. The factors leading to the degradation of the system BER performance can be attributed to the limited capabilities of the Tx LUT or Rx LUT-MLSE, which are insufficient to completely alleviate the nonlinear distortion problem during the high-rate transmission of high-order modulation signals, thereby affecting the ability of the algorithm for the receiver to recover the transmitted signal. Summary of the Invention

[0007] The present invention aims to overcome the above-mentioned defects in the prior art and provides a nonlinear compensation method that jointly uses LUTs at the transmitter and receiver in the MLSE detection scheme, which can more effectively eliminate severe Inter-Symbol Interference (ISI) and alleviate nonlinear distortion.

[0008] To solve the above technical problems, the technical solution adopted by the present invention is: a non-linear compensation method for jointly using a look-up table (LUT) at the transmitter and receiver in an MLSE detection scheme. Among them, before transmitting the signal, a training sequence is used to statistically obtain the look-up tables (LUTs) that will be used at the receiver and transmitter; in the transmitter DSP, first, the bit stream sequence is mapped into a PAM-8 signal, then pre-distortion of the transmitter LUT is performed, and then the signal is Nyquist-shaped using a root-raised cosine filter; after that, the shaped signal is downsampled, and finally the signal is input into a digital-to-analog converter; in the receiver DSP, the signal obtained by the optical receiver PD detection is successively upsampled, matched filtered, synchronized, forward feedback equalized, post-filtered, MLSE based on the receiver look-up (LUT) table, PAM-8 signal demapped, and bit error rate calculated.

[0009] Further, the generation of the transmitter look-up table (LUT) includes the following steps: First, a training sequence is transmitted in the system, and a sliding window is used to move forward one symbol at a time in the training sequence. The content of the training sequence contained in the sliding window is denoted as: S t (k)=[S t (k - l)…S t (k)…S t (k + l)]; Using S t (k), the corresponding table index i is calculated at each moment. The sequence after being processed by the forward feedback equalizer FFE is named S e , and the amplitude difference between S e (k) and S t (k) is denoted as e(k)=S e (k)-S t (k). This value will be accumulated in the corresponding LUT value; as the sliding window moves, each time an amplitude difference is added, the corresponding LUT value will be accumulated and updated, and N(i) records the update times of the corresponding LUT value; when the sliding window moves to the end of the training sequence, the generation of the LUT value is completed.

[0010] Further, the generation of the receiver look-up table (LUT) includes the following steps: First, a training sequence is transmitted in the system, and a sliding window is used to move forward one symbol at a time in the training sequence. The content of the training sequence contained in the sliding window is denoted as: S t (k)=[S t (k - l)…S t (k)…S t (k + l)]; Using S t (k), the corresponding table index i is calculated at each moment, and then the output S of the post-filter p-fAccumulate in the corresponding table; as the sliding window moves, for each additional amplitude difference, the corresponding LUT value is accumulated and updated, and N(i) records the update times of the corresponding LUT value; when the sliding window moves to the end of the training sequence, the generation of the LUT value is completed.

[0011] Further, both the receiving - end lookup table and the transmitting - end lookup table take the average value.

[0012] Further, the transmitting - end LUT predistortion includes the following steps: calculate the LUT index i using a sliding window of the same length, search for the corresponding LUT value according to the index i, and the predistorted symbol can be written as: S p-d (k) = S(k) - LUT(i), where S(k) is the PAM - 8 signal to be transmitted.

[0013] Further, the MLSE based on the receiving - end lookup table includes the following steps: the receiving - end lookup table replaces the dot - product in the calculation of the MLSE branch metric, calculates the BM using the output of the post - filter and the generated receiving - end lookup table (LUT), and the formula for constructing the BM is defined as: BM(m,k)=|S p-f (k)-f(S m (k))| 2 ; where S m (k) is the m - th pattern generated based on the MLSE trellis diagram, and f(S m (k)) is the LUT value corresponding to S m (k); then the calculated branch metric is sent to the Viterbi algorithm module to select the surviving path, and after reaching the traceback length, start to traceback one by one to find the most likely transmitted signal.

[0014] The present invention also provides a non - linear compensation system that jointly uses LUTs at the transmitting and receiving ends in an MLSE detection scheme, including a transmitting - end DSP module and a receiving - end DSP module;

[0015] The transmitting - end DSP module is used to map the bit - stream sequence into a PAM - 8 signal, then perform transmitting - end LUT predistortion, and then use a root - raised - cosine filter to perform Nyquist shaping on the signal; afterwards, down - sample the shaped signal and finally input the signal into a digital - to - analog converter;

[0016] The receiving - end DSP module detects the signal obtained through the optical receiver PD and sequentially performs up - sampling, matched filtering, synchronization, forward feedback equalization, post - filtering, MLSE based on the receiving - end lookup table (LUT), PAM - 8 signal demapping, and bit - error rate calculation;

[0017] Among them, before transmitting the signal, a lookup table (LUT) used in the receiving - end DSP module and the transmitting - end DSP module is statistically obtained by using a training sequence.

[0018] Furthermore, the generation of the lookup table (LUT) of the transmitting - end DSP module includes the following steps: First, a training sequence is transmitted in the system. Using a sliding window, it advances symbol - by - symbol forward in the training sequence. The content of the training sequence contained in the sliding window is denoted as: S t (k)=[S t (k - l)…S t (k)…S t (k + l)]; Using S t (k), the corresponding table index i is calculated at each moment. The sequence after being processed by the forward - feedback equalizer FFE is named S e , S e (k) and S t (k) The amplitude difference between them is denoted as e(k)=S e (k)-S t (k). This value will be accumulated in the corresponding LUT value. As the sliding window moves, for each additional amplitude difference, the corresponding LUT value is accumulated and updated. N(i) records the update times of the corresponding LUT value. When the sliding window moves to the end of the training sequence, the accumulated LUT value is averaged and stored, thus obtaining the LUT of the transmitting - end DSP module. The generation of the lookup table (LUT) of the receiving - end DSP module includes the following steps: First, a training sequence is transmitted in the system. Using a sliding window, it advances symbol - by - symbol forward in the training sequence. The content of the training sequence contained in the sliding window is denoted as: S t (k)=[S t (k - l)…S t (k)…S t (k + l)]; Using S t (k), the corresponding table index i is calculated at each moment, and then the output S p-f of the post - filter is accumulated in the corresponding table. As the sliding window moves, for each additional amplitude difference, the corresponding LUT value is accumulated and updated. N(i) records the update times of the corresponding LUT value. When the sliding window moves to the end of the training sequence, the accumulated LUT value is averaged and stored, thus obtaining the LUT of the receiving - end DSP module.

[0019] Furthermore, in the transmitting - end DSP module, when performing LUT pre - distortion processing, a sliding window of the same length is used to calculate the LUT index i. According to the index i, the corresponding LUT value is searched. The pre - distorted symbol can be written as: S p-d(k) = S(k) - LUT(i), where S(k) is the PAM-8 signal to be transmitted.

[0020] Further, in the receiving-end DSP module, when performing MLSE processing based on the receiving-end look-up table, the branch metric (BM) is calculated using the output of the post-filter and the generated receiving-end look-up table. The formula for constructing BM is defined as: BM(m,k) = |S p-f (k) - f(S m (k))| 2 ; where S m (k) is the m-th pattern generated based on the MLSE trellis diagram, and f(S m (k)) is the LUT value corresponding to S m (k); then the calculated branch metric is sent to the Viterbi algorithm module to select the surviving path. After reaching the traceback length, the most likely transmitted signal is searched for one by one through traceback.

[0021] Compared with the prior art, the beneficial effects are as follows: A non-linear compensation method for jointly using LUTs at the transmitter and receiver in an MLSE detection scheme proposed by the present invention. The present invention first proposes to jointly use the LUT technology (Joint Tx and Rx LUT-MLSE) at the transmitter and receiver and use MLSE detection to effectively mitigate the non-linear distortion of the system and eliminate the inter-symbol interference problem. The present invention can significantly improve the sensitivity of the system receiver and is more suitable for application in intensity modulation direct detection systems with high-rate high-order modulation formats (PAM-8) with limited bandwidth. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 is a schematic diagram of the method framework of the present invention.

[0023] Figure 2 is a schematic diagram of the device structure provided in Embodiment 2.

[0024] Figure 3 is the signal histogram after FFE in Embodiment 2. Among them, (a)-(c) are the signal histograms of 150 / 165 / 180 Gbit / s PAM-8 after FFE (without using Tx LUT), and (d)-(e) are the signal histograms of 150 / 165 / 180 Gbit / s PAM-8 after FFE (using Tx LUT).

[0025] Figure 4It is the statistics of the LUT symbol distortion at the receiving end in Embodiment 2. Among them, (a)-(c) are the statistics of the LUT symbol distortion at the receiving end of 150 / 165 / 180 Gbit / s PAM-8 signals (without using Tx LUT), and (d)-(e) are the statistics of the LUT symbol distortion at the receiving end of 150 / 165 / 180 Gbit / s PAM-8 signals (using Tx LUT).

[0026] Figure 5 It is the bit error rate versus received optical power graph for optical back-to-back transmission (OBTB) in Embodiment 2 for (a) 150 Gbit / s, (b) 165 Gbit / s, and (c) 180 Gbit / s PAM-8 signal transmissions.

[0027] Figure 6 It is the bit error rate versus received optical power graph for 20 km standard single-mode fiber (SSMF) transmission in Embodiment 2 for (a) 150 Gbit / s, (b) 165 Gbit / s, and (c) 180 Gbit / s PAM-8 signal transmissions. Detailed implementation manners

[0028] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. The present invention will be described below in conjunction with the specific implementation manners in one of the embodiments. Among them, the accompanying drawings are only for illustrative purposes, showing only schematic diagrams, rather than physical diagrams, and should not be construed as a limitation to this patent; in order to better illustrate the embodiments of the present invention, some components in the accompanying drawings will be omitted, enlarged or reduced, and do not represent the dimensions of the actual product; for those skilled in the art, it is understandable that some well-known structures and their descriptions in the accompanying drawings may be omitted.

[0029] In the description of the present invention, it should be understood that if there are terms such as "upper", "lower", "left", "right", etc., the orientation or positional relationship indicated is based on the orientation or positional relationship shown in the drawings. This is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, the terms describing the positional relationship in the drawings are only for illustrative purposes and cannot be understood as a limitation of this patent. For those of ordinary skill in the art, the specific meanings of the above terms can be understood according to specific circumstances. In addition, if there is a description involving "first", "second", etc. in the embodiments of the present invention, the description of "first", "second", etc. is only for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one such feature. In addition, the meaning of "and / or" appearing throughout the text is that it includes three parallel solutions. Taking "A and / or B" as an example, it includes solution A, or solution B, or a solution where A and B are satisfied simultaneously.

[0030] Embodiment 1:

[0031] This embodiment provides a non - linear compensation method for jointly using a look - up table (LUT) at the transceiver in an MLSE detection scheme. Among them, before transmitting the signal, a training sequence is used to statistically obtain the look - up tables (LUTs) that will be used at the receiving end and the transmitting end; in the transmitting - end DSP, first, the bit - stream sequence is mapped into a PAM - 8 signal, then pre - distortion of the transmitting - end LUT is performed, and then the signal is Nyquist - shaped using a root - raised - cosine filter; after that, the shaped signal is down - sampled, and finally the signal is input into a digital - to - analog converter; in the receiving - end DSP, the signal obtained by the optical receiver PD detection is successively up - sampled, matched - filtered, synchronized, forward - feedback equalized, post - filtered, MLSE based on the receiving - end look - up table, PAM - 8 signal demapped, and bit - error rate calculated.

[0032] As Figure 1 shown, the generation mechanism of the LUT is as shown in the upper - left part of Figure 1 . The generation of the transmitting - end look - up table (LUT) includes the following steps: First, it is inevitable to transmit a training sequence in the system. The sliding window is advanced symbol - by - symbol in the training sequence. The length of the sliding window is 2l + 1, and the content of the training sequence included in the sliding window is denoted as: S t (k)=[S t (k - l)…S t (k)…S t (k + l)]; Using S t (k), the corresponding table index i is calculated at each moment. The sequence after being processed by the forward - feedback equalizer FFE is named Se , S e (k) and S t (k), the amplitude difference between them is denoted as e(k) = S e (k) - S t (k), and this value will be accumulated in the corresponding LUT value; as the sliding window moves, each time an amplitude difference is added, the corresponding LUT value will be accumulated and updated, and N(i) records the update times of the corresponding LUT value; when the sliding window moves to the end of the training sequence, the accumulated LUT value is averaged and stored, thus obtaining the look-up table for transmitter pre-distortion. Averaging calculation can minimize the interference of additive noise.

[0033] The generation of the receiver look-up table is carried out simultaneously with the generation of the transmitter look-up table. As shown in Figure 1 the upper right part, the generation of the receiver look-up table (LUT) includes the following steps: First, a training sequence is transmitted in the system. Using a sliding window, it advances symbol by symbol in the training sequence. The content of the training sequence included in the sliding window is denoted as: S t (k) = [S t (k - l)…S t (k)…S t (k + l)]; Using S t (k), the corresponding table index i is calculated at each moment, and then the output S p-f of the post-filter is accumulated in the corresponding table; as the sliding window moves, each time an amplitude difference is added, the corresponding LUT value will be accumulated and updated, and N(i) records the update times of the corresponding LUT value; when the sliding window moves to the end of the training sequence, the accumulated LUT value is averaged and stored, thus obtaining the receiver LUT.

[0034] LUT pre-distortion is performed on the transmitter side. As shown in Figure 1 the lower left part, a sliding window of the same length is used to calculate the LUT index i. According to the index i, the corresponding LUT value is searched. The pre-distorted symbol can be written as: S p-d (k) = S(k) - LUT(i), where S(k) is the PAM-8 signal to be transmitted.

[0035] MLSE based on the receiver look-up table is as shown in Figure 1 the lower right part: The receiver look-up table replaces the dot product in the calculation of the MLSE branch metric. Using the output of the post-filter and the generated receiver look-up table, the BM is calculated. The formula for constructing BM is defined as: BM(m,k) = |S p-f (k) - f(S m (k))| 2 ; where S m(k) is the m-th pattern generated based on the MLSE trellis diagram, and f(S m (k)) is the LUT value corresponding to S m (k); then the calculated branch metrics are sent to the Viterbi algorithm module to select the surviving path. After reaching the traceback length, the most likely transmitted signal is searched for one by one through traceback.

[0036] Compared with the existing methods, this method can more significantly eliminate the non-linear problem and more effectively mitigate ISI. Therefore, this method can significantly improve the BER performance of the system and enhance the sensitivity of the receiver. Using this method, the 150 / 165 / 180 Gbit / s PAM-8 signal transmission under bandwidth-limited conditions is successfully achieved, and the statistically calculated BER after transmission can be lower than the hard decision forward error correction threshold (HD-FEC), that is, 3.8×10 -3 . And under the transmission of 150 / 165 Gbit / s, compared with the single transmitter LUT combined with MLSE (Tx LUT-MLS) and the single receiver LUT (Rx LUT-MLSE), this method can improve the receiver sensitivity by 1 - 2 dB. Under the transmission of 180 Gbit / s, Joint Tx and Rx LUT-MLSE is the only solution that can achieve a BER lower than the HD-FEC threshold.

[0037] Embodiment 2

[0038] Figure 2 The digital signal processing (DSP) flowchart and the experimental setup diagram of the present invention are given.

[0039] First, before the transmitted signal, the LUT to be used at the transmitter and receiver is statistically obtained using the training sequence.

[0040] (1) Transmitter DSP

[0041] In the DSP at the transmitter, first, the 0-1 bit stream sequence is mapped into a PAM-8 signal. Then, pre-distortion of the transmitter LUT is performed, and then the signal is Nyquist-shaped using a root raised cosine filter with a roll-off factor of 0.1. After that, the shaped signal is downsampled and then input into a digital-to-analog converter (DAC) with a bandwidth of 16 GHz and a sampling rate of 80 GSa / s.

[0042] (2) Receiver DSP

[0043] In the DSP at the receiver, the signals obtained by PD detection are mainly upsampled, matched filtered, synchronized, LMS equalized, Rx LUT-MLSE, PAM-8 signal demapped, and bit error rate (BER) calculated.

[0044] (3) Experimental setup

[0045] This experimental system is a bandwidth-limited IM / DD system. First, at the transmitter, the offline data generated by MATLAB is loaded into a DAC with a sampling rate of 80 GSa / s and a 3-dB bandwidth of 16 GHz for digital-to-analog conversion. Then, an electrical attenuator is used to mitigate the nonlinear distortion. After that, a signal is amplified by an electrical amplifier (EA) with a gain of 23 dB. The transmitted signal after the electrical amplifier, a 1310-nm laser, and a DC bias are fed into a Mach-Zehnder modulator (MZM) to achieve electro-optic conversion (E / O). The optical carrier of the modulator comes from a distributed feedback laser (DFB). Then, the optical signal is transmitted through BTB (Back-to-Back) or a 20-km standard single-mode fiber (SSMF). After the transmission is completed, a variable optical attenuator (VOA) is used to adjust the received optical power (ROP) of the received signal. The attenuated optical signal is subjected to optoelectronic conversion (O / E) using a PD (PIN-TIA). The analog electrical signal output by the PD is collected by an oscilloscope with a sampling rate of 128 GSa / s and a cut-off bandwidth of 33 GHz and subjected to offline DSP.

[0046] Before the experiment started, the frequency response of the system was measured using training symbols, as Figure 2 shown in the accompanying figure. The bandwidths at -3 dB and -10 dB are 3.3 GHz and 20.6 GHz respectively, and it can be seen that the bandwidth limitation of the system is very serious.

[0047] 2. Result Analysis

[0048] First, the ISI and nonlinear distortion levels of different bit-rate signals with and without a Tx LUT were analyzed. Figure 3 is the histogram of different bit-rate signals after FFE. In Figure 3 (a)-(c), obvious nonlinear distortion can be observed at higher amplitudes. However, when the transmitter LUT is implemented, the nonlinear distortion is significantly mitigated, as Figure 3 (d)-(f) shows. At the same time, it can be seen from Figure 3 (d)-(f) that as the bit rate increases, the ISI problem becomes more and more serious.

[0049] Then, the distortion levels of the symbols in the receiver LUT at different bit rates were statistically analyzed, and the results are as Figure 4 shown. Figure 4(a)-(c) are the symbol distortions of 150 / 165 / 180 Gbit / s PAM-8 signals without using Tx LUT. It is worth noting that some of the symbol distortions are very close to zero, while others are larger, which is due to the high (or low) peak amplitudes in the corresponding patterns. These obvious distortions become more serious with the increase of the bit rate. When the bit rate is 180 Gbit / s, the distortion can even reach +2 or -1.5, seriously affecting the system performance. Therefore, Tx LUT can alleviate the distortion to a certain extent. The symbol distortion of the PAM-8 signal under the condition of using Tx LUT is as shown in Figure 4 (d)-(f). More distortions tend to zero, and the remaining distortions will also decrease accordingly, which benefits from the adoption of Tx LUT. However, when transmitting 180 Gbit / s PAM-8 signals, the remaining symbol distortions cannot be ignored, and the distortion is between ±1. Therefore, it is necessary to adopt Rx LUT-MLSE to further eliminate the remaining nonlinear distortions.

[0050] Under the same channel conditions, the bit error rate performance of 150 / 165 / 180 Gbit / s PAM-8 signals in OBTB transmission was evaluated and compared with the scheme of not using LUT at both the transmitter and receiver combined with MLSE (w / o LUT-MLSE), the scheme of only using Tx LUT combined with MLSE (Tx LUT-MLSE), and the scheme of only using Rx LUT-MLSE (Rx LUT-MLSE). As shown in Figure 5 . Comparing Figure 5 (a), Figure 5 (b) and Figure 5 (c), when only using w / o LUT-MLSE, the bit error rate curve can never reach the HD-FEC threshold, indicating that the system nonlinear problem is very serious, and simply solving the linear ISI problem cannot make the system reach the ideal BER. For the 150 Gbit / s PAM-8 signal shown in Figure 6 (a), adopting the Joint Tx and Rx LUT-MLSE scheme proposed by the present invention can significantly reduce the BER. At the HD-FEC threshold, compared with Rx LUT-MLSE and Tx LUT-MLSE, the receiving sensitivities are increased by 1.19 dB and 1.57 dB respectively. As shown in Figure 6As shown in (b), when the bit rate is increased to 165 Gbit / s, by adopting the Joint Tx and Rx LUT-MLSE scheme proposed in the present invention, the sensitivity of the receiver at the HD-FEC threshold is increased by 1.08 dB and 1.22 dB respectively. When the bit rate is further increased to 180 Gbit / s, when using Tx LUT-MLSE and Rx LUT-MLSE, the BER can no longer reach the HD-FEC threshold. Only the Joint Tx and Rx LUT-MLSE scheme proposed in the present invention can achieve a BER lower than the HD-FEC threshold, which is a significant improvement. In addition, the transmission performance of 150 / 165 / 180 Gbit / s PAM-8 signals on 20 km SSMF is also studied. The experimental results are as Figure 6 shown. The trend of the BER is similar to the propagation of OBTB. In the 150 Gbit / s transmission, compared with Rx LUT-MLSE and Tx LUT-MLSE, the received sensitivity at the HD-FEC threshold is increased by 1.26 dB and 2.04 dB respectively. When the bit rate is increased to 165 Gbit / s, by adopting the Joint Tx and Rx LUT-MLSE scheme proposed in the present invention, the received sensitivity can be increased by 1.15 dB and 1.26 dB. The Joint Tx and Rx LUT-MLSE scheme proposed in the present invention is still the only scheme that enables the BER of the PAM-8 signal to be lower than the HD-FEC threshold at a bit rate of 180 Gbit / s.

[0051] In the intensity modulation direct detection system with a bandwidth-limited high-rate high-order modulation format (PAM-8), the present invention proposes a scheme that can more effectively eliminate ISI and mitigate nonlinear distortion. The experimental results show that in the 150 Gbit / s PAM-8 optical back-to-back (OBTB) transmission, the proposed Joint Tx and Rx LUT-MLSE has superior performance. At the HD-FEC threshold, compared with Rx LUT-MLSE and Tx LUT-MLSE, it can achieve a received sensitivity improvement of 1.19 dB and 1.57 dB. When the bit rate is increased to 165 Gbit / s, the receiver sensitivity is still increased by 1.08 dB and 1.22 dB. It should be noted that Joint Tx and Rx LUT-MLSE is the only scheme that can achieve a BER lower than the HD-FEC threshold at a bit rate of 180 Gbit / s. The results of this scheme after propagating 20 km in standard single-mode fiber (SSMF) are basically the same as those of OBTB transmission, and it can achieve a significant improvement in receiver sensitivity.

[0052] Example 3

[0053] This embodiment provides a non - linear compensation system that jointly uses LUTs at the transmitter and receiver in an MLSE detection scheme, including a transmitter DSP module and a receiver DSP module;

[0054] The transmitter DSP module is used to map the bit - stream sequence into a PAM - 8 signal, then perform pre - distortion of the transmitter LUT, and then use a root - raised cosine filter to perform Nyquist shaping on the signal; after that, down - sample the shaped signal, and finally input the signal into a digital - to - analog converter;

[0055] The receiver DSP module detects the signal obtained by the optical receiver PD and sequentially performs up - sampling, matched filtering, synchronization, feed - forward equalization, post - filtering, MLSE based on the receiver lookup table, PAM - 8 signal demapping, and bit - error rate calculation;

[0056] Among them, before transmitting the signal, a training sequence is used to statistically obtain the lookup tables (LUTs) that will be used in the receiver DSP module and the transmitter DSP module.

[0057] Specifically, the generation of the transmitter DSP module lookup table (LUT) includes the following steps: First, transmit a training sequence in the system. Use a sliding window to move forward one symbol at a time in the training sequence. The content of the training sequence contained in the sliding window is denoted as: S t (k)=[S t (k - l)…S t (k)…S t (k + l)]; Use S t (k) to calculate the corresponding table index i at each moment. The sequence after being processed by the feed - forward equalizer FFE is named S e , and the amplitude difference between S e (k) and S t (k) is denoted as e(k)=S e (k)-S t (k). This value will be accumulated in the corresponding LUT value; as the sliding window moves, each time an amplitude difference is added, the corresponding LUT value will be accumulated and updated. N(i) records the update times of the corresponding LUT value; when the sliding window moves to the end of the training sequence, take the average value of the accumulated LUT value and store it to obtain the LUT of the transmitter DSP module. The generation of the receiver DSP module lookup table (LUT) includes the following steps: First, transmit a training sequence in the system. Use a sliding window to move forward one symbol at a time in the training sequence. The content of the training sequence contained in the sliding window is denoted as: S t (k)=[S t (k - l)…S t (k)…S t(k + l)); Using S t (k) calculates the corresponding table index i at each moment, and then accumulates the output S of the post-filter p-f in the corresponding table; as the sliding window moves, for each additional amplitude difference, the corresponding LUT value is accumulated and updated, and N(i) records the update times of the corresponding LUT value; when the sliding window moves to the end of the training sequence, the accumulated LUT value is averaged and stored to obtain the LUT of the receiving-end DSP module.

[0058] Among them, in the transmitting-end DSP module, when performing LUT predistortion processing, a sliding window of the same length is used to calculate the LUT index i, and the corresponding LUT value is searched according to the index i. The predistorted symbol can be written as: S p-d (k) = S(k) - LUT(i), where S(k) is the PAM-8 signal to be transmitted.

[0059] In addition, in the receiving-end DSP module, when performing MLSE processing based on the receiving-end lookup table, the branch metric (BM) is calculated using the output of the post-filter and the generated receiving-end lookup table. The formula for constructing BM is defined as: BM(m,k) = |S p-f (k) - f(S m (k))| 2 ; where S m (k) is the m-th mode generated based on the MLSE trellis diagram, and f(S m (k)) is the LUT value corresponding to S m (k); then the calculated branch metric is sent to the Viterbi algorithm module to select the surviving path. After reaching the traceback length, start to traceback one by one to find the most likely transmitted signal

[0060] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, rather than limiting the implementation manners of the present invention. For those of ordinary skill in the art, other different forms of changes or variations can be made based on the above description. It is not necessary and impossible to list all the implementation manners here. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention shall be included in the protection scope of the claims of the present invention.

Claims

1. A non - linear compensation method for jointly using a LUT at the transceiver end in an MLSE detection scheme, characterized in that, Before transmitting the signal, use the training sequence to separately calculate the lookup tables LUTs to be used at the receiving end and the transmitting end; in the transmitting-end DSP, first map the bitstream sequence into a PAM-8 signal, then perform pre-distortion of the transmitting-end LUT, and then use the root raised cosine filter to perform Nyquist shaping on the signal; after that, downsample the shaped signal, and finally input the signal into the digital-to-analog converter; In the receiving-end DSP, the optical receiver PD detects the obtained signal and successively performs upsampling, matched filtering, synchronization, forward feedback equalization, post-filtering, MLSE based on the receiving-end lookup table, PAM-8 signal demapping, and bit error rate calculation; The generation of the receiver lookup table LUT includes the following steps: First, a training sequence is transmitted in the system, and a sliding window is used to move forward symbol by symbol in the training sequence. The content of the training sequence contained in the sliding window is denoted as: S t (k)=[S t (k - l)…S t (k)…S t (k + l)]; Using S t (k), the corresponding table index i is calculated at each moment, and then the output S of the post-filter p-f is accumulated in the corresponding table; As the sliding window moves, for each additional amplitude difference, the corresponding LUT value is accumulated and updated, and N(i) records the update times of the corresponding LUT value; When the sliding window moves to the end of the training sequence, the generation of the LUT value is completed; The MLSE based on the receiver lookup table includes the following steps: The receiver lookup table replaces the dot product in the MLSE branch metric calculation. The BM is calculated using the output of the post-filter and the generated receiver lookup table. The formula for constructing the BM is defined as: BM(m,k) = |S p-f (k) - f(S m (k))| 2 ; where S m (k) is the m-th pattern generated based on the MLSE trellis diagram, and f(S m (k)) is the LUT value corresponding to S m (k); then the calculated branch metrics are sent to the Viterbi algorithm module to select the surviving path. After reaching the traceback length, the traceback is started one by one to find the most likely transmitted signal.

2. The non - linear compensation method for jointly using a LUT at the transceiver end in an MLSE detection scheme according to claim 1, characterized in that, The generation of the lookup table LUT at the transmitting end includes the following steps: First, transmit a training sequence in the system, and use a sliding window to move forward symbol by symbol in the training sequence. The content of the training sequence contained in the sliding window is denoted as: S t (k) = [S t (k - l)…S t (k)…S t (k + l)]; Use S t (k) to calculate the corresponding table index i at each moment. The sequence after being processed by the forward feedback equalizer FFE is named S e , S e (k) and S t (k) The amplitude difference between them is denoted as e(k) = S e (k) - S t (k), and this value will be accumulated in the corresponding LUT value; as the sliding window moves, for each additional amplitude difference, the corresponding LUT value will be accumulated and updated, and N(i) records the update times of the corresponding LUT value; when the sliding window moves to the end of the training sequence, the generation of the LUT value is completed.

3. The non - linear compensation method for jointly using a LUT at the transceiver end in an MLSE detection scheme according to claim 2, characterized in that, The average value is taken for both the receiving-end lookup table and the transmitting-end lookup table.

4. The non - linear compensation method for jointly using a LUT at the transceiver end in an MLSE detection scheme according to claim 2, characterized in that, The described transmitter LUT predistortion includes the following steps: calculating the LUT index i using a sliding window of the same length, searching for the corresponding LUT value according to the index i, and the predistorted symbol can be written as: S p-d (k) = S(k) - LUT(i), where S(k) is the PAM-8 signal to be transmitted.

5. A non - linear compensation system for jointly using a LUT at the transceiver end in an MLSE detection scheme, characterized in that, It includes a transmitting-end DSP module and a receiving-end DSP module; The described transmitting-end DSP module is used to map the bitstream sequence into a PAM-8 signal, then perform pre-distortion of the transmitting-end LUT, and then use the root raised cosine filter to perform Nyquist shaping on the signal; after that, downsample the shaped signal, and finally input the signal into the digital-to-analog converter; The described receiving-end DSP module detects the obtained signal through the optical receiver PD and successively performs upsampling, matched filtering, synchronization, forward feedback equalization, post-filtering, MLSE based on the receiving-end lookup table, PAM-8 signal demapping, and bit error rate calculation; Among them, before transmitting the signal, use the training sequence to calculate the lookup table LUTs to be used in the receiving-end DSP module and the transmitting-end DSP module; The generation of the lookup table (LUT) of the receiving-end DSP module includes the following steps: First, transmit a training sequence in the system, and use a sliding window to move forward symbol by symbol in the training sequence. The content of the training sequence contained in the sliding window is denoted as: S t (k)=[S t (k - l)…S t (k)…S t (k + l)]; Calculate the corresponding table index i at each moment using S t (k), and then accumulate the output S p-f of the post-filter in the corresponding table; As the sliding window moves, for each additional amplitude difference, the corresponding LUT value will be accumulated and updated, and N(i) records the update times of the corresponding LUT value; When the sliding window moves to the end of the training sequence, take the average of the accumulated LUT values and store them to obtain the LUT of the receiving-end DSP module; In the receiving - end DSP module, when performing MLSE processing based on the receiving - end lookup table, the branch metric (BM) is calculated using the output of the post - filter and the generated receiving - end lookup table. The formula for constructing BM is defined as: BM(m,k) = |S p-f (k)-f(S m (k))| 2 ; where S m (k) is the m - th pattern generated based on the MLSE trellis diagram, and f(S m (k)) is the LUT value corresponding to S m (k). Then, the calculated branch metrics are sent to the Viterbi algorithm module to select the surviving path. After reaching the traceback length, the traceback is started one by one to find the most likely transmitted signal.

6. The non - linear compensation system for jointly using a LUT at the transceiver end in an MLSE detection scheme according to claim 5, characterized in that, The generation of the lookup table LUT of the transmitting end DSP module includes the following steps: First, transmit a training sequence in the system, and use a sliding window to advance symbol by symbol in the training sequence. The content of the training sequence included in the sliding window is denoted as: S t (k) = [S t (k - l)…S t (k)…S t (k + l)]; Calculate the corresponding table index i at each moment using S t (k). The sequence after being processed by the forward feedback equalizer FFE is named S e , and the amplitude difference between S e (k) and S t (k) is denoted as e(k) = S e (k) - S t (k). This value will be accumulated in the corresponding LUT value; as the sliding window moves, each time an amplitude difference is added, the corresponding LUT value will be accumulated and updated, and N(i) records the update times of the corresponding LUT value; when the sliding window moves to the end of the training sequence, take the average value of the accumulated LUT values and store it to obtain the LUT of the transmitting end DSP module.

7. The non - linear compensation system for jointly using a LUT at the transceiver end in an MLSE detection scheme according to claim 6, characterized in that, In the transmitting - end DSP module, when performing LUT predistortion processing, a sliding window of the same length is used to calculate the LUT index i, and the corresponding LUT value is searched according to the index i. The predistorted symbol can be written as: S p-d (k)=S(k)-LUT(i), where S(k) is the PAM - 8 signal to be transmitted.

Images