Error propagation suppression method for decision feedback equalizer based on error correlation

Abstract

The present invention provides an error propagation suppression method for a decision feedback equalizer based on error correlation, and the receiving end digital signal processing includes: using a threshold detector to detect the input signal before the DFE decision, when the input signal before the DFE decision is greater than a preset threshold, the current sampling point is judged to be the end point of the burst error, and the error correction step is activated; the output signal after the DFE decision is input to a post filter, and then the post filter output signal is subtracted from the DFE input signal through an adder to obtain the DFE-induced error information, and the DFE-induced error information corresponding to the burst error end point is respectively multiplied with the response of the error correlation filter corresponding to the error events of different lengths to obtain different error correlation values; the maximum effective error correlation value is output; and finally, the error event corresponding to the maximum effective error correlation value is used to correct the burst error caused by the DFE. The present invention can effectively suppress the error propagation introduced by the DFE, significantly improve the receiver sensitivity, and improve the transmission performance of the system.

Description

Technical Field

[0001] The present invention relates to optical fiber communication technology, and in particular to an equalizer error propagation suppression technology for direct detection optical fiber communication systems. Background Art

[0002] Driven by various bandwidth-intensive services, data center networks (DCNs) continue to evolve to meet the demand for larger capacity links. For cost-sensitive data center networks, intensity modulation direct detection (IM / DD) systems are the simplest and most effective solution. The next generation of Ethernet links is expected to reach 1.6Tb / s, and single-channel over 200Gb / s is expected to reduce integration complexity. Among different modulation formats, PAM-4 signals can effectively balance bandwidth limitations and noise tolerance, and are the most promising solution for achieving single-wavelength 200Gb / s transmission. However, with the increase in transmission rates, commercial optical transceivers bring severe bandwidth limitations, resulting in severe inter-symbol interference (ISI). Therefore, in order to simultaneously meet the transmission performance and power consumption requirements of the system, a reasonable method is needed to effectively solve the problem of ISI.

[0003] At present, many advanced digital signal processing (DSP) technologies have been proposed to alleviate severe ISI in bandwidth-limited systems. The feedforward equalizer (FFE) can eliminate inter-symbol interference caused by bandwidth limitation and dispersion (CD), but FFE enhances high-frequency noise, causing the colorization of white Gaussian noise. The use of maximum likelihood sequence estimation (MLSE) and decision feedback equalizer (DFE) after FFE equalization can effectively suppress noise enhancement. However, high-baud-rate IM / DD channels have long memory lengths, which increases the computational complexity of MLSE. On the other hand, the decision feedback operation of DFE leads to error propagation, which limits the performance of DFE. Therefore, how to effectively suppress the error propagation of DFE is of great research significance for achieving high-speed optical transmission. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide an effective method for suppressing the propagation of DFE errors.

[0005] The technical solution adopted by the present invention to solve the above technical problems is an error propagation suppression method of a decision feedback equalizer based on error correlation, and the digital signal processing at the receiving end includes the steps of:

[0006] Burst error detection step: before the decision feedback equalizer DFE makes a decision, a threshold detector is used to detect the input signal before the DFE makes a decision. When the input signal before the DFE makes a decision is greater than a preset threshold, the current sampling point is determined to be the end point of the burst error, and the error correction step is activated;

[0007] Error correction steps: input the output signal after DFE judgment to the post-filter whose response is G(D)=1+αD, where α represents the filter coefficient and D represents the delay of a single symbol interval, then subtract the DFE input signal from the post-filter output signal through an adder to obtain the DFE-induced error information, multiply the DFE-induced error information corresponding to the burst error endpoint with the response of the error correlation filter corresponding to the error events of different lengths to obtain the error correlation value of the current sampling point corresponding to the error events of different lengths, and calculate the error correlation value with the largest absolute value; then determine whether the maximum error correlation value is valid, if so, output the maximum error correlation value as the maximum effective error correlation value, and use the error event corresponding to the maximum effective error correlation value to correct the burst error caused by DFE. If not, do not modify the DFE output.

[0008] Since the threshold detector can directly detect the end point of the burst error, in the error correction module, it is only necessary to multiply the error information brought by the decision feedback equalizer DFE corresponding to the position of the burst error end point with the response of the error correlation filter corresponding to the error events of different lengths to obtain the maximum error correlation value, and this method has a low computational complexity. At the same time, the present invention proposes that the common terms of the error correlation filters corresponding to the error events of different lengths only need to be calculated once, thereby further reducing the computational complexity.

[0009] The beneficial effects of the present invention are as follows: the method proposed by the present invention can effectively suppress the error propagation introduced by DFE, significantly improve the receiver sensitivity, and improve the transmission performance of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 It is a block diagram of the error propagation suppression principle of the DFE based on error correlation of the present invention;

[0011] Figure 2 Schematic diagram of the structure of an error correlation filter group with L≤3; L≤3 refers to a filter group composed of error correlation filters with three error lengths of L=1, L=2, and L=3;

[0012] Figure 3 Figure 1. Experimental setup for PAM-4 intensity modulation direct detection optical fiber communication system. DETAILED DESCRIPTION

[0013] The receiving end digital signal processing DSP implements error propagation suppression based on error correlation decision feedback equalizer EC-DFE by performing the following steps:

[0014] The received signal is equalized by the feed-forward equalizer FFE to eliminate the inter-symbol interference (ISI), and then the first post-filter PF with a response of G(D)=1+αD is used to perform noise whitening. D represents the delay of a single symbol interval, and α represents the filter coefficient. The error correlation-based DFE (EC-DFE) is then used to eliminate the ISI introduced by the PF and suppress the error propagation of the DFE. The specific process is as follows: Figure 1 shown.

[0015] EC-DFE consists of two steps:

[0016] In the first step, a threshold detector is used to detect the end point of the burst error before the DFE decision. Assuming a PAM-4 signal, the upper threshold of the threshold detector is 3+βd, and the lower threshold is -3+βd, where d is the minimum Euclidean distance of the PAM-4 signal, d=2, and β is a constant for adjusting the threshold. The performance of the system will depend on the value of β, so the value of β needs to be optimized.

[0017] In the second step, when the symbol before the DFE decision exceeds the threshold, the error correction module is activated to correct the burst error caused by the DFE.

[0018] The feedforward equalizer FFE in the DSP at the receiving end is connected to the first post-filter, and the output end of the first post-filter is connected to the decision feedback equalizer DFE and the error correction module respectively.

[0019] Taking a single-tap DFE as an example, the DFE includes a decider, an adder, a single symbol interval delay device D and a multiplier. One input end of the adder in the DFE is connected to the output end of the first post-filter, and the other input end is connected to the output end of the multiplier. One output end of the adder in the DFE is respectively connected to the input end of the decider and the input end of the threshold detector. The output end of the decider is connected to the input end of the single symbol interval delay device D, the output end of the single symbol interval delay device D is connected to the input end of the multiplier, and the output end of the decider serves as the output end of the DFE.

[0020] The error correction module includes a post filter with a response of G(D), an adder, L max The error correlation filter, maximum value calculation unit and error correction unit corresponding to the error events of different lengths are provided. One input end of the adder in the error correction module is connected to the output end of the first post filter, and the other input end is connected to the output end of the post filter in the error correction module. The input end of the post filter is connected to the output end of the DFE. The output ends of the adders in the error correction module are respectively connected to L max The error correlation filter corresponding to the error events of different lengths, L maxThe output end of each error correlation filter is connected to the input end of the maximum value calculation unit, the output end of the maximum value calculation unit is connected to one input end of the error correction unit, and the other input end of the error correction unit is connected to the output end of the DFE. The output end of the error correction unit outputs a signal after correcting the burst error caused by the DFE to the output end of the DFE.

[0021] Signal Input decision feedback equalizer DFE, x(k)*G(D) means the transmission symbol x(k) passes through the post filter with response G(D), k means the kth sampling point of the signal, Represents noise. The output signal after DFE judgment can be expressed as the transmission symbol x(k) plus the error e(k). The error correction module inputs the output signal after DFE judgment into the post-filter with a response of G(D) to obtain the signal x(k)*G(D)+e(k)*G(D), and then subtracts the input signal of DFE through the adder. Get the error information E(k) induced by DFE, When the threshold detector detects the position k of the burst error end point e When E(k e ) are multiplied with the responses of the error correlation filters corresponding to the error events of different lengths, and the error correlation value of the position can be calculated. Finally, the maximum value calculation unit calculates the maximum effective error correlation value, and the error correction unit uses the error event corresponding to the maximum effective error correlation value to correct the burst error caused by DFE.

[0022] The error correlation filter is composed of the error event and the response G(D -1 )=α+D. In a bandwidth-limited system, DFE causes burst errors to alternate between -2 and 2. Therefore, the possible error events for a length of L are {+2,–2,+2,…} or {–2,+2,–2,…}. Since the error events in the two cases differ only by a positive or negative sign, we only need to use an error correlation filter for the error event of length L, and its expression is

[0023] Figure 2 3 is a schematic diagram of the structure of the error correlation filter group L≤3, L≤3 includes a filter group composed of error correlation filters of three error lengths L=1L=2L=3. max To represent the maximum error length considered. Assuming that we only consider error events with a length of less than 3, then L max The maximum length of the error event is L max =3. maxThe same is true for the error correlation filter group corresponding to error events of different lengths greater than 3. As can be seen from the figure, the error correlation filter expressions corresponding to error events of different lengths contain common terms, so the common terms only need to be calculated once, which can avoid repeated calculations and reduce computational complexity.

[0024] When the error event corresponding to the obtained maximum error correlation value is unreasonable, that is, the DFE symbol corrected according to the error event exceeds the maximum or minimum level value, we believe that this maximum error correlation value is invalid. At this time, there is no need to modify the output of DFE. On the contrary, the error event corresponding to the maximum effective error correlation value is used to correct the burst error caused by DFE. That is, the method for judging whether the maximum error correlation value is valid is: the DFE symbol corrected by the error event corresponding to the obtained maximum error correlation value is valid within the effective level range.

[0025] The embodiment takes a four-level pulse amplitude modulation PAM-4 direct detection optical fiber communication system as an example to illustrate the specific working principle of the present invention. Figure 3 shown.

[0026] In the digital signal processing at the transmitting end, the original data (binary bit stream) is first mapped into a PAM-4 signal and up-sampled. Then, a root raised cosine filter is used to perform Nyquist shaping on the signal. The shaped signal is resampled and input into the digital-to-analog converter DAC for digital-to-analog conversion to obtain an analog electrical signal. Then, the electrical signal is amplified by an electrical amplifier EA and input into the Mach-Zehnder MZM modulator together with the laser output by the laser for modulation to obtain the transmitted optical signal.

[0027] After the transmitted signal has been transmitted over a certain distance of optical fiber, a variable optical attenuator (VOA) is used at the receiving end to adjust the received optical power (ROP). The attenuated optical signal is converted into electrical and optical signals through a photodiode (PD). The electrical signal received by the PD is collected by an oscilloscope and then processed by the offline receiving end digital signal processing (DSP). This process mainly includes resampling, matched filtering, downsampling, feedforward equalization, post-filtering, EC-DFE, PAM-4 demapping, and bit error rate calculation.

[0028] In the offline receiving end digital signal processing DSP, the received data is equalized by the feedforward equalizer FFE to eliminate ISI, and the signal after FFE equalization is subjected to noise whitening processing by the post filter PF. Then, the error-correlation-based DFE is used to eliminate the ISI introduced by the post filter, and a threshold detector is introduced before the DFE decision to detect the end point of the burst error. When the input symbol does not exceed the threshold, the DFE directly outputs the result after the decision. Once a symbol exceeding the threshold is detected, the error correction module is activated to correct the burst error caused by the DFE. The error information E(k) induced by the DFE can be obtained by subtracting the input signal of the DFE from the signal obtained by convolving the DFE decision output and G(D). The end point position of the burst error detected by the threshold detector is denoted by k. e Indicates that E(k e ) and the response of the error correlation filter corresponding to the error events of different lengths can be multiplied to calculate the error correlation value corresponding to the end position of the burst error. Then, the maximum effective error correlation value is calculated, and its corresponding error event is used to correct the burst error caused by DFE. Finally, the corrected signal is demapped into a binary bit stream through PAM-4, and the bit error rate is calculated by comparing it with the transmitted data.

[0029] This invention only uses the PAM-4 intensity modulation direct detection optical fiber communication system as an example to illustrate the main technical concept and features of the invention, but it is not a limitation on the implementation methods of the invention. Any modifications, equivalent transformations and improvements made in this application should be included in the protection scope of the claims of the invention.

Claims

1. A method for suppressing error propagation of a decision feedback equalizer based on error correlation, characterized in that: The receiving end digital signal processing includes the following steps: Burst error detection step: before the decision feedback equalizer DFE makes a decision, a threshold detector is used to detect the input signal before the DFE makes a decision. When the input signal before the DFE makes a decision is greater than a preset threshold, the current sampling point is determined to be the end point of the burst error, and the error correction step is activated; Error correction steps: input the output signal after DFE judgment to a post-filter whose response is G(D)=1+αD, where α represents the filter coefficient and D represents the delay of a single symbol interval, and then subtract the DFE input signal from the post-filter output signal through an adder to obtain the DFE-induced error information, and multiply the DFE-induced error information corresponding to the burst error endpoint with the response of the error correlation filter corresponding to the error events of different lengths to obtain the error correlation value of the current sampling point corresponding to the error events of different lengths, and calculate the error correlation value with the largest absolute value; Then determine whether the maximum error correlation value is valid. If so, output the maximum error correlation value as the maximum effective error correlation value, and use the error event corresponding to the maximum effective error correlation value to correct the burst error caused by DFE. If not, do not modify the output of DFE.

2. The method according to claim 1, characterized in that: The response of the error correlation filter corresponding to error events of different lengths is expressed as: L represents the burst error length, and k represents the length variable.

3. The method according to claim 1, characterized in that: The upper threshold of the threshold detector is (M-1)+βd, and the lower threshold is -(M–1)+βd; Where M is the modulation order, d is the minimum Euclidean distance of the input signal, and β is a constant for adjusting the threshold.

4. The method according to claim 3, characterized in that: When the input signal is a PAM-4 signal, that is, M=4, the threshold detector takes d=2.

5. The method according to claim 1, characterized in that: When calculating the responses of the error correlation filters corresponding to error events of different lengths in the error correction step, the common terms between the expressions of the responses of the error correlation filters corresponding to error events of different lengths only need to be calculated once.

6. The method according to claim 1, characterized in that: The method for judging whether the maximum error correlation value is valid is that the DFE symbol after the error event correction corresponding to the obtained maximum error correlation value is valid within the valid level range.

Images